The Hopf Fibration: Linking the Sphere in Circles

The map: from four dimensions down to two

The construction lives most naturally in complex coordinates. Write the 3-sphere as the unit sphere inside two complex dimensions, S³ = { (z₁, z₂) ∈ ℂ² : |z₁|² + |z₂|² = 1 }. Because ℂ² is the same as ℝ⁴, this is a genuine three-dimensional surface curving through four-dimensional space. The 2-sphere in the target is identified with the Riemann sphere ℂ ∪ {∞}, equivalently the complex projective line ℂP¹.

The Hopf map p: S³ → S² is then astonishingly simple — it just takes the ratio of the two coordinates: p(z₁, z₂) = z₁ / z₂. In honest real coordinates on the unit 2-sphere in ℝ³ the same map reads p(z₁, z₂) = ( 2·Re(z₁ z̄₂), 2·Im(z₁ z̄₂), |z₁|² − |z₂|² ). A one-line check confirms the image lands on the unit sphere: the squared length of that vector is 4|z₁|²|z₂|² + (|z₁|² − |z₂|²)² = (|z₁|² + |z₂|²)² = 1. This same formula reappears verbatim in quantum mechanics as the map from a qubit state to its point on the Bloch sphere, which is one reason the Hopf fibration is not just a curiosity.

Every point pulls back to a circle

Now ask which points of S³ the Hopf map sends to the same place. Two points (z₁, z₂) and (z₁′, z₂′) have the same ratio precisely when they differ by a common complex number of modulus one: (z₁′, z₂′) = (e^{iθ} z₁, e^{iθ} z₂). Multiplying by e^{iθ} cancels in the ratio and leaves the real formula untouched, since |e^{iθ}|² = 1. So the preimage of a single point — the fiber — is the entire orbit { (e^{iθ} z₁, e^{iθ} z₂) : 0 ≤ θ < 2π }.

Geometrically this orbit is the unit circle inside the two-real-dimensional plane spanned by the vectors v = (z₁, z₂) and iv = (iz₁, iz₂). Those two vectors are orthonormal in ℝ⁴ (their real inner product is Re(−i(|z₁|²+|z₂|²)) = 0), so the fiber is a great circle of S³ — a circle of the maximum possible radius, 1. Every fiber is disjoint from every other, because p is a well-defined function. Packaging this up, the Hopf map is a smooth fiber bundle written S¹ ↪ S³ → S²: locally it looks like the product S² × S¹, with the circle fibers as the ‘strands.’ Because the circle acts as a group of unit phases U(1), it is more precisely a principal U(1)-bundle, and its topological twisting is measured by a first Chern number equal to 1.

Why the circles are linked — and why the bundle is not trivial

Here is the heart of the matter. Take any two distinct fibers. They never touch, yet they are linked like two rings of a chain: their linking number is exactly 1, so together they form the classic Hopf link. You cannot slide one clear of the other without one passing through the other. This linking is not decoration — it is the nontriviality of the bundle made visible.

To see the bundle cannot be a mere product, note a clean topological obstruction. If S³ were homeomorphic to the product S² × S¹, its fundamental group would be that of the circle factor, π₁ = ℤ, and it would have first homology ℤ. But S³ is simply connected — every loop contracts, π₁(S³) = 0. So S³ ≅ S² × S¹ is impossible, and the Hopf bundle is genuinely twisted.

You can watch this happen. Apply stereographic projection from S³ into ordinary ℝ³. Each great-circle fiber maps to a round circle (the one fiber through the projection point becomes a straight line, a circle through infinity). All the fibers sitting over one line of latitude on S² sweep out a single torus, and on that torus they are Villarceau circles — the perfectly round slices you get by cutting a doughnut with a diagonal bilateral tangent plane. Each is a (1,1) curve, winding once the long way and once the short way. As the latitude sweeps from pole to pole these tori nest inside one another and fill all of space, a structure called the Hopf foliation. The two polar fibers, a central circle and the straight axis, are themselves a Hopf link at the core of the whole picture.

The Hopf invariant and the shock of π₃(S²)

Before 1931 there was a reasonable expectation that a sphere could not be wrapped nontrivially onto a lower-dimensional sphere. Homology says as much: the reduced homology of S² vanishes in degree 3, and any map S³ → S² is zero on homology. The homotopy groups πₙ(Sᵐ), which count maps up to continuous deformation rather than up to boundaries, were widely assumed to follow suit and vanish for n > m. Hopf demolished that assumption: his map is not deformable to a constant, and in fact π₃(S²) ≅ ℤ, the full group of integers, generated by the Hopf map.

