There is a particular kind of vertigo that comes from noticing coincidences. When two calculations, performed by entirely different methods on entirely different objects, produce the same numbers, we suspect something deeper is at work. Mathematics is, in one sense, the discipline of taking such suspicions seriously.

In the middle of the twentieth century, Alexander Grothendieck found himself surrounded by such coincidences. Algebraic varieties, those geometric objects cut out by polynomial equations, admitted several different cohomology theories: de Rham cohomology built from differential forms, étale cohomology built from covering spaces in a Grothendieckian sense, Betti cohomology built from ordinary topology. These theories were constructed from wildly different materials, yet they agreed on Betti numbers, on traces of Frobenius, on Euler characteristics. Something was orchestrating the agreement.

Grothendieck proposed that behind these theories stood a common source—a category of motives—from which each cohomology theory would be a mere shadow. The word itself, borrowed from music and painting, suggests the recurring theme that a variety expresses in different keys. It remains, six decades later, one of mathematics' most beautiful conjectures and one of its most stubborn open problems.

A cohomology theory, at its most abstract, is a functor. It takes an algebraic variety and returns a graded vector space, or a family of groups, in a way that respects the underlying geometric maps. What made Grothendieck's situation extraordinary was that several such functors existed simultaneously, each with its own peculiar strengths and defects.

De Rham cohomology, in characteristic zero, captures the calculus of differential forms on a variety. Étale cohomology, working with any prime $\ell$ different from the characteristic, gives access to arithmetic information—Galois actions, zeta functions, the whole subterranean life of a variety over a finite field. Betti cohomology, defined for complex varieties, records the ordinary topology visible when we forget the algebraic structure entirely.

These theories should not agree. They are built from incompatible ingredients using incompatible constructions. Yet comparison theorems keep proving they do agree, at least numerically, and often more strongly. The Weil conjectures rested on this mysterious harmony; their proof by Deligne used étale cohomology while treating it as a stand-in for something more fundamental.

Grothendieck's proposal was to construct a category $\mathcal{M}$ of motives together with a functor $h$ from varieties to $\mathcal{M}$, such that every reasonable cohomology theory factors through $h$. The variety $X$ would have a motive $h(X)$, and each cohomology theory would be a realization functor extracting one aspect of it. The coincidences would not be coincidences; they would be consequences of a shared origin.

This is a characteristically Grothendieckian move. Where others saw parallel constructions, he saw a missing object whose absence created the illusion of parallelism. Find the object, and the parallel structures become projections of a single reality.

Takeaway

When several independent constructions keep producing the same answer, the correct response is not to admire the coincidence but to search for the object of which they are all shadows.

To construct motives, one needs morphisms. What should count as a map between motives? Grothendieck's answer was radical: not just ordinary morphisms of varieties, but algebraic correspondences—cycles on the product $X \times Y$, considered up to a suitable equivalence relation.

The intuition comes from linear algebra. A linear map between vector spaces is determined by its graph, a subspace of the product. Analogously, a correspondence from $X$ to $Y$ is a formal linear combination of subvarieties of $X \times Y$, and composition of correspondences is defined by pulling back, intersecting, and pushing forward. This construction produces a category enriched over abelian groups, with each variety appearing as an object.

The choice of equivalence relation on cycles is delicate. Rational equivalence yields the Chow groups $\mathrm{CH}^*(X)$, the finest invariant. Numerical equivalence is coarsest, identifying cycles that pair identically against all others. Homological equivalence, defined via a cohomology theory, sits between them. Each choice yields a different candidate category of motives, and the relationships among them encode some of the deepest questions in algebraic geometry.

Working with Chow correspondences, one obtains the category of Chow motives. Formally inverting the Lefschetz motive and taking a pseudo-abelian envelope produces a category that behaves like a linear-algebraic universe attached to algebraic geometry. Direct sums exist, tensor products exist, duals exist. The category of motives becomes a Tannakian category—or would, if certain conjectures held.

The philosophical shift here deserves attention. We normally think of maps as functions between sets of points. Correspondences liberate us from this: a map is now a relation, weighted and geometric, capable of describing rich transformations that no function could. Algebraic geometry, seen through motives, becomes a form of linear algebra over an unusually generous base.

Takeaway

Enlarging what counts as a morphism can be more revolutionary than enlarging what counts as an object; correspondences show that the category itself is where the mathematics lives.

For Grothendieck's picture to work as envisioned, certain statements must be true. He formulated them explicitly as the standard conjectures on algebraic cycles. Roughly, they assert that certain operations naturally defined in cohomology—Lefschetz operators, Künneth projectors—are themselves induced by algebraic correspondences.

The conjecture $C(X)$ asserts that the Künneth components of the diagonal are algebraic. The conjecture $D(X)$ asserts that numerical and homological equivalence coincide. The Hodge-type conjecture asserts a positivity property for the pairing on primitive cycles. Together, they would guarantee that the category of numerical motives, built from cycles modulo numerical equivalence, is a semisimple Tannakian category over $\mathbb{Q}$—a mathematical universe as clean as anyone could wish.

Half a century later, these conjectures remain open in essentially their original form. Certain cases are known: for abelian varieties, for varieties over finite fields under strong hypotheses, for a scattering of low-dimensional examples. The Hodge conjecture and the Tate conjecture, both close relatives, are equally open and equally central. We do not know whether every Hodge class on a smooth projective complex variety is algebraic. This is not a technicality; it is a chasm.

Meanwhile, workarounds have flourished. Voevodsky constructed a triangulated category of motives that captures much of Grothendieck's vision without resolving the standard conjectures. Mixed motives, motivic cohomology, and motivic homotopy theory have grown into vast enterprises with their own applications, including Voevodsky's proof of the Milnor conjecture. The dream persists in modified forms.

One must be honest about the situation. Motives, in the pure form Grothendieck imagined, remain conjectural objects. Yet the framework has already reshaped algebraic geometry, guiding intuition and organizing results even where its foundations remain incomplete. It is a theory whose usefulness has run considerably ahead of its proof.

Takeaway

A mathematical framework can be productive long before it is proven consistent; sometimes the vision organizes the work, and the foundations follow when we finally learn how to build them.

Motives illustrate something peculiar about mathematical progress. Grothendieck did not need to construct the category to change the field; the mere articulation of what such a category should be redirected decades of research. Étale cohomology, derived categories, six-functor formalisms—these were built partly to serve a vision whose central object may still elude us.

There is a lesson here about the role of conjecture. A well-posed dream can be more valuable than a proved theorem, because it names a target that a community can approach from many directions. The standard conjectures organize algebraic geometry the way an unseen planet organizes the orbits around it, its existence inferred from perturbations in what we can observe.

Whether motives will eventually be constructed as Grothendieck imagined, or whether the theory will settle into some modified form that captures the essential insight while abandoning the original scaffolding, remains genuinely uncertain. What seems settled is that the search itself has become a permanent feature of mathematics—a reminder that the deepest patterns often reveal themselves slowly, and only to those willing to keep looking.