An introduction to renormalization
\(\phi^4\)-modelasymptotic freedimensional renormalizationFeynman graphforestperturbative expansionperturbative renormalizationrenormalization groupSchwinger function
Gamma, beta and polygamma functions (33B15) Research exposition (monographs, survey articles) pertaining to quantum theory (81-02) Model quantum field theories (81T10) Perturbative methods of renormalization applied to problems in quantum field theory (81T15) Nonperturbative methods of renormalization applied to problems in quantum field theory (81T16) Renormalization group methods applied to problems in quantum field theory (81T17) Statistical mechanics of polymers (82D60)
Reviews on perturbative renormalization (Sec. 2) and renormalization group (Sec. 3) are presented. The main problem of perturbative renormalization is how to give counterterms to diverging terms of the perturbative expansions of Schwinger functions (Green functions or time ordered vacuum expectation values, continued to Euclidean points) so that expansions give finite values. In Sect. 2, the \(\phi^4_d\)-model is taken as the example, perturbative renormalization is explained. The Schwinger functions of the \(\phi^4\)-model are the normalized moments \[ S_N(z_1,\dots, z_N)+ \int\phi(z_1)\cdots \phi(z_N)\,d\nu(\phi),\quad d\nu={1\over Z} e^{-{g\over 4!}\int\phi^4(x)\,dx} d\mu(\phi), \] where \(Z\) is the normalization factor and \(d\mu(\phi)\) is the Lebesgue measure at \(\phi\). The formal expansion of \(S_N\) with respect to the coupling constant \(g\) is \[ S_N(z_1,\dots, z_N)= {1\over Z} \sum^\infty_{n=0} {(-g)^n\over n!} \int \Biggl[{\phi^4(x)\,dx\over 4!}\Biggr]^n \phi(z_1)\cdots \phi(z_N)\,d\mu(\phi), \] (it is well known that we can expect only Borel summability for this expansion). Exact calculation of the order \(n\) term of this expansion is performed using Feynman graph as follows: \[ \begin{gathered} ZS_N= \sum_{\phi^4\text{ graphs }G\text{ with }N(G)= N} {(-g)^{n(G)}\over S(G)} A_G,\\ A_G(z_1,\dots, z_N)= \int \prod^n_{i=1} dx_i \prod_{l\in G} C(x_i, y_l),\end{gathered} \] where \(N(G)\) is the number of external vertices of \(G\), \(S(G)\) is the symmetry factor of the diagram and \(C(x,y)= C(x- y)\) is the translation invariant propagator whose Fourier transform is \(C(p)= {1\over (2\pi)^d} {1\over p^2+ m^2}\). Then using ``Feynman parametric representation, for the graph \(G_0\) which involves only one total external moment \(q\), the divergence term is shown to have the form \[ \begin{aligned} A^R_{G_0}(q) &= \int d^4p{1\over (p^2+ m^2)((p+ q)^2+ m^2)},\\ &= \int^\infty_0 \int^\infty_0 {d\alpha_1d\alpha_2\over (\alpha_1+ \alpha_2)^2}\,e^{-(\alpha_1+ \alpha_2)m^2- {\alpha_1\alpha_2\over \alpha+1+ \alpha_2} q^2},\end{aligned} \] (cf. \textit{N. Nakanishi}, Graph theory and Feynman integrals (Mathematics and its Applications 11, Gordon and Breach Science Publishers, New York-London-Paris) (1971; Zbl 0212.29203)) which diverges at large \(p\) or small \(\alpha\) and is called the ultraviolet divergence. In the case of the graph \(G_0\), the renormalization amplitude is \[ \begin{aligned} A^R_{G_0} &= \int\Biggl[{1\over (p^2+ m^2)((p+ q)^2+ m^2)}- {1\over (p^2+ m^2)^2}\Biggr]\,d^4p\\ &= \int^\infty_0 \int^\infty_0 {d\alpha_1 d\alpha_2 e^{-(\alpha_1+ \alpha_2)m^2}\over (\alpha_1+ \alpha_2)^2} [e^{-{\alpha_1\alpha_2\over \alpha_1+ \alpha_2}q^2}- 1].\end{aligned} \] The \(d= 2,\,3\), \(\phi^4\)-models have only a finite number of ``primitively divergent subgraphs which are called superrenormalizable. In \(d=4\), degree of divergence only depends on the number of external legs and called renormalizable. For \(d> 3\), we have infinitely many primitive divergent graphs with arbitrary large number of external legs, and the theory is called nonrenormalizable (Subsec. 2.4). In the renormalizable case, we need to answer how to introduce a proper set of counterterms which are local and remove all the ultraviolet divergence of every graph, not only the main global primitive divergences but also all the divergences associated to subgraphs. The answer is given using the forest; a subset of subgraphs such that for any pair \(S_1\), \(S_2\) of the forest, either \(S_1\subset S_2\) or \(S_2\subset S_1\) or \(S_1\) and \(S_2\) are disjoint, by Bogoliubov and Parasiuk, Hepp, and Zimmermann [\textit{W. Zimmermann}, Commun. Math. Phys. 15, 208--234 (1969; Zbl 0212.29203)]. Later this method is reinterpreted from the viewpoint of Hopf algebra [\textit{A. Connes} and \textit{D. Kreimer}, Commun. Math. Phys. 199, No. 1, 203--242 (1998; Zbl 0192.61203)]. These are reviewed in Subsec. 2.5. In Subsec. 2.6, the dimensional renormalization, which was used to show the renormalizability of non-Abelian gauge theory, is stated [\textit{G. 