Grothendieck Period Conjecture

Grothendieck Period Conjecture

Definitions and mathematical statement

For a Nori motive $M/{\mathbb{Q}}$, its Betti and de Rham realizations are finite-dimensional rational vector spaces compared over ${\mathbb{C}}$. Their comparison matrix, in rational bases, supplies the periods of $M$. Its motivic Galois group $G_{\rm mot}(M)$ is the affine algebraic group of tensor automorphisms of the fiber functor on the tensor category generated by $M$. The conjecture is \[ \operatorname{trdeg}_{{\mathbb{Q}}}{\mathbb{Q}}(\text{entries of the period matrix of }M) =\dim G_{\rm mot}(M). \] Changes of rational bases do not change the generated period field. The Nori-motive convention and the category generated by $M$ are those of the source.

Formulation and concept references: [S1].

Short English statement

Are all algebraic relations among a motive's periods exactly those forced by its motivic symmetries?

Research attempt

Current frontier.

The target is the arithmetic statement for Nori motives over ${\mathbb{Q}}$, not a statement about period functions over a complex function field. Bakker and Tsimerman formulate the former as Conjecture 1.1 and prove a functional analogue in Theorem 1.1, made precise in Theorem 4.25. Their Sections 4.2--4.3 fix the Nori category and period-field conventions used here. The difference between the constant and functional settings is essential. [S1], [S2] Kreutz--Shen--Vial, Corollary 6.17(i), prove the period-conjecture statement for K3 surfaces of Picard rank $20$; Kawabe's self-product CM Kummer case is a subcase, as his revised Remark 3.1 acknowledges. Neither result covers all Nori motives. Statements for motivated classes or for de Rham--Betti classes must also be distinguished from the full arithmetic equality. [E1], [S3] Bertolin's work on $1$-motives explicitly separates general conjectures of Grothendieck--André type from the CM/torsion cases treated there. [E3]

Attack route.

Three routes deserve separate tests. One can try to make the ideal of numerical period relations invariant under the motivic group; transitivity would then force the desired genericity. One can try to transfer functional transcendence to every arithmetic specialization, but this needs a specialization theorem with exceptional fibers controlled. Finally, one can attempt a bounded-degree search for all period relations, which needs an a priori relation-degree bound, not just Noetherianity. The calculations below expose the exact gaps in these proposals. They also test the desired invariance on a small logarithmic motive where the outstanding arithmetic independence is already substantial.

Attempt.

1. A component-aware genericity criterion. Let $G=G_{\rm mot}(M)$ and let \[ \mathcal P=\operatorname{Isom}^{\otimes} (\omega_{\rm dR},\omega_B)\big|_{\langle M\rangle} \] be the comparison torsor. These are the fiber functors on the tensor category generated by $M$, including duals and subquotients. The comparison isomorphism determines a complex point $p\in\mathcal P({\mathbb{C}})$. The torsor formalism and its relation to the period field are those in [S2], Sections 4.2, 4.3, and 4.6. The following commutative-algebra argument is supplied here to track the direction of the implication.

Writing $A={\mathbb{Q}}[\mathcal P]$, evaluation defines a prime ideal $I=\ker(A\to{\mathbb{C}})$. Coordinates of $\mathcal P$ are generated by matrix entries of the comparison on $M$ and the inverse determinant, subject to the tensor relations. Inverting a nonzero determinant does not change the field generated by its entries. Thus \[ \operatorname{Frac}(A/I)={\mathbb{Q}}(\mathrm{periods}(M)),\qquad \operatorname{trdeg}_{{\mathbb{Q}}}{\mathbb{Q}}(\mathrm{periods}(M)) =\dim(A/I)\le\dim\mathcal P=\dim G. \tag{49.1} \] This proves the easy upper bound, not the required lower bound.

There is a useful precaution when $G$ is not connected. Choose a finite number field $K\subset{\mathbb{C}}$ over which the geometric components of $\mathcal P$ are defined, and let $C$ be the component containing $p$. In characteristic zero the torsor is smooth; its geometric connected components are integral, all of dimension $\dim G$. Hence $K[C]$ is a domain, and, putting $J=\ker(K[C]\to{\mathbb{C}})$, we obtain \[ \operatorname{trdeg}_{{\mathbb{Q}}}{\mathbb{Q}}(\mathrm{periods}(M))=\dim G \quad\Longleftrightarrow\quad J=0. \tag{49.2} \] Indeed, adjoining $K$ changes no transcendence degree; a nonzero prime ideal in the coordinate ring of an integral finite-type variety cuts out a proper subvariety of strictly smaller dimension. The fraction field of the image here is $K(\mathrm{periods}(M))$. A formulation using injectivity on the whole coordinate ring without addressing components would require an extra irreducibility justification.

