Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 11 results · all verified · 2 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 9 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Regular Homotopy and Sphere Eversion

1 · Prerequisites

2 · Summary

This page applies the Smale--Hirsch theory of formal immersions. The derivative of an immersion into Euclidean space is a section of the monomorphism bundle Mono⁡(TM,εn). Once a tangent metric is supplied, fibrewise polar normalization gives a section of the separate orthonormal Stiefel bundle V(TM,εn). The proposition below proves the deformation retraction between these section-space models; it retains the positive-definite factor rather than identifying all injections with orthonormal frames. Formal Euclidean immersions consist of a base map together with a monomorphism section. Contracting the base-map factor and normalizing the section show that their homotopy theory is that of Stiefel-bundle sections. The differential of a regular homotopy is a path of formal data, giving the necessity half of each classification.

The page computes the two instances the design requires. For the circle in the plane the tangent bundle is trivial, the section space has π0≅Z by degree, and the resulting invariant of an immersion is the rotation number: this is the Whitney--Graustein classification π0Imm⁡(S1,R2)≅Z, with the nonzero-fold round circles realising every nonzero integer and the Gerono lemniscate realising zero. For spheres in positive codimension the basepoint evaluation of the section space is a Hurewicz fibration whose fibre is the based section space; its long exact sequence, together with the connectivity of Stiefel manifolds and the vanishing π2(SO(3))=0 proved here from the quaternion double cover, gives Smale's classification: for n≥m+2 regular homotopy classes of immersions Sm→Rn correspond to πm(Vm(Rn)), realised by the clutching difference class of the tangent framings, and in codimension one the target-rotation loop argument upgrades the surjection to a non-canonical bijection from πm(SO(m+1)). The vanishing for S2 in R3 makes Imm⁡(S2,R3) path connected, so the standard embedding is regularly homotopic to its inside-out reflection: this is sphere eversion.

Two closing remarks separate the phenomena. Eversion cannot be an isotopy through embeddings, since the sign of the parametrisation relative to the bounded complementary region changes and is computed by a continuous flux integral; and regular homotopy permits self-intersections but never a rank drop. The page contributes only the differential-topological reduction: the homotopy groups of Stiefel manifolds remain algebraic-topology inputs, as the closing remark records. Countable choice supplies the general smooth tangent structures and metrics, and is inherited through the Smale--Hirsch and smoothing suppliers, while the non-isotopy argument additionally inherits the axiom of choice used by Jordan--Brouwer separation.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Gauss frame map of an immersion into Euclidean space

Definition

Supply the canonical smooth tangent-bundle structures and smooth global differential, established under ACω by Assuming countable choice, the tangent bundle has a canonical smooth 2n-manifold structure and Assuming countable choice, the global differential of a smooth map is smooth. Once these are supplied, the following constructions use no further choice.

Let f:Mm→Rn be a smooth immersion with n≥m. Canonically trivializing the target tangent bundle makes its differential a smooth section x⟼dfxofM=Mono⁡(TM,εn)⟶M, whose fibre consists of injective linear maps TxM→Rn, with the subspace smooth structure in the space of linear maps. This is the unnormalized Gauss data of f.

For a smooth local tangent frame (s1,…,sm), the local Gauss frame map is the full-rank matrix gU(x)=(dfxs1(x)∣⋯∣dfxsm(x)). A change of frame by A(x)∈GLm(R) changes this matrix to gU(x)A(x). These frame changes define the monomorphism bundle, and the smoothness of df gives a smooth section in every such trivialization.

If a smooth tangent metric is supplied, write E=V(TM,εn) for the separate bundle of isometric linear injections into the Euclidean target. Its fibre in an orthonormal tangent frame is the Stiefel manifold Vm(Rn). The normalized Gauss section is the fibrewise polar part Qf=df (df∗df)−1/2∈Γ(E). Positivity of df∗df, smoothness of its positive square root, and independence of orthonormal tangent frames are proved in Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections ↗. For an orthogonal change U of tangent frame the normalized matrix changes to QfU. Ordinary Gram–Schmidt in an arbitrary frame does not supply this equivariant formula; polar normalization is the convention here.

A global tangent frame identifies M with the product bundle with fibre Mono⁡(Rm,Rn). Orthonormalizing that frame in the supplied metric identifies E with the product bundle with fibre Vm(Rn), so the normalized section becomes a global map into that Stiefel manifold. Without a global frame it remains a section of E. Polar decomposition retains a positive-definite factor in the monomorphism fibre, and the cited proposition proves the resulting deformation retraction onto the Stiefel fibre. For m=0 both fibres are points; for n=m they are respectively GLm(R) and O(m).

No tangent metric is part of the unnormalized datum; normalization uses the supplied metric. No orientation, properness or normal framing is required. The normal bundle is the quotient εn/df(TM) and is not part of these Gauss data. The model assertions in this definition are justified by the cited proposition, whose proof uses the raw full-rank matrix and frame-change definitions without assuming those assertions.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections

Statement

Let Mm be a smooth manifold and n≥m. Supply its canonical smooth tangent bundle and smooth global differential, and fix a smooth Riemannian metric on TM. These smooth constructions and such a metric exist under ACω; the conclusions below require no additional choice once they are supplied. Give εn=M×Rn its Euclidean metric. Distinguish the monomorphism bundle M=Mono⁡(TM,εn) from its orthonormal Stiefel subbundle E=V(TM,εn), whose fibre is Vm(Rn), associated to the orthonormal frame bundle of TM. All section spaces carry the weak compact-open C∞ topology. Then:

  1. Smooth bundle monomorphisms B:TM→εn over the identity correspond homeomorphically to Γ(M). Fibrewise polar normalization B↦B(B∗B)−1/2 gives an O(m)-equivariant strong deformation retraction Γ(M)→Γ(E). In an orthonormal frame, the polar map is a diffeomorphism Mono⁡(Rm,Rn)≅Vm(Rn)×Sym⁡m+, where the second factor consists of positive-definite self-adjoint m×m matrices; thus the monomorphism fibre and the Stiefel fibre have the same homotopy type, rather than being identified homeomorphically.
  2. The canonical target tangent trivialization gives the actual homeomorphism FImm⁡(M,Rn)≅C∞(M,Rn)×Γ(M). Contracting the first factor and polar-normalizing the second give a homotopy equivalence FImm⁡(M,Rn)≃Γ(E). In particular the two spaces have the same path components and homotopy invariants. Under the actual homeomorphism, the derivative map is f↦(f,df); in the normalized model it sends df to its polar normalization.
  3. For the standard metric on Sm, the bundle E=V(TSm,εn) is the pullback along x↦TxSm∈Grm(Rm+1) of the bundle of isometric injections from the tautological m-plane to Rn. Whenever TM is trivial, E is trivial; in particular V(TS1,ε2)≅S1×S1.

No orientation of M or global tangent frame is needed. The normalization uses the supplied metric; the canonical smooth tangent constructions and metric existence for general M are the stated countable-choice uses. The section-space maps use no further choice once those data are supplied.

Facts & Assumptions

Given: A smooth manifold Mm with a supplied smooth tangent metric, n≥m, the trivial Euclidean bundle εn, the monomorphism bundle M, and the orthonormal Stiefel subbundle E.

[F1]

In a local tangent frame, a fibrewise injection is a full-rank matrix, and changes of tangent frame act by right multiplication; these matrices represent the formal Gauss data. Gauss frame map of an immersion into Euclidean space

[F2]

A formal immersion (f,F) is a smooth map f and a smooth bundle monomorphism covering f; the formal immersion spaces and smooth mapping spaces have the weak compact-open C∞ topology, generated by finitely many compact chart pieces and derivative bounds. Formal immersion between smooth manifolds, Space of immersions and space of formal immersions, The weak compact-open C-infinity topology on mapping spaces

[F3]

A smooth section is a smooth map into the bundle whose projection is the identity. Smooth sections, local sections, and support

[F4]

A bundle map over the identity restricts to a linear map on each fibre; smoothness is checked in local bundle charts. Vector bundle maps over a smooth base map, Smooth vector bundles, rank, fibres, and trivial bundles

[F5]

A locally trivial fibre bundle has local product charts. Locally trivial fiber bundle

[F6]

Vm(Rn) consists of ordered orthonormal frames; the Grassmannian has graph charts, and its tautological bundle has fibre the represented plane. Stiefel spaces, Grassmannians, and tautological bundles

[F7]

The orthonormal frame bundle, using the supplied metric, is a principal O(m)-bundle; its associated bundles use the given group action. Frame bundles and associated vector bundles

[F8]

For a regular level set its tangent space is the kernel of the differential. The tangent space of a regular level set is the kernel

[F9]

A smooth vector bundle is trivial if and only if it admits a smooth global frame. A vector bundle is trivial if and only if it has a global frame

[F10]

Under countable choice every smooth manifold admits a Riemannian metric. Every smooth manifold admits a riemannian metric, The Axiom of Countable Choice (ACω)

[F11]

A smooth positive-definite self-adjoint bundle endomorphism has a unique smooth positive square root. Locally, the derivative of matrix squaring at a positive matrix R is H↦RH+HR; its eigenvalues on symmetric matrices are ri+rj>0, so the root is smooth in the matrix entries. Positive-definite bundle endomorphisms have smooth positive square roots

[F12]

Under ACω, the canonical tangent-bundle atlas gives smooth local product charts linear on each fibre, and the global differential of a smooth map is smooth. Assuming countable choice, the tangent bundle has a canonical smooth 2n-manifold structure, Assuming countable choice, the global differential of a smooth map is smooth. These structures are supplied here.

Proof

1.1F1F2F3F4F5F6F7F10F12construct

Use the supplied smooth tangent-bundle structures of [F12] and the supplied metric; [F10] supplies one under ACω if needed. In any local tangent frame the condition that a matrix have rank m is open, so [F1] and [F4] give the smooth monomorphism bundle M. Orthonormal tangent frames give the associated bundle with fibre Vm(Rn) in [F6] and [F7], which is exactly its isometric-injection subbundle E. The map B↦(x↦Bx) is a bijection onto Γ(M) by [F3]. It is a homeomorphism: section coordinates are precisely the matrix coefficients on each compact base chart piece. A compact piece of TM has compact base projection and bounded vector coordinates, so controlling coefficient jets there controls every jet of the fibre-linear total-space map. Conversely evaluating that map along the finitely many local basis-vector sections over a compact base piece recovers each coefficient and all its derivatives. These two estimates show that the weak total-space and coefficient topologies agree with the section topology in [F2].

