Formal loops. III. Additive functions and the Radon transform (Q952423)

Let \(X\) be a \(C^\infty\)-manifold, and let \(\eta\) be a smooth \(1\)-form on \(X\). Its Radon transform is the function \(\tau(\eta)\) on the free loop space \(L(X)=C^\infty(S^1,X)\) whose value at \(\gamma:S^1\rightarrow X\) is \(\int_{S^1}\gamma^\ast\eta.\) The problem of describing the range of the transform \(\tau\) was studied by Brylinski who, in the case \(X=\mathbb{R}^N,\) characterized it by a system of differential equations. One part of these equations expresses that \(\tau(\eta)\) is invariant under reparametrizations of \(S^1.\) If \(X\) is a complex manifold and \(\eta\) is holomorphic, then \(\tau(\eta)\) satisfies two more properties, that \(\tau(\eta)=0\) on \(L^0(X)\), the space of loops extending holomorphically into the unit disc, and that \(\tau(\eta)\) is additive in holomorphic pairs of pants; the value on \(\tau(\eta)\) on the ``waist'' circle is equal to the sum of values on the ``leg'' circles. In the present paper the authors assume that \(X\) is a smooth algebraic variety over \(\mathbb{C}\) and relate the Radon transform to the theory of vertex operator algebras and factorization algebras. \(L^0(X)\) and \(L(X)\) are replaced by the scheme \(\mathcal{L}^0(X)\) of formal arcs, and the ind-scheme \(\mathcal{L}(X)\) of formal loops. They are obtained by replacing the unit disk by \(\operatorname{Spec}\mathbb{C}[[t]]\) and the unit circle by \(\operatorname{Spec}\mathbb{C}((t)).\) For a regular \(1\)-form \(\eta\) there is a function \(\tau(\eta)\) on \(\mathcal{L}(X)\) vanishing on \(\mathcal{L}^0(X)\). For any closed \(2\)-form \(\omega\) the Radon transform \(\tau(d^{-1}(\omega))\) of any analytic primitive \(d^{-1}(\omega)\) of \(\omega\) is defined algebraically and depends only on \(\omega\). If \(\omega\) is nondegenerate, then \(\tau(d^{-1}(\omega))\) is a version of the symplectic action functional on the space of loops. In an earlier paper in this series [Ann. Sci. Éc. Norm. Supér. (4) 40, No. 1, 113--133 (2007; Zbl 1129.14022)], the authors showed how the space \(\mathcal{L}(X)\) provides a geometric construction of a particular sheaf of vertex algebras on \(X\), the chiral de Rham complex \(\Omega_X^{\operatorname{ch}}\). It was realized as a semiinfinite de Rham complex of a particular \(D\)-module on \(\mathcal{L}(X)\) of distributions supported on \(\mathcal{L}^0(X)\), and the sheaf \(\mathcal{O}_{\mathcal{L}(X)}\) of functions on \(\mathcal{L}(X)\) acts on \(\Omega_X^{\operatorname{ch}}\) by multiplication. The main result of the article is the following: Let \(f\) be a function on \(\mathcal{L}(X)\) vanishing on \(\mathcal{L}^0(X)\). Then the following are equivalent: (i) \(f\) has the form \(\tau(d^{-1}(\omega))\) for a closed \(2\)-form \(\omega\). (ii) The operator of multiplication by \(\exp(f)\) in \(\Omega_X^{\operatorname{ch}}\) is an automorphism of vertex algebras. The article is a part of a series devoted to interpreting, in geometric terms involving \(\mathcal{L}(X)\), the gerbe of chiral differential operators (CDO) on \(X\). Objects of this gerbe are sheaves of vertex algebras similar to \(\Omega_X^{\operatorname{ch}}\) but withouth the fermionic variables. For any such object \(\mathcal{A}\) the sheaf \(\underline{\operatorname{Aut}}(\mathcal{A})\) was found to be canonically isomorphic to \(\Omega^{2,\operatorname{cl}}_X\), the sheaf of closed \(2\)-forms. This is precisely the domain of definition of the Radon transform \(\tau\circ d^{-1}.\) In a subsequent paper the authors will show that the anomaly inherent in the construction of CDO is related to the determinantal anomaly for the loop space by the Radon transform. Sheaves of CDO on \(X\) form a gerbe with lien \(\Omega^{2,\operatorname{cl}}_X\) while the determinantal gerbe of \(\mathcal{L}(X)\) has the lien \(\mathcal{O}^\times_{\mathcal{L}(X)}.\) On the level of liens the relation between the two gerbes associates to a \(2\)-form \(\omega\in\Omega^{2,\operatorname{cl}}_X\) the invertible function \(S(\omega)=\exp(\tau d^{-1}(\omega))\in\mathcal{O}_{\mathcal{L}(X)}^\times.\) The authors give nice and thorough treatments of the essential, elementary parts. Most of the article is used to prove the main theorem, and the techniques are of general interest to a lot of applications of this theory, and also to other fields of geometry.