Dynamical group approach to conformal field theory G. A. Kerimov Abstract 1 Introduction The key role of symmetry in modern physics is widely recognized. In the late 1990s, we proposed a group-theoretical approach [ 1 ] to describe S -matrices of quantum mechanical systems with dynamical symmetries. The S -matrices for all these problems are related to the intertwining operators between equivalent representations of the underlining dynamical groups ( see, e.g. [ 2 ] , [ 3 ] ). This approach naturally generalizes to quantum field theories with dynamical symmetry. It can be done by using the Heisenberg picture in the definition of the S -matrix. In the Heisenberg picture [ 4 ] , [ 5 ] the S -matrix is given by S_{\alpha_{\beta}}=\left(\Psi_{\alpha}^{-},\Psi_{\beta}^{+}\right)=\left(\Psi_{\alpha}^{-},S\Psi_{\beta}^{-}\right) (1.1) where \Psi_{\beta}^{+} and \Psi_{\alpha}^{-} are the time-independent ‘in’ and ‘out’ states, and Greek letters \alpha and \beta are abbreviations indicating the particle content. The Hilbert spaces \mathcal{H}^{in} and \mathcal{H}^{out} spanned by the ‘in’ and ‘out’ states are Fock spaces. Since \mathcal{H}^{in}=\mathcal{H}^{out} , the Heisenberg S -matrix operator S is a unitary operator between \mathcal{H}^{in} and \mathcal{H}^{out} S\colon\quad\mathcal{H}^{out}\longrightarrow\mathcal{H}^{in} (1.2) Let G be a dynamical group containing the Poincaré group \mathcal{P} , the kinematical group of Minkowski spacetime. By analogy with the non-relativistic case, we assume that the representations of the group G acting on the spaces \mathcal{H}^{in} and \mathcal{H}^{out} are not the same. Namely, the unitary representations T^{in}\left(g\right) and T^{out}\left(g\right) of the group G , acting in the spaces \mathcal{H}^{in} and \mathcal{H}^{out} , are induced by equivalent irreducible representations of the group G . Thus, the S -matrix is an intertwining operator between the representations T^{in}\left(g\right) and T^{out}\left(g\right) of the dynamical group G ST^{out}\left(g\right)=T^{in}\left(g\right)S,\quad\forall g\in G (1.3) or SdT^{out}\left(X\right)=dT^{in}\left(X\right)S,\quad\forall X\in\mathfrak{g,} (1.4) where dT^{in} and dT^{out} are the corresponding representations of the algebra \mathfrak{g} of G . It should be emphasized that [ 4 ] the same representation U\left(\Lambda,\mathbf{a}\right) of the Poincaré group \mathcal{P} acts on both spaces \mathcal{H}^{in} and \mathcal{H}^{out} . Hence, SU\left(\Lambda,\mathbf{a}\right)=U\left(\Lambda,\mathbf{a}\right)S,\quad\left(\Lambda,\mathbf{a}\right)\in\mathcal{P} (1.5) for arbitrary Lorentz transformations \Lambda and translations \mathbf{a} . This means that the S -matrix is Poincare invariant. In this paper we discuss the application of this approach to Euclidean quantum field theories whose dynamics are governed by the Euclidean conformal group. The objects considered in Euclidean quantum field theory [ 6 ] are not scattering amplitudes, but the Schwinger functions or Euclidean Green’s functions. As we shall see, the kernel of the intertwining operator is related to the Schwinger function. In particular, the 2-point Schwinger function is the kernel of the intertwining operator between equivalent irreducible representations of the Euclidean conformal group, as already pointed out by Koller [ 7 ] . The remaining of this paper is arranged as follows. In Sec. 1 we present basic facts about the Euclidean conformal group and notations necessary for subsequent sections. In Sec. 2 we consider d-dimensional Euclidean conformal field theory for scalar bosons. The intertwining property of S leads to powerful constraints on the n-point Schwinger functions, which allows the definition of two- and three-point functions. Moreover, the Clebsch-Gordan decomposition for the Euclidean conformal group can be used to define n-point Schwinger functions in terms of Clebsch-Gordan kernels. For simplicity of description, in Sec. 3 we consider only four-point functions in two dimensions. The appendix presents Naimark’s [ 8 ] results on the decomposition of tensor products of scalar principal series representations, which are necessary for evaluating four-point functions. 