---
title: 'Motivic Fourier Transform: Concepts & Applications'
url: https://www.emergentmind.com/topics/motivic-fourier-transform
type: topic
---

# Motivic Fourier Transform: Concepts & Applications

Searching arXiv for fresh relevant papers on motivic Fourier transform and adjacent frameworks.
arxiv_search(query="motivic Fourier transform", max_results=10)
Motivic Fourier transform denotes a family of algebraic and motivic analogues of Fourier analysis in which the classical analytic kernel is replaced by geometric or motivic data. In one major incarnation, for an abelian variety \(A\) over a field \(k\), the transform is induced by the Poincaré line bundle on \(A\times \widehat A\); in another, it is a motivic integral transform on exponential constructible functions over valued fields; in a third, it becomes a multiplicative Poisson formalism for split algebraic tori; and in a fourth, it appears as a Fourier operator on Schwartz spaces of spherical varieties obtained from Braverman–Kazhdan spaces [2410.21094], [2412.17764], [2604.03162], [2009.11490]. Across these settings, the common theme is the replacement of transcendental Fourier analysis by correspondences, character objects, and functorial integral transforms in algebraic or motivic categories.

## 1. Classical origin: abelian varieties, correspondences, and the Poincaré bundle

For an abelian variety \(A/k\) with dual \(\widehat A\), the classical Fourier transform on cycles is the correspondence defined by the Poincaré line bundle \(\mathcal P_A\) on \(A\times \widehat A\). On Chow groups with rational coefficients it is
\[
\mathcal F_A \colon CH_*(A)_\mathbf Q \longrightarrow CH_*(\widehat A)_\mathbf Q,
\qquad
\mathcal F_A(x):={p_2}_*\big(\operatorname{ch}(\mathcal P_A)\cdot p_1^*x\big).
\]
Its kernel is therefore not an additive character but the Chern character of a universal biextension, and the first Chern class
\[
\ell=c_1(\mathcal P_A)\in CH^1(A\times \widehat A)_\mathbf Q
\]
generates the relevant correspondences. The transform is an isomorphism, its inverse is expressed via the Poincaré bundle on \(\widehat A\times A\), and it yields an isomorphism of motives. It is central in the classical proofs of the Chow–Künneth decomposition of the motive of \(A\) due to Shermenev and Deninger–Murre [2410.21094].

A defining structural property is the exchange of the Pontryagin product
\[
x*y:=\mu_*(x\times y),\qquad \mu\colon A\times A\to A,
\]
with the usual intersection product:
\[
\mathcal F_A(x*y)=\mathcal F_A(x)\cdot \mathcal F_A(y),
\qquad
\mathcal F_A(x\cdot y)=(-1)^{\dim A}\,\mathcal F_A(x)*\mathcal F_A(y).
\]
This duality explains why the Fourier transform is not merely a correspondence between Chow groups but a mechanism for transferring multiplicative structure between two different products [2410.21094].

The classical theory is closely tied to Beauville’s eigenspace decomposition
\[
CH^i(A)_\mathbf Q=\bigoplus_j CH^i(A)_{(j),\mathbf Q},
\]
where the summands are eigenspaces for pullback under multiplication-by-\(n\). Moonen–Polishchuk showed that the Fourier transform respects this decomposition and constructed a divided power structure on the Pontryagin ideal
\[
CH_{>0}(A)\subset CH_*(A),
\]
and, over algebraically closed fields, on the augmentation ideal
\[
I=CH_{>0}(A)\oplus CH_0(A)^{\deg=0}.
\]
Beckmann–de Gaay Fortman then revisited the integral problem: rather than requiring a literal integral equality \(\Gamma=\operatorname{ch}(\mathcal P_A)\), they characterized when the rational correspondence can be lifted integrally up to torsion [2410.21094].

## 2. Étale motivic cohomology and the étale Chow ring

The étale-motivic extension replaces ordinary Chow groups by étale motivic Chow groups defined in the framework of Cisinski–Déglise’s \(h\)-motives. For a finite type \(k\)-scheme \(X\),
\[
CH^n_{\mathrm{ét}}(X)
=
\operatorname{Hom}_{\underline{DM_h(k,\mathbf Z)}}\big(M_h(X),\mathbf 1_k(n)[2n]\big),
\]
and
\[
CH^{\mathrm{ét}}_n(X)
=
\operatorname{Hom}_{\underline{DM_h(k,\mathbf Z)}}\big(M_h(X)(n)[2n],\mathbf 1_k\big).
\]
When \(X\) is smooth projective of dimension \(d\), these groups identify with classical motivic cohomology, compatibly with pushforward and pullback. For an abelian variety \(A\), the étale Chow ring \(CH_*^{\mathrm{ét}}(A)\) therefore serves as an “étale” version of the Chow ring defined at the motivic level [2410.21094].

