--- title: Cohomotopy url: https://www.emergentmind.com/topics/cohomotopy type: topic --- # Cohomotopy Cohomotopy is the contravariant homotopy theory whose degree-\(n\) objects are homotopy classes of maps into the \(n\)-sphere, \[ \pi^n(X)=[X,S^n], \] or, in the pointed setting, \(\pi^n(X)=[X,S^n]_{*}\). Unlike ordinary cohomology, represented by Eilenberg–Mac Lane spaces \(K(A,n)\), cohomotopy is generally unstable and non-abelian: its coefficient object is a finite sphere rather than an infinite loop space. Consequently, \(\pi^n(X)\) is generally a pointed set, although it may carry group, torsor, or stable-cohomotopy structures in suitable dimensions. Cohomotopy retains nonlinear homotopy information, including Hopf and Whitehead products, torsion, representation-theoretic data, and framed-submanifold information that can disappear under passage to ordinary cohomology or stabilization. ## 1. Representability, instability, and basic structures Ordinary cohomology with coefficients in an abelian group \(A\) is represented by \[ \widetilde H^n(X;A)\cong [X,K(A,n)]. \] The spaces \(K(A,n)\) form an \(\Omega\)-spectrum, producing suspension isomorphisms and abelian groups. Cohomotopy instead uses spheres: \[ \pi^n(X)=\pi_0\operatorname{Maps}^{*/}(X,S^n). \] The full mapping space \[ {}^n(X):=\operatorname{Maps}^{*/}(X,S^n) \] contains more information than its set of components. Its connected components are the cohomotopy classes, while its loop spaces encode higher monodromy and, in proposed physical applications, higher observables [1912.10425]. The terminology “non-abelian generalized cohomology” emphasizes that a connected pointed space \(\mathcal A\) may serve as a coefficient object through \[ H^1(X;\Omega\mathcal A):=\pi_0\operatorname{Maps}(X,\mathcal A). \] For \(\mathcal A=S^n\), this gives unstable cohomotopy. The analogy with homotopy groups is \[ \pi_n(X)\simeq \pi_0\operatorname{Maps}^{*}(S^n,X), \qquad \pi^n(X)\simeq \pi_0\operatorname{Maps}^{*}(X,S^n). \] Cohomotopy is contravariant in \(X\), invariant under homotopy equivalence, and naturally associated with pointed spaces. Its algebraic structure depends on the target. For example, multiplication on \(S^3\) induces a group structure on \([X,S^3]\), while \([X,S^2]\) is generally only a pointed set equipped with an action by \([X,S^3]\). In higher-dimensional calculations, cohomotopy sets may fit into exact sequences, extensions, or torsor descriptions rather than canonical group structures [1203.1608]. Stable cohomotopy replaces spheres by the sphere spectrum: \[ \pi_s^n(X)=[\Sigma^\infty X,\Sigma^n\mathbb S]. \] It is a genuine generalized cohomology theory with groups, long exact sequences, and suspension stability. The relationship among the three theories is schematically \[ \text{ordinary cohomology} \longleftarrow \text{stable cohomotopy} \longleftarrow \text{unstable cohomotopy}. \] The arrows indicate loss of information: ordinary cohomology forgets nonlinear and torsion data, while stabilization can suppress unstable Hopf and Pontryagin–Thom phenomena. ## 2. Geometric interpretation through Pontryagin–Thom theory For a closed smooth \(D\)-manifold \(X^D\), the unstable Pontryagin–Thom theorem identifies degree-\(n\) cohomotopy with framed cobordism classes of codimension-\(n\) submanifolds: \[ \pi^n(X^D)\simeq \left\{ \begin{array}{c} \text{closed submanifolds }\Sigma^{D-n}\hookrightarrow X^D\\ \text{with framed normal bundle} \end{array} \right\}\bigg/\text{framed cobordism}. \] A map \(c:X\to S^n\), after choosing a regular value, determines the framed inverse image \[ \Sigma=c^{-1}(\mathrm{pt}). \] Conversely, a framed submanifold determines a Pontryagin–Thom collapse map. The normal framing is a trivialization \[ N\Sigma\simeq \Sigma\times\mathbb R^n. \] This interpretation explains why cohomotopy can refine ordinary characteristic classes. The underlying submanifold determines a homology class, but the framing contributes additional information. In degree \(n\), cohomotopy therefore records both a codimension-\(n\) geometric cycle and its normal-framing data. For maps from a closed oriented \(4\)-manifold \(X\), the classical cases are \[ [X,S^n]\cong H^n(X;\mathbb Z) \qquad(n=1\text{ or }n\geq 4), \] whereas \(n=2,3\) contain genuinely nonclassical information. Pontryagin–Thom identifies \[ [X,S^3]\cong F_1(X), \qquad [X,S^2]\cong F_2(X), \] where \(F_k(X)\) denotes normally framed bordism classes of \(k\)-dimensional submanifolds [1203.1608]. For \([X,S^3]\), forgetting the framing gives a surjection \[ h_1:[X,S^3]\longrightarrow H_1(X;\mathbb Z). \] The kernel depends on the parity of the intersection form. If \(X\) is odd, then \[ [X,S^3]\cong H_1(X;\mathbb Z). \] If \(X\) is even, there is an exact sequence \[ 0\longrightarrow\mathbb Z/2 \longrightarrow [X,S^3] \overset{h_1}{\longrightarrow}H_1(X;\mathbb Z) \longrightarrow0. \] The extension class is determined by \(w_2(X)\); it splits precisely when \(X\) is spin. Thus, for a spin \(X\), \[ [X,S^3]\cong H_1(X;\mathbb Z)\oplus\mathbb Z/2. \] For \([X,S^2]\), forgetting the framing maps to \[ h_2:[X,S^2]\longrightarrow H_2(X;\mathbb Z). \] A surface admits a normal framing precisely when its self-intersection vanishes. Hence the image is \[ H_2^\circ(X)= \{\alpha\in H_2(X;\mathbb Z)\mid \alpha\cdot\alpha=0\}. \] The group \([X,S^3]\) acts transitively on each fiber of \(h_2\), and the stabilizer is twice the framed intersection map: \[ \operatorname{Stab}(F_\nu)=\operatorname{im}(\tau_\alpha), \qquad \tau_\alpha=2i_{\alpha,\nu}. \] Consequently, each fiber is a torsor for \[ [X,S^3]/\operatorname{im}(\tau_\alpha). \] The factor \(2\) arises both from resolving double points in framed bordisms and from the degree of the conjugation map \(S^2\times S^1\to S^3\). The co-degree-one calculation generalizes this picture. For a closed connected smooth \((n+1)\)-manifold \(M\), \(n\geq3\), \[ [M,S^n]\cong F_1(M), \] and forgetting the framing gives \[ 0\longrightarrow\ker(h)\longrightarrow F_1(M) \overset{h}{\longrightarrow}H_1(M;o_M)\longrightarrow0, \] where \(o_M\) is the orientation local system. The obstruction controlling the extension is \[ w_1(M)^2+w_2(M)\in H^2(M;\mathbb Z_2), \] the obstruction to a \(Pin^-\)-structure. Type I manifolds have \[ [M,S^n]\cong H_1(M;o_M), \] whereas type II manifolds have an extension by \(\mathbb Z_2\), splitting precisely when a \(Pin^-\)-structure exists [2307.03805]. For an oriented spin rank-\(n\) vector bundle \(E\to M\), a transverse section has a framed one-dimensional zero locus \(L\), defining a refined Euler class \[ e_{\mathrm{coh}}(E):=[L,\varphi]\in[M,S^n]. \] Its image under the cohomotopy Hurewicz map is the Poincaré dual of the ordinary Euler class. The refined class vanishes