2 The group SO_{0}\left(d+1,1\right) and its Class 1 representations In this section we give, mainly to establish notation, a brief review of class 1 (scalar) representations [ 9 ] of the group SO_{0}\left(d+1,1\right) . 2.1 The group SO_{0}\left(d+1,1\right) and its subgroups Let R^{d+1,1} be a \left(d+2\right) - dimensional real linear space with bilinear form \left[\xi,\eta\right]=-\xi_{1}\eta_{1}-\xi_{2}\eta_{2}-...-\xi_{d+1}\eta_{d+1}+\xi_{d+2}\eta_{d+2}. (2.1) The upper sheet of the cone \left[\xi,\xi\right]=0,\,\xi_{d+2}>0 will be denoted by C_{+}^{d+1} . By SO_{0}\left(d+1,1\right) we denote the group of all linear transformations g on R^{d+1,1} that preserve the bilinear form \left[\xi,\eta\right] and for which \det g=1,\qquad g_{d+2\;d+2}\geq 1. (2.2) The Euclidean conformal group is isomorphic to SO_{0}\left(d+1,1\right) . Denoting by N,\overline{N},A,M the subgroups of the group SO_{0}\left(d+1,1\right) , consisting, respectively, of the matrices \displaystyle n\left(\mathbf{b}\right) \displaystyle=\left(\begin{array}[c]{ccc}I_{d}&-\mathbf{b}^{t}&\mathbf{b}^{t}\\ \mathbf{b}&1-\frac{\mathbf{b}^{2}}{2}&\frac{\mathbf{b}^{2}}{2}\\ \mathbf{b}&-\frac{\mathbf{b}^{2}}{2}&1+\frac{\mathbf{b}^{2}}{2}\end{array}\right)\,,\quad\overline{n}\left(\mathbf{c}\right)=\left(\begin{array}[c]{ccc}I_{d}&\mathbf{c}^{t}&\mathbf{c}^{t}\\ -\mathbf{c}&1-\frac{\mathbf{c}^{2}}{2}&-\frac{\mathbf{c}^{2}}{2}\\ \mathbf{c}&\frac{\mathbf{c}^{2}}{2}&1+\frac{\mathbf{c}^{2}}{2}\end{array}\right),\, (2.3) \displaystyle a_{\lambda} \displaystyle=\left(\begin{array}[c]{ccc}I_{d}&0&0\\ 0&\frac{\lambda^{2}+1}{2\lambda}&\frac{\lambda^{2}-1}{2\lambda}\\ 0&\frac{\lambda^{2}-1}{2\lambda}&\frac{\lambda^{2}+1}{2\lambda}\end{array}\,\right),\qquad m=\left(\begin{array}[c]{ccc}m&0&0\\ 0&1&0\\ 0&0&1\end{array}\right),\,m\in SO\left(d\right),\quad we have SO_{0}\left(d+1,1\right)=NAM\overline{N} (2.4) (the equality holds almost everywhere). Here I_{d} is a d -dimensional identity matrix, \mathbf{b,c} are d -dimensional row vectors, and \mathbf{b}^{t} , \mathbf{c}^{t} are the corresponding column vectors. We denote for simplicity an element of M and the corresponding element of SO\left(d\right) by the same letter. The decomposition ( 2.4 ) is called the Gauss decomposition of SO_{0}\left(d+1,1\right) . It is worth noting that A is a dilation subgroup, the subgroup \overline{N} consists of special conformal transformations, and the subgroup MN=NM is isomorphic to the inhomogeneous rotation group ISO(d) (Euclidean group). The isomorphism MN\cong ISO(d) is given by mn\left(\mathbf{b}\right)\,\longrightarrow\,\left(m,\mathbf{b}\right) (2.5) where \left(m,\mathbf{b}\right) is an element of ISO(d) . 2.2 Class 1 representations of SO_{0}\left(d+1,1\right) Class 1 representations T^{\sigma}\left(g\right) of the group SO_{0}\left(d+1,1\right) can be realized in the space of smooth functions on C_{+}^{d+1} such that F\left(\omega\xi\right)=\omega^{\sigma}F\left(\xi\right),\quad\omega>0,\quad\xi\in C_{+}^{d+1}, (2.6) where \sigma is an arbitrary complex number. This representations are given by \left[T^{\sigma}\left(g\right)F\right]\left(\xi\right)=F\left(g^{-1}\xi\right). (2.7) where g\in SO_{0}\left(d+1,1\right) . Almost every \xi\in C_{+}^{d+1} can be represented as \xi=\omega\zeta,\quad\zeta=\left(\mathbf{x,}\frac{1}{2}\left(-1+\mathbf{x}^{2}\right),\frac{1}{2}\left(1+\mathbf{x}^{2}\right)\right), (2.8) where \omega>0 , \mathbf{x}\in R^{d} . Consequently, the class 1 representations of SO_{0}\left(d+1,1\right) can be realized on the space \mathcal{C}^{\infty}\left(R^{d}\right) of smooth functions on R^{d} . In this realization the representations T^{\sigma}\left(g\right) of SO_{0}\left(d+1,1\right) are given by \left[T^{\sigma}\left(g\right)f\right]\left(\mathbf{x}\right)=\omega_{g}\left(\mathbf{x}\right)^{\sigma}f\left(\mathbf{x}_{g}\right), (2.9) where \omega_{g}\left(\mathbf{x}\right) and \mathbf{x}_{g} are defined from g^{-1}\zeta=\omega_{g}\left(\mathbf{x}\right)\zeta_{g}, (2.10) with \omega_{g}\left(\mathbf{x}\right)=\omega\left(\mathbf{x,}g\right),\quad\zeta_{g}=\left(\mathbf{x}_{g}\mathbf{,}\frac{1}{2}\left(-1+\mathbf{x}_{g}^{2}\right),\frac{1}{2}\left(1+\mathbf{x}_{g}^{2}\right)\right). (2.11) Below we give the expressions of \omega_{g}\left(\mathbf{x}\right) and \mathbf{x}_{g} for subgroups in ( 2.4 ) : 1. Translation subgroup N\ni g=n\left(\mathbf{b}\right) \omega_{g}\left(\mathbf{x}\right)=1,\quad\mathbf{x}_{g}=\mathbf{x}-\mathbf{b.