This setting is compared in the paper with Lichtenbaum cohomology, defined via the étale sheafification of Bloch’s cycle complex. For smooth projective \(X\) of characteristic \(p\), Cisinski–Déglise show that
\[
\mathbb H^m_{\acute et}(X,\mathbf Z(n))[1/p]\simeq H^m_M(X,\mathbf Z(n))[1/p].
\]
Thus the étale motivic Chow groups used in the construction are compatible with Lichtenbaum cohomology after inverting the characteristic. The ring \(CH_*^{\mathrm{ét}}(A)\) carries both the intersection product and, because \(A\) is a commutative monoid, the Pontryagin product \(x*y=\mu_*(x\times y)\). Its positive-degree part is
\[
CH^{\mathrm{ét}}_{>0}(A)=\bigoplus_{i>0} CH_i^{\mathrm{ét}}(A),
\]
and the augmentation ideal is
\[
I^{\mathrm{ét}}=CH_{>0}^{\mathrm{ét}}(A)\oplus CH_0^{\mathrm{ét}}(A)^{\deg=0}.
\]
These are the principal ideals on which the divided power theory is developed [2410.21094].

Within this category the rational Fourier transform is defined by the same formula as in the classical Chow-theoretic setting, using \(\operatorname{ch}(\mathcal P_A)\). The paper then introduces two integral notions. A homomorphism
\[
\mathcal F^{\mathrm{ét}}\colon CH_*^{\mathrm{ét}}(A)\to CH_*^{\mathrm{ét}}(\widehat A)
\]
is a **weak integral étale Fourier transform** if it lifts the rational transform after tensoring with \(\mathbf Q\); it is **algebraic** if it is induced by a cycle \(\Gamma\in CH_*^{\mathrm{ét}}(A\times \widehat A)\) whose rational image is \(\operatorname{ch}(\mathcal P_A)\). There is also an **integral étale Fourier transform up to homology**, defined by compatibility with the cohomological Fourier transform
\[
\mathfrak F_A\colon H^\bullet_{\acute et}(A_{k^s},\mathbf Q_\ell(\bullet))
\longrightarrow
H^\bullet_{\acute et}(\widehat A_{k^s},\mathbf Q_\ell(\bullet)),
\]
which preserves integral \(\mathbf Z_\ell\)-lattices [2410.21094].

A central role is played by the rational classes
\[
\mathcal R_A=\frac{\ell^{2g-1}}{(2g-1)!}\in CH^{2g-1}_{\mathrm{ét}}(A\times \widehat A)_\mathbf Q
\]
and, under a principal polarization \(\lambda\colon A\to \widehat A\) with symmetric ample class \(\Theta\),
\[
\Gamma_\Theta=\frac{\Theta^{g-1}}{(g-1)!}\in CH^{g-1}_{\mathrm{ét}}(A)_\mathbf Q.
\]
The paper also uses the Pontryagin exponential
\[
E(a)=\sum_{n\ge 0}\frac{a^{*n}}{n!},
\]
together with the fact that \(E((-1)^g\mathcal R_A)\) recovers \(\operatorname{ch}(\mathcal P_A)\). The integrality of these classes controls whether the rational Fourier correspondence can be realized integrally [2410.21094].

## 3. Divided powers, integral equivalences, and algebraically closed fields

A major structural contribution is the construction of a divided power structure for Pontryagin products in the étale-motivic setting. For a commutative ring \(R\) and ideal \(I\subset R\), a PD-structure is a family \(\gamma_d\colon I\to I\) satisfying the standard divided-power identities. In the geometric context, if \(M\) is a quasi-projective commutative monoid scheme over \(k\) with proper multiplication \(\mu\), one defines \(\gamma_d^M\) using symmetric powers \(\operatorname{Sym}^d(M)\), the induced map \(p_d\colon \operatorname{Sym}^d(M)\to M\), and combinatorial formulas analogous to those of Moonen–Polishchuk. The theorem states that the maps
\[
\gamma_d^M\colon CH_{>0}^{\mathrm{ét}}(M)\to CH_*^{\mathrm{ét}}(M),\qquad
x\mapsto (p_d)_*\gamma_d(x),
\]
extend uniquely to a PD-structure on \(CH_{>0}^{\mathrm{ét}}(M)\) for the Pontryagin product, functorially in proper morphisms of graded monoids [2410.21094].