2.1F2F6F7F11step 1.1algebra

For an injective B put R=(B∗B)1/2 and Q=BR−1. By [F11] these operations are smooth, and Q∗Q=I. Conversely (Q,R) with Q isometric and R positive gives B=QR, uniquely, so this is the asserted local polar diffeomorphism. Under a change of orthonormal tangent frame U∈O(m), B changes to BU, R to U∗RU, and Q to QU; hence it descends globally without a global frame. The path Bt=B((1−t)I+tR−1) remains injective because its second factor is positive definite, has B0=B and B1=Q, and fixes every isometric B. The operations and this path are continuous for weak C∞ section topologies: on finitely many compact chart pieces the positive spectra stay bounded away from zero, and the smooth matrix operations and all chain-rule derivatives vary continuously there. Thus it is an equivariant strong deformation retraction of section spaces. Rank zero gives the unique empty matrix and the same formulas.

3.1F2F4F12step 1.1step 2.1construct

The canonical identification TRn≅Rn×Rn sends a formal immersion (f,F) to (f,B) with Bx:TxM→Rn the same fibre map. This is a bijection onto C∞(M,Rn)×Γ(M) and a homeomorphism for the identical local matrix/derivative neighbourhoods, as in step 1.1. The contraction (f,B)↦((1−t)f,B) deforms the first factor to zero; combining it with step 2.1 on the second factor gives the homotopy equivalence to Γ(E). Its homotopy inverse sends an isometric section Q to (0,Q). For an immersion the smooth formal pair is (f,df) by [F2] and [F12], so the normalized section is the polar part of df, as stated. No compactness of M is needed because every basic neighbourhood controls only finitely many compact chart pieces.

4.1F6F7F8F9step 2.1step 3.1∎

If TM has a smooth global frame by [F9], orthonormalizing it in the supplied metric gives a global orthonormal frame and identifies E with M×Vm(Rn). For M=Sm with its standard metric, [F8] gives TxSm=x⊥; its projection I−xx∗ varies smoothly, so the Gauss map to the Grassmannian is smooth in its graph charts. The tautological-plane pullback of [F6] therefore is TSm, and its isometric-injection bundle pulls back to E. On S1 the standard unit angular field is a global orthonormal frame, giving E≅S1×V1(R2)=S1×S1. When m=0 the Stiefel and monomorphism fibres are points, and when n=m the normalized fibre is O(m) while the monomorphism fibre retains its positive-definite polar factor. These are included in the same construction.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

The basepoint evaluation of the Stiefel section space is a fibration

Statement

Let 1≤m≤n, let E=V(TSm,εn)→Sm be the orthonormal Stiefel model of the section-space proposition with fibre F=Vm(Rn), let x0∈Sm, and let Γ be the space of smooth sections of E with the weak compact-open C∞ topology and Γ∗ the subspace of sections that take a fixed chosen value s0(x0)∈Ex0. Then:

  1. The evaluation map ev⁡:Γ→Ex0, s↦s(x0), is a Hurewicz fibration with fibre Γ∗.
  2. If Γ∗≠∅, a trivialisation of the pullback of E to the closed characteristic disk and a reference section identify Γ∗, up to homotopy, with the space of continuous based maps (Sm,x0)→(Vm(Rn),e0); hence π0(Γ∗)≅πm(Vm(Rn)) whenever Γ∗≠∅; if Γ≠∅ and n≥m+1 then Γ∗≠∅ as well, because the fibre F is path connected and evaluation is surjective on path components. Via this bijection the class of a section in π0(Γ∗) is its difference class: if two sections agree at x0, the resulting transported disk models give based maps Sm→F, and two sections are homotopic through sections fixed at x0 exactly when these based maps are based-homotopic.
  3. When Γ∗≠∅, base the fibration at a chosen section in Γ∗. Its long exact sequence yields an exact sequence of pointed sets π1(Vm(Rn))⟶π0(Γ∗)⟶π0(Γ)⟶π0(Vm(Rn)); in particular, if Vm(Rn) is simply connected and Γ≠∅ then the difference class induces a non-canonical bijection π0(Γ)≅πm(Vm(Rn)), and if πm(Vm(Rn))=0, n≥m+1 and Γ≠∅ then Γ is path connected. The last conclusion requires a path-connected fibre; it is not asserted for n=m.

Facts & Assumptions

Given: Integers 1≤m≤n, the basepoint x0∈Sm, the bundle E=V(TSm,εn)→Sm with fibre F=Vm(Rn), a chosen value s0(x0)∈Ex0, and the smooth section spaces Γ, Γ∗ with the weak compact-open C∞ topology. Write Γ0, Γ∗0 for continuous sections with the compact-open topology.

[F1]

The Stiefel model of E=V(TSm,εn) has fibre the orthonormal injections TxSm→Rn; a frame of TSm trivialises this bundle. Fibrewise polar normalization of arbitrary monomorphisms is a deformation retraction to this model, so it gives the same section homotopy type. Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections

[F2]

Vm(Rn) is the space of ordered orthonormal m-frames and is path connected for n≥m+1. Stiefel spaces, Grassmannians, and tautological bundles, Stiefel manifolds are connected in positive codimension and simply connected in codimension at least two

[F3]

Assuming AC, a numerable locally trivial fibre bundle is a Hurewicz fibration with its supplied charts and partition of unity. The supplier proof uses AC only to well-order the set of finite chart words (its well-order construction). Locally trivial fiber bundle, Numerable fiber bundles are hurewicz fibrations

[F4]

A Hurewicz fibration in CGWH has the homotopy lifting property relative to every closed cofibration pair, and lifts paths with prescribed initial points; that relative clause is choice-free. A fibration has path lifting and homotopy lifting relative to a subspace

[F5]

The pair (Sm,x0) is a relative CW pair, Sm arising from x0 by attaching one m-cell, and CW pairs are cofibration pairs with the homotopy extension property; products with I are taken with their ordinary topology. CW complex with closure finiteness and weak topology, Cofibration and homotopy extension property

[F6]
[F7]

For a based Serre fibration the long exact sequence of homotopy groups is exact in all degrees, with pointed sets in degree zero; π0 is the pointed set of path components and πn is computed by based cubes. Long exact sequence of homotopy groups of a fibration, Higher homotopy group by based cubes

[F8]

Cubical and spherical models agree: a fixed orientation-preserving homeomorphism induces πn(X,x0)≅[Sn,X]∗. Cubical and spherical models of higher homotopy agree

[F9]

The fibre over a basepoint is the inverse image with its subspace topology, and based homotopy equivalences induce isomorphisms on all homotopy groups. Fiber and fiber homotopy equivalence, Higher homotopy groups are functorial and based homotopy invariant

[F10]

A continuous linear matrix equation has a unique solution on its prescribed compact time interval; for a smooth coefficient depending on parameters, local solution maps depend smoothly on those parameters. Linear matrix ODEs have unique global solutions on a fixed interval, Smooth dependence of ODE solutions on parameters

[F11]

A smooth positive-definite self-adjoint bundle endomorphism has a unique smooth positive square root; locally the matrix squaring derivative is the invertible Sylvester map H↦RH+HR, whose eigenvalues are ri+rj>0, so the square root is smooth in the matrix parameters. Positive-definite bundle endomorphisms have smooth positive square roots

Proof

1.1F1F2F3F9construct

Evaluation on smooth sections is locally trivial. The group O(n) acts on every section by target multiplication. For any frame a∈F, complete it to an orthonormal basis; Gram-Schmidt applied to a nearby frame b followed by the remaining fixed basis vectors gives a smooth orthogonal matrix Ra(b) with Ra(a)=I and Ra(b)a=b. Thus s↦(ev⁡(s),Ra(ev⁡(s))−1s) identifies ev⁡−1(Ua) with Ua×Γa, including its inverse (b,s)↦Ra(b)s. These maps are continuous for the weak C∞ topology because target multiplication multiplies each derivative by the same finite matrix. The same argument works for continuous sections. If either section space is nonempty the transitive O(n) action makes evaluation surjective; otherwise its fibres are all empty. If a section space is empty, its evaluation has the homotopy lifting property vacuously, so it is a Hurewicz fibration with empty fibres without invoking a surjective bundle convention. For a nonempty section space, compactness of F gives finitely many such neighbourhoods, with a subordinate continuous partition obtained from finitely many ambient bump functions. Order these finitely many charts once, and well-order their finite words first by length and then lexicographically. Substituting this explicit well-order in the well-order construction of the proof of [F3] supplies its sole use of AC; all remaining constructions there apply to the supplied finite numeration without choice. Thus both evaluations are Hurewicz fibrations; their fibres at s0(x0) are the prescribed fixed-value section spaces. This argument applies also when m=n and F is disconnected.

1.2F11givenconstruct

Polar normalization is well defined for any fibrewise injection B: in the global matrix presentation put T=B∗B+xx∗, which is positive definite, and normalize by BT−1/2. By [F11] this operation is smooth on smooth sections and continuous on continuous sections. The fibrewise path B((1−t)I+tT−1/2) remains injective and fixes orthonormal sections, giving the deformation retraction to the Stiefel model used in [F1]. We now compare its smooth and continuous fixed-value section spaces constructively. Represent an orthonormal section by matrices A(x):Rm+1→Rn with A(x)=A(x)Px and A(x)∗A(x)=Px, where Px=I−xx∗, and put a=A(x0). Extend A radially to an annulus, multiply by a fixed radial cutoff equal to one near the unit sphere, and convolve with a fixed smooth compactly supported Euclidean kernel of radius ε<1/8. Restriction to the sphere and right multiplication by Px gives a smooth bundle map Bε(A), converging uniformly to A. Pin its value by adding χ(x)(a−Bε(A)(x0))Px, for a fixed smooth bump χ with χ(x0)=1; denote the result by Cε(A). It equals a at x0, converges uniformly to A, depends continuously on (A,ε) into the smooth topology for ε>0, and these operators have a common uniform norm bound.

