The Hodge Conjecture

Take a smooth surface drawn by polynomial equations — a projective variety. Inside it live all sorts of loops, sheets and higher-dimensional "shapes." Some of these shapes are special: they are themselves carved out by equations, like a circle traced on a sphere. Mathematicians call those algebraic cycles.

But topology — the floppy, stretchy study of holes — sees many more shapes than that. It packages them into objects called cohomology classes, and among these a refined filter (the Hodge decomposition) singles out the well-behaved ones, the Hodge classes.

The Hodge conjecture asks a deceptively simple question: is every Hodge class secretly built from algebraic cycles? In other words, whenever topology says a shape could be algebraic, is it? Nobody knows. This is one of the seven Millennium Prize Problems, and a correct proof or disproof is worth one million dollars.

Rigor here needs years of algebraic geometry, so instead let's build the feeling. On the surface below, loops appear. Algebraic loops are the ones a polynomial equation can cut out cleanly; transcendental loops are shapes topology can see but no equation traces. The Hodge conjecture is the bet that, among the "Hodge-type" shapes, only the algebraic ones survive.

<p class="hint">{{hint_p1}}</p>
<div class="stage">
  <svg id="surface" viewBox="0 0 300 200" aria-label="{{aria_surface}}"></svg>
</div>
<div class="status" id="status">{{status_initial}}</div>
<div class="witness" id="witness"></div>
<div class="btns">
  <button id="drop" type="button">{{btn_drop}}</button>
  <button id="reveal" type="button">{{btn_reveal}}</button>
  <button id="reset" type="button" class="ghost">{{btn_reset}}</button>
</div>

* { box-sizing: border-box; }
body { font-family: system-ui, sans-serif; color: #222; margin: 0; }
.hint { font-size: .9rem; color: #444; margin: 0 0 .7rem; line-height: 1.45; }
.stage { background: #f3f6f9; border: 1px solid #cdd9e3; border-radius: 10px; padding: 6px; }
svg { width: 100%; height: auto; display: block; }
.surface-body { fill: #dde7ef; stroke: #9fb2c8; stroke-width: 1.5; }
.loop-alg { fill: none; stroke: #0a7d33; stroke-width: 2.6; }
.loop-tr  { fill: none; stroke: #c92f3c; stroke-width: 2.6; stroke-dasharray: 5 4; }
.eqline   { stroke: #1d3557; stroke-width: 1.4; stroke-dasharray: 3 3; opacity: .85; }
.status { font-size: 1rem; font-weight: 600; margin: .6rem 0 .2rem; min-height: 1.4em; }
.status.ok  { color: #0a7d33; }
.status.bad { color: #c92f3c; }
.witness { font: 500 .85rem ui-monospace, monospace; color: #1d3557; min-height: 1.2em; margin-bottom: .5rem; }
.btns { display: flex; gap: .5rem; flex-wrap: wrap; }
button { font: 600 14px system-ui, sans-serif; padding: .45rem .9rem; border: 1px solid #1d3557;
         background: #1d3557; color: #fff; border-radius: 8px; cursor: pointer; }
button.ghost { background: #fff; color: #1d3557; }

// Code not found

Click Test a loop to drop a candidate shape, then Is it algebraic? to reveal its verdict. Notice the asymmetry: when a loop is algebraic you can point to the equation that draws it — that's a witness anyone can check. When it is only topological, no such certificate exists, and deciding that in full generality is exactly what no one has managed to do.

How hard is the Hodge conjecture, really?

• It is OPEN. First articulated by W. V. D. Hodge in his 1950 address to the International Congress of Mathematicians, it was named one of the seven Millennium Prize Problems by the Clay Mathematics Institute in 2000. No proof and no counterexample is known.
• The statement is precise. For a smooth projective complex variety, every rational Hodge class of degree $(p,p)$ should be a rational combination of the classes of algebraic subvarieties. Easy to write, brutal to verify.
• We only know fragments. The Lefschetz $(1,1)$ theorem settles the very first case — divisors, degree-2 classes — completely. Beyond that, the conjecture is proven only in scattered special families; the general case is wide open.
• The obstacle is the gap itself. Topology hands you a Hodge class with no recipe attached. Turning "this class is allowed to be algebraic" into "here is the actual subvariety" requires constructing geometry out of pure algebra of cohomology — and there is no general machine for that.

Unlike a problem in P vs NP, where we can at least bound the difficulty, here we lack even a strategy. The Hodge conjecture sits at the frontier where topology, algebra and geometry refuse to fully agree.

"Which abstract classes come from real geometry?" sounds esoteric, but it is load-bearing for huge parts of modern mathematics:

• The theory of motives: Grothendieck's dream of a universal cohomology rests on understanding algebraic cycles — exactly what Hodge governs.
• Arithmetic geometry: bridges between topology and number theory, including how solutions to equations are counted, lean on cycle-class maps being well understood.
• The Standard Conjectures: a web of deep statements about algebraic geometry would partly follow from, or interlock with, the Hodge conjecture.
• Computation and certificates: the asymmetry between exhibiting an algebraic cycle (a checkable witness) and proving none exists is the same verifier-versus-search gap that animates complexity theory.

Even unsolved, the conjecture organizes how mathematicians think about the boundary between the algebraic and the merely topological — the same line you met informally in the demo, and a cousin of the ideas behind SAT and verifiable certificates.

The Hodge conjecture distills a single haunting idea: when topology whispers that a shape could be drawn by equations, is it always so? For divisors the answer is yes; everywhere else, after seventy years, we genuinely do not know.

It remains open, one of the Millennium Prize Problems, with a million dollars waiting for whoever can finally tell algebra apart from topology — or prove the two part ways. Until then, the cleanest summary is the one the demo leaves you with: spotting an algebraic shape is easy once you hold its equation; proving none exists may be out of reach, much as it is for P vs NP.