The integer attached to a map f: S³ → S² is its Hopf invariant H(f). One clean definition: pick two regular values on S², look at their preimages (each a link of circles in S³), and take their linking number — that number is H(f), independent of the choices. For the Hopf map itself the fibers link once, so H = 1. An equivalent analytic definition uses differential forms: choosing a 1-form α with dα = f*ω (ω the area form on S²), the Hopf invariant is the integral of α ∧ dα over S³, a purely cohomological cup-product quantity. Because H is additive and takes the value 1 on the Hopf map, that map generates the whole group ℤ.

A family of four: division algebras and Adams' theorem

The complex construction has siblings. Replace ℂ by the other real normed division algebras and repeat the ‘take the ratio’ idea, and you get exactly four Hopf fibrations: the real one S⁰ ↪ S¹ → S¹, the complex one S¹ ↪ S³ → S², the quaternionic one S³ ↪ S⁷ → S⁴, and the octonionic one S⁷ ↪ S¹⁵ → S⁸. In each, total space, base, and fiber are all spheres — a coincidence that provably occurs for no other dimensions.

Why exactly these four? A map S^{2n−1} → Sⁿ of Hopf invariant one exists only for n = 1, 2, 4, 8. This is Adams' Hopf-invariant-one theorem (J. F. Adams, 1960), one of the deepest results of mid-century topology. It is equivalent to several famous facts that once looked unrelated: that ℝ, ℂ, ℍ (quaternions), and ᴼ (octonions) are the only real division algebras with the right structure; that only S¹, S³, and S⁷ among spheres are parallelizable; and that you cannot comb a sphere of any dimension other than these without singularities in the relevant sense. The Hopf fibrations are the geometric face of that rigid arithmetic of dimensions 1, 2, 4, 8.

The 3-sphere as a group: quaternions and SU(2)

The complex Hopf fibration hides an even tidier algebraic story. The 3-sphere is not just a manifold; it is a Lie group. Identify ℝ⁴ with the quaternions ℍ; then S³ is exactly the set of unit quaternions, which multiply among themselves and so form a group. That group is isomorphic to SU(2), the 2×2 special unitary matrices. Inside it, the phases e^{iθ} form a circle subgroup U(1) ≅ S¹ — precisely the fibers.

From this angle the Hopf map is a coset projection: S² = SU(2) / U(1), the space of left cosets of the circle inside the 3-sphere. Each coset is a fiber, each point of the quotient a point of the base. (A close cousin: SU(2)/{±1} = SO(3) = ℝP³ is rotation space itself, and S³ = SU(2) is its double cover, which is why unit quaternions parametrize 3D rotations two-to-one — a related but distinct quotient.) This group-theoretic description generalizes the Hopf fibration to homogeneous spaces throughout geometry and physics.

Where the Hopf fibration shows up

The abstraction pays off in physics. A qubit is a unit vector (α, β) ∈ ℂ² with |α|² + |β|² = 1 — a point of S³. Its overall phase e^{iθ} is physically invisible, so the true physical state is the Hopf image on the Bloch sphere S². The Hopf fibration is literally the map from the state vector to the Bloch sphere, its fibers being the unobservable global phase. Transport a state around a loop and the leftover phase is the Berry (geometric) phase, the holonomy of this very bundle.

The same bundle, as a principal U(1)-bundle with Chern number 1, is the field of a Dirac magnetic monopole; the integrality of the Chern number is exactly why magnetic charge would be quantized. Configurations of a field whose linked level sets carry a nonzero Hopf invariant are called hopfions — knotted, particle-like solitons that appear in the Faddeev–Skyrme model of field theory, in magnetic materials (magnetic hopfions), in structured and knotted light in optics, and in fluid vortices. In every case the conserved ‘knottedness’ is the integer Hopf first met in 1931: the number of times two circles link.

The four Hopf fibrations, one for each real normed division algebra. In every case the total space, base, and fiber are all spheres — a coincidence that happens for no other fiber dimensions.

Division algebra	Total space	Base	Fiber
Real ℝ	S¹	S¹	S⁰ (two points)
Complex ℂ (the classic)	S³	S²	S¹
Quaternionic ℍ	S⁷	S⁴	S³
Octonionic ᴼ	S¹⁵	S⁸	S⁷