1.3F1F10givenconstructalgebra

Choose an orthonormal basis e1,…,em of x0⊥. For the closed unit disk Dm, write y=ru and define q(y)=sin⁡(πr)∑iuiei−cos⁡(πr)x0, with q(0)=−x0. The functions sin⁡(πr)/r and cos⁡(πr) are smooth at r=0 by their power series, so q is smooth on the closed disk. Its boundary maps to x0, its interior maps homeomorphically to Sm∖{x0}, and the induced map Dm/∂Dm→Sm is a homeomorphism, since it is a continuous bijection from a compact space to a Hausdorff space. Put P(t,y)=I−q(ty)q(ty)∗ and K=[∂tP,P]. Solve ∂tO=KO, O(0,y)=I. By [F10] the solution exists throughout [0,1] and is jointly smooth in (t,y): the local smooth solution maps patch along the compact time interval by uniqueness. Since K∗=−K, O∗O=I; differentiating P2=P gives ∂tP=[K,P], and therefore PO−OP(0) solves the linear equation with zero initial value and is zero. Consequently O(1,y)e1,…,O(1,y)em is a smooth orthonormal frame of q∗TSm on the whole closed disk, including its boundary. By [F1] it trivialises q∗E.

2.1step 1.2algebraconstruct

There is a continuous positive smoothing radius on the entire metric space Γ∗0. For every integer k≥1, put fk(A)=sup⁡0<ε≤1/(8k)∥Cε(A)−A∥∞ is continuous, since the uniformly bounded operators make these functions uniformly Lipschitz, and fk(A)→0. Put wk(A)=2−kmax⁡(0,1−8fk(A)) and ε(A)=(∑k=1∞wk(A)/(8k))/∑k=1∞wk(A). Both series converge uniformly, the denominator is positive, and if k0 is the first positive weight then ε(A)≤1/(8k0) and ∥Cε(A)(A)−A∥∞<1/8. Every convex combination Lt=(1−t)A+tCε(A)(A) is therefore injective on the tangent fibres. Normalize it by Lt(Lt∗Lt)−1/2 on those fibres; the inverse square root is given by the convergent binomial series near the identity, since ∥Lt∗Lt−I∥<17/64. This normalization is continuous, smooth in x when A is smooth, and fixes a at x0. It gives a homotopy from the identity to a continuous smoothing map S:Γ∗0→Γ∗; restricted to smooth sections it is continuous in the weak C∞ topology as well. Thus inclusion and S are homotopy inverses. The same construction without the pinning correction compares the unrestricted section spaces. No family of charts or approximation choices is selected: the kernel, cutoffs and series are fixed.

2.2F6step 1.3construct

In this frame the prescribed fibre value s0(x0) determines a boundary map g:∂Dm→F, generally nonconstant. A continuous fixed-value section pulls back to a map γ:Dm→F with γ∣∂Dm=g. Conversely such a map gives a section of q∗E whose images in E all equal s0(x0) on the collapsed boundary, so it descends to a unique continuous section of E. This bijection is a homeomorphism Γ∗0≅Mg:={γ:γ∣∂Dm=g}. For compact-open continuity, pullback is continuous, and if K⊆Sm is compact then q−1(K) is compact, so the inverse image of the section neighbourhood s(K)⊆U is the corresponding pullback neighbourhood over q−1(K); composing with the bundle frame and its inverse is continuous by the exponential correspondence. The same quotient reasoning applies to parametrized homotopies.

2.3F2F4step 1.1

If Γ≠∅ and n≥m+1, [F2] makes F path connected. A path from any evaluated value to s0(x0) lifts under the smooth evaluation fibration of step 1.1, producing a section in Γ∗. The same lifting moves a representative of every component of Γ into Γ∗, so the map of component sets π0(Γ∗)→π0(Γ) is surjective.

3.1F4F5F6step 2.2construct

If Γ∗≠∅, choose its disk model γ0∈Mg and one y0∈∂Dm, and put e0=g(y0). The formula gt(u)=γ0((1−t)u+ty0) contracts g to the constant map e0 while fixing y0. Restriction C(Dm,F)→C(∂Dm,F) is a Hurewicz fibration: ∂Dm⊂Dm is a closed cofibration, as its radial collar gives the usual homotopy extension retraction of Dm×I onto Dm×{0}∪∂Dm×I; for every test space, compose that retraction with the prescribed disk map and boundary homotopy and transpose by [F6]. Lifting the path gt, with all points of its starting fibre as parameters, gives transport Mg→Me0. Transport along the reversed path is a homotopy inverse: the two concatenations retrace the same path and contract to constant paths by shortening their excursion; relative homotopy lifting [F4] lifts these contractions to fibre homotopies, with prescribed initial maps. Thus Mg≃Me0.

4.1F6F8F9step 2.1step 2.2step 3.1

The constant-boundary maps descend to the based mapping space C∗((Dm/∂Dm,∗),(F,e0)), again homeomorphically for compact-open topologies. Combining steps 2.1, 2.2 and 3.1 gives Γ∗≃C∗(Sm,F) and hence π0(Γ∗)≅πm(F) by [F8]. A homotopy of sections fixed at x0 gives a path in Mg and, under transport, a based homotopy in Me0. Conversely a based homotopy can be transported back, and the fibre homotopies between the composites and the identities provide a fixed-value section homotopy; the smoothing comparison makes it a homotopy of smooth sections when the endpoints are smooth. These are both directions of the difference-class criterion. The identification depends on the disk frame, reference section and contraction; it is not canonical.

5.1F2F4F7F9step 1.1step 4.1step 2.3∎

The homotopy exact sequence of evaluation, based at a chosen section in Γ∗, is the displayed sequence of [F7]. When F is simply connected, two fibre components that become connected in Γ differ by transport around a loop in F; a nullhomotopy of that loop and relative lifting show they were already connected in the fibre. Thus the component map is injective, and step 2.3 makes it surjective, giving π0(Γ)≅πm(F) by step 4.1. If instead πm(F)=0, n≥m+1 and Γ≠∅, steps 4.1 and 2.3 show that π0(Γ) is a quotient of a singleton and hence itself a singleton. The connected-fibre hypothesis cannot be omitted: for m=n=1 the two orientations of an everywhere nonzero circle field give two section components even though π1(O(1))=0.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Stiefel manifolds are connected in positive codimension and simply connected in codimension at least two

Statement

For 1≤m≤n: (a) Vm(Rn) is path connected when n≥m+1; (b) π1(Vm(Rn),e0)=0 when n≥m+2, where e0=(e1,…,em) is the standard frame. In particular Vm(Rn) is simply connected for n≥m+2, and Vm(Rm+1) is homeomorphic to SO(m+1) and connected, with V1(Rk)=Sk−1. No orientation of the frames is involved: Vm(Rn) is the space of ordered orthonormal m-frames.

Facts & Assumptions

Given: Integers 1≤m≤n, the standard frame e0=(e1,…,em)∈Vm(Rn), and the two-letter alphabet {+,−}.

[F1]

Vm(Rn)={(v1,…,vm):⟨vi,vj⟩=δij}⊆(Rn)m with the subspace topology; in particular V1(Rk) is the unit sphere Sk−1. Stiefel spaces, Grassmannians, and tautological bundles

[F2]

A locally trivial fibre bundle has product charts θi:p−1(Ui)≅Ui×F; it is numerable when the data include a locally finite partition of unity whose closed supports lie in the chart domains. Locally trivial fiber bundle

[F3]

Every numerable fibre bundle is a Hurewicz fibration, hence a Serre fibration. Numerable fiber bundles are hurewicz fibrations The only use of AC in its proof is the well-order of the set of finite chart words; for a two-element chart family the finite words in two letters are enumerated explicitly by length and binary expansion, so the instance used below needs no choice principle.

[F4]

For a based Serre fibration p:(E,x0)→(B,b0) with fibre F, the long exact sequence of homotopy groups is exact in every degree, with pointed sets in degree zero and groups from degree one on. Long exact sequence of homotopy groups of a fibration

[F5]

πn(X,x0) is the set of based cubes modulo boundary-fixed homotopies and π0(X,x0) is the pointed set of path components of X. Higher homotopy group by based cubes

[F6]

A based homeomorphism induces bijections on π0 and isomorphisms on all πn, n≥1. Higher homotopy groups are functorial and based homotopy invariant

[F7]

For 0≤k<r every continuous based map (Sk,a)→(Sr,b) is nullhomotopic through maps fixing a; consequently πk(Sr)=0 for k<r. Lower-dimensional sphere maps are based nullhomotopic

[F8]

For n≥2 the unit sphere Sn−1 is path connected. For n≥2, the sphere Sn−1 is path-connected and connected

[F9]

SO(n)={A∈Rn×n:ATA=I, det⁡A=1} is the special orthogonal group. Orthogonal and special orthogonal Lie groups The ACω in the cited example supplies the Lie group structure on O(n) and SO(n) and is not used here; only the displayed matrix set is needed.

[F10]

A space is simply connected exactly when it is 1-connected, i.e. nonempty, path connected, and has trivial fundamental group at every basepoint. N connected space and n connected map

Proof

1.1F1F2

For 1≤m≤n the first-vector map π:Vm(Rn)→V1(Rn)=Sn−1, π(v1,…,vm)=v1, is a locally trivial fibre bundle with fibre Vm−1(Rn−1). The open sets U+={u∈Sn−1:⟨u,en⟩>−23} and U−={u∈Sn−1:⟨u,en⟩<23} cover Sn−1, and on U± the denominators ∥en±u∥2=2(1±⟨u,en⟩) are bounded below by 23, so the Householder reflections Hu±v=v−2⟨v, en±u⟩∥en±u∥2(en±u) depend continuously on u and are orthogonal; they satisfy Hu+(en)=−u and Hu−(en)=u, hence map the hyperplane en⊥≅Rn−1 isometrically onto u⊥. Therefore Θ±(u,f2,…,fm)=(u,Hu±f2,…,Hu±fm) are homeomorphisms U±×Vm−1(Rn−1)→π−1(U±) over U±, with inverses (v1,…,vm)↦(v1,(Hv1±)−1v2,…,(Hv1±)−1vm).

1.2F1F5F7F8

Base case m=1: V1(Rn)=Sn−1 by [F1], which is nonempty and path connected for n≥2=m+1 by [F8]; and for n≥3=m+2 every based loop S1→Sn−1 is nullhomotopic by [F7] with k=1<r=n−1, so π1(V1(Rn),e1)=0.