2. The proposed group-invariance step. Suppose the $K$-Zariski closure $Z$ of $p$ inside $C$ were stable under the connected group acting transitively on $C$. After extension to ${\mathbb{C}}$, choose any point of the nonempty closed set $Z_{{\mathbb{C}}}$. Its group orbit is all of $C_{{\mathbb{C}}}$, so $Z_{{\mathbb{C}}}=C_{{\mathbb{C}}}$, and therefore $J=0$. In ideal language, a sufficient missing lemma is that the numerical relation ideal $J$ is a subcomodule for this group action.

No proof of that lemma is obtained. A motivic automorphism acts on the space of tensor-compatible comparison isomorphisms; it has not thereby been shown to induce an automorphism of the field of their actual complex values. To infer $R(gp)=0$ from $R(p)=0$ would be to assume the very stability being sought. The torsor gives an algebraic space in which the periods lie, but it does not make the distinguished complex point automatically generic. For a transitive component, this stability condition is equivalent to genericity itself, not a known independently weaker arithmetic theorem.

3. A finite-degree obstruction to a shortcut. The following exact model shows why checking a large finite collection of tensor or polynomial relations cannot be promoted to unrestricted algebraic independence without a further theorem. Fix $D\ge1$, choose any transcendental complex number $\tau$, and set \[ T=\mathbb G_m^2,\qquad p_D=(\tau,\tau^{2D+1}). \] For Laurent monomials $X^aY^b$ with $|a|,|b|\le D$, evaluation gives $\tau^{a+(2D+1)b}$. These exponents are all distinct: equality would imply \[ (2D+1)(b-b')=a'-a, \] whose right side has absolute value at most $2D$, forcing $b=b'$ and then $a=a'$. Consequently no nonzero Laurent polynomial supported in that box vanishes at $p_D$, even with coefficients in a fixed number field. Multiply by a sufficiently large power of $\tau$ and use its transcendence to see this explicitly. Nevertheless \[ Y-X^{2D+1}=0\quad\hbox{at }p_D,\qquad \operatorname{trdeg}_{{\mathbb{Q}}}{\mathbb{Q}}(p_D)=1<2=\dim T. \tag{49.3} \] The model passes every such degree-$D$ test and still fails genericity. It is not asserted to be the comparison torsor and period point of any Nori motive. It falsifies only a general algebraic inference used in the proposed proof method. Noetherianity says that an already given ideal has finitely many generators; it supplies neither their numerical identification nor the needed uniform degree bound.

4. What functional transcendence does and does not supply. For a basic test, the germ $\log t$ is transcendental over ${\mathbb{C}}(t)$. If it were algebraic, continuation around a base point outside the finite branch locus of its polynomial equation could produce only finitely many values. Continuation around $0$ instead gives $\log t+2\pi i k$ for all integers $k$, a contradiction. Yet the branch with $\log1=0$ has an algebraic value at $t=1$.

This is an obstruction to the inference "a function is transcendental, therefore every specified algebraic specialization has the full expected transcendence degree." It is not a counterexample to the arithmetic period conjecture. In particular, the simple relative-pair representation using marked points $\{1,t\}$ changes when the points collide at $t=1$; one may not silently preserve that motive's group dimension at the collision. The example isolates the missing specialization control, rather than claiming that a theorem about nondegenerate families has been contradicted. Bakker--Tsimerman's functional theorem is used with its stated quantifiers, not as such an arithmetic specialization theorem. [S2]

5. An actual logarithmic motive as an arithmetic stress test. Take the Nori motive \[ M_0=\mathcal H^1(\mathbb G_m,\{1,2,3\}). \] Its period field is ${\mathbb{Q}}(2\pi i,\log2,\log3)$, and its motivic Galois group is an extension of $\mathbb G_m$ by $\mathbb G_a^2$, hence has dimension $3$. This is the case $\alpha_1=2,\alpha_2=3$ of the explicit calculation in [S2], Example 4.4. Thus this particular instance requires \[ \operatorname{trdeg}_{{\mathbb{Q}}}{\mathbb{Q}}(2\pi i,\log2,\log3)=3. \tag{49.4} \] The elementary linear check does go through. If $a(2\pi i)+b\log2+c\log3=0$ with rational coefficients, imaginary parts give $a=0$. Clearing denominators and exponentiating gives $2^B3^C=1$; unique prime factorization forces $B=C=0$. This proves linear, not algebraic, independence.