exactly when \(E\) admits a nowhere-vanishing section. Thus ordinary Euler-class vanishing can be insufficient: the cohomotopy refinement detects a residual framing obstruction. ## 3. Postnikov towers and explicit calculations The difference between cohomotopy and ordinary cohomology is controlled by the Postnikov tower of the sphere. The bottom stage of \(S^4\) is \[ (S^4)_0=K(\mathbb Z,4), \] and successive stages impose \(k\)-invariant obstructions. The first \(2\)-primary obstruction is \[ Sq^2\rho_2\iota_4\in H^6(K(\mathbb Z,4);\mathbb Z_2). \] Further stages involve secondary operations, mod-\(8\) lifts of \(Sq^4\), \(P_{11}\), and higher operations whose mod-\(2\) reductions include \(Sq^8\iota_4\). At odd primes, the tower contains operations such as \[ P^1_3\rho_3\iota_4 \quad\text{and}\quad P^1_5\rho_5\iota_4. \] For an integral class \(c\in H^4(Y^{11};\mathbb Z)\), lifting \(c\) to degree-four cohomotopy requires successive obstruction vanishing. The primary conditions include \[ Sq^2(c)=0\in H^6(Y;\mathbb Z_2), \qquad P^1_3(c)=0\in H^8(Y;\mathbb Z_3), \] followed by the vanishing of secondary and tertiary obstructions such as \(\alpha_7\), \(\beta_8\), and \(P_{11}\). In particular, \[ Sq^4(c)=c\smile c=0\in H^8(Y;\mathbb Z_2) \] appears as a necessary condition. These operations generally depend on choices of preceding lifts; they are not all primary operations on \(c\) alone [2001.07640]. The rational distinction is already visible for \(S^4\). Its Sullivan model has generators \(g_4,g_7\) with \[ dg_4=0, \qquad dg_7=g_4\wedge g_4. \] A rational degree-four cohomotopy cocycle therefore contains a closed \(4\)-form \(G_4\) together with a \(7\)-form \(G_7\) trivializing its square: \[ dG_4=0, \qquad dG_7=G_4\wedge G_4. \] Thus rational cohomotopy is not merely rational cohomology; it includes a nonlinear trivialization of the quadratic operation \(G_4\wedge G_4\). Recent calculations for closed orientable \((n-1)\)-connected \((2n+2)\)-manifolds with \(n=2,3,4\) and \(2\)-torsion-free homology combine these Postnikov towers with suspension decompositions. For simply connected \(6\)-manifolds, \[ \pi^6(M)\cong\mathbb Z, \qquad \pi^5(M)\cong\mathbb Z/(1-\varepsilon), \] and \(\pi^4(M)\) fits into \[ 0\longrightarrow\mathbb Z/(1-\varepsilon) \longrightarrow\pi^4(M) \longrightarrow \ker\!\left(Sq^2_{\mathbb Z}:H^4(M;\mathbb Z)\to H^6(M;\mathbb Z)\right) \longrightarrow0. \] For \(2\)-connected \(8\)-manifolds, \[ \pi^7(M)\cong\mathbb Z, \qquad \pi^8(M)\cong\mathbb Z, \qquad \pi^5(M)\cong H^5(M)\oplus G_{24}, \] where \(G_{24}\subseteq\mathbb Z/24\). For \(3\)-connected \(10\)-manifolds, \[ \pi^9(M)\cong\mathbb Z, \qquad \pi^{10}(M)\cong\mathbb Z, \qquad \pi^7(M)\cong G_{24}, \] and \[ \pi^6(M)\cong \ker\!\left(\mathcal P^3_{\mathbb Z}:H^6(M;\mathbb Z)\to H^{10}(M;\mathbb Z/3)\right). \] The groups \(\pi^4(M)\) remain unresolved in the \(8\)- and \(10\)-dimensional cases under consideration [2410.16055]. ## 4. Equivariant and stable cohomotopy For a finite group \(G\), equivariant cohomotopy uses \(G\)-equivariant maps into representation spheres. If \(V\in RO(G)\) is a real \(G\)-representation, its one-point compactification is \[ S^V=(\mathbb R^V)^{\mathrm{cpt}}, \] and \[ \pi_G^V(X)=[X,S^V]^G_*. \] The equivariant degree is representation-valued rather than integer-valued. Fixed-point geometry is organized by the