} (2.12) 2. Dilatation subgroup. A\ni g=a\left(\lambda\right) \omega_{g}\left(\mathbf{x}\right)=\lambda,\quad\mathbf{x}_{g}=\frac{\mathbf{x}}{\lambda}. (2.13) 3. Rotation subgroup. M\ni g=m \omega_{g}\left(\mathbf{x}\right)=1,\quad\mathbf{x}_{g}=m^{-1}\mathbf{x.} (2.14) 4. The subgroup of special conformal transformations \overline{N}\ni g=\overline{n}\left(\mathbf{c}\right) \omega_{g}\left(\mathbf{x}\right)=1+\mathbf{c}^{2}\mathbf{x}^{2}-2\mathbf{cx},\quad\mathbf{x}_{g}=\frac{\mathbf{x-cx}^{2}}{1+\mathbf{c}^{2}\mathbf{x}^{2}-2\mathbf{cx}}. (2.15) The representation T^{\sigma}\left(g\right) of SO_{0}\left(d+1,1\right) is irreducible if \sigma is a non-integer or \sigma is an integer from the interval -d-1\leq\sigma<0 . The irreducible representations T^{\sigma}\left(g\right) can be extended to unitary ( irreducible ) representations of SO_{0}\left(d+1,1\right) for the following values of \sigma : 1. If \sigma=-\frac{d}{2}+i\rho, \rho\in R (principal series). 2. If -d<\sigma<0 (complementary series). 3. If \sigma=-d-p with p\in Z_{+} (discrete series). We restrict the discussion to the principal series of class 1 unitary irreducible representations of SO_{0}\left(d+1,1\right) , that are relevant to our purpose. The principal series representations T^{\sigma}\left(g\right),\sigma=-\frac{d}{2}+i\rho act on the space L^{2}(R^{d}) of square-integrable functions on R^{d} . The representations T^{\sigma} and T^{\widetilde{\sigma}} , where \widetilde{\sigma}=-d-\sigma , are equivalent. The principal series representation T^{\sigma} can also be realized in the space L^{2}(R^{d}) of square-integrable functions on the d -dimensional momentum space R^{d} . This realization is obtained by passing from functions f\left(\mathbf{x}\right) to their Fourier transform \varphi\left(\mathbf{p}\right)=\int_{R^{d}}e^{i\mathbf{px}}f\left(\mathbf{x}\right)d\mathbf{x.} (2.16) We find that \left(T^{\sigma}\left(g\right)\varphi\right)\left(\mathbf{p}\right)=\int_{R^{d}}K^{\sigma}\left(\mathbf{p,p}^{{}^{\prime}},g\right)\varphi\left(\mathbf{p}^{{}^{\prime}}\right)d\mathbf{p}^{{}^{\prime}}, (2.17) where K^{\sigma}\left(\mathbf{p,p}^{{}^{\prime}},g\right)=\frac{1}{\left(2\pi\right)^{d}}\int_{R^{d}}e^{i\mathbf{px-}i\mathbf{p}^{{}^{\prime}}\mathbf{x}_{g}}\omega_{g}\left(\mathbf{x}\right)^{\sigma}d\mathbf{x.} (2.18) Restricting T^{\sigma}\left(g\right) to the subgroup ISO(d) , we obtain \displaystyle\left(T^{\sigma}\left(n\left(\mathbf{b}\right)\right)\varphi\right)\left(\mathbf{p}\right) \displaystyle=e^{i\mathbf{pb}}\varphi\left(\mathbf{p}\right), \displaystyle\left(T^{\sigma}\left(m\right)\varphi\right)\left(\mathbf{p}\right) \displaystyle=\varphi\left(m^{-1}\mathbf{p}\right), where n\left(\mathbf{b}\right)\in N and m\in M . Thus, the restriction of the representation T^{\sigma}\left(g\right) of the group SO_{0}\left(d+1,1\right) to the subgroup ISO(d) is a reducible representation of this subgroup. 3 Euclidean field theory with conformal symmetry We begin this section by defining the Euclidean Fock space \mathcal{E} for scalar bosons [ 10 ] , [ 11 ] . The Euclidean one-particle space \mathcal{E}_{1} is represented as L^{2}(R^{d}) and the Euclidean Fock space \mathcal{E} is \mathcal{E=\oplus}_{n=0}^{\infty}\;\mathcal{E}_{n}, (3.1) where \mathcal{E}_{0}=C and \mathcal{E}_{n} is the n -fold symmetric tensor product of \mathcal{E}_{1} \mathcal{E}_{n}=\left(\mathcal{E}_{1}\otimes\cdots\otimes\mathcal{E}_{1}\right)_{sym}. (3.2) A unitary representation of the Euclidean group ISO(d) is defined on \mathcal{E} by \left(U\left(m,\mathbf{b}\right)\varphi\right)_{n}\left(\mathbf{p}_{1},\ldots,\mathbf{p}_{n}\right)=e^{i\left(\mathbf{p}_{1}+\ldots+\mathbf{p}_{n}\right)\mathbf{b}}\varphi_{n}\left(m^{-1}\mathbf{p}_{1},\ldots,m^{-1}\mathbf{p}_{n}\right), (3.3) where \left(m.