For an abelian variety over an algebraically closed field, the result strengthens to the augmentation ideal:
\[
I^{\mathrm{ét}}=CH_{>0}^{\mathrm{ét}}(A)\oplus CH_0^{\mathrm{ét}}(A)^{\deg=0}\subset CH_*^{\mathrm{ét}}(A).
\]
This is presented as the exact étale motivic analogue of Moonen–Polishchuk’s PD-structure on the usual Chow ring. The key input is that over algebraically closed fields one has \(CH_0(A)\simeq CH_0^L(A)\), so the étale version behaves like the classical zero-cycle theory [2410.21094].

The Fourier-transform existence theorem identifies integrality, divided powers, and torsion-free preservation as equivalent phenomena. For an abelian variety \(A/k\) of dimension \(g\), the following are equivalent: the cycle \(\mathcal R_A\) lifts integrally; \(A\) admits an étale motivic weak integral Fourier transform; \(A\times \widehat A\) admits one as well; and, under a principal polarization, equivalent conditions include the lifting of
\[
\frac{c_1(\mathcal P_A)^{2g-2}}{(2g-2)!}
\quad\text{and}\quad
\Gamma_\Theta,
\]
the existence of a principally polarized weak integral étale Fourier transform, preservation of the torsion-free étale Chow groups by the rational Fourier transform, and existence of a PD-structure on
\[
CH^{\mathrm{ét}}_{>0}(A)_{\mathrm{tf}}\subset CH_*^{\mathrm{ét}}(A)_{\mathrm{tf}}.
\]
A parallel equivalence is proved for \(\mathbf Z_\ell\)-cohomology, where “lifting to the étale Chow group” is replaced by algebraicity of the relevant cohomology class [2410.21094].

The paper then derives existence results over several classes of fields. Over \(\mathbf C\), finitely generated fields over \(\mathbf Q\), and finite fields, one obtains algebraic integral Fourier transforms up to homology in the appropriate Betti or \(\ell\)-adic sense. Most strikingly, for an algebraically closed field \(k\), an abelian variety \(A/k\) admits an integral algebraic étale Fourier transform if and only if it admits algebraic \(\ell\)-adic integral Fourier transforms up to homology for all primes \(\ell\neq \mathrm{char}(k)\); in particular, such an integral algebraic étale Fourier transform always exists. In this sense, the integrality obstruction disappears over algebraically closed fields. The same framework is used to relate the Fourier symmetry of \(A\) and \(\widehat A\) to integral Hodge and Tate-type statements, and to explicit Chow–Künneth projectors for the étale motive \(M_h(A)\) [2410.21094].

## 4. Additive motivic Fourier transform in motivic integration

A different but complementary theory places Fourier transform inside motivic integration over valued fields. In the framework of the paper on motivic Mellin transforms, the language is an expansion of the Denef–Pas language, with valued-field sort \(VF\), value-group sort \(VG\) endowed with Presburger structure, and residue-ring sorts \(R_t\). The motivic functions of interest live in rings \(\mathcal C(X)\) and \(\mathcal C^{\exp}(X)\), whose generators encode geometric classes, additive-character data through imaginary functions \(g\colon Z\to VF/O\), and residue-ring characters through functions \(h\colon Z\to R_i\) [2412.17764].

In this setting the motivic Fourier transform is defined for integrable \(F\in \mathcal C(VF^n)\) by
\[
\hat F(y)=\int_{x\in VF^n} F(x)\,E(x\cdot y),
\]
where \(E(g)\) is the motivic additive-character factor attached to the imaginary function \(g(x,y)=x\cdot y\). The same formalism simultaneously supports Mellin transforms, but the Fourier kernel is specifically additive. Stability under integration implies that if \(F\in \mathcal C^{\exp}(VF^n)\) is integrable, then \(\hat F\) again lies in \(\mathcal C^{\exp}(VF^n)\). Thus the transform is closed on the class of exponential constructible motivic functions [2412.17764].