2.1F2F3F6step 1.1

The explicit functions σ±(u)=max⁡(0,±⟨u,en⟩+13) satisfy σ++σ−≥13 on Sn−1 and have closed supports supp⁡σ±={±⟨u,en⟩≥−13}⊆U±, so ρ±=σ±/(σ++σ−) is a partition of unity subordinate to the two-element cover of step 1.1. With these charts the bundle of step 1.1 is numerable, so by [F3] it is a Hurewicz fibration and in particular a Serre fibration; the only choice-like step of the cited proof is the well-order of finite words over the chart alphabet, and for the two-element alphabet the words are explicitly enumerated by their binary digits, so this application uses no choice. The identification of the fibre over u with Vm−1(Rn−1) is the homeomorphism Hu± restricted to en⊥, so π0 and π1 of the fibre are those of Vm−1(Rn−1) by [F6].

3.1F4F5F6F8step 2.1

Step (a) for m≥2: assume Vm−1(Rn−1) is path connected, which by the induction hypothesis holds because n≥m+1 gives n−1≥(m−1)+1. The bundle of steps 1.1 and 2.1 is a based Serre fibration with path-connected base Sn−1 and fibre Vm−1(Rn−1) over e1, and its exact sequence in degree zero reads π0(F)→π0(Vm(Rn),e0)→π0(Sn−1,e1), a sequence of pointed sets whose two outer terms are singletons; exactness makes the middle term a singleton as well, that is, Vm(Rn) is path connected.

4.1F4F5F6F7step 2.1step 3.1

Step (b) for m≥2: assume π1(Vm−1(Rn−1),⋅)=0, which by the induction hypothesis holds because n≥m+2 gives n−1≥(m−1)+2. The exact sequence of the same based Serre fibration reads π1(F)→π1(Vm(Rn),e0)→π1(Sn−1,e1), and the target is trivial by [F7] with k=1<r=n−1 because n≥3; exactness makes the first map surjective, so the triviality of π1(F) forces π1(Vm(Rn),e0)=0. Since Vm(Rn) is path connected by step 3.1, triviality at one basepoint gives triviality at every basepoint.

5.1F1F5F6F9F10step 3.1∎

For the final identifications: the map Φ:Vm(Rm+1)→SO(m+1) that sends (v1,…,vm) to the matrix whose first m columns are v1,…,vm and whose last column is the unique unit vector w orthogonal to all vi with det⁡(v1,…,vm,w)=+1 is a bijection onto SO(m+1): the orthogonal complement of span⁡{v1,…,vm} is a line containing exactly two unit vectors, and exactly one of them gives determinant +1; the coordinates of w are the m×m minors of the matrix (v1∣⋯∣vm), namely the coefficients of the Hodge dual, which are polynomial in the entries of the vi, and the inverse is the continuous projection to the first m columns, so Φ is a homeomorphism. Hence by [F6] the homotopy invariants of Vm(Rm+1) and SO(m+1) agree, and Vm(Rm+1) is connected by step 3.1 applied with n=m+1. The case m=1 gives V1(R2)=S1≅SO(2), consistent with V1(Rk)=Sk−1. Steps 1.2, 3.1 and 4.1 cover m=1 and all m≥2 in the stated ranges, and together with the definition of simple connectivity [F10] they give that Vm(Rn) is simply connected whenever n≥m+2.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

The second homotopy group of SO(3) vanishes

Statement

π2(SO(3),I)=0. More precisely, for the two-sheeted covering homomorphism ρ:S3→SO(3), ρ(q)(v)=qvq−1, from the unit quaternions onto the rotations of Im⁡H, the induced homomorphism ρ∗:π2(S3,1)→π2(SO(3),I) is a bijection and π2(S3,1)=0. Consequently also π2(O(3))=0 and π2(V2(R3))=0: the map (u,v)↦(u,v,u×v) is a homeomorphism V2(R3)→SO(3), and O(3) is the disjoint union of the two cosets of SO(3), each homeomorphic to SO(3).

Facts & Assumptions

Given: The quaternion double cover ρ:S3→SO(3), the identity matrix I∈SO(3), and the standard frames (e1,e2) of V2(R3), (e1,e2,e3) of R3.

[F1]

ρ(q)(v)=qvq−1 defines a continuous surjective group homomorphism with kernel {±1}, it is a two-sheeted covering map, and for every covering p:E→B, every e0 and every n≥2, the induced map p∗:πn(E,e0)→πn(B,p(e0)) is an isomorphism. The quaternion double cover generates the third homotopy group of SO(3)

[F3]

πn is computed by based cubes and π0 is the pointed set of path components; π2 is a group and based homotopy equivalences induce isomorphisms. Cubical classes agree with based sphere-map classes. Cubical and spherical models of higher homotopy agree, Higher homotopy group by based cubes, Higher homotopy groups are functorial and based homotopy invariant

[F4]

For 0≤k<r, every continuous based map (Sk,a)→(Sr,b) is nullhomotopic through maps fixing a. Lower-dimensional sphere maps are based nullhomotopic

[F5]

V2(R3)={(u,v):⟨u,u⟩=⟨v,v⟩=1, ⟨u,v⟩=0} with the subspace topology; SO(3) is the group of real 3×3 matrices with RTR=I and det⁡R=1. Stiefel spaces, Grassmannians, and tautological bundles, The quaternion double cover generates the third homotopy group of SO(3). Write O(3)={A:ATA=I} with its matrix subspace topology.

[F6]

The cross product is bilinear, alternating, orthogonal to both factors, and satisfies the scalar triple product identity ⟨x×y,z⟩=det⁡[x y z]. The cross product in R3, The cross product is bilinear, alternating, and orthogonal to both factors

Proof

1.1F5F6

The map ϕ:V2(R3)→SO(3), ϕ(u,v)=(u∣v∣u×v), is a homeomorphism. Its image lies in SO(3): for orthonormal u,v the vector u×v is orthogonal to u and v by [F6] and is unit, since expanding its coordinates gives ∥u×v∥2=∥u∥2∥v∥2−⟨u,v⟩2=1, so the three columns are orthonormal, and det⁡(u∣v∣u×v)=⟨u×v,u×v⟩=1 by the triple product identity [F6]. It is injective because the first two columns determine the argument. It is surjective: for R∈SO(3) with columns c1,c2,c3, the vector c1×c2 is a unit vector orthogonal to c1 and c2 by [F6], hence equals ±c3, and the sign is + because ⟨c1×c2,c3⟩=det⁡(c1∣c2∣c3)=det⁡R=1; thus R=ϕ(c1,c2) with (c1,c2)∈V2(R3). Both ϕ and the projection R↦(c1,c2) to the first two columns are continuous, so ϕ is a homeomorphism.

1.2F1

ρ∗:π2(S3,1)→π2(SO(3),I) is an isomorphism: ρ is a two-sheeted covering map by [F1], and covering projections induce isomorphisms on πn for n≥2 by the second clause of [F1] applied with n=2, p=ρ, e0=1.

1.3F3F4

π2(S3,1)=0: every continuous based map S2→S3 is nullhomotopic through based maps by [F4] with k=2<r=3, so every element of π2(S3,1) equals the class of the constant map, the distinguished element of the group [F3].

2.1F3step 1.2step 1.3

Hence π2(SO(3),I)=0: an isomorphism of groups carries the distinguished element to the distinguished element, so the triviality of the source in step 1.3 forces the triviality of the target.

3.1F3F5step 2.1

O(3) is the disjoint union of its two cosets: every A∈O(3) has det⁡A=±1 by ATA=I, so A∈SO(3) or A∈R0SO(3) for the reflection R0=diag⁡(1,1,−1), and the two cosets are disjoint and each is homeomorphic to SO(3) by left translation. Since the square I2 is connected, every based cube I2→O(3) and every boundary-fixed homotopy of such cubes lies in the single component of the basepoint, so evaluating cubical representatives identifies π2(O(3),J) with π2 of the component of J, which after left translation is π2(SO(3),I) and hence is 0 by step 2.1.

4.1F1F3step 1.1step 2.1step 3.1∎

π2(V2(R3),e)=0 at every basepoint e=(u,v): the homeomorphism of step 1.1 satisfies ϕ(u,v)=R and is a based homotopy equivalence, so it induces an isomorphism π2(V2(R3),e)≅π2(SO(3),R) [F3]; the left translation A↦R−1A is a homeomorphism of SO(3) carrying R to I, hence induces an isomorphism π2(SO(3),R)≅π2(SO(3),I), which is 0 by step 2.1. Together with steps 2.1 and 3.1 this proves all three claimed vanishings at every basepoint, the argument uses the unconditional topological covering statement [F1] and no Lie-group structure or choice principle.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Rotation number of an immersed oriented circle in the plane

Definition

Let S1 be oriented and parametrised as R/2πZ with the positive orientation, let ∂θ be the corresponding positively oriented unit tangent vector field of S1, and let f:S1→R2 be a smooth immersion, that is, a regular closed curve (Immersions, submersions, and constant-rank maps, Smooth manifolds and their smooth charts). Writing TR2≅R2×R2 for the canonical identification, the differential (The differential of a smooth map, The tangent bundle as a disjoint union) gives for each θ the velocity vector dfθ(∂θ), which is nonzero because f is an immersion.

The normalised velocity, or unit tangent, of f is the map τf:S1⟶S1,τf(θ)=dfθ(∂θ)∣dfθ(∂θ)∣, which is continuous because the velocity never vanishes and every nonzero vector is a positive multiple of a unique unit vector.

The rotation number (also rotation index) of f is rot⁡(f):=deg⁡(τf)∈Z, the degree of the unit tangent map. Concretely, after choosing the base point θ0=[0], identifying the unit circle with R/Z by the standard parametrisation and rotating the target circle by a constant so that τf(θ0) corresponds to [0], the class of τf is a based loop in the sense of The degree of a based circle loop, and its degree is the integer rot⁡(f); the constant rotation of the target plane changes neither the winding number nor the degree, and the degree descends to based loop classes (Degree defines a function Deg⁡:π1(S1,[0])→Z), so the value is independent of the choice of θ0 and of the lift used to compute it (Two based circle loops are path-homotopic if and only if they have equal degree).

Equivalent formulations, all taking the same value:

  • The velocity curve t↦f′(t) is a closed C1, hence rectifiable, loop in C×=R2∖{0}, and rot⁡(f) is its winding number about the origin: the normalised velocity t↦f′(t)/∣f′(t)∣ is exactly τf, and after the constant nonzero complex multiplication γ↦γ/γ(t0) of the target plane, which changes neither the winding number about 0 nor the degree of the normalised loop, For loops in C times, the winding number about 0 equals the circle degree identifies n(γ,0) with the degree of the normalised loop, while Winding number identifies the fundamental group of C times with the integers identifies that winding number with the class in π1(C×,1)≅Z.
  • rot⁡(f) equals 12π times the total signed turning angle of the tangent, i.e. the rotation index of the regular closed plane curve f in the sense of Rotation index of a regular closed plane curve, where the tangent angle is lifted continuously and the corner jumps are zero for a smooth curve.