\mathbf{b}\right)\in\, ISO(d) with m\in SO(d) and \mathbf{b\in\,}R^{d} . It should be noted that the representation of the group ISO(d) defined on \mathcal{E}_{1} is nothing but the unitary reducible representation (2.19). In Euclidean field theory with dynamic symmetry SO_{0}\left(d+1,1\right) we assume that the representations of SO_{0}\left(d+1,1\right) acting on \mathcal{E}^{in} and \mathcal{E}^{out} are induced by principal series representations T^{\sigma}\left(g\right) and T^{\widetilde{\sigma}}\left(g\right),\,\widetilde{\sigma}=-d-\sigma , respectively, i.e. \displaystyle\left(T^{\sigma_{1}\ldots\sigma_{n}}\left(g\right)\varphi\right)_{n}\left(\mathbf{p}_{1},\ldots,\mathbf{p}_{n}\right) \displaystyle=\int_{R^{d}}\cdots\int_{R^{d}}K^{\sigma_{1}}\left(\mathbf{p}_{1}\mathbf{,p}_{1}^{{}^{\prime}},g\right)\cdots\,K^{\sigma_{n}}\left(\mathbf{p}_{n}\mathbf{,p}_{n}^{{}^{\prime}},g\right)\varphi_{n}\left(\mathbf{p}_{1}^{\prime},\ldots,\mathbf{p}_{n}^{\prime}\right)d\mathbf{p}_{1}^{{}^{\prime}}\cdots\,d\mathbf{p}_{n}^{{}^{\prime}} and \displaystyle\left(T^{\widetilde{\sigma}_{1}\ldots\widetilde{\sigma}_{n}}\left(g\right)\varphi\right)_{n}\left(\mathbf{p}_{1},\ldots,\mathbf{p}_{n}\right) \displaystyle=\int_{R^{d}}\cdots\int_{R^{d}}K^{\widetilde{\sigma}_{1}}\left(\mathbf{p}_{1}\mathbf{,p}_{1}^{{}^{\prime}},g\right)\cdots\,K^{\widetilde{\sigma}_{n}}\left(\mathbf{p}_{n}\mathbf{,p}_{n}^{{}^{\prime}},g\right)\varphi_{n}\left(\mathbf{p}_{1}^{\prime},\ldots,\mathbf{p}_{n}^{\prime}\right)d\mathbf{p}_{1}^{{}^{\prime}}\cdots\,d\mathbf{p}_{n}^{{}^{\prime}}. In other words, T^{\sigma_{1}\ldots\sigma_{n}}= T^{\sigma_{1}}\otimes\cdots\otimes T^{\sigma_{n}} and T^{\widetilde{\sigma}_{1}\ldots\widetilde{\sigma}_{n}}=T^{\widetilde{\sigma}_{1}}\otimes\cdots\otimes T^{\widetilde{\sigma}_{n}} . Since L^{2}(R^{d}) is the carrier space of the principal series representations, the representations of the group SO_{0}\left(d+1,1\right) defined on \mathcal{E}^{in} and \mathcal{E}^{out} are unitary. Moreover, the equality \mathcal{E=E}^{in}=\mathcal{E}^{out} holds. The aim of this section is to study the intertwining operator between the representations T^{\sigma_{1}\ldots\sigma_{n_{1}}}\left(g\right) and T^{\widetilde{\sigma}_{n_{1}+1}\ldots\widetilde{\sigma}_{n}}\left(g\right) , where n=n_{1}+n_{2} . In momentum space realization, the operators of these representations are integral operators with kernel. Therefore, it is more convenient to work with the position space realization of these representations instead of dealing with the momentum space realization. Let \mathcal{S} be an intertwining operator between the representations \displaystyle\left(T^{\sigma_{1}\ldots\sigma_{n_{1}}}\left(g\right)f\right)_{n_{1}}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n_{1}}\right) \displaystyle=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\cdots\,\omega_{g}\left(\mathbf{x}_{n_{1}}\right)^{\sigma_{n_{1}}}f_{n_{1}}\left(\mathbf{x}_{1g},\ldots,\mathbf{x}_{n_{1}g}\right) (3.6) and \displaystyle\left(T^{\widetilde{\sigma}_{n_{1}+1}\ldots\widetilde{\sigma}_{n}}\left(g\right)f\right)_{n_{2}}\left(\mathbf{x}_{n_{1}+1},\ldots,\mathbf{x}_{n}\right) \displaystyle=\omega_{g}\left(\mathbf{x}_{n_{1}+1}\right)^{\widetilde{\sigma}_{n_{1}+1}}\cdots\,\omega_{g}\left(\mathbf{x}_{n}\right)^{\widetilde{\sigma}_{n}}f_{n_{2}}\left(\mathbf{x}_{n_{1}+1g},\ldots,\mathbf{x}_{ng}\right), (3.7) which means that \mathcal{S}T^{\widetilde{\sigma}_{n_{1}+1}\ldots\widetilde{\sigma}_{n}}\left(g\right)=T^{\sigma_{1}\ldots\sigma_{n_{1}}}\left(g\right)\mathcal{S}\text{\thinspace},\text{\quad}\forall g\in SO_{0}\left(d+1,1\right). (3.8) Such an operator is defined by a kernel \displaystyle\left(\mathcal{S}f\right)_{n_{1}}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n_{1}}\right) \displaystyle=\int_{R^{d}}\cdots\int_{R^{d}}S_{n}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n}\right)f_{n_{2}}\left(\mathbf{x}_{n_{1}+1},\ldots,\mathbf{x}_{n}\right)d\mathbf{x}_{n_{1}+1}\cdots\,d\mathbf{x}_{n}\,. (3.9) The intertwining relation ( 3.8 ) implies \left(\mathcal{S}T^{\widetilde{\sigma}_{n_{1}+1}\ldots\widetilde{\sigma}_{n}}\left(g\right)f\right)_{n_{1}}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n_{1}}\right)=\left(T^{\sigma_{1}\ldots\sigma_{n_{1}}}\left(g\right)\mathcal{S}f\right)_{n_{1}}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n_{1}}\right). (3.10) It follows from here that \displaystyle\int_{R^{d}}\cdots\int_{R^{d}}S_{n}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n}\right)\omega_{g}\left(\mathbf{x}_{n_{1}+1}\right)^{\widetilde{\sigma}_{n_{1}+1}}\cdots\,\omega_{g}\left(\mathbf{x}_{n}\right)^{\widetilde{\sigma}_{n}} \displaystyle\times