Schanuel's conjecture applied to the three linearly independent numbers $2\pi i,\log2,\log3$ would imply $49.4$, since their exponentials are $1,2,3$. This flags a dependency on the separate unresolved conjecture, not a proof of $49.4$. Conversely, nothing here claims that the original ${\mathbb{Q}}$-motive period conjecture implies full Schanuel for arbitrary complex numbers. The broader Grothendieck--André conjecture over arbitrary subfields of ${\mathbb{C}}$ is a different statement; the distinction is explicit in [S2], Section 4.3, and [E3], Section 6.

6. A sector where both sides can actually be calculated. For comparison, let $M$ be a direct sum of an Artin motive and finitely many pure Tate motives ${\mathbb{Q}}(m_j)$. This does not include arbitrary extensions of these objects. Artin periods are algebraic, while the Tate periods are integral powers of $2\pi i$ or their inverses. If every $m_j=0$, the period field is algebraic and the motivic group is finite, so both quantities are zero. If some $m_j\ne0$, the group has a one-dimensional torus image, with only a finite Artin part, and the period field has transcendence degree exactly one. Here the lower bound follows from the transcendence of $\pi$. [E2] These are familiar special cases, included as a check of dimensions, finite components, and twists, not as a newly discovered theorem.

Stress test.

The upper bound $49.1$ survives changes of rational bases and adjoining a finite coefficient field. The component argument does not assume a rational point on the torsor. The proposed invariance lemma, however, is not supplied by the definition of a motivic automorphism. The finite-degree model $49.3$ shows exactly how hidden nonlinear relations can escape every prescribed finite test; it says nothing against the additional geometric structure of genuine motives. Likewise, the logarithm specialization is a test of an inference, not a motive with a verified dimension discrepancy.

No numerical independence computation is reported as evidence for a proof: a small nonzero residual cannot establish algebraic independence, and an approximate integer relation cannot establish an exact one. The proof of linear independence in the genuine test $49.4$ does not control quadratic or higher relations. Extending the Artin--Tate calculation to nontrivial mixed extensions already encounters such logarithmic periods.

Outcome.

Outcome: Exploratory approach with unresolved gap.

The attempt supplies a component-aware genericity formulation, spells out the exact unproved group-invariance step, and proves a finite-degree countermodel to a possible shortcut. It also reduces a concrete logarithmic motive to the explicit algebraic-independence requirement $49.4$ and checks a known direct-sum Artin--Tate sector. It produces neither a general arithmetic lower bound nor a Nori-motive counterexample. A resolution by this route needs a genuinely arithmetic theorem preventing proper numerical relation loci, not just the formal symmetries of the torsor, functional transcendence, or a finite number of numerical tests. No claim of historical novelty is made for these auxiliary deductions.

Sources

• S1 Formal statement, Conjecture 1.1. https://www.cambridge.org/core/journals/forum-of-mathematics-sigma/article/functional-transcendence-of-periods-and-the-geometric-andregrothendieck-period-conjecture/6B3E39BB9BC68DC2BC6E793D15B0735B. Catalogue formulation source.
• S2 Functional Transcendence of Periods and the Geometric André--Grothendieck Period Conjecture. https://arxiv.org/html/2208.05182v2. Catalogue research/status reference; not a proof endorsement.
• S3 Grothendieck's period conjecture for Kummer surfaces of self-product CM type. https://arxiv.org/abs/2303.05030. Catalogue research/status reference; not a proof endorsement.
• E1 Tobias Kreutz, Mingmin Shen, and Charles Vial, De Rham--Betti classes with coefficients, arXiv:2206.08618v3, especially Corollary 6.17(i) and Remark 6.18. https://arxiv.org/html/2206.08618v3. Added primary source; the Picard-rank-20 K3 result is distinguished from the general conjecture. Consulted 27 September 2026.
• E2 Ferdinand Lindemann, Ueber die Zahl $\pi$, Mathematische Annalen 20 $1882$, 213--225, DOI:10.1007/BF01446522. https://doi.org/10.1007/BF01446522. Classical transcendence of $\pi$, used only for the Tate-sector check. Publisher bibliographic record checked 27 September 2026; the subscription full text was not inspected.
• E3 Cristiana Bertolin, Schanuel conjecture for 1-motives, arXiv:2509.08700v1, Sections 6--7. https://arxiv.org/html/2509.08700v1. Added primary source; the broader Grothendieck--André formulation and the restricted CM/torsion result, not a proof of the present universal statement. Consulted 27 September 2026.