Twisted Periods: Theory and Applications

• Twisted periods are period integrals modified by extra data (e.g., a character or quadratic extension) that encode deep arithmetic, geometric, and analytic structures.
• They appear in diverse areas such as automorphic forms, modular symbols, and twisted (co)homology, linking special L-values with hypergeometric and Feynman integrals.
• This framework unifies concepts across mathematics and physics, providing explicit formulas for central L-values, period relations, and symmetry operations in graph theory.

Twisted periods are period integrals in which the underlying subgroup, cycle, or integrand is modified by additional data such as a character, a quadratic extension, or a flat connection. In the automorphic setting they are integrals of automorphic forms over tori or symmetric subgroups with a character inserted in the integrand; in the theory of modular forms they encode twisted critical values $L(f,\chi,s)$; in twisted (co)homology they are pairings between twisted Betti homology and twisted de Rham cohomology; and in mathematical physics they appear as exponential periods and as equalities among graph periods produced by twist operations (Martin, 2010, Goto, 2013, Massidda, 31 Mar 2026, Schnetz, 5 May 2025).

1. General pattern of the notion

A period, in the basic automorphic sense, is an integral of an automorphic form over a closed subgroup, typically a torus. A twisted period is obtained by inserting a character in that integral. For $\mathrm{GL}_2$, the prototype is the toric integral

$$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$

where $T(F)\cong E^\times$, $D$ is a quaternion algebra over $F$ containing $E$, and $\pi_D$ is the Jacquet–Langlands transfer of $\pi$ (Martin, 2010).

A related local representation-theoretic usage occurs for

$$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$

whose elements are called twisted local linear periods when $\mathrm{GL}_2$0 (Chen et al., 2017). In the geometric-analytic setting of rank one locally symmetric spaces, the twisted geodesic period is

$\mathrm{GL}_2$1

where the restriction of $\mathrm{GL}_2$2 to the totally geodesic cycle $\mathrm{GL}_2$3 is tested against another Laplace eigenfunction $\mathrm{GL}_2$4 (Möllers et al., 2017).

In twisted (co)homology, the same term refers to integrals such as

$\mathrm{GL}_2$5

interpreted as pairings between twisted Betti homology and twisted de Rham cohomology, with differential $\mathrm{GL}_2$6 or $\mathrm{GL}_2$7 (Massidda, 31 Mar 2026). Hypergeometric examples use multivalued integrands $\mathrm{GL}_2$8 and twisted cycles $\mathrm{GL}_2$9, so that Lauricella functions or the Wirtinger integral become twisted periods in the same sense (Goto, 2013, Goto et al., 21 Nov 2025).

2. Toric periods, $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$0, and central $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$1-values

For a number field $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$2, a cuspidal automorphic representation $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$3 of $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$4 with trivial central character, a quadratic extension $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$5, and a unitary Hecke character

$$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$6

the twisted Rankin–Selberg $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$7-function is

$$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$8

Its center is $$P_D(\phi)=\int_{Z(\mathbb{A}_E)\,T(\mathbb{A}_F)\backslash T(\mathbb{A}_E)} \phi(t)\,\chi^{-1}(t)\,dt,$$9, and under the sign condition

$T(F)\cong E^\times$0

the central value is expressed by a Waldspurger-type formula as the square of a twisted toric period (Martin, 2010).

The basic schematic identity is

$T(F)\cong E^\times$1

with $T(F)\cong E^\times$2 a suitable test vector. In Waldspurger’s formula one has

$T(F)\cong E^\times$3

and in the Martin–Whitehouse refinement the Gross–Prasad test vector yields a fully explicit constant in terms of conductor, discriminants, ramification degrees, adjoint $T(F)\cong E^\times$4-values, and archimedean factors (Martin, 2010).

This circle of ideas is simultaneously global and local. The global twisted period factors as

$T(F)\cong E^\times$5

and Tunnell’s theorem gives

$T(F)\cong E^\times$6

with the relevant quaternion algebra $T(F)\cong E^\times$7 characterized by local epsilon factors (Martin, 2010). The relative trace formula then identifies the spectral side built from twisted periods with the split side containing $T(F)\cong E^\times$8, and this produces the explicit central-value formula.

For modular functions on $T(F)\cong E^\times$9, the same toric theme leads to Dirichlet series generated by twisted periods. Fixing a quadratic number field $D$0, Reznikov considers torus periods twisted by Hecke characters of $D$1, and for a Hecke–Maass form $D$2 with coefficients $D$3 in a torus-adapted expansion defines

$D$4

The normalized series $D$5 extends holomorphically to $D$6, while the Eisenstein-series analogue $D$7 extends meromorphically to $D$8 (Reznikov, 2010).