The theory includes structural analogues of classical harmonic analysis. Theorem 5.18 asserts injectivity up to measure-zero sets: if two integrable functions have the same Fourier transform, then they coincide almost everywhere. Proposition 6.9 gives the inversion formula
\[
\hat{\hat F}=\mathbb L^{-n}F^\vee
\]
almost everywhere, where \(F^\vee(x)=F(-x)\). Convolution is defined motivically and satisfies the expected identity \((F*G)\,\widehat{}=\hat F\,\hat G\). Fubini and change of variables are proved in general for the underlying integration theory, so iterated Fourier-type integrals and parameter-dependent transforms are handled in the same framework [2412.17764].

A further feature is specialization. Motivically integrable functions specialize to \(L^1\) \(p\)-adic functions for local fields of sufficiently large residue characteristic, and motivic integrals commute with specialization. Consequently, the motivic Fourier transform specializes to the classical \(p\)-adic Fourier transform
\[
(\hat f)(y)=\int_{x\in K^n} f(x)\,\psi(x\cdot y)
\]
for a chosen additive character \(\psi\). The framework also supports a transfer principle between local fields with isomorphic residue rings up to finite depth. The paper presents this as a unification and extension of earlier exponential motivic integration, now enlarged so that Fourier and Mellin transforms are treated in the same algebraic formalism [2412.17764].

## 5. Multiplicative and toric motivic Fourier analysis

For split algebraic tori, motivic Fourier analysis takes a multiplicative form. The paper on the motivic Poisson formula for split algebraic tori constructs a Grothendieck ring \(K_0^{\mathrm{Char}_M}\mathrm{Var}_S\) of varieties with multiplicative characters, where the geometric factor and the character factor are kept simultaneously. After localizing one obtains
\[
\mathscr C^{\mathrm{Char}_M}\mathrm{Var}_S
=
K_0^{\mathrm{Char}_M}\mathrm{Var}_S[\mathbf L_S^{-1}],
\]
and there is a canonical integration-over-characters operator
\[
\int_{D(\underline M_S)}\colon
K_0^{\mathrm{Char}_M}\mathrm{Var}_S\to K_0\mathrm{Var}_S
\]
characterized by orthogonality:
\[
\int_{D(\underline M_S)} \ev(m,\chi)\,d\chi
=
\begin{cases}
1 & m=0,\\
0 & m\neq 0.
\end{cases}
\]
This is the multiplicative analogue of the cancellation property underlying Poisson summation [2604.03162].

Given a finitely generated abelian group \(M\), a finite-support motivic function \(\psi\in \mathscr M_{\underline M_S}\) has Fourier transform
\[
\mathcal F\psi=\psi\cdot \ev,
\]
equivalently
\[
(\mathcal F\psi)(\chi)=\sum_{m\in M}\psi(m)\chi(m).
\]
The inversion formula integrates \(\mathcal F\psi\) against \(\ev(-x,\chi)\), and the local motivic Poisson formula expresses sums over a subgroup \(H\subset G\) as integrals over the subgroup of characters trivial on \(H\):
\[
\sum_{h\in H}\psi(g+h)
=
\int_{D(G/H)_S}\mathcal F\psi(\chi)\,\ev(-g,\chi)\,d\chi.
\]
Here there is no Haar measure or locally compact topology; the paper emphasizes that integration is realized as pushforward in Grothendieck rings and convergence is controlled by the dimensional filtration [2604.03162].

The global form is built for a split torus \(U\simeq \mathbb G_m^n\) over the function field \(F=k(C)\) of a smooth projective geometrically irreducible curve \(C/k\). Local \(U(\mathcal O_v)\)-invariant functions factor through the cocharacter lattice \(\mathcal X_*(U)\), and global adelic families are encoded by symmetric products of \(C\) and of constant group schemes \(\underline{\mathcal X_*(U)}_C\). Rational points \(U(F)\) are modeled via principal divisors \(\PDiv(C)^n\), and the global motivic multiplicative Poisson formula states that for a family \(f\) of global \(K(U)\)-invariant functions of level \(\pi\),
\[
\sum_{\PDiv^\pi} f
=
\int_{(\PDiv^\pi)^\perp}\!\four f.
\]
This is the motivic Poisson formula for the split torus \(U\) [2604.03162].