The normalisation is fixed by the round unit circle: the immersion θ↦(cos⁡θ,sin⁡θ) traversed once in the positive direction has unit tangent (−sin⁡θ,cos⁡θ) winding once positively, hence rotation number +1, and its reverse has rotation number −1. Reversing the orientation of the domain negates the rotation number, while a regular (orientation-preserving) reparametrisation leaves it unchanged. Indeed, if h is a positively oriented circle diffeomorphism with increasing lift H satisfying H(θ+2π)=H(θ)+2π, then the chain rule gives τf∘h=τf∘h: composing a tangent-angle lift with H preserves its total increment. For the reversal a(θ)=−θ, τf∘a(θ)=−τf(−θ), so an angle lift is α(−θ)+π when α lifts τf; its total increment is the negative of that of α.

No convexity, simplicity, self-intersection restriction or properness is imposed; rot⁡ is an invariant of the oriented immersed circle, not of its image.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Formal immersions of the circle in the plane are classified by the winding number

Statement

Let E=V(TS1,ε2)→S1 be the Stiefel bundle of the section-space proposition for M=S1 and n=2, with fibre V1(R2)=S1. Then E≅S1×S1 is trivial, its section space Γ is homeomorphic to C∞(S1,S1) with the weak smooth topology, and π0(Γ)≅Z by degree. The resulting winding invariant w(f,F)∈Z of a formal immersion (f,F) is the degree of the section expressed in the angular trivialisation defined by ∂θ; for the derivative (f,df) of an immersion it equals the rotation number rot⁡(f) of the preceding definition. Two formal immersions of S1 into R2 lie in the same path component of FImm⁡(S1,R2) if and only if their winding invariants are equal.

Facts & Assumptions

Given: The oriented circle S1=R/2πZ, its positively oriented unit tangent field ∂θ, the angular frame s(θ)=∂θ, and the bundle E=V(TS1,ε2)→S1 with fibre S1=V1(R2).

[F1]

With the standard angular metric, the normalized Stiefel bundle E=V(TS1,ε2) is S1×S1. The monomorphism section space retracts to its smooth isometric section space Γ by polar normalization, and FImm⁡(S1,R2)≃C∞(S1,R2)×Γ; the contractible first factor gives the same path components. Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections

[F2]

rot⁡(f)=deg⁡(τf) for an immersion f:S1→R2, with τf the normalised velocity. Rotation number of an immersed oriented circle in the plane

[F3]

A smooth rank-r vector bundle is trivial if and only if it has a global frame; (∂θ) is a global frame of TS1; global frames trivialise the frame bundle and every associated bundle. A vector bundle is trivial if and only if it has a global frame, Local and global frames of a vector bundle, Frame bundles and associated vector bundles

[F4]

V1(R2) is the unit circle S1 and the fibres of E are the isometric injections TθS1→R2. Stiefel spaces, Grassmannians, and tautological bundles

[F5]

Degree descends to path-homotopy classes of based circle loops and identifies them up to homotopy: two based circle loops are path-homotopic exactly when their degrees agree; equivalently the winding number of a closed rectifiable loop in C× about 0 is the degree of its normalised circle loop and classifies its loop class. Two based circle loops are path-homotopic if and only if they have equal degree, Degree defines a function Deg⁡:π1(S1,[0])→Z, For loops in C times, the winding number about 0 equals the circle degree, Winding number identifies the fundamental group of C times with the integers

[F6]

For S1 and R2, use their finite standard atlases: their derivative transitions and fixed rational-ball bases give smooth tangent total spaces without choice. The angular frame identifies TS1=S1×R, and TR2=R2×R2; these explicit structures supply the tangent-space topology used here. Define the concrete formal space directly as the set of smooth pairs (f,F) with πR2F=fπS1 and each Fx linear and injective, with the subspace topology from C∞(S1,R2)×C∞(TS1,TR2) (The weak compact-open C-infinity topology on mapping spaces). The tangent total spaces are the explicit products just constructed, so this instance uses no general tangent-bundle existence premise.

Proof

1.1F1F3F4

TS1 is trivial with the global frame (∂θ): the field is smooth and nowhere zero at every point of the circle, so it is a global frame by [F3]. Hence E=V(TS1,ε2)≅S1×S1 by the triviality clause of [F1], and its fibres are the isometric injections TθS1→R2, identified with S1 by [F4].

2.1F1F3

Under the trivialisation of step 1.1, a smooth section of E is exactly a smooth map S1→S1, so Γ≅C∞(S1,S1); a section F corresponds to θ↦ the coordinate of Fθ(∂θ) in the trivialisation, a nowhere-zero continuous function for a monomorphism. Writing v(θ)=Fθ(∂θ) for a bundle monomorphism over the identity, normalisation v↦v/∣v∣ is a homotopy of nowhere-zero maps, because the straight segment from v(θ) to v(θ)/∣v(θ)∣ stays in the open ray through v(θ) and misses 0.

3.1F5step 2.1construct

Degree classifies the smooth section components. A smooth circle map has a smooth angular lift α:R→R with α(θ+2π)=α(θ)+2πd, where d is its degree: the continuous lift in the circle-loop model is smooth on each local inverse branch of the exponential. The linear interpolation (1−t)α(θ)+tdθ exponentiates to a smooth path of circle maps to the standard degree-d map, continuous in the weak C∞ topology. Thus equal degrees give a path of smooth sections; conversely any such path is a continuous homotopy and preserves degree by [F5]. Every integer occurs via θ↦eidθ, so π0(Γ)≅Z.

4.1F2step 2.1step 3.1

The winding invariant w(f,F) is the degree of the section of (f,F) in the fixed trivialisation of step 1.1, so w is constant on path components and induces the bijection π0(Γ)≅Z of step 3.1. For the derivative (f,df) of an immersion, the corresponding section is the velocity map v(θ)=dfθ(∂θ), whose normalisation is τf; by step 2.1 v and τf are homotopic through nowhere-zero maps, so they have the same degree, and that degree is rot⁡(f) by [F2].

5.1F1F2F6step 3.1step 4.1∎

Two formal immersions (f,F), a smooth f with a smooth bundle monomorphism F over f in the sense of [F6], lie in the same path component of FImm⁡(S1,R2) exactly when w agrees: by [F1], FImm⁡(S1,R2) is homotopy equivalent to C∞(S1,R2)×Γ via polar normalization, and C∞(S1,R2) is contractible, so path components of the product correspond bijectively to path components of Γ, which are classified by the degree by step 3.1; the winding invariant is that degree. This fixes the normalisation: the round unit circle traversed once in the positive direction has w=1 and its reverse has w=−1, matching the sign convention of [F2].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Regular homotopy preserves the formal Gauss class

Statement

Assume ACω for the canonical smooth tangent-bundle structures and global differentials. Let Mm,Nn be smooth manifolds with m≤n (no compactness of M is needed for this invariance) and let H:M×[0,1]→N be a regular homotopy, that is, a smooth map whose restriction Ht to every slice is an immersion. Then t↦(Ht,dHt) is a continuous path in FImm⁡(M,N) for the weak compact-open C∞ topology, and consequently every homotopy invariant of formal data is constant along a regular homotopy: the path component of (Ht,dHt) in FImm⁡(M,N) does not depend on t. For M=Sm,N=Rn this invariant is the Gauss frame class of the preceding items (the difference class in πm(Vm(Rn)) when n≥m+2); for M=S1,N=R2 it is the winding number, i.e. the rotation number. In particular two regularly homotopic immersions have homotopic formal data, which is the necessity half of every classification on this page.

Facts & Assumptions

Given: ACω, smooth manifolds Mm,Nn with m≤n and a regular homotopy H:M×[0,1]→N.

[F1]

A regular homotopy is a smooth H with every slice Ht an immersion, with H0 and H1 the prescribed immersions; the space Imm⁡(M,N) carries the subspace topology of the weak compact-open C∞ topology. Regular homotopy of immersions, Space of immersions and space of formal immersions

[F2]

The weak compact-open C∞ topology is generated by finitely many chart data, compact chart pieces and derivative tolerances. The weak compact-open C-infinity topology on mapping spaces

[F3]

Under ACω for the smooth tangent-bundle structures, the derivative map D:Imm⁡(M,N)→FImm⁡(M,N), f↦(f,df), is continuous for the weak topologies. The derivative map is continuous

[F4]

For compact M, path components of Imm⁡(M,N) are exactly the regular homotopy classes. A continuous family of immersions on a compact parameter domain can be smoothed through immersions relative to the parameter boundary provided its adjoint is smooth on a neighbourhood of that boundary times M; these assertions inherit the countable-choice assumption of the relative approximation theorem. Smooth families and path components in the weak topology, The Axiom of Countable Choice (ACω)

[F5]

Points joined by a continuous path lie in the same path component, and path components are the equivalence classes of the relation "joined by a path". Homotopies of continuous maps, homotopies relative to a subspace, and path homotopies relative to the endpoints

[F6]

The adjoint and evaluation correspondences for smooth families of maps identify a smooth H:M×[0,1]→N with a map [0,1]→C∞(M,N). Smooth families of maps and their evaluation maps, The exponential law: for a locally compact metric X and any spaces Z and Y, transposition is a bijection between C(X×Z,Y) and C(Z,C(X,Y)) with the compact-open topology

[F7]

The angular winding invariant of formal circle immersions equals rotation number on derivatives. Formal immersions of the circle in the plane are classified by the winding number.

[F8]

For sphere sections with simply connected Stiefel fibre, the evaluation lemma identifies components by the difference class. The basepoint evaluation of the Stiefel section space is a fibration.

Proof

1.1F1F2F6

The adjoint H^:[0,1]→C∞(M,N), H^(t)=Ht, is continuous for the weak compact-open C∞ topology directly from the definition: a basic weak neighbourhood of Ht0 is determined by finitely many charts, compact pieces Ki⊆M and tolerances εi on derivatives of order at most ri, and compactness with continuity gives a closed time interval J about t0 relative to [0,1] on which H(Ki×J) lies in every prescribed target chart. For each of the finitely many occurring multi-indices α the local expression Dα(ψi∘Ht∘φi−1)(x) is continuous on the compact set φi(Ki)×J, hence uniformly continuous there, so for t close to t0 every one of the finitely many derivatives stays within its tolerance; therefore H^ is continuous. Its image lies in Imm⁡(M,N) because every slice of H is an immersion by [F1].