f_{n_{2}}\left(\mathbf{x}_{n_{1}+1g},\ldots,\mathbf{x}_{ng}\right)d\mathbf{x}_{n_{1}+1}\cdots\,d\mathbf{x}_{n}\, \displaystyle=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\cdots\,\omega_{g}\left(\mathbf{x}_{n_{1}}\right)^{\sigma_{n_{1}}}\int_{R^{d}}\cdots\int_{R^{d}}S_{n}\left(\mathbf{x}_{1g},\ldots,\mathbf{x}_{n_{1}g},\mathbf{x}_{n_{1}+1}^{\prime},\ldots,\mathbf{x}_{n}^{\prime}\right) \displaystyle f_{n_{2}}\left(\mathbf{x}_{n_{1}+1}^{\prime},\ldots,\mathbf{x}_{n}^{\prime}\right)d\mathbf{x}_{n_{1}+1}^{\prime}\cdots\,d\mathbf{x}_{n}^{\prime}. (3.11) Performing the change of variables \mathbf{x}_{i}^{\prime}=\mathbf{x}_{ig}\,,\quad d\mathbf{x}_{i}^{\prime}=\omega_{g}\left(\mathbf{x}_{i}\right)^{-d}d\mathbf{x}_{i}\,,\quad i=n_{1}+1,\ldots,n, (3.12) on the right side of (3.11) and taking into account the arbitrariness of the function f_{n_{2}} we find the following constraint on the Green functions S_{n} : S_{n}\left(\mathbf{x}_{1},\ldots,\mathbf{x}_{n}\right)=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\cdots\,\omega_{g}\left(\mathbf{x}_{n}\right)^{\sigma_{n}}S_{n}\left(\mathbf{x}_{1g},\ldots,\mathbf{x}_{ng}\right). (3.13) It is well known [ 12 ] that conformal symmetry allows one to determine the two- and three-point functions up to a constant factor. Below we provide a step-by-step calculation of these Green functions. Let’s start with the two-point function. In this case, S_{2}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right)=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\omega_{g}\left(\mathbf{x}_{2}\right)^{\sigma_{2}}S_{2}\left(\mathbf{x}_{1g},\mathbf{x}_{2g}\right). (3.14) By choosing g=n\left(\mathbf{b}\right) and letting \mathbf{b}=\mathbf{x}_{2} , we see that S_{2}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right) is a function of \mathbf{x}_{1}-\mathbf{x}_{2} only S_{2}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right)=S_{2}\left(\mathbf{x}_{1}-\mathbf{x}_{2}\right), (3.15) where S_{2}\left(\mathbf{x}\right)=S_{2}\left(\mathbf{x},\mathbf{0}\right) . Inserting ( 3.15 ) into ( 3.14 ), we obtain S_{2}\left(\mathbf{x}_{1}-\mathbf{x}_{2}\right)=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\omega_{g}\left(\mathbf{x}_{2}\right)^{\sigma_{2}}S_{2}\left(\mathbf{x}_{1g}-\mathbf{x}_{2g}\right). (3.16) Let now g=a\left(\lambda\right) . By putting \lambda=\left\|\mathbf{x}_{1}-\mathbf{x}_{2}\right\| we find S_{2}\left(\mathbf{x}\right)=\left\|\mathbf{x}\right\|^{\sigma_{1}+\sigma_{2}}S_{2}\left(\frac{\mathbf{x}}{\left\|\mathbf{x}\right\|}\right), (3.17) where \left\|\mathbf{x}\right\|=\sqrt{x_{1}^{2}+x_{2}^{2}+\cdots+x_{n}^{2}} . If we insert ( 3.17 ) into ( 3.16 ) and put g=m , we have S_{2}\left(\mathbf{x}\right)=\left\|\mathbf{x}\right\|^{\sigma_{1}+\sigma_{2}}S_{2}\left(\frac{m^{-1}\mathbf{x}}{\left\|\mathbf{x}\right\|}\right). (3.18) Let us recall that every vector \mathbf{x}\in R^{d} is represented as \mathbf{x}=\left\|\mathbf{x}\right\|m_{\mathbf{x}}\mathbf{e} , where m_{\mathbf{x}}\in SO\left(d\right) and \mathbf{e=}\left(0,\ldots,0,1\right) . The matrix m_{\mathbf{x}} is independent of the length of \mathbf{x} , i.e. it depends only on \mathbf{x}\,/\left\|\mathbf{x}\right\| . Choosing m=m_{\mathbf{x}} , we have S_{2}\left(\mathbf{x}\right)=S_{2}\left(\mathbf{e}\right)\left\|\mathbf{x}\right\|^{\sigma_{1}+\sigma_{2}}. (3.19) All that remains is to use special conformal transformations. Inserting ( 3.19 ) in ( 3.16 ) and choosing g=\overline{n}\left(\mathbf{c}\right) , we obtain \left\|\mathbf{x}_{1}-\mathbf{x}_{2}\right\|^{\sigma_{1}+\sigma_{2}}=\left(\frac{1+\mathbf{c}^{2}\mathbf{x}_{1}^{2}-2\mathbf{cx}_{1}}{1+\mathbf{c}^{2}\mathbf{x}_{2}^{2}-2\mathbf{cx}_{2}}\right)^{\frac{\sigma_{1}-\sigma_{2}}{2}}\left\|\mathbf{x}_{1}-\mathbf{x}_{2}\right\|^{\sigma_{1}+\sigma_{2}}. (3.20) This is only possible if \sigma_{1}=\sigma_{2} . Hence S_{2}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right)=C\left\|\mathbf{x}_{1}-\mathbf{x}_{2}\right\|^{2\sigma}\text{\qquad if\quad\ }\sigma_{1}=\sigma_{2}=\sigma, (3.21) where C is a constant coefficient, which is not determined by conformal symmetry. A similar