isotropy subgroups \(H\leq G\), with stages \[ c^H:X^H\longrightarrow (S^V)^H. \] Equivariant stable cohomotopy is obtained by stabilization over representations. For compact Lie groups, the classical \(RO(G)\)-graded theory studies stabilized equivariant maps between representation spheres. For finite groups, the stable equivariant sphere ring is identified with the Burnside ring: \[ \mathbb S_G^0\simeq A(G). \] The Boardman homomorphism linearizes finite \(G\)-sets into real permutation representations: \[ \beta:\mathbb S_G^0\longrightarrow KO_G^0, \qquad \beta(S)=\mathbb R[S]. \] This passage from unstable cohomotopy to \(K\)-theory loses information, especially fixed-point subset data. The ghost kernel illustrates a different limitation of fixed-point detection. For pointed \(G\)-spaces \(X,Y\), the ghost map is \[ [X,Y]^G\longrightarrow H^0(G;[X,Y])\oplus [X^G,Y^G], \qquad f\longmapsto(f,f^G). \] Its kernel consists of equivariant stable maps that become null after forgetting equivariance and are null on \(G\)-fixed points. For \(G=C_p\), \(p\) odd, and \[ V_a=a\mathbb C G, \qquad W_a=a\mathbb C G/(\mathbb C\oplus\mathbb R G), \] the equivariant stable cohomotopy group \[ [\mathbb CP(V_a)_+,S^{W_a}]^G \] has nontrivial ghost kernel for \(p>5\). More precisely, for \(p>3\), its \(p\)-power torsion is elementary abelian of rank \(r\) satisfying \[ 1\leq r\leq\frac{p+1}{2}. \] The proof uses the Borel cohomology Adams spectral sequence \[ E_2^{s,t} = \operatorname{Ext}_{b^*b^*}^{s,t} \bigl(b^*S^{W_a},b^*\mathbb CP(V_a)_+\bigr) \Longrightarrow [\mathbb CP(V_a)_+,S^{W_a}]^G_{t-s,\widehat p}, \] together with Steenrod-algebra calculations and a filtration of the equivariant projective space [1110.2283]. For proper actions of Lie groups, finite-dimensional representation spheres may be inadequate. An analytic extension replaces them with \(G\)-Hilbert bundles, Fredholm morphisms, compact nonlinear perturbations, and fiberwise one-point compactifications. A cocycle is a quadruple \[ (E,F,l,c), \] where \(l:E\to F\) is Fredholm and \(c:E\to F\) is an equivariant compact perturbation such that \(l+c\) extends to a pointed map \[ S^E\longrightarrow S^F. \] The resulting theory is graded by equivariant real \(K\)-theory, \[ KO_G^0(X), \] rather than only by \(RO(G)\). It satisfies functoriality, excision, homotopy invariance, relative exactness, induction and restriction, products, and suspension. For discrete groups acting on proper \(G\)-CW pairs, the parametrized Schwartz index identifies this analytic theory with Lück’s finite-dimensional equivariant stable cohomotopy under the stated hypotheses [1302.1712]. ## 5. Parametrized cohomotopy and gauge-theoretic constructions Cohomotopy can be parametrized over a base space. If \(p:E\to B\) is a fiberwise pointed space, stable cohomotopy is formed in the category of spectra over \(B\). For parametrized sphere spectra \(S^\xi\) associated with vector or virtual bundles, \[ [S^\xi,S^\eta]_B \] denotes stable fiberwise homotopy classes. This framework supports family versions of gauge-theoretic invariants. For a smooth family of closed oriented \(4\)-manifolds with complex spin structure, \[ p:Y\to B, \] the family Seiberg–Witten monopole map is a fiber-preserving \(\mathbb T\)-equivariant map \[ \mu:\mathcal U\longrightarrow\mathcal V, \] with linearization \[ \lambda:\mathcal U\longrightarrow\mathcal