2.1F3step 1.1

Composition with the continuous derivative map [F3] gives the continuous path t↦D(Ht)=(Ht,dHt) in FImm⁡(M,N), since the composite of continuous maps is continuous and H^ takes values in Imm⁡(M,N) by step 1.1.

3.1F5step 2.1

A continuous path has all its points in a single path component [F5], so the path component of (Ht,dHt) in FImm⁡(M,N) is independent of t; hence every invariant of formal data that is constant on path components takes the same value at (H0,dH0) and (H1,dH1), which is the necessity half of the classifications: regularly homotopic immersions have homotopic formal data.

4.1F4F7F8step 3.1∎

The two instances: for M=Sm, N=Rn with n≥m+2, the path component of the formal data is detected by the difference class of the basepoint-evaluation lemma through the section-space description of FImm⁡(Sm,Rn), so the class in πm(Vm(Rn)) is constant along a regular homotopy; for M=S1, N=R2 the corresponding invariant of a path component is the winding number, which for the derivative (f,df) is the rotation number of the immersion, as established in Formal immersions of the circle in the plane are classified by the winding number. The sphere difference-class conclusion is The basepoint evaluation of the Stiefel section space is a fibration. The compact-source identification of path components with regular homotopy classes used in the converse direction is [F4] with its countable-choice hypothesis; the invariance proved here uses steps 1.1–3.1 and inherits the structural countable-choice hypothesis of [F3]. No approximation or additional choice is used in this direct argument.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Whitney–Graustein classification of plane circle immersions

Statement

Assume ACω. Let f,g:S1→R2 be oriented immersions of the circle (regular closed curves). Then f and g are regularly homotopic through immersions if and only if their rotation numbers agree, rot⁡(f)=rot⁡(g). Equivalently, the rotation number induces a bijection π0Imm⁡(S1,R2)≅Z; every integer is realised by a k-fold round circle for k≠0, and by the Gerono lemniscate for k=0. Reversing the domain negates the rotation number, so the once-traversed round circle cannot be turned inside out in the plane through immersions. A zero-rotation immersed circle is regularly homotopic to its reversal.

Facts & Assumptions

Given: Oriented immersions f,g:S1→R2 and their rotation numbers rot⁡(f),rot⁡(g)∈Z.

[F1]

A regular homotopy is a smooth family of immersions with prescribed ends; the rotation number is deg⁡(τf) for the normalised velocity, is invariant under regular reparametrisation and negates under reversal of the orientation of the domain. Regular homotopy of immersions, Rotation number of an immersed oriented circle in the plane

[F2]

A regular homotopy gives a continuous path of formal data in FImm⁡, so every homotopy invariant of formal data, in particular the winding invariant, is constant along it. Regular homotopy preserves the formal Gauss class

[F3]

For S1⊆R2 the winding invariant classifies formal immersions: two formal immersions lie in the same path component of FImm⁡(S1,R2) exactly when their winding invariants agree, π0Γ≅Z by degree, and the winding invariant of the derivative of an immersion is its rotation number. Formal immersions of the circle in the plane are classified by the winding number

[F4]

The derivative map D:Imm⁡(S1,R2)→FImm⁡(S1,R2) is a weak homotopy equivalence (1<2), hence induces a bijection on path components; for compact sources path components of Imm⁡ are regular homotopy classes, and the bijection matches regular homotopy classes with homotopy classes of formal immersions. The Smale–Hirsch immersion theorem, Regular homotopy classes of immersions are formal homotopy classes, Weak homotopy equivalence

[F5]

Path components are the classes of the relation "joined by a path"; a continuous path stays in one path component. Homotopies of continuous maps, homotopies relative to a subspace, and path homotopies relative to the endpoints, Space of immersions and space of formal immersions

Proof

1.1F2F3F5

Necessity: let H be a regular homotopy from f to g. By [F2] the formal data (Ht,dHt) form a continuous path in FImm⁡(S1,R2), so the winding invariant is constant along it by [F5]; for the derivative of an immersion the winding invariant equals the rotation number by [F3]. Hence rot⁡(f)=rot⁡(g), and two curves with different rotation numbers are not regularly homotopic.

2.1F3F4step 1.1

Sufficiency: assume rot⁡(f)=rot⁡(g). By [F3] the winding invariants of the derivatives (f,df) and (g,dg) are equal, so these formal immersions lie in the same path component of FImm⁡(S1,R2), i.e. are homotopic through formal immersions. The derivative map is a weak homotopy equivalence by [F4], hence a bijection on path components, so f and g lie in the same path component of Imm⁡(S1,R2); by [F4] path components of Imm⁡ for the compact source S1 are exactly the regular homotopy classes, so f and g are regularly homotopic.

3.1F1F3step 1.1step 2.1construct

Consequently rotation number gives a bijection π0Imm⁡(S1,R2)≅Z: necessity and sufficiency give well-definedness and injectivity. For k≠0, fk(θ)=(cos⁡kθ,sin⁡kθ) has velocity of norm ∣k∣ and unit tangent sgn⁡(k)(−sin⁡kθ,cos⁡kθ), a constant rotation of eikθ, hence degree k. For k=0, take δ(θ)=(cos⁡θ,sin⁡2θ). Its velocity is p(sin⁡θ), where p(u)=(−u,2−4u2) never vanishes for real u: if its first coordinate is zero its second is 2. The homotopy p((1−t)sin⁡θ) contracts this velocity loop to (0,2) through nonzero vectors, so its degree is zero. Thus every integer occurs.

4.1F1F3F4step 2.1step 3.1∎

Reversing the domain gives τf∘a(θ)=−τf(−θ) for a(θ)=−θ. The constant target rotation by π preserves degree and domain reversal negates it, so the reverse of a curve of rotation number k has rotation number −k. In particular the once-traversed round circle and its reverse have values 1 and −1 and are not regularly homotopic. The zero class is nonempty by step 3.1; its representatives are regularly homotopic to their reversals by step 2.1. Countable choice is inherited from [F4].

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Smale's classification of sphere immersions in Euclidean space

Statement

Assume ACω. Let 1≤m≤n, let F=Vm(Rn)≅O(n)/O(n−m) be the Stiefel manifold of orthonormal m-frames, and let E=V(TSm,εn)→Sm be the Stiefel bundle of the section-space proposition. Its sections are the fibrewise-injection part of the formal non-holonomic data, and the projection FImm⁡(Sm,Rn)→Γ(E) forgetting the underlying map and polar-normalizing the fibrewise injection is a homotopy equivalence, so path components of Γ(E) and of FImm⁡(Sm,Rn) agree.

  1. If n≥m+2, then F is simply connected, E admits sections (because TSm⊕εn−m≅εn), and the difference class of the evaluation lemma induces a non-canonical bijection π0Γ(E)≅πm(Vm(Rn))=πm(O(n)/O(n−m)). Since the Smale–Hirsch derivative map is a weak homotopy equivalence, the same set classifies regular homotopy classes of immersions: there is a non-canonical bijection between regular homotopy classes of immersions Sm→Rn and πm(O(n)/O(n−m)), realised by the clutching difference class of the tangent framings.
  2. If n=m+1, then Vm(Rm+1)≅SO(m+1) and the difference class gives a non-canonical bijection πm(SO(m+1))→π0Γ(E)≅π0Imm⁡(Sm,Rm+1); in particular, if πm(SO(m+1))=0 then all immersions Sm→Rm+1 are regularly homotopic.
  3. The instances used on this page: m=1, n=2, where π0Imm⁡(S1,R2)≅Z with the rotation number as invariant; and m=2, n=3, where π2(SO(3))=0 and hence all immersions S2→R3 are regularly homotopic.

Facts & Assumptions

Given: Integers 1≤m≤n, the sphere Sm, the Stiefel bundle E=V(TSm,εn) with fibre Vm(Rn), and the space Γ(E) of its sections.

[F1]

E has fibre Vm(Rn) and its sections are isometric injections. Arbitrary smooth bundle monomorphisms TSm→εn over the identity correspond homeomorphically to sections of M=Mono⁡(TSm,εn), whose section space strongly deformation retracts to Γ(E) by polar normalization. There is an actual homeomorphism FImm⁡(Sm,Rn)≅C∞(Sm,Rn)×Γ(M); contracting the first factor and normalizing the second give a homotopy equivalence to Γ(E) and a bijection of path components. Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections

[F2]

Evaluation at a basepoint of the section space is a Hurewicz fibration; if the fibre Vm(Rn) is simply connected and Γ(E)≠∅, the difference class gives a non-canonical bijection π0Γ(E)≅πm(Vm(Rn)); if n≥m+1, πm(Vm(Rn))=0 and Γ(E)≠∅, then Γ(E) is path connected. The basepoint evaluation of the Stiefel section space is a fibration

[F3]

Vm(Rn) is path connected for n≥m+1 and simply connected for n≥m+2, and Vm(Rm+1)≅SO(m+1). Stiefel manifolds are connected in positive codimension and simply connected in codimension at least two, Stiefel spaces, Grassmannians, and tautological bundles

[F4]

π2(SO(3))=0. The second homotopy group of SO(3) vanishes

[F5]

For m<n the derivative map D:Imm⁡(Sm,Rn)→FImm⁡(Sm,Rn) is a weak homotopy equivalence, and for compact sources it induces a bijection between regular homotopy classes of immersions and homotopy classes of formal immersions; a weak homotopy equivalence induces a bijection on path components. The Smale–Hirsch immersion theorem, Regular homotopy classes of immersions are formal homotopy classes, Weak homotopy equivalence, Formal immersion between smooth manifolds, Space of immersions and space of formal immersions

[F6]

The normal bundle of the standard sphere Sm⊆Rm+1 is trivial with global frame x↦x (the radial field is nowhere zero and normal, since TxSm=ker⁡d(∣x∣2−1)x), and the tangent-normal identity gives TSm⊕ε1≅εm+1; adding trivial summands gives TSm⊕εn−m≅εn for n≥m+1. Formal immersion gives the tangent normal-bundle identity, Whitney sums of vector bundles, A vector bundle is trivial if and only if it has a global frame, The tangent space of a regular level set is the kernel

[F7]

Vm(Rn)≅O(n)/O(n−m), with O(n−m) embedded as diag⁡(Im,Q); the quotient map takes the first m columns, and SO and O are the special and full orthogonal groups. Stiefel spaces, Grassmannians, and tautological bundles, Orthogonal and special orthogonal Lie groups

[F8]

For m=1, n=2: two formal immersions of S1 into R2 are in the same path component exactly when their winding invariants agree, and π0Γ(E)≅Z by degree. Formal immersions of the circle in the plane are classified by the winding number

Proof

1.1F1F3F6F7algebra

For the quotient identification in [F7], every orthonormal m-frame extends to an orthonormal basis by finite-dimensional Gram–Schmidt. Two matrices have the same first m columns exactly when they differ on the right by diag⁡(Im,Q) with Q∈O(n−m). Thus the first-column map induces a continuous bijection O(n)/O(n−m)→Vm(Rn); it is a homeomorphism because the source is compact and the target Hausdorff. When n≥m+1, [F6] gives TSm⊕εn−m≅εn. Inclusion of the tangent summand is a smooth monomorphism; polar-normalizing it by [F1] gives an isometric section of E, so Γ(E)≠∅.