analysis can be done on three-point functions. In this case, S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2},\mathbf{x}_{3}\right)=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\omega_{g}\left(\mathbf{x}_{2}\right)^{\sigma_{2}}\omega_{g}\left(\mathbf{x}_{3}\right)^{\sigma_{3}}S_{3}\left(\mathbf{x}_{1g},\mathbf{x}_{2g},\mathbf{x}_{3g}\right). (3.22) Translation covariance imposes the following constraint on S_{3} S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2},\mathbf{x}_{3}\right)=S_{3}\left(\mathbf{x}_{1}-\mathbf{x}_{3},\mathbf{x}_{2}-\mathbf{x}_{3}\right), (3.23) where S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right)=S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2},\mathbf{0}\right) . Inserting ( 3.23 ) into ( 3.22 ), we obtain \displaystyle S_{3}\left(\mathbf{x}_{1}-\mathbf{x}_{3},\mathbf{x}_{2}-\mathbf{x}_{3}\right) \displaystyle=\omega_{g}\left(\mathbf{x}_{1}\right)^{\sigma_{1}}\omega_{g}\left(\mathbf{x}_{2}\right)^{\sigma_{2}}\omega_{g}\left(\mathbf{x}_{3}\right)^{\sigma_{3}}S_{3}\left(\mathbf{x}_{1g}-\mathbf{x}_{3g},\mathbf{x}_{2g}-\mathbf{x}_{3g}\right). (3.24) Scale and rotation covariance implies that S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2}\right)=\left\|\mathbf{x}_{2}\right\|^{\sigma_{1}+\sigma_{2}+\sigma_{3}}S_{3}\left(\frac{m_{\mathbf{x}_{2}}^{-1}\mathbf{x}_{1}}{\left\|\mathbf{x}_{2}\right\|}\right), (3.25) where S_{3}\left(\mathbf{x}\right)=S_{3}\left(\mathbf{x},\mathbf{e}\right) . We now insert ( 3.25 ) in (3.24) and make use of covariance under special conformal transformations g=\overline{n}\left(\mathbf{c}\right) . By putting \mathbf{x}_{3}=\mathbf{0,\hskip 11.49994ptx}_{2}=\mathbf{e,\hskip 11.49994ptc}=\frac{2\mathbf{x}_{1}}{\mathbf{x}_{1}^{2}}-\mathbf{e,} (3.26) we obtain S_{3}\left(\mathbf{x}\right)=2^{-\sigma_{1}+\sigma_{2}-\sigma_{3}}S_{3}\left(2\mathbf{e}\right)\left\|\mathbf{x}\right\|^{\sigma_{1}-\sigma_{2}+\sigma_{3}}\left\|\mathbf{e-x}\right\|^{\sigma_{1}+\sigma_{2}-\sigma_{3}} (3.27) ( with \mathbf{x}_{1} replaced by \mathbf{x} ). Then according to ( 3.27 ), ( 3.25 ) and ( 3.23 ) we have S_{3}\left(\mathbf{x}_{1},\mathbf{x}_{2},\mathbf{x}_{3}\right)=C_{3}\left\|\mathbf{x}_{1}-\mathbf{x}_{3}\right\|^{\sigma_{1}-\sigma_{2}+\sigma_{3}}\left\|\mathbf{x}_{2}-\mathbf{x}_{3}\right\|^{-\sigma_{1}+\sigma_{2}+\sigma_{3}}\left\|\mathbf{x}_{1}-\mathbf{x}_{2}\right\|^{\sigma_{1}+\sigma_{2}-\sigma_{3}}, (3.28) where C_{3} is a three-point structure constant depending on \sigma_{1},\sigma_{2},\sigma_{3} . It has long been known that three-point functions of conformal field theory are related to the Clebsch-Gordan kernels of the conformal group (see [ 13 ] and references therein). This fact is more transparent in the definition ( 3.8 ). Moreover, by decomposing the tensor product of two principal series representations into irreducibles, n -point functions can be written in terms of Clebsch-Gordan kernels. This will be given explicitly for four-point functions in two dimensions in the next section. 4 Four-point functions in two dimensions To outline our technique, we restrict ourselves, for the sake of simplicity, to the four point functions in two dimensions. Instead of working with the group SO_{0}\left(3,1\right) , we use its universal covering group SL(2,C) , consisting of all complex unimodular 2\times 2 matrices. It is known that any representation of SL(2,C) is a single- or double- valued representation of SO_{0}\left(3,1\right) . We denote an element of SL(2,C) by g=\left(\begin{array}[c]{cc}g_{11}&g_{12}\\ g_{21}&g_{22}\end{array}\right),\qquad g_{11}g_{22}-g_{12}g_{21}=1. (4.1) The principal series representations of SL(2,C) [ 14 ] are characterized by the pairs of numbers \chi= \left(s,\sigma\right) where s is an integer or half- integer, and \sigma=-1+i\rho,\quad\rho\in R . These representations can be realized in a Hilbert space L^{2}(C) \left(T^{\chi}\left(g\right)f\right)\left(z\right)=\left|g_{12}z+g_{22}\right|^{2\sigma-2s}\left(g_{12}z+g_{22}\right)^{2s}f\left(\frac{g_{11}z+g_{21}}{g_{12}z+g_{22}}\right). (4.2) The representations \chi= \left(s,\sigma\right) and \widetilde{\chi}= \left(-s,\widetilde{\sigma}\right) , \widetilde{\sigma}=-2-\sigma , are equivalent. We will restrict ourselves to