V. \] After finite-dimensional approximation and one-point compactification, it gives \[ \operatorname{BF}(Y/B,\mathfrak s_{Y/B}) \in \left[ S^{\ker(\lambda)},S^{\operatorname{coker}(\lambda)} \right]^\mathbb T_B. \] This is the family Bauer–Furuta invariant. It restricts on every fiber to the ordinary Bauer–Furuta invariant and is natural under pullback: \[ \operatorname{BF}(g^*Y/B',g^*\mathfrak s) = g^*\operatorname{BF}(Y/B,\mathfrak s). \] A family of Fredholm operators also determines a Thom spectrum \(B^\lambda\). The parametrized monopole map yields an ordinary stable cohomotopy class \[ \operatorname{BF}(Y/B,\mathfrak s) \in \pi^0(B^\lambda). \] For a fixed complex spin \(4\)-manifold \((X,\mathfrak s_X)\), the universal family over \[ B\operatorname{Spiff}^c(X,\mathfrak s_X) \] defines a universal characteristic cohomotopy class. Pullback along a classifying map of a family recovers that family’s invariant [2002.01738]. The K3 example demonstrates that family stable cohomotopy contains information not determined by the invariant of an individual fiber. For a family over \(S^2\), the Dirac index bundle has complex rank \(2\), while the plus-harmonic bundle has rank \(3\). The universal Kummer family realizes Dirac-bundle degree \(-2\), and the family invariant constrains the Dirac index bundle to have even degree. Mapping tori of diffeomorphisms yield invariants depending only on isotopy class; subtracting the invariant of the identity gives a reduced mapping-torus invariant that vanishes for diffeomorphisms isotopic to the identity. Cohomotopy also underlies nonlinear Fredholm constructions for proper group actions. The monopole map in Bauer–Furuta theory is a proper compact perturbation of a Fredholm operator and therefore defines an analytic cohomotopy cocycle. Restriction along the central \(S^1\) recovers the usual \(S^1\)-equivariant Bauer–Furuta invariant within the proper-action framework [1302.1712]. ## 6. Twisted and differential cohomotopy Twisting replaces the fixed sphere \(S^n\) by a sphere bundle. If \(V\to X\) is a rank-\((n+1)\) vector bundle, its unit sphere bundle \(S(V)\to X\) is the coefficient object for \(V\)-twisted degree-\(n\) cohomotopy. The \(J\)-twist is induced by the action of orthogonal groups on spheres: \[ J_n:O(n)\longrightarrow\operatorname{Aut}(S^{n-1}). \] For a tangent bundle, this produces a spherical fibration whose sections represent twisted cohomotopy classes. In the M-theory applications, a tangentially twisted degree-four cocycle is represented schematically by a map into \[ S^4\sslash\operatorname{Sp}(2)\longrightarrow B\operatorname{Sp}(2). \] The quaternionic Hopf fibration \[ S^3\longrightarrow S^7\longrightarrow S^4 \] relates degree-four and degree-seven data. Its rational model includes a degree-three generator \(h_3\): \[ dh_3=\omega_4, \qquad d\omega_7=-\omega_4\wedge\omega_4. \] In the twisted setting, the corresponding relation becomes \[ dH_3=G_4-\frac14p_1, \] up to normalization conventions. The shifted flux \[ \widetilde G_4=G_4+\frac14p_1 \] is integrally quantized under the cohomotopical character map. In compatible \(\operatorname{Sp}(2)\cdot\operatorname{Sp}(1)\)-structures on \(8\)-manifolds, the resulting degree-seven equation includes the nonlinear \(G_4\wedge G_4\) term and gravitational corrections \(I_8\) or \(X_8\). Twistorial cohomotopy refines the \(S^4\)-coefficient by the twistor fibration \[ S^7\longrightarrow\mathbb CP^3\longrightarrow S^4. \] The additional degree-two generator \(f_2\) produces a differential-form system \[ dF_2=0, \] \[ dH_3=G_4-\frac14p_1-F_2\wedge F_2, \] together with the degree-seven relation. Integral classes satisfy \[ [F_2]\in H^2(X;\mathbb Z), \qquad \left[G_4+\frac14p_1\right]\in H^4(X;\mathbb Z), \] and \[ \left[G_4-\frac14p_1\right]=[F_2\wedge F_2]. \] This realizes a factorized Green–Schwarz-type relation associated with an \(S(U(1)^2)\)-sector rather than arbitrary \(E_8\)-gauge data [2008.08544]. Differential cohomotopy combines sphere-valued topological data, differential-form data, and a compatibility homotopy. For \(S^4\), a differential cocycle can be represented schematically by \[ (c,h,\omega), \] where \(c:X\to S^4\), \(\omega\) consists of forms \(\omega_4,\omega_7\) with \[ d\omega_4=0, \qquad d\omega_7=\omega_4^2, \] and \(h\) identifies the realification of \(c\) with the differential-form data. Differential Postnikov towers combine the ordinary Postnikov tower, its real localization, and a Sullivan-model tower. Torsion homotopy groups remain purely topological in the differential refinement, so Steenrod and secondary obstructions remain visible even though differential forms do not detect torsion [2001.07640]. For the M5-brane, a \(3\)-form \(H_3\) on the worldvolume is interpreted as a homotopy lift through the quaternionic Hopf fibration. The full Hopf–Wess–Zumino anomaly is represented by a twisted homotopy Whitehead integral involving both \(H_3\wedge \widetilde G_4\) and \(G_7\). Under integral \(J\)-twisted cohomotopy quantization, twice the anomaly is integral: \[ 2\widetilde S\in\mathbb Z. \] This ensures that the exponentiated Wess–Zumino action is well-defined. For multiple M5-branes, the level is multiplied by \(N(N+1)\), which is even for every integer \(N\) [1906.07417]. The restriction of the equivariant quaternionic Hopf fibration to a heterotic M5-brane yields an \(\operatorname{Sp}(1)\cong\operatorname{SU}(2)\) gauge field. The induced structure factors through \[ B\operatorname{String}^{c_2}_2(4) = B\operatorname{Spin}(4)\times_{B^3U(1)}B\operatorname{Sp}(1), \] and the corresponding differential relation is \[ dH_3=\operatorname{Tr}(R\wedge R)-\operatorname{Tr}(F\wedge F). \] Thus the induced field exhibits a \(c_2\)-twisted String structure [2002.11093]. ## 7. Applications, extensions, and limitations Equivariant cohomotopy assigns representation spheres to group actions and fixed-point stages to isotropy subgroups. For a finite group \(G\), the unstable equivariant group \[ \pi_G^V(X)=[X,S^V]^G_* \] can distinguish positive-dimensional stages, with integer Hopf degrees, from zero-dimensional fixed stages, with subset-valued data. The unstable equivariant Hopf–tom Dieck theorem yields a stagewise description involving factors of \(\mathbb Z\) and \(\{0,1\}\). For ADE groups acting on the quaternionic representation \(\mathbf 4_{\mathbb H}\), mobile branes occur in regular-representation multiples, while fixed-point contributions are finite unstable data. Stabilization and the Boardman map retain Chan–Paton representation information but lose the unstable distinctions among fixed-plane charges [1909.12277]. Differential cohomotopy on Penrose-diagram spacetimes can be modeled by configuration spaces. For \(d