Bott periodicity for O via iterated minimal geodesics

Intuition Beginner

Imagine standing at the north pole of a round sphere and walking to the south pole. Every line of longitude is a shortest path, and they are all the same length. The collection of these tied shortest routes is itself a shape: it is the equator, a smaller sphere one dimension down.

Bott periodicity for the rotations works by repeating this idea. Start with a large space of rotations. Pick two points as far apart as possible, like the two poles. Look at all the tied shortest paths between them. That collection is a new, smaller space. Do it again on the new space, and again.

After eight repetitions the pattern of these spaces comes back to where it started, scaled down. That eightfold return is the heart of the theorem.

Visual Beginner

The picture shows a tall spiral staircase. Each landing is one of the eight stages. At every landing you stand between two far-apart points and the shortest routes between them form the next, smaller landing below.

The eighth landing has the same layout as the first, only smaller. That repetition is what the word periodicity means here. The image records the rhythm, not the proof.

Worked example Beginner

Take the simplest case: the round sphere with two poles. The shortest paths between the poles are the lines of longitude. Picking one line of longitude is the same as picking a point on the equator, so the space of shortest paths is the equator.

A two-dimensional sphere gives a one-dimensional equator. A three-dimensional sphere gives a two-dimensional equator. Each time, the space of shortest paths is a sphere one dimension smaller.

This is the toy version of the whole story. In Bott's setting the big space is a space of rotations rather than a sphere, but the move is identical: shortest paths between two poles assemble into a smaller, well-understood space.

Check your understanding Beginner

Formal definition Intermediate+

Let $M$ be a compact connected Riemannian symmetric space and let $p,q\in M$ be a pair of antipodal points, meaning $q$ is the image of $p$ under the geodesic symmetry and the two are joined by minimal geodesics realising the diameter. Write

$${\Omega }_{p,q}^{\min }(M)\subseteq {\Omega }_{p,q}(M)$$

for the subspace of the path space consisting of those geodesics from $p$ to $q$ that have minimal length. The classification of irreducible Riemannian symmetric spaces 07.04.13 supplies, for the chain relevant to the orthogonal group, an explicit identification: each ${\Omega }_{p,q}^{\min }$ is again a symmetric space, one stage further along a fixed chain.

The orthogonal chain runs through the eight stages

$$O\supset O/U\supset U/Sp\supset Sp\supset Sp/U\supset U/O\supset O/O\times O\supset O,$$

written stably (for $O=O(16n)$ with $n\to \infty$ and the matching block sizes). At each stage the inclusion of one factor into the previous is realised as the embedding of ${\Omega }_{p,q}^{\min }$ of the previous stage. The iterated minimal-geodesic theorem asserts that for each stage $M_i$ in the chain there is a homotopy equivalence

$${\Omega }_{p,q}^{\min }(M_i)\simeq M_{i+1}$$

through a range of dimensions growing without bound as the block size $n\to \infty$, and that the composite of eight stages returns to $O$ up to homotopy. Combined with the standard fact ${\Omega }_{p,q}^{\min }(M)\hookrightarrow {\Omega }_{p,q}(M)$ being a homotopy equivalence in low degrees, this yields ${\Omega }^8(O)\simeq O\times \mathbb{Z}$ stably.

Key theorem with proof Intermediate+

Theorem (Morse-theoretic real periodicity, Milnor §24). For the stable orthogonal group there are isomorphisms

$${\pi }_i(O)\cong {\pi }_{i+8}(O)(i\ge 0),$$

and the eight base values are

$${\pi }_0(O),\dots ,{\pi }_7(O)=\mathbb{Z}/2,\mathbb{Z}/2,0,\mathbb{Z},0,0,0,\mathbb{Z}.$$

Sketch with the key estimate. Fix antipodal $p,q$ in a stage $M_i$. By the Morse index theorem 03.02.19, a geodesic from $p$ to $q$ has index equal to the number of conjugate points along it, counted with multiplicity. The minimal geodesics have index $0$; every non-minimal geodesic from $p$ to $q$ has index at least ${\lambda }_i$, where ${\lambda }_i\to \infty$ with the block size $n$. The path space ${\Omega }_{p,q}(M_i)$ therefore admits a Morse function (the energy) whose minimum locus is exactly ${\Omega }_{p,q}^{\min }(M_i)$ and whose next critical value sits at index $\ge {\lambda }_i$.