3. Local models, distinction, and multiplicity one

For a local field $D$9 of characteristic zero and $F$0, twisted linear periods are the $F$1-equivariant functionals

$F$2

The main uniqueness statement is that for all but countably many characters $F$3 of $F$4,

$F$5

and in the non-archimedean case the exceptional set is finite (Chen et al., 2017). The proof uses a twisted Gelfand–Kazhdan criterion, invariant distributions satisfying left and right $F$6-equivariance, reduction to matrix spaces and the Lie algebra, Harish–Chandra descent, and a Fourier-analytic vanishing statement on the nilpotent cone (Chen et al., 2017).

The same paper proves uniqueness of twisted Shalika models. For the Shalika subgroup

$F$7

and a character $F$8 allowing a twist on

$F$9

one has

$E$0

for every irreducible admissible smooth representation $E$1 of $E$2 (Chen et al., 2017).

A broader inner-form framework is provided by distinction problems for automorphic representations of general linear groups over division algebras. The relevant periods are linear periods, twisted-linear periods, and Galois periods, and the local-global principle is stated in terms of local distinction, a further local obstruction, and poles of certain global $E$3-functions associated to the underlying involution via the Jacquet–Langlands correspondence (Matringe et al., 30 Aug 2025). In this setting, twisted-linear periods are not defined by inserting a character into the integrand; instead, the subgroup $E$4 is the centralizer of a quadratic $E$5-structure, so the twist is encoded in the involution and the corresponding base-change geometry (Matringe et al., 30 Aug 2025).

4. Modular forms, modular symbols, and arithmetic special values

For $E$6 and an even primitive Dirichlet character $E$7 of conductor $E$8, the twisted cusp form

$E$9

lies in $\pi_D$0, and the twisted periods are

$\pi_D$1

The twisted period polynomial $\pi_D$2 and its two-variable refinement $\pi_D$3 are then organized by the main identity

$\pi_D$4

where $\pi_D$5 is the twisted Kronecker series (Blakestad et al., 2024). For $\pi_D$6 and $\pi_D$7, this recovers Zagier’s identity.

For harmonic weak Maass forms of weight $\pi_D$8, Bruinier shows that the holomorphic Fourier coefficients are periods of algebraic differentials of the third kind. If $\pi_D$9 is a normalized newform, $\pi$0 satisfies $\pi$1, and $\pi$2 is a fundamental discriminant with $\pi$3, then

$\pi$4

where $\pi$5 is the normalized differential of the third kind attached to the twisted Heegner divisor $\pi$6 (Bruinier, 2011). On the elliptic-curve side this becomes a period formula on the quadratic twist $\pi$7, up to rational factors.

For level one cusp forms, twisted periods are also studied as functionals

$\pi$8

If $\pi$9 is sufficiently large relative to $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$0 and $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$1, then any $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$2 periods with the same twist but different indices are linearly independent; if $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$3 is sufficiently large relative to $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$4, then any $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$5 periods with the same index but different twists mod $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$6 are linearly independent (Ni et al., 22 Jul 2025). The proof passes through traces of products and Rankin–Cohen brackets of Eisenstein series of level $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$7 with nebentypus.

A $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$8-adic variant appears for elliptic curves with split multiplicative reduction at $$\mathrm{Hom}_H(\pi,\chi), \qquad H=\mathrm{GL}_n(k)\times \mathrm{GL}_n(k)\subset \mathrm{GL}_{2n}(k),$$9. Darmon’s automorphic period $\mathrm{GL}_2$00 is compared with the Tate period $\mathrm{GL}_2$01, and the paper proves an equality of refined $\mathrm{GL}_2$02-invariants using twisted versions of refined exceptional zero conjectures: $\mathrm{GL}_2$03 When the conductor is exactly $\mathrm{GL}_2$04 and $\mathrm{GL}_2$05 has conductor $\mathrm{GL}_2$06, the equality is proved unconditionally using de Shalit’s work (Salaza et al., 1 Jun 2026).

5. Twisted homology, hypergeometric functions, and period relations

For Lauricella’s hypergeometric function $\mathrm{GL}_2$07, the Euler-type integral has integrand $\mathrm{GL}_2$08 with multivalued $\mathrm{GL}_2$09, and the associated twisted homology $\mathrm{GL}_2$10 and twisted cohomology $\mathrm{GL}_2$11 satisfy

$\mathrm{GL}_2$12

Goto constructs twisted cycles $\mathrm{GL}_2$13 indexed by subsets $\mathrm{GL}_2$14 such that

$\mathrm{GL}_2$15

gives the canonical basis of $\mathrm{GL}_2$16 local solutions of the differential system $\mathrm{GL}_2$17 (Goto, 2013). Intersection pairings $\mathrm{GL}_2$18 on twisted homology and $\mathrm{GL}_2$19 on twisted cohomology then yield twisted period relations of the form

$\mathrm{GL}_2$20

which become explicit quadratic relations among Lauricella $\mathrm{GL}_2$21 functions (Goto, 2013).