The principal application is to motivic height zeta functions of smooth projective split toric compactifications \(X_\Sigma\). The motivic multivariate height zeta function
\[
\zeta_H^{\mathrm{mot}}(\mathbf T)
=
\sum_{\delta\in \Pic(X_\Sigma)^\vee}
\big[\Hom_k^\delta(C,X_\Sigma)_U\big]\mathbf T^\delta
\]
is rewritten as a motivic adelic sum, then transformed by the global Poisson formula into an integral over characters. The local Fourier transform of the height factor is computed explicitly in terms of the fan \(\Sigma\) and the toric polynomial \(Q_\Sigma\), producing motivic \(L\)-factors and polyhedral \(\mathfrak L\)-functions. The resulting analysis yields meromorphic continuation, an explicit leading term at the main pole, and a multivariate stabilization theorem described as the motivic Batyrev–Manin–Peyre principle for split toric varieties [2604.03162].

## 6. Spherical varieties, Braverman–Kazhdan descent, and the scope of the term

A further development appears in the theory of spherical varieties attached to triples of quadratic spaces. Let \(V_1,V_2,V_3\) be even-dimensional quadratic spaces over a number field \(F\), and let \(Y\subset V_1\oplus V_2\oplus V_3\) be the closed subscheme cut out by
\[
\mathcal Q_1(v_1)=\mathcal Q_2(v_2)=\mathcal Q_3(v_3).
\]
The space \(Y\) is a spherical variety for a reductive group \(H\) defined by equal similitude factors. The relevant Fourier transform is not defined directly by an explicit kernel on \(Y\); instead, \(Y\) is obtained from a Braverman–Kazhdan space \(X\) for \(\mathrm{Sp}_6\) by an integral transform
\[
I(f)(\xi)
=
\int_{G_{\gamma_b}(\mathbb A_F)\backslash G(\mathbb A_F)}
f_1(\gamma_b g)\,\rho(g)f_2(\xi)\,dg,
\]
where \(G\simeq SL_2^3\) and \(\rho\) is the corresponding Weil representation. The local and global Schwartz spaces on \(Y\) are defined as the images of the corresponding maps \(I_v\) and \(I\) [2009.11490].

The central result is that the Braverman–Kazhdan Fourier transform on \(X\) descends to a genuine operator on \(Y\). Assuming \(Y^{\mathrm{sm}}(\mathbb A_F)\neq\emptyset\), there is a unique \(\mathbf C\)-linear isomorphism
\[
\mathcal F_Y\colon \mathcal S(Y(\mathbb A_F))\xrightarrow{\sim}\mathcal S(Y(\mathbb A_F))
\]
such that
\[
I\circ \mathcal F_X=\mathcal F_Y\circ I,
\]
and similarly at each place \(v\). The paper describes this as the first time a summation formula with boundary terms has been proven for a spherical variety that is not a Braverman–Kazhdan space [2009.11490].

The corresponding Poisson formula contains explicit boundary terms indexed by non-open orbits. Besides the main sum over \(Y^{\mathrm{sm}}(F)\), there are contributions from quotient varieties \(Y_0\) and \(Y_i\) together with residues at \(s=1\) of auxiliary integrals \(I_i(f\otimes \Phi)(\xi,s)\). Under stronger local cusp-type conditions, these boundary terms vanish and one obtains an intrinsic Poisson summation formula
\[
\sum_{y\in Y^{\mathrm{sm}}(F)} f(y)
=
\sum_{y\in Y^{\mathrm{sm}}(F)} \mathcal F_Y(f)(y)
\]
for pure tensors in \(\mathcal S(Y(\mathbb A_F))\). The construction is therefore a prototype for Fourier transforms on non-BK spherical varieties via descent from a BK parent space [2009.11490].

A common misconception is that “motivic Fourier transform” denotes a single operator. The current literature instead exhibits several technically distinct constructions: a correspondence on Chow or étale motivic Chow groups for abelian varieties, an additive integral transform on exponential constructible motivic functions, a multiplicative Poisson formalism on cocharacter lattices and adelic symmetric products for split tori, and a descended Schwartz-space operator on spherical varieties [2410.21094], [2412.17764], [2604.03162], [2009.11490]. This suggests a unifying principle—Fourier duality encoded by algebraic kernels, motivic integration, and Poisson summation—while also showing that the dual objects, functional spaces, and integrality questions depend sharply on the ambient category.

Source: https://www.emergentmind.com/topics/motivic-fourier-transform