2.1F1F2F3F5F7step 1.1

Clause 1: if n≥m+2, then by [F3] the fibre Vm(Rn) is simply connected, and Γ(E)≠∅ by step 1.1; hence [F2] gives the non-canonical bijection π0Γ(E)≅πm(Vm(Rn)), the non-canonicity coming from the choice of trivialisation and of the section used to identify the difference classes. Passing to immersions: the derivative map is a weak homotopy equivalence by [F5], so it induces a bijection π0Imm⁡(Sm,Rn)≅π0FImm⁡(Sm,Rn), and [F1] identifies the latter with π0Γ(E); the resulting bijection between regular homotopy classes of immersions and πm(Vm(Rn))=πm(O(n)/O(n−m)) is realised by the clutching difference class of the tangent framings.

2.2F2F3F5F7step 1.1

Clause 2: if n=m+1, then Vm(Rm+1)≅SO(m+1) by [F3], using the unique final normal vector that completes a frame to a positive orthonormal basis. In particular the fibre is path connected, so [F2] and step 1.1 give the surjection πm(F)→π0Γ(E). Its only possible identifications are the evaluation-loop action. Given a loop of evaluated frames et based at e0, write C(e) for the uniquely completed oriented matrix and put At=C(et)C(e0)−1. These matrices define a loop in SO(m+1) with A0=A1=I and Ate0=et. For every section s with s(x0)=e0, the sections st(x)=Ats(x) lift that loop and return to the same section s. Thus every evaluation loop acts trivially on every component of the fixed-value section space. The exact-sequence component map is therefore injective as well as surjective, giving the asserted non-canonical bijection πm(SO(m+1))≅π0Γ(E). By [F5] it also classifies regular homotopy components of immersions; in particular vanishing of this group gives a single component.

3.1F4F5F8step 2.1step 2.2∎

Clause 3: for m=1, n=2, clause 2 applies with SO(2)=S1; the winding invariant of [F8] is a surjection π0Γ(E)→Z that is also injective by the classification of formal immersions of the circle, so π0Γ(E)≅Z and [F5] gives π0Imm⁡(S1,R2)≅Z with the rotation number as invariant. For m=2, n=3, clause 2 applies with V2(R3)≅SO(3) and π2(SO(3))=0 by [F4], so Γ(E) is path connected and all immersions S2→R3 are regularly homotopic. The Smale–Hirsch input carries its countable-choice hypothesis.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Standard and reflected two-sphere immersions have homotopic formal data in R^3

Statement

Let ι:S2↪R3 be the standard inclusion of the unit sphere and let a:S2→S2, a(x)=−x, be the antipodal map; equivalently replace ι by r∘ι for a reflection r of R3, which differs from ι∘a by an orientation-preserving rotation of the target. Then the formal immersions (ι,dι) and (ι∘a,d(ι∘a)) lie in the same path component of FImm⁡(S2,R3). More precisely, after moving their sections to a common basepoint value, the characteristic-disk model and transport in the evaluation lemma give based maps into V2(R3)≅SO(3). Their difference class in π2(SO(3)) vanishes. This vanishing is the algebraic content of eversion.

Facts & Assumptions

Given: The unit sphere S2⊆R3, the standard inclusion ι, the antipodal map a, and the Stiefel bundle E=V(TS2,ε3)→S2 with fibre V2(R3).

[F1]

ι is an immersion (its differential is injective at every point), and a is a diffeomorphism, so ι∘a is an immersion with derivative d(ι∘a)=dι∘da; the pairs (ι,dι) and (ι∘a,d(ι∘a)) are formal immersions S2→R3. Immersions, submersions, and constant-rank maps, Formal immersion between smooth manifolds

[F2]

For M=S2, n=3: E has fibre V2(R3); bundle monomorphisms over the identity correspond to sections; FImm⁡(S2,R3) is homotopy equivalent to C∞(S2,R3)×Γ(E) via fibrewise polar normalization and the projection forgetting f is a homotopy equivalence, so path components of FImm⁡ correspond to path components of Γ(E). Euclidean formal immersions are homotopy equivalent to Stiefel-bundle sections

[F3]

The basepoint-evaluation lemma: for E=V(TSm,εn), evaluation at x0 is a Hurewicz fibration on the section space; if the fibre is path connected, has πm=0 and the section space is nonempty, then the section space is path connected. The basepoint evaluation of the Stiefel section space is a fibration

[F4]

π2(SO(3))=0 and π2(O(3))=0; π2 is computed by based cubes, and V2(R3) is homeomorphic to SO(3) while O(3) is the disjoint union of its two cosets of SO(3). The second homotopy group of SO(3) vanishes, Higher homotopy group by based cubes

[F6]

Path components are the equivalence classes of the relation "joined by a continuous path", and a homotopy of formal data is a path in FImm⁡. Homotopies of continuous maps, homotopies relative to a subspace, and path homotopies relative to the endpoints

Proof

1.1F1F2F4F5

E has a section: the differential dι of the standard inclusion is a bundle monomorphism TS2→ε3 over ι, hence a section sι of E by [F2]: for the standard metrics dιx(v)=v is already isometric. Thus Γ(E)≠∅. The fibre V2(R3) is path connected by [F5] and π2(V2(R3))=0 by [F4].

2.1F3F4step 1.1

The section space Γ(E) is path connected: by the evaluation-fibration lemma [F3] applied with m=2, n=3, the long exact sequence makes π0Γ(E) a quotient of π2(V2(R3)) as soon as Γ(E)≠∅, and that group is trivial by [F4]; equivalently the evaluation fibration is surjective on path components and its based section space has π0≅π2(V2(R3))=0. Hence any two sections of E are joined by a path of sections.

3.1F1F2F6step 2.1

Consequently any two formal immersions S2→R3 lie in the same path component of FImm⁡(S2,R3): path components are the classes of the relation "joined by a continuous path" and a homotopy of formal data is a path in FImm⁡ [F6], so it suffices that by [F2] the space is homotopy equivalent to C∞(S2,R3)×Γ(E) via fibrewise polar normalization and the first factor is contractible, so its path components are exactly those of Γ(E), a single point by step 2.1. Applying this to the two formal immersions of [F1] gives that (ι,dι) and (ι∘a,d(ι∘a)) are homotopic through formal immersions.

4.1F1F2F3F4F5step 3.1∎

For the difference-class description, first move both frame sections to one prescribed basepoint value by evaluation path lifting, using the path-connected fibre in [F5]. The evaluation lemma [F3] pulls them to the closed characteristic disk with the same, possibly nonconstant boundary map, and transports both disk models along a contraction supplied by a reference section. They then descend to based maps into V2(R3)≅SO(3); subtracting their classes in π2 gives the difference obstruction. By [F4] this group vanishes, so the difference map is nullhomotopic and the two sections are homotopic, as already established in step 3.1. The two components of O(3) are homeomorphic to SO(3), so their second homotopy groups vanish too. Replacing ι by r∘ι for a reflection changes the target by a rotation relative to ι∘a: for r(x,y,z)=(x,y,−z) take A=diag⁡(−1,−1,1)∈SO(3), and for any reflection A=−r is likewise a rotation. A path of target rotations from I to A gives a homotopy of the corresponding formal data. Hence the reflected embedding has the same formal component as ι∘a.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Sphere eversion

Statement

Assume ACω for the existence and classification assertions. The standard embedding ι:S2↪R3 is regularly homotopic, through immersions S2→R3, to its inside-out reflection ι∘a (equivalently to r∘ι for a reflection r of R3). More precisely, every two immersions S2→R3 are regularly homotopic: the space Imm⁡(S2,R3) is path connected, and a regular homotopy from ι to ι∘a cannot be chosen through embeddings (this last assertion uses AC, through the Jordan–Brouwer separation theorem).

Facts & Assumptions

Given: The unit sphere S2, the standard embedding ι, the antipodal map a(x)=−x, a reflection r of R3, the space Imm⁡(S2,R3) with the weak compact-open C∞ topology, and the formal-immersion space FImm⁡(S2,R3).