scalar principal series representations \chi= \left(0,\sigma\right) and will denote the corresponding operators by T^{\sigma} . A brief overview of Naimark’s results on the decomposition of tensor products of scalar principal series representations is given in the Appendix. We shall construct intertwining operators between representations \displaystyle\left(T^{\sigma_{1}\sigma_{2}}\left(g\right)f\right)\left(z_{1},z_{2}\right) \displaystyle=\left|g_{12}z_{1}+g_{22}\right|^{2\sigma_{1}}\left|g_{12}z_{2}+g_{22}\right|^{2\sigma_{2}}f\left(\frac{g_{11}z_{1}+g_{21}}{g_{12}z_{1}+g_{22}},\frac{g_{11}z_{2}+g_{21}}{g_{12}z_{2}+g_{22}}\right) (4.3) and \displaystyle\left(T^{\widetilde{\sigma}_{3}\widetilde{\sigma}_{4}}\left(g\right)f\right)\left(z_{3},z_{4}\right) \displaystyle=\left|g_{12}z_{3}+g_{22}\right|^{2\widetilde{\sigma}_{3}}\left|g_{12}z_{4}+g_{22}\right|^{2\widetilde{\sigma}_{4}}f\left(\frac{g_{11}z_{3}+g_{21}}{g_{12}z_{3}+g_{22}},\frac{g_{11}z_{4}+g_{21}}{g_{12}z_{4}+g_{22}}\right). (4.4) By decomposing T^{\sigma_{1}\sigma_{2}}\left(g\right) into irreducibles (see Appendix), the problem reduces to the much simpler problem of finding an intertwining operator between the representations T^{\widetilde{\sigma}_{3}\widetilde{\sigma}_{4}} and T^{\chi} . Thus, we get that the operator \mathcal{S} with the intertwining property \mathcal{S}T^{\widetilde{\sigma}_{3}\widetilde{\sigma}_{4}}\left(g\right)=T^{\sigma_{1}\sigma_{2}}\left(g\right)\mathcal{S}, (4.5) can be written as an integral operator \left(\mathcal{S}f\right)\left(z_{1},z_{2}\right)=\int_{C^{2}}S_{4}\left(z_{1},z_{2},z_{3},z_{4}\right)f\left(z_{3},z_{4}\right)dz_{3}dz_{4}, (4.6) with kernel \displaystyle S_{4}\left(z_{1},z_{2},z_{3},z_{4}\right) \displaystyle=\frac{1}{4\pi^{4}}{\textstyle\sum\limits_{s=-\infty}^{\infty}}{\displaystyle\int\limits_{-\infty}^{\infty}}d\rho\left(s^{2}+\rho^{2}\right)\overline{C(\sigma_{1},\sigma_{2},\chi)}C(\widetilde{\sigma}_{3},\widetilde{\sigma}_{4},\chi) \displaystyle\times S_{4}^{\chi}\left(z_{1},z_{2},z_{3},z_{4}\right) (4.7) (the bar denotes complex conjugate). Here C is the three-point structure constant, and S_{4}^{\chi} is an s -channel four-point conformal block S_{4}^{\chi}\left(z_{1},z_{2},z_{3},z_{4}\right)=\int_{C}\overline{N\left(z_{1}\sigma_{1}z_{2}\sigma_{2};z\chi\right)}N\left(z_{3}\widetilde{\sigma}_{3}z_{4}\widetilde{\sigma}_{4};z\chi\right)dz. (4.8) Inserting for the kernels N in ( 4.8 ) its expression (A.3), we have \displaystyle S_{4}^{\chi}\left(z_{1},z_{2},z_{3},z_{4}\right) \displaystyle=\left|z_{3}-z_{4}\right|^{s-\sigma+\sigma_{3}+\sigma_{4}}\left(z_{3}-z_{4}\right)^{-s} \displaystyle\times\left|z_{1}-z_{2}\right|^{s+\sigma+\sigma_{1}+\sigma_{2}+2}\overline{\left(z_{1}-z_{2}\right)}^{-s}J, (4.9) where J=\int_{C}dz{\textstyle\prod\limits_{i=1}^{4}}\left(z-z_{i}\right)^{-\alpha_{i}}\overline{\left(z-z_{i}\right)}^{-\nu_{i}}, (4.10) with \displaystyle 2\alpha_{1} \displaystyle=s+\sigma-\sigma_{1}+\sigma_{2}+2,\qquad 2\nu_{1}=-s+\sigma-\sigma_{1}+\sigma_{2}+2, \displaystyle 2\alpha_{2} \displaystyle=s+\sigma+\sigma_{1}-\sigma_{2}+2,\qquad 2\nu_{2}=-s+\sigma+\sigma_{1}-\sigma_{2}+2, \displaystyle 2\alpha_{3} \displaystyle=-s-\sigma-\sigma_{3}+\sigma_{4},\qquad\quad 2\nu_{3}=s-\sigma-\sigma_{3}+\sigma_{4}, \displaystyle 2\alpha_{4} \displaystyle=-s-\sigma+\sigma_{3}-\sigma_{4},\qquad\quad 2\nu_{4}=s-\sigma+\sigma_{3}-\sigma_{4}. These parameters are subject to the condition {\textstyle\sum\limits_{i=1}^{4}}\alpha_{i}={\textstyle\sum\limits_{i=1}^{4}}\nu_{i}=2,\quad\alpha_{i}-\nu_{i}\in Z. (4.12) From (4.9) we obtain \displaystyle S_{4}^{\chi}\left(z_{1},z_{2},z_{3},z_{4}\right) \displaystyle=\left|g_{12}z_{1}+g_{22}\right|^{2\sigma_{1}}\cdots\left|g_{12}z_{4}+g_{22}\right|^{2\sigma_{4}} \displaystyle\times S_{4}^{\chi}\left(\frac{g_{11}z_{1}+g_{21}}{g_{12}z_{1}+g_{22}},\ldots,\frac{g_{11}z_{4}+g_{21}}{g_{12}z_{4}+g_{22}}\right), (4.13) i.e. conformal blocks S_{4}^{\chi} are subject to the same conformal symmetry constraints as the four-point functions. The integral ( 4.10 ) was evaluated in [ 15 ] , and hence we have \displaystyle S_{4}^{\chi}\left(z_{1},z_{2},z_{3},z_{4}\right) \displaystyle=\pi