't Hooft} and \textit{M. Kreimer}, Combinatorics of gauge fields, Nuclear Phys., B 50, 318--353 (1972)] and it is remarked that this is at the core of the Riemann-Hilbert interpretation of renormalization [\textit{A. Connes} and \textit{D. Kreimer}, Commun. Math. Phys. 210, No. 1, 249--273 (2000; Zbl 1032.81026) and ibid. 216, No. 1, 215--241 (2001; Zbl 1042.81059)]. In Subsec. 2.7, the complication of the forest formula because it becomes unreasonably large in the size of renormalized amplitudes when the size of the graph increases (the renormalon problem), are pointed out as the difficulties of perturbative renormalization. Models considered in this section are taken from quantum field theory, but the author remarks there is a deep analogy between the Feynman-Kac formula and correlation functions in classical statistical mechanics (Subsec. 2.1). The \(\alpha\)-representation of \(A_{G_0}(q)\) has an interpretation in terms of Brownian motion, which leads to the relation between a polymer model and a \([(\phi)^2]^2\) field theory with \(O(N)\) symmetry in the \(N\to 0\) limit. This is remarked in Subsec. 2.3. The renormalization group provides a method to overcome the difficulties of perturbative renormalization. For a \(\phi^4\)-model, it slices the propagator as follows: \[ C_i= \int^{M^{-2(i-1)}}_{M^{-2i}} e^{-m^2\alpha- {| x-y|^2\over 4\alpha}}{d\alpha\over \alpha^{d/2}},\;i\geq 1,\quad C_0= \int^\infty_1 e^{-m^2\alpha- {| x-y|^2\over 4\alpha}}{d\alpha\over \alpha^{d/2}}. \] The full propagator with ultraviolet cutoff \(M^\rho\) is \(C_{\leq\rho}= \sum^\rho_{i=0} C_i\). Decomposition of a covariance into a Gaussian process corresponds to a decomposition of the field into independent Gaussian random variables \(\phi^i\), each distributed with a measure \(d\mu_i\) of covariance \(C_i\). We set \(\Phi_i= \sum^i_{j=0} \phi_j\). Then, if \[ \begin{aligned} S_{i-1}(\Phi_{i-1}) &= -\log[Z_{i-1}(\Phi_{i-1})],\quad\text{i.e., }Z_{i-1}= e^{-S_{i-1}},\\ Z_{i-1}(\Phi_{i-1}) &= \int d\mu_i(\phi_i) e^{-S(\phi_i+ \Phi_{i-1})},\end{aligned} \] we can iterate. This is the renormalization group method (Subsec. 3.1). This step does not render polynomial action, even when it is started from a polynomial action, and there are efforts to overcome this difficulty (Subsec. 3.4). On the other hand, the coupling constants \(g_i\) and \(g_{i-1}\) relate \(g_{i-1}\simeq g_i- \beta g^2_i\), \[ \lim_{i\to \infty}\, \sum_{(j,j')|\text{inf}(j, j')= i} \int d^4 yC_j(x, y)C_{j'}(x,y)\simeq\beta> 0. \] Therefore renormalized coupling constants seem to diverge. But in non-Abelian gauge, \(g_i\simeq g_{i-1}- \beta g_i g_{i-1}\), and precise computation leads to \[ g_i\simeq {g_0\over 1+ g_0(\beta_i+ \gamma\log i+ O(1))}. \] Hence the effective coupling tends to \(0\) with the cutoff for a finite fixed small renormalized coupling and is called asymptotically free (Subsect. 3.3). The paper finally states recent progress of the renormalization group in the study of interacting Fermion models [\textit{J. Feldman}, \textit{H. Knörrer} and \textit{E. Trubowitz}, A two-dimensional Fermi liquid. I: Overview, II: Convergence, III: The Fermi surface, Commun. Math. Phys. 247, No. 1, 1--47, 49--111, 113--177 (2004); Particle -- hole ladders, Commun. Math. Phys. 247, No. 1, 179--194 (2004); Convergence of perturbation expansions in Fermionic models. I: Nonperturbative bounds, Commun. Math. Phys. 247, No. 1, 195--242 (2004)]. Topics related to the renormalization group equation are not explained in this lecture [\textit{S.-I. Ei}, \textit{K. Fujii} and \textit{T. Kunihiro}, Ann. Phys. 280, No. 2, 236--298 (2000; Zbl 1052.81073)] and references therein. For the entire collection see [Zbl 1012.00029].
- Running coupling expansion for the renormalized \(\phi_4^4\)-trajectory form renormalization invariance
- Summation and renormalization of bubble graphs to all orders
- Renormalizing an initial state
- Perturbative and constructive renormalization: An overview
- Constructive renormalization theory
- Causality and the perturbative renormalization group
- On renormalization in nonrelativistic interaction models with infinities
- Natural renormalization
- Exact renormalization group
- Introduction to a Renormalisation Group Method
- Some aspects of the non-perturbative renormalization of the \(\varphi ^{4}\) model
- Renormalization
- Renormalization in spherical field theory
- Approximate differential equations for renormalization group functions in models free of vertex divergencies
- How to take particle physics seriously: a further defence of axiomatic quantum field theory
- Proof of the Bogoliubov-Parasiuk theorem on renormalization
- A renormalization group proof of perturbative renormalizability
This page was built for publication: An introduction to renormalization
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4461788)
