Divergences in quantum field theory on the noncommutative two-dimensional Minkowski space with Grosse-Wulkenhaar potential (Q540003)

Adding a harmonic oscillator potential, Grosse-Wulkenhaar showed noncommutative \(\phi^4\)-models on two- and four-dimensional Euclidean spaces to be renormalisable [\textit{H. Grosse} and \textit{R. Wulkenhaar}, Commun. Math. Phys. 256, No. 2, 305--374 (2005; Zbl 1075.82005)]. In this paper, the Grosse-Wulkenhaar model on the noncommutative two-dimensional Minkowski space is discussed, and it is shown that there is a new type of divergence in planar graphs. The author argues that this new kind divergence might make the construction of a Minkowski space version of the Grosse-Wulkenhaar model impossible. Let \([x^\mu, x^\nu]= i\theta^{\mu\nu}= i\lambda^2_{nc} \varepsilon^{\mu\nu}\), \(\varepsilon= \left(\begin{smallmatrix} 0 & 1\\ -1 & 0\end{smallmatrix}\right)\) be the commutation relation. Then the Grosse-Wulkenhaar potential is introduced by \[ D^{\pm}_\mu=- i\partial_\mu\mp 2\Omega\theta^{-1}_{\mu\nu} x^\nu= -i\partial_\mu\pm 2\lambda^{-2} \varepsilon_{\mu\nu} x^\nu, \quad \lambda= \Omega^{-{1\over 2}}\lambda_{nc}. \] The choice \(\Omega= 1\), i.e., \(\lambda= \lambda_{nc}\), corresponds to the self-dual point. The wave equation for a scalar field \(\phi\) of mass \(\mu\) in the quadratic potential is \[ \big(-\tfrac12(D^+_\mu D^{+\mu}+ D^-_\mu D^{-\mu})+ \mu^2\big) \phi= (\partial_\mu \partial^\mu+ 4\lambda^{-4} x_\mu x^\mu+ \mu^2)\phi \qquad(\S2). \] Due to the lack of translation invariance, the definition of a Feynman propagator depends on the quantum state. To avoid this ambiguity, the retarded propagator is considered. In the massless case, the wave operator for the Grosse-Wulkenhaar portential is \(4\partial_u \partial_v+ 4\lambda^{-4} uv\), \(u= x_0- x_1\), \(v= x_0+ x_1\). Its retarded propagator is \[ \Delta_{\text{ret}}(u_s, v_s, u_t, v_t)= \tfrac12 H(u_t) H(v_t) J_0(\lambda^{-2} \sqrt{u_tu_sv_tv_s}), \] \(u_s= u_1+ u_2\), \(u_t= u_1- u_2\) and \(H\) is the Heaviside function (Propositions 5.1, 5.4, 5.5). In Appendix B, the Fourier transform of this propagator is expressed in terms of the Bessel function \(K_0\). Since \[ |J_0(\lambda^{-2}\sqrt{u_tu_sv_tv_s})|\leq Ce^{{1\over 3\lambda^2}(u^2_t+ u^2_s+ v^2_t+ v^2_s)}, \] the retarded propagator is well-defined as a generalized function on the Gel'fand-Shilov space \({\mathcal S}_{\alpha, A}\), \(\alpha= {1\over 2}\), \(A= {\sqrt{2}(\lambda- \varepsilon)\over\sqrt{e}}\), \(\varepsilon\) is arbitrarily small (cf. [\textit{I. M. Gel'fand} and \textit{G. E. Shilov}, Generalized functions. Vol. I: Properties and operations. New York and London: Academic Press (1964; Zbl 0115.33101)]; for the use of Gel'fand-Shilov space in noncommutative field theory, see [\textit{M. Chaichian}, \textit{M. Mnatsakanova}, \textit{A. Tureanu} and \textit{Yu. Vernov}, ``Test functions space in noncommutative quantum field theory'', J. High Energy Phys. 2008, No. 9, 125 (2008)]). Since the planar fish graph is given by \(\Delta_F(x,y)*_x\overline*_y \Delta_F(x,y)\), a computation of \[ \Delta_{\text{ret}}(x, y)*_x\overline *_y \Delta_{\text{ret}}(x,y)\tag{1} \] is tried in \S6. This product is not well-defined in the sense of Hörmander's product of distributions [\textit{L. Hörmander}, The analysis of linear partial differential operators. I. Distribution theory and Fourier analysis. 