For the Wirtinger integral, the multivalued function

$\mathrm{GL}_2$22

defines local systems $\mathrm{GL}_2$23 and $\mathrm{GL}_2$24 on the elliptic curve $\mathrm{GL}_2$25 (Goto et al., 21 Nov 2025). The twisted periods are

$\mathrm{GL}_2$26

with $\mathrm{GL}_2$27 and $\mathrm{GL}_2$28, and the period matrices satisfy

$\mathrm{GL}_2$29

Using the involution $\mathrm{GL}_2$30, the twisted homology and cohomology decompose into $\mathrm{GL}_2$31-eigenspaces, and the $\mathrm{GL}_2$32 period relation splits into two $\mathrm{GL}_2$33 relations (Goto et al., 21 Nov 2025). This suggests that, in the elliptic uniformization of $\mathrm{GL}_2$34, twisted period relations retain the classical hypergeometric structure but with an additional symmetry decomposition.

6. Geometric analysis, Kloosterman connections, and twisted symmetric squares

On rank one locally symmetric spaces, the twisted geodesic period

$\mathrm{GL}_2$35

is the central object. For $\mathrm{GL}_2$36 and a totally geodesic cycle $\mathrm{GL}_2$37, Theorem A gives a second-moment bound for the modified periods $\mathrm{GL}_2$38, and Corollary B yields pointwise bounds for $\mathrm{GL}_2$39 in terms of Laplace eigenvalues (Möllers et al., 2017). Representation-theoretically, one has

$\mathrm{GL}_2$40

where $\mathrm{GL}_2$41 is an explicit $\mathrm{GL}_2$42-invariant bilinear form and $\mathrm{GL}_2$43 is a global proportionality constant (Möllers et al., 2017).

For the rank-two Kloosterman connection, the twisted symmetric powers

$\mathrm{GL}_2$44

have periods identified with Bessel moments of even degree (Chuang et al., 2023). The period matrix entries are

$\mathrm{GL}_2$45

and the rational structures on Betti homology and de Rham cohomology produce both $\mathrm{GL}_2$46-linear and quadratic relations among these Bessel moments (Chuang et al., 2023).

A different but related use of twisting occurs for Picard–Fuchs operators. The symmetric square of a second-order elliptic Picard–Fuchs operator yields a third-order operator governing K3 periods, and generalized Clausen identities show that

$\mathrm{GL}_2$47

for Apéry-like elliptic operators (Álvarez-García et al., 2021). In this sense the K3 periods are twisted symmetric squares of elliptic periods, and the resulting expressions are globally valid throughout moduli space (Álvarez-García et al., 2021).

7. Exponential periods, Feynman identities, and broader extensions

In the thesis framework for integrals in physics, twisted periods are pairings between twisted Betti homology and twisted de Rham cohomology for exponential or multivalued integrands. Typical examples are

$\mathrm{GL}_2$48

with twisted differential

$\mathrm{GL}_2$49

Baikov Feynman integrals are rewritten as exponential periods by setting $\mathrm{GL}_2$50, and the resulting twisted (co)homology controls master integrals, intersection pairings, wall crossing, and Stokes phenomena (Massidda, 31 Mar 2026).

At the level of period identities, a graph-theoretic twist can preserve Feynman periods. The five-twist identity acts on five-vertex cuts of completed primitive Feynman graphs and produces a new graph with the same period: $\mathrm{GL}_2$51 In $\mathrm{GL}_2$52 theory this identity is independent from the twist, the Fourier identity, and the Fourier split (Schnetz, 5 May 2025). This is a different mechanism from twisted cohomology, but it preserves a period through a combinatorial twist operation.

In a still broader, non-integral usage, Cayley–Dickson algebras may be regarded as twisted group algebras with multiplication

$\mathrm{GL}_2$53

and the paper on Cayley–Dickson twists studies periodicity properties of the twist $\mathrm{GL}_2$54 under shifts by powers of $\mathrm{GL}_2$55 (Bales, 2016). This suggests that the phrase “twisted periods” can also denote periodic behavior of the twist itself rather than a period integral.

Across these settings, twisted periods consistently mediate between symmetry and special values. In automorphic theory they are geometric avatars of twisted central $\mathrm{GL}_2$56-values; in modular and $\mathrm{GL}_2$57-adic arithmetic they package twisted $\mathrm{GL}_2$58-values, modular symbols, and refined $\mathrm{GL}_2$59-invariants; in twisted (co)homology they are period pairings for multivalued or exponential integrands; and in physics they organize both master integrals and nontrivial identities among Feynman periods (Martin, 2010, Blakestad et al., 2024, Massidda, 31 Mar 2026, Schnetz, 5 May 2025). This suggests that “twisted period” is best understood not as a single definition, but as a unifying pattern in which a period acquires arithmetic, geometric, or analytic structure through an additional twist.