[F1]

The formal data of ι and of ι∘a are homotopic: they lie in the same path component of FImm⁡(S2,R3), and the difference class in π2(V2(R3))≅π2(SO(3)) vanishes. Standard and reflected two-sphere immersions have homotopic formal data in R^3

[F2]

The derivative map Imm⁡(S2,R3)→FImm⁡(S2,R3) is a weak homotopy equivalence (2<3, compact closed source), hence induces a bijection on path components; for the compact source S2, path components of Imm⁡ are the regular homotopy classes. The Smale–Hirsch immersion theorem, Regular homotopy classes of immersions are formal homotopy classes, Weak homotopy equivalence

[F3]

All immersions S2→R3 are regularly homotopic, since π2(SO(3))=0 and V2(R3)≅SO(3); equivalently the immersion space is path connected by the classification theorem. Smale's classification of sphere immersions in Euclidean space, The second homotopy group of SO(3) vanishes

[F4]

A regular homotopy is a smooth family whose every slice is an immersion; a homotopy through embeddings is a smooth family whose every slice is injective and immersive, hence an embedding of the compact sphere. Regular homotopy of immersions, Immersions, submersions, and constant-rank maps

[F5]

Jordan–Brouwer separation (AC): the image of every embedding S2↪R3 has exactly two complementary components, one bounded and one unbounded, with common boundary the image. Jordan–Brouwer separation, The Axiom of Choice

[F6]

The divergence theorem for bounded C1 Euclidean domains: for a bounded domain B with C1 boundary and the field X=13x, ∫Bdiv⁡X dV=∫∂BX⋅n dA, so the flux of 13x through the outward-oriented boundary equals vol⁡(B); with the opposite orientation the flux is −vol⁡(B). Divergence on a bounded C1 Euclidean domain

Proof

1.1F1F2F4

ι and ι∘a are regularly homotopic: by [F1] their formal data lie in one path component of FImm⁡(S2,R3), and the derivative map is a weak homotopy equivalence, hence a bijection on path components by [F2]; path components of Imm⁡(S2,R3) are the regular homotopy classes by [F2], so there is a regular homotopy S2×[0,1]→R3 from ι to ι∘a. Follow this homotopy by t↦At∘(ι∘a), where At is a smooth rotation path from I to A=−r. For a reflection in a plane with unit normal u, A fixes u and rotates u⊥ by π, so rotation through πt supplies this path. Reparametrizing both paths to be constant near their endpoints makes their concatenation smooth, with initial map ι and final map r∘ι.

1.2F3F4

Imm⁡(S2,R3) is path connected: by [F3] every two immersions of S2 into R3 are regularly homotopic, and regular homotopies are paths in the immersion space by [F4].

1.3F4F5F6

No regular homotopy from ι to r∘ι can be chosen through embeddings. Suppose H:S2×[0,1]→R3 were such a family with every slice an embedding. Define the flux S(t)=∫S2Ht∗ω, where ω is the 2-form of the field 13x, that is, the integral over the parametrised surface of 13H⋅(∂1H×∂2H) in positively oriented local coordinates. The integrand depends continuously on (x,t) and S2×[0,1] is compact, so S is continuous. For each t, Ht is a smooth embedding of the compact sphere, so by [F5] its image bounds a compact region Bt; the divergence theorem in the form of [F6] identifies S(t) with ±vol⁡(Bt), the sign being + or − according to the orientation of the parametrisation, so S(t)≠0 for every t; a continuous nonzero function on [0,1] has constant sign.

2.1F5F6step 1.1step 1.2step 1.3∎

Evaluating the two ends: for H0=ι with the positively oriented coordinates of S2 as the boundary of the unit ball, ι∗ω is the outward-oriented flux form of 13x through the unit sphere, whose integral is the volume 4π3 of the unit ball by [F6]. For H1=r∘ι or H1=ι∘a=(−I)∘ι, the chain rule gives ∂i(r∘ι)=r∘∂iι, and the identity (Au)×(Av)=det⁡(A)A(u×v) for either orthogonal map with determinant −1 shows that the pulled-back flux form changes sign: ∫S2(r∘ι)∗ω=−∫S2ι∗ω=−4π3. Hence S(0)>0>S(1), contradicting the constant sign forced in step 1.3. Therefore no homotopy from ι to r∘ι through embeddings exists, every regular homotopy between them has non-injective slices, and the inside/outside labelling necessarily changes along any eversion. The existence assertion of the theorem is step 1.1 and the path-connectedness is step 1.2; the Jordan–Brouwer input of step 1.3 carries AC, while the existence and classification assertions inherit countable choice from Smale–Hirsch.

RemarkRemark: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Sphere eversion cannot be an isotopy through embeddings

Remark

Assume AC, including the countable-choice hypotheses of the eversion and divergence suppliers. No homotopy from ι to ι∘a (or to r∘ι) through embeddings of S2 into R3 exists, so every eversion contains slices that are not injective; self-intersections are unavoidable. The invariant is the coorientation: an embedding of S2 into R3 bounds a compact complementary region (indeed a ball, by Alexander's theorem, which the argument below does not need; Jordan–Brouwer separation supplies only the existence of the bounded region), and the sign of the parametrisation relative to the boundary orientation of that region is locally constant along a continuous family of embeddings: the proof below exhibits it as the sign of a continuous flux integral, so continuity of the regions themselves is never invoked. Thus ι has the positively oriented parametrisation with respect to the outward normal while ι∘a and r∘ι have the opposite sign, and the two lie in different path components of the space of embeddings Emb⁡(S2,R3)⊆C∞(S2,R3). Equivalently, an ambient isotopy of R3 preserves the side of the image, whereas eversion reverses the inside/outside labelling. Regular homotopy is strictly coarser than isotopy here: the two maps are in one path component of Imm⁡(S2,R3) by the eversion theorem and in different path components of the embedding space.

Facts & Assumptions

Given: The unit sphere S2, the standard embedding ι, the antipodal map a, a reflection r of R3 with r∘ι=A∘(ι∘a) for a rotation A, and the family of slices of a hypothetical homotopy through embeddings.

[F1]

By the eversion theorem, ι and ι∘a are regularly homotopic through immersions, and every regular homotopy between them has non-injective slices; a homotopy through embeddings is a homotopy whose slices are injective immersions of the compact sphere, hence embeddings. Sphere eversion, Smooth embeddings, An injective immersion from a compact manifold is an embedding

[F2]

A regular homotopy has every slice immersive; embeddings are the injective immersions, and the orientation of the parametrisation relative to the bounded side is the coorientation sign of the remark. Regular homotopy of immersions, Immersions, submersions, and constant-rank maps, Orientable manifolds

[F3]

Jordan–Brouwer separation (AC): the image of an embedding S2↪R3 has exactly two complementary components, one bounded and one unbounded, with common boundary the image. Jordan–Brouwer separation, The Axiom of Choice

[F4]

The divergence theorem on a bounded C1 Euclidean domain: the flux of the field 13x through the boundary equals vol⁡(B) for the outward orientation and −vol⁡(B) for the inward orientation. Divergence on a bounded C1 Euclidean domain

Proof

1.1F2F3F4

Suppose t↦Ht is a continuous path in the weak C∞ space of embeddings from H0=ι to H1=r∘ι or H1=ι∘a. For each t the slice Ht is an embedding of the compact sphere and bounds a bounded region Bt by [F3]; define the flux S(t) of the field 13x through the parametrised surface Ht. Continuity of this path controls the values and first spatial derivatives uniformly on a finite chart cover of the compact sphere. Therefore the integrand 13Ht⋅(∂1Ht×∂2Ht) is continuous in (x,t) and hence bounded and uniformly continuous, so t↦S(t) is continuous; and by [F4], S(t)=±vol⁡(Bt) with the sign given by the orientation of the parametrisation relative to the outward normal of Bt, so S(t)≠0 for all t and the sign of S is constant.

2.1F1step 1.1∎

At the ends: for H0=ι in positively oriented coordinates of the unit sphere, the flux of 13x is the volume 4π3 of the unit ball, so S(0)>0; for H1=T∘ι with T=r or T=−I, the chain rule and (Tu)×(Tv)=det⁡(T)T(u×v) with det⁡T=−1 give S(1)=−S(0)<0. This contradicts the constant sign of step 1.1, so no homotopy through embeddings from ι to r∘ι exists, and by [F1] every regular homotopy from ι to ι∘a has non-injective slices: eversion necessarily produces self-intersections. AC is inherited from Jordan–Brouwer and from the eversion assertion; it also implies the countable-choice assumption of the divergence theorem. The sign computation is elementary.

The two path components just separated are components of the space of embeddings, while the eversion theorem puts the two maps in one component of the space of immersions; this is the precise sense in which regular homotopy is coarser than isotopy for the sphere in R3.

RemarkRemark: Literature-sourcedProof: Not applicableOpen item page →

Regular homotopy allows self-intersections but never rank drop

Remark

A regular homotopy H:M×[0,1]→N of immersions requires every slice Ht to be immersive (Regular homotopy of immersions, Smooth families of maps and their evaluation maps): the rank of dHt is dim⁡M at every point of every slice (Immersions, submersions, and constant-rank maps), and no slice may contain a point of rank drop, a cusp of the parametrised family or a point whose derivative degenerates.

Self-intersections, by contrast, are permitted: an immersion may identify two distinct points x≠y with Ht(x)=Ht(y) as long as the differential is injective at each point. Their tangent images may coincide; transversality is not required by the immersion condition. During an eversion of S2 in R3 the slices must acquire self-intersections and cannot be embeddings (Sphere eversion cannot be an isotopy through embeddings), while no slice may have a rank drop, so the two phenomena are logically independent.

For a fixed smooth map, the set of source points where its differential has full rank is open (The immersion and submersion loci are open). This is a statement about the source, rather than about a topology on a space of maps. A smooth family is a regular homotopy exactly when every slice is an immersion; immersive slices for t<1 do not guarantee an immersive final slice.

RemarkRemark: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

The sphere immersion groups are algebraic-topology computations, not differential-topology constructions

Remark

The homotopy-theoretic inputs used on this page — π2(SO(3))=0 (The second homotopy group of SO(3) vanishes) via the quaternion double cover and the covering isomorphism on πn for n≥2; π1(SO(2))≅Z via circle degree (Formal immersions of the circle in the plane are classified by the winding number); the Stiefel connectivities π0(Vm(Rn))=0 for n≥m+1 and π1(Vm(Rn))=0 for n≥m+2 (Stiefel manifolds are connected in positive codimension and simply connected in codimension at least two), and the higher homotopy exact sequences of the fibrations Vm(Rn)→Sn−1 — are algebraic-topology computations owned by the prerequisite pages on higher homotopy groups and cofiber sequences, fibrations and homotopy exact sequences, covering spaces and lifting, and the Hurewicz, Whitehead, Freudenthal and CW-approximation theorems. Here they are consumed as the values of πm(Vm(Rn)) and π2(SO(3)) in the sense of Higher homotopy group by based cubes; Smale's classification of sphere immersions in Euclidean space states the classification in terms of them.

This page contributes only the differential-topological reduction: the Smale–Hirsch weak homotopy equivalence (from the predecessor page), the identification of formal data with Stiefel sections, and the clutching difference class that turns the classification into these groups. No new homotopy-theoretic machinery is minted here, and conversely the groups πm(Vm(Rn)) remain algebraic-topology inputs; each use is recorded in the dependencies of the items above.

5 · Examples, counterexamples and false statements

None yet.

Sources