2nd ed. Berlin etc.: Springer-Verlag (1990; Zbl 0712.35001)]. So the definition of \(*\)-product by duality is used; cf. [\textit{M. A. Soloviev}, Theor. Math. Phys. 153, No. 1, 1351--1363 (2007); translation from Teor. Mat. Fiz. 153, No. 1, 3--17 (2007; Zbl 1136.81408)]. It is shown that, if \(\sqrt{e}\lambda_{nc}< (\lambda-\varepsilon)\), then there exists a nontrivial subset \({\mathcal S}\) of \({\mathcal S}_{\alpha,A}(\mathbb{R}^4)\) such that the above equation (1) is defined as a generalized function on \({\mathcal S}\), but if \(\lambda_{nc}\geq \lambda\), then such \({\mathcal S}\) does not exist (Propositions 6.1 and 6.2). The author remarks that whether we can get rid the factor \(\sqrt{e}\) is an interesting problem). This is the main result of this paper but seems rather abstract. To give concrete understanding, precise calculation of equation (1) by using Proposition 5.1, are also performed and show the divergence of above \(*\)-product is dominated by \[ \sum^\infty_{m=0} \Omega^{4m}= {1\over 1-\Omega^4}, \] i.e., if \(\Omega= 1-\varepsilon\), then we find a divergence \(\varepsilon^{-1}\) as \(\varepsilon\to 0\). This is not the problem, but the loop integral over the momenta is divergent. So, as the author concludes, this is no ordinary UV divergence. In actual loop calculation, equation (1) is not complete. A supplement on this point is given in the rest of \S6. Then it is remarked that the fish loop calculated in this section also occurs in the four-point function of the \(\phi^4\) model, so the problem is not specific to the \(\phi^3\) model. Before giving these direct discussion in position space, assuming \(\lambda= \lambda_{nc}\), a discussion of generalized eigenfunctions of the wave equation of the Grosse-Wulkenhaar model and its quantization in terms of the eigenfunctions is given in \S3 and \S4 and the same conclusion is obtained. If \(\lambda= \lambda_{nc}\), we have \[ H*f= {\lambda^2\over 2} D_\mu^- D^{-\mu} f,\qquad f* H= {\lambda^2\over 2} D^+_\mu D^{+\mu} f, \] where \(H= -{2\over \lambda^2} x_\mu x^\mu\). By using these expressions, generalized eigenfunctions of the wave equation are computed in terms of Kummer's confluent hypergeometric functions (for \S3 cf. [\textit{A. Fischer} and \textit{R. J. Szabo}, ``Duality covariant quantum field theory on noncommutative Minkowski space'', J. High Energy Phys. 2009, No. 2, 031 (2009)]). Adopting these computations, the fish graph in the \(\phi^3\) model is computed. The loop integral of its planar graph diverges due to the presence of the square of a \(\delta\) distribution. While the imaginary part of the loop integral of nonplanar graph contributes to violate the unitarity (\S4). Hence we have same conclusion as in \S6. In \S7, the last section, the author tells us how to achieve more detailed studies on the above divergence; to represent the retarded propagator in terms of the generalized eigenfunctions of the wave equation (\S3) should be useful. In Appendix A, a connection to the matrix model setting of [Fischer and Szabo, loc. cit.] is clarified. In this paper, noncommutative fields are considered to be generalized functions on the Gel'fand-Shilov space, whereas in [Fischer and Szabo, loc. cit.] noncommutative fields are considered to be elements of Gel'fand-Shilov space. After investigating the difference of these two arguments, a counterexample to Theorem 4.2 of [Fischer and Szabo, loc. cit.] is presented in this appendix, too.