z_{12}^{\alpha_{5}}z_{23}^{\alpha_{1}+\alpha_{4}-1}z_{31}^{\alpha_{2}-1}z_{24}^{-\alpha_{4}}z_{34}^{\alpha_{6}}\overline{z}_{12}^{\nu_{5}}\overline{z}_{23}^{\nu_{1}+\nu_{4}-1}\overline{z}_{31}^{\nu_{2}-1}\overline{z}_{24}^{-\nu_{4}}\overline{z}_{34}^{\nu_{6}}\Phi\left(\eta,\overline{\eta}\right), where z_{ij}=z_{i}-z_{j} and \displaystyle 2\alpha_{5} \displaystyle=-s-\sigma+\sigma_{1}+\sigma_{2},\qquad\ 2\nu_{5}=s-\sigma+\sigma_{1}+\sigma_{2}, \displaystyle 2\alpha_{6} \displaystyle=-s-\sigma+\sigma_{3}+\sigma_{4},\qquad\ 2\nu_{6}=s-\sigma+\sigma_{3}+\sigma_{4}. The function \Phi is expressed in terms of products of ordinary hypergeometric functions \displaystyle\Phi\left(\eta,\overline{\eta}\right) \displaystyle=AF\left(1-\alpha_{2},\alpha_{4};\alpha_{3}+\alpha_{4};\eta\right)F\left(1-\nu_{2},\nu_{4};\nu_{3}+\nu_{4};\overline{\eta}\right) \displaystyle+B\eta^{\alpha_{1}+\alpha_{2}-1}\overline{\eta}^{\nu_{1}+\nu_{2}-1}F\left(1-\alpha_{3},\alpha_{1};\alpha_{1}+\alpha_{2};\eta\right) \displaystyle\times F\left(1-\nu_{3},\nu_{1};\nu_{1}+\nu_{2};\overline{\eta}\right), (4.16) with \eta=\frac{z_{12}z_{34}}{z_{13}z_{24}}, (4.18) and \displaystyle A \displaystyle=\frac{\Gamma\left(1-\alpha_{1}\right)\Gamma\left(1-\alpha_{2}\right)\Gamma\left(\alpha_{1}+\alpha_{2}-1\right)}{\Gamma\left(\nu_{1}\right)\Gamma\left(\nu_{2}\right)\Gamma\left(2-\nu_{1}-\nu_{2}\right)}, (4.19) \displaystyle B \displaystyle=\frac{\Gamma\left(1-\alpha_{3}\right)\Gamma\left(1-\alpha_{4}\right)\Gamma\left(\alpha_{3}+\alpha_{4}-1\right)}{\Gamma\left(\nu_{3}\right)\Gamma\left(\nu_{4}\right)\Gamma\left(2-\nu_{3}-\nu_{4}\right)}. (4.20) It should be noted that the four-point conformal blocks for different channels are related by the Racah coefficients [ 16 ] , [ 17 ] for principal series of representations of SL(2,C) . Appendix A Decomposition of tensor products of scalar principal series representations The problem of decomposing the tensor product of two irreducible representations of the group SL(2,C) into irreducible ones was completely solved by Naimark in [ 8 ] . Here we give a brief overview of Naimark’s results on the decomposition of tensor products of scalar principal series representations. The tensor product T^{\sigma_{1}}\otimes T^{\sigma_{2}} can be realized in the Hilbert space L^{2}(C^{2}) of functions f\left(z_{1},z_{2}\right) of two variables z_{1} and z_{2} \displaystyle\left[T^{\sigma_{1}}\otimes T^{\sigma_{2}}\left(g\right)f\right]\left(z_{1},z_{2}\right) \displaystyle=\left|g_{12}z_{1}+g_{22}\right|^{2\sigma_{1}}\left|g_{12}z_{2}+g_{22}\right|^{2\sigma_{2}}f\left(\frac{g_{11}z_{1}+g_{21}}{g_{12}z_{1}+g_{22}},\frac{g_{11}z_{2}+g_{21}}{g_{12}z_{2}+g_{22}}\right). (A.1) This representation is decomposed only into principal series representations T^{\chi} , and into those and only those for which s is an integer. The components of this expansion, transforming according to the representation T^{\chi} , are determined by the formula f\left(z\chi\right)=\int_{C^{2}}N\left(z_{1}\sigma_{1}z_{2}\sigma_{2};z\chi\right)f\left(z_{1},z_{2}\right)dz_{1}dz_{2}, (A.2) where \displaystyle N\left(z_{1}\sigma_{1}z_{2}\sigma_{2};z\chi\right) \displaystyle=\left|z_{2}-z_{1}\right|^{s-\sigma-\sigma_{1}-\sigma_{2}-4}\left(z_{2}-z_{1}\right)^{-s}\left|z-z_{1}\right|^{-s+\sigma-\sigma_{1}+\sigma_{2}}\left(z-z_{1}\right)^{s} \displaystyle\times\left|z_{2}-z\right|^{-s+\sigma+\sigma_{1}-\sigma_{2}}\left(z_{2}-z\right)^{s}. (A.3) The function f\left(z_{1},z_{2}\right) is expressed through these components by the formula f\left(z_{1},z_{2}\right)=\frac{1}{4\pi^{4}}{\displaystyle\sum\limits_{s=-\infty}^{\infty}}{\displaystyle\int\limits_{-\infty}^{\infty}}\left(s^{2}+\rho^{2}\right)d\rho\int\overline{N\left(z_{1}\sigma_{1}z_{2}\sigma_{2};z\chi\right)}f\left(z\chi\right)dz. (A.4) In this case, the Plancherel formula is \int_{C^{2}}\left|f\left(z_{1},z_{2}\right)\right|^{2}dz_{1}dz_{2}=\frac{1}{4\pi^{4}}{\displaystyle\sum\limits_{s=-\infty}^{\infty}}{\displaystyle\int\limits_{-\infty}^{\infty}}\left(s^{2}+\rho^{2}\right)d\rho\int_{C}\left|f\left(z\chi\right)\right|^{2}dz. (A.5) It should be noted that the kernel N is uniquely determined up to a constant factor C(\sigma_{1},\sigma_{2},\chi) . Plancherel’s formula ( A.5 ) fixes the absolute value of this factor. The phase of C(\sigma_{1},\sigma_{2},\chi) is arbitrary.