The fundamental Morse-theory comparison then states: a manifold of paths whose non-minimal critical points all have index $\ge \lambda$ is homotopy equivalent to its minimum locus through dimension $\lambda -1$. Applying this, ${\Omega }_{p,q}(M_i)\simeq {\Omega }_{p,q}^{\min }(M_i)=M_{i+1}$ in a range that grows without bound. Since ${\Omega }_{p,q}(M_i)\simeq \Omega (M_i)$ for connected $M_i$, one stage gives $\Omega (M_i)\simeq M_{i+1}$. Iterating eight times and tracking the chain back to $O$ gives ${\Omega }^8(O)\simeq O\times \mathbb{Z}$, hence the homotopy-group periodicity. Reading the homotopy of the eight stages — each a known symmetric space whose low homotopy is computed directly — produces the table.

Bridge. This Morse proof builds toward the index-theoretic uses of $KO$ that appear again in 03.09.10, where the eightfold pattern of $K{O}^{\bullet }(\mathrm{pt})$ controls which Dirac indices are integers, mod-$2$ invariants, or forced to vanish. The central insight is that the geometry of shortest paths is the periodicity: putting these together, the foundational reason $O$ is eight-periodic is exactly that its symmetric-space chain closes after eight minimal-geodesic passes, and this is exactly the same eight that the Clifford-module proof produces by a wholly different route. The bridge is the classification of symmetric spaces, which generalises the round-sphere example into a finite, repeating ladder.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — the manifold of minimal geodesics, the conjugate-point index count, and the Morse connectivity estimate are all absent from Mathlib, so the iterated construction cannot yet be stated.

-- Pseudocode only: none of these objects exist in Mathlib yet.
axiom min_geodesic_space
    (M : SymmetricSpace) (p q : M) (h : Antipodal p q) :
    Manifold

axiom stage_equivalence
    (M : SymmetricSpace) (p q : M) (h : Antipodal p q) :
    LoopSpace M ≃ₕ NextStage M  -- one minimal-geodesic pass

axiom real_bott_periodicity_O
    (i : Nat) :
    HomotopyGroup StableO i ≃ HomotopyGroup StableO (i + 8)

The genuine content — that the energy functional on the path space has its first non-minimal critical value at an index diverging with the block size — has no Mathlib counterpart.

Advanced results Master

The eight-stage chain is best read through the classification of compact irreducible symmetric spaces 07.04.13. Each inclusion in the chain

$$O\supset O/U\supset U/Sp\supset Sp\supset Sp/U\supset U/O\supset O/(O\times O)\supset O$$

is realised as a totally geodesic embedding of the minimal-geodesic space of the prior stage. Bott's 1959 computation identifies, stage by stage, which symmetric space appears: starting from $O=O(16n)$, the minimal geodesics from $I$ to $-I$ form $O(16n)/U(8n)$ (a choice of compatible complex structure), then the minimal geodesics there form $U(8n)/Sp(4n)$ (a compatible quaternionic structure), and so on, each step pinning down one further structure on the same Euclidean space until, after eight steps, one is choosing an orthogonal splitting and lands back in a copy of $O$ of smaller block size.

The diverging-index phenomenon is the analytic heart. Along a non-minimal geodesic from $p$ to $q$ in stage $M_i$, the conjugate points are computed from the eigenvalues of the curvature operator of the symmetric space. For the chain above these eigenvalues are bounded below away from zero, and the number of conjugate points before the endpoint scales linearly with $n$. The Morse index theorem 03.02.19 then forces the index of every non-minimal geodesic up with $n$, which is exactly the input the connectivity estimate consumes.

There is a clean way to see why exactly eight stages close the loop. Choosing the successive structures — complex, quaternionic, then back down — is the same datum as a representation of a Clifford algebra on the ambient real space, and the real Clifford algebras repeat after eight generators. The Morse chain and the Clifford ladder are the same ladder viewed geometrically versus algebraically. This is the bridge to the contrast with 03.08.08 and the KR comparison of 03.09.12.

