nLab
additive analytic geometry

Contents

Contents

Idea

Additive analytic geometry over the complex number is an analog of classical (called here multiplicative) complex analytic geometry obtained by replacing the convolution algebra 1(,)\ell^1(\mathbb{N},\mathbb{C}) of converging power series of radius 11 (discrete Mellin transforms) by the algebra L 1( +,)L^1(\mathbb{R}_+,\mathbb{C}) of integrable functions (continuous Mellin transforms) or other variants. The analog of the complex unit disc D m(0,1)=D(0,1)D_m(0,1)=D(0,1) is the hyperbolic/additive unit disc

D a(,0)={s,Re(s)0}{},D_a(-\infty,0)=\{s,Re(s)\leq 0\}\cup \{-\infty\},

(if one choses the right class of functions/distributions, since the Fourier transform of an L 1L^1 function vanishes at infinity) and the analog of the affine line 𝔸 m 1=\mathbb{A}^1_m=\mathbb{C} is the additive affine line

𝔸 a 1={}.\mathbb{A}^1_a=\{-\infty\}\cup \mathbb{C}.

The main interest of this new geometry is that it has a built in flow (action of the additive monoid +\mathbb{R}_+ on itself) and its functions are locally given by continuous Mellin transforms.

One may work, if needed with other kinds of functions/distributions on +\mathbb{R}_+, for example if one wants to study distributional trace formulas. In any case, one needs to use Dirac measures/distributions to relate directly the above two types of analytic geometries, by the relation on coordinates given by z=e sz=e^s in an archimedean context, and z=p sz=p^{-s} in a non-archimedean context.

Higher dimensional spaces may also be considered, but the lack of a unit element in L 1L^1 makes the extension of the semi and sub-analytic topologies to this geometry a bit tricky. Unitarization may be necessary.

A full complex geometry with various Grothendieck topologies seems to emerge from the consideration of the simple analogy/link with classical complex analysis given by the complex exponential map, but it is not so easy to find the right class of functions/distributions to be used here. One may simply try out with L 1L^1 functions, and see where one arrives, but a regularity condition on an additive disk may also be imposed by asking that (logx) nf(x)x δ(\log x)^n f(x)x^\delta is L 1L^1 for every nn, in order to obtain smoothness of functions on the boundary; using schwartz functions or overconvergent ones may also be a solution to the problem of having a nice notion of differential calculus on rational domains, that one must also solve in classical (multiplicative) analytic geometry.

Additive analytic geometry also has a non-commutative and automorphic extension (formulated in spectral terms), that may be useful to study and formulate some problems of automorphic representation theory in a geometric context. For example, the results of Godement and Jacquet looks like a kind of pasting of two “Mellin automorphic spectra” along the usual automorphic spectrum, in this context.

A simple example

One may paste two additive unit discs along their boundary \mathbb{R}, that is the spectrum of L 1()L^1(\mathbb{R}) by sss\mapsto -s on \mathbb{C}, to get the additive version of the projective space a 1()={}{+}\mathbb{P}^1_a(\mathbb{C})=\{-\infty\}\cup \mathbb{C}\cup \{+\infty\} and the exponential map sends a 1()\mathbb{P}^1_a(\mathbb{C}) to the usual (multiplicative) projective space 1()\mathbb{P}^1(\mathbb{C}). One may also paste two different discs along an annulus. Remark that there is also a way to use other pastings, such as s1ss\mapsto 1-s. In any case, there is an analog of the homogeneous coordinates on projective space that correspond to the projection from

𝔸 a 1×𝔸 a 1{(,)}\mathbb{A}^1_a\times \mathbb{A}^1_a-\{(-\infty,-\infty)\}

to its quotient by the relation given by (s 1,s 2)(λ+s 1,λ+s 2)(s_1,s_2)\sim (\lambda+s_1,\lambda+s_2) for λ𝔾 m a():=\lambda\in \mathbb{G}_m^a(\mathbb{C}):=\mathbb{C}.

Remarks

To be able to work with LL-functions, seen as additive functions, it seems one needs to pass to the setting of distributions of a certain kind, that may be analytic on a half space and distributional on the boundary. The study of the spectrum of distribution convolution algebras probably requires further work.

If such a function may be extended to a type of additive projective space using a functional equation, then the complement of its set SS of zeroes gives an open subspace UU of an additive projective space. In good situations, this gives an additive subspace that is equipped with the given flow, and its embedding into projective space induces an identification between the associated additive de Rham cohomology spaces in degree 00 and 22. The Poincaré group of this space is generated by looping around each of the given points in the set SS, that may be infinite.

It seems that the L-function has to be automorphic so that one may study it using additive geometry, because the functional equation of theta functions is at the heart of the above construction. Indeed, even if higher dimensional complex varieties allow one to define various types of generalized multiplicative spaces (using Mellin transforms of δ\delta distributions, for example), there is yet no clear additive versions of global arithmetic schemes, because there is no convolution operation on L 1( +,)L^1(\mathbb{R}_+,\mathbb{Z}).

A possible approach to this problem of finding a higher dimensional version of the above construction is to first try to give a spectral interpretation for theta functions, so that one may study more general LL functions. This corresponds to decomposing the LL-function of an arithmetic scheme with respect to the motivic weight (cohomology relative to the spectrum of \mathbb{Z}), and not its absolute motivic weight (which would allow to interpret, for example, the factor ss in the ξ\xi function).

Proving that the given spectral θ\theta functions fulfils a functional equation is not an easy task and can’t be solved directly without showing that they are decomposed in LL-functions of cuspidal automorphic type for some GL m\GL_m, and using Godement and Jacquet’s generalization of Tate’s thesis.

Additive Hodge theory

One may develop an analog of Hodge theory in this context (working with real analytic additive functions in the ss and s¯\bar{s} coordinates, and the corresponding analog of the Laplacian; the formal theory with Sobolev spaces seems to carry through to this context as one may see on wikipedia). This will not give finite dimensionality of de Rham cohomology, because of the above remark on the Poincaré group. However, such a Hodge theory may still have interesting applications related to its interaction with Poincaré duality in additive de Rham cohomology, since, in good cases, all the given construction are compatible with the given flow.

Last revised on May 4, 2019 at 17:11:19. See the history of this page for a list of all contributions to it.