Why the index estimate is sharp

The connectivity range cannot be improved beyond the first non-minimal index, because the next critical manifold contributes genuine homotopy exactly at its index. In Milnor's treatment the minimal-geodesic space ${\Omega }^{\min }$ is a deformation retract of the sublevel set just below the first non-minimal critical value, and attaching the next critical cells changes homotopy starting precisely in degree equal to that index. Thus the range ${\lambda }_i-1$ is the best possible at finite $n$, and only the limit $n\to \infty$ delivers periodicity in every fixed degree.

Synthesis. Putting these together, the iterated minimal-geodesic proof shows that the eightfold periodicity of $O$ is the foundational reason the real $K$-theory table reads $\mathbb{Z}/2,\mathbb{Z}/2,0,\mathbb{Z},0,0,0,\mathbb{Z}$: each entry is the low homotopy of one stage of the symmetric-space chain. This is exactly the same period that the Clifford-module argument generalises from the algebra ${\mathrm{Cl}}_{p,q}$, and the two proofs are dual to one another — geometry of geodesics versus algebra of modules — so that the central insight is a single ladder admitting two descriptions. The bridge is the symmetric-space classification, which appears again in every place the orthogonal group's stable homotopy is needed, from index theory to surgery. The geometric proof builds toward the analytic uses of these groups while the algebraic proof feeds the module-theoretic ones, and the agreement of their answers is the structural fact the rest of $KO$-theory rests on.

Full proof set Master

Proposition (one-stage minimal-geodesic equivalence). Let $M$ be a compact symmetric space in the orthogonal chain, with antipodal points $p,q$, and suppose every non-minimal geodesic from $p$ to $q$ has Morse index $\ge \lambda$. Then the inclusion ${\Omega }_{p,q}^{\min }(M)\hookrightarrow {\Omega }_{p,q}(M)$ is a homotopy equivalence through dimension $\lambda -1$, and ${\Omega }_{p,q}(M)\simeq \Omega (M)$.

Proof. Equip ${\Omega }_{p,q}(M)$ with the energy functional $E(\gamma )=\frac{1}{2}{\int }_0^1|\dot{\gamma }{|}^{2}dt$. Its critical points are the geodesics from $p$ to $q$, and by the Morse index theorem 03.02.19 the index of each equals the number of interior conjugate points counted with multiplicity. The minimum value of $E$ is attained exactly on ${\Omega }_{p,q}^{\min }(M)$, where the index is $0$. By hypothesis every other critical point has index $\ge \lambda$.

Using the standard finite-dimensional approximation of the path space by broken geodesics, $E$ becomes a Morse-Bott function on a finite-dimensional manifold. The sublevel set $\{E\le c\}$ for $c$ just below the first non-minimal critical value deformation-retracts onto the minimum locus ${\Omega }_{p,q}^{\min }(M)$. Passing the first non-minimal critical value attaches cells of dimension $\ge \lambda$, by the fundamental theorem of Morse theory. Cell attachment in dimension $\ge \lambda$ does not change homotopy groups below dimension $\lambda$, so ${\Omega }_{p,q}^{\min }(M)\hookrightarrow {\Omega }_{p,q}(M)$ induces isomorphisms on ${\pi }_k$ for $k<\lambda$ and a surjection on ${\pi }_{\lambda }$; this is a homotopy equivalence through dimension $\lambda -1$. Finally, for connected $M$ the based path space ${\Omega }_{p,q}(M)$ is homotopy equivalent to the based loop space $\Omega (M)$, since any choice of minimal geodesic from $p$ to $q$ furnishes a basepoint-translation equivalence. $\square$

Proposition (closure of the eight-stage chain). With $M_0=O(16n)$ and $M_{i+1}={\Omega }_{p,q}^{\min }(M_i)$ for the antipodal pair at stage $i$, one has $M_8\cong O(n)$, and the composite gives ${\Omega }^8(O)\simeq O\times \mathbb{Z}$ stably.

Proof. Bott's stage-by-stage identification ^{[Bott Annals 70]} gives $M_1=O/U$, $M_2=U/Sp$, $M_3=Sp$, $M_4=Sp/U$, $M_5=U/O$, $M_6=O/(O\times O)$, and $M_7=O$, each realised as the minimal-geodesic manifold of the previous stage by the choice-of-structure description. The eighth stage returns to an orthogonal group of reduced block size. Applying the one-stage equivalence at each step gives $\Omega (M_i)\simeq M_{i+1}$ through a range ${\lambda }_i-1$, with ${\lambda }_i\to \infty$ as $n\to \infty$ because the conjugate-point count scales with $n$ ^{[Milnor §24]}. Composing the eight equivalences yields ${\Omega }^8(O)\simeq M_8\times \mathbb{Z}\simeq O\times \mathbb{Z}$ in every fixed degree, where the $\mathbb{Z}$ records the discrete component data of the $O/(O\times O)$ stage. Reading ${\pi }_0$ through ${\pi }_7$ off the eight stages produces the table $\mathbb{Z}/2,\mathbb{Z}/2,0,\mathbb{Z},0,0,0,\mathbb{Z}$. $\square$

Connections Master

• Bott periodicity (complex and real overview) 03.08.07 — that unit states the eightfold real periodicity $K{O}^q\cong K{O}^{q+8}$ and the Clifford-module route; the present unit supplies the independent geometric proof via iterated minimal geodesics, so the two together show the period $8$ from both the algebraic and the Morse-theoretic side.

• Bott periodicity for U via minimal geodesics 03.08.08 — the unitary case is the model for the construction here: the minimal geodesics from $I$ to $-I$ in $U(2n)$ form $U(2n)/(U(n)\times U(n))$, a Grassmannian, and a single further pass closes the two-stage complex chain. The orthogonal argument of this unit is exactly this same mechanism iterated eight times instead of two, which is the foundational reason real periodicity is eightfold rather than twofold.

• Classification tables of irreducible symmetric spaces 07.04.13 — the eight stages $O,O/U,U/Sp,Sp,Sp/U,U/O,O/(O\times O),O$ are read directly off Cartan's list; without the classification the chain would be a sequence of unidentified spaces rather than a finite ladder that demonstrably closes.

• Jacobi fields, conjugate points, Morse index theorem 03.02.19 — the index estimate that drives every stage equivalence is the Morse index theorem applied to geodesics in each symmetric space, converting conjugate-point counts into the connectivity range ${\lambda }_i$.

• KR-theory and real Clifford-module periodicity 03.09.12 — the algebraic shadow of the geometric chain: choosing complex then quaternionic structures along the geodesic stages is the same datum as a real Clifford-module structure, and the $(1,1)$-periodicity there is dual to the minimal-geodesic closure here.

Historical & philosophical context Master

Bott discovered the periodicity of the stable classical groups in the late 1950s while studying the geometry of geodesics on Lie groups and symmetric spaces; the orthogonal case, with its eight-stage chain, was the deepest part of the 1959 announcement ^{[Bott Annals 70]}. The argument was striking because it derived a purely homotopy-theoretic periodicity from Riemannian geometry — counting conjugate points along geodesics — rather than from algebraic topology directly. Milnor's Morse Theory §24 turned the announcement into a textbook proof, isolating the minimal-geodesic manifolds as the engine and making the connectivity estimate precise ^{[Milnor §24]}.

Philosophically, the two proofs of real periodicity — geometric here, algebraic via Clifford modules in 03.08.07 and 03.09.12 — are a case study in how a single structural fact can have genuinely different explanations. The Morse proof says the period is eight because a chain of symmetric spaces closes after eight minimal-geodesic passes; the Clifford proof says it is eight because the real Clifford algebras stabilise after eight generators. That these meet is not a coincidence but a reflection of the same underlying ladder of orthogonal, complex, and quaternionic structures, a theme Lawson and Michelsohn develop in the contrasting algebraic treatment ^{[Lawson-Michelsohn §I.9]}.

Bibliography Master

• Bott, R., "The stable homotopy of the classical groups", Annals of Mathematics 70 (1959), 313–337.
• Milnor, J., Morse Theory, Annals of Mathematics Studies 51, Princeton University Press, 1963. §24.
• Bott, R., "The space of loops on a Lie group", Michigan Mathematical Journal 5 (1958), 35–61.
• Atiyah, M. F., Bott, R. & Shapiro, A., "Clifford Modules", Topology 3 Suppl. 1 (1964), 3–38.
• Lawson, H. B. & Michelsohn, M.-L., Spin Geometry, Princeton University Press, 1989. §I.9.