Geometri approa h towards stable homotopy groups of spheres. The Steenrod-Hopf invariant 8 0 0 ∗ 2 Pyotr M. Akhmet'ev n a J 9 Àííîòàöèÿ ] T Inthispaper ageometri approa h toward stablehomotopy groups G ofspheres,basedonthePontrjagin-Thom [P℄ onstru tionisproposed. . h From this approa h a new proof of Hopf1I5n,v3a1ri,a6n3t,1O2n7e Theorem by t a J.F.Adams [A1℄ for all dimnens>ion1s2e7x ept is obtained. m It is proΠved that for in the stable homotopy group [ of spheres n there is no elements with Hopf invariant one. The 1 new proof is based on geometri topology methods. The Pontrjagin- v Thom Theorem (in the form proposed by R.Wells [W℄) about the 2 representationofstablehomotopygroupsoftherealproje tivein(cid:28)nite- 1 4 dimensional spa e (this groups ismapped onto 2- omponents of stable 1 homotopy groups of spheres by the Khan-Priddy Theorem [A2℄) by . 1 obordism lasses of immersions of odimension 1 of losed manifolds 0 (generally speaking, non-orientable) is onsidered. The Hopf Invariant 8 0 is expressed as a hara teristi number of the dihedral group for the : self-interse tion manifold of an immersed odimension 1 manifold that v i represents the given element in the stable homotopy group. In the X new proof the Geometri Control Prin iple (by M.Gromov)[Gr℄ for r a immersions inagiven regularhomotopy lassesbasedonSmale-Hirs h Immersion Theorem [H℄ is required. f : Mn−1 # Rn n = 2l −1 Let , , be a smooth immersion of odimension 1. The hara teristi number (cid:10)wn−1(M);[Mn−1](cid:11) = h(f) 1 ∗ This work was supported in part by the London Royal So iety (1998-2000), RFBR 08-01-00663,INTAS 05-1000008-7805. 1 is alled the stable Hopf invariant or the Steenrod-Hopf invaMrian−n1t. This hara teristi number depends only on the immersed manifold itself. The relationship with the de(cid:28)nition of the Steenrod-Hopf invariant in algebrai topology is onsidered in [E℄,[K℄,[L℄. Theorem (by J.F.Adams), [A℄ l ≥ 4 h(f) = 0 For , . Skew-framed immersions f : Mn−k # Rn k κ : E(κ) → Mn−k Let beMann−ikmmersionΞof: koκd→imeνn(sfi)on .Let be a line bundle over and let be an isomorphism of the f k normal bundle of the immersion with the Whitney sum of opies of the κ line bundle . (f,κ,Ξ) We shall all tκhe∈trHip1l(eMn−k;Z/2)a sknew-framed immwer(sκio)n with 1 hara teristi lassMn−k . (If is odd then is the orientation lass of , see [A-E℄ for more details). The Steenrod-Hopf invariants for skew-framed immersions (cid:10)w (κ)n−k;[Mn−k](cid:11) = h(f,κ,Ξ) 1 The hara teristi lass is alled the (f,κ,Ξ) Steenrod-Hopf invariant of the skew-framed immersion . The Main Theorem 1 f : Mn−k # Rn,κ,Ξ) n = 2l − 1 Let ( be a skew-framed immersion, , dim(M) = n−k = n+1 +7 n ≥ 255 l ≥ 8 2 . Then for (i.e. for ) h(f,κ,Ξ) = (cid:10)w (κ)dim(M);[M](cid:11) = 0 (mod 2). 1 Corollary n ≥ 255 Adams' Theorem for . D 8 4 Let be the dihedral group of order , D = {a,b|a4 = b2 = e,[a,b] = a2}. 4 This is the group of symmetries of the two oordinate axes in the plane. Let I = {e,a,a2,a3},I = {e,b,a2,a2b},I = {e,ab,a2,a3b} a b c 2 D Z/4 I 4 a be the subgroups in of index 2. The y liπ (cid:21)sZub/g2r⊕ouZp/2 is generateId bIy rotation of the plane through the angle 2. The (cid:21)subgroup b c ( ) is generated by symmetrZie/s2(or re(cid:29)e tioIns)∩wIit=h r{ees,pae2 }t to the bise tors b c of the oordinate axes. The (cid:21)subgroup is generated by the entrfal :syMmmn−ektry#. Rn,κ,Ξ) Let ( Nn−2k be a skew-fraRmnedg g:eNnenr−i2 k i#mmRenrsion. Its double points mνa(ngi)fold is immersed into , Ψ : ν(g) = kη,∗and the nη∗ormal bundle admits a anoniN anl−d2ek ompoDsition , wheηre∗ 4 is a two-dimensional bundle over E(Dw)i→th K(-Dstr,u1 )ture. The bundle 4 4 is a pηul:l-Nban −k2ko→f thKe(uDniv,e1r)sal bundle , via the lassifying 4 map . N¯n−2k → Nn−2k Let us onsider the anoni al 2-fold overing over the g double poiInt⊂mDanifoIld=of{et,hae2,imabm,ae3rbs}ion . This overing orresponds to the c 4 c subgrouκ¯p ∈ H1(N¯,n−2k;Z/2) . Let I → I be the oIho=m{oelo,gay2 ≃ laasbs} =orrZe/sp2onding to the c d d epimorphism witah3tbhe image κ¯ = i∗(κ) i : N¯n(−th2ke#kerMnenl−iks generatedbytheelement ).Bythede(cid:28)nition , is the anoni al immersion of the double point overing. Let us de(cid:28)ne the following hara teristi number h¯(g,η,Ψ) = (cid:10)κ¯n−2k;[N¯n−2k](cid:11). Lemma 2 ¯ h(f,κ,Ξ) = h(g,η,Ψ). Proof of Lemma 2 ImmediatefromHerbert's Theorem.(Con erningHerbert's Theorem,see e.g. [E-G℄.) De(cid:28)nition (Cy li stru ture for skew-framed immersions) (f,κ,Ξ) Nn−2k Let be a skew-framed immersion, be the (odd-dimensional) f double self-interse tion point manifold of . A mapping µ : Nn−2k → K(I ,1) a I = {e,a,a2,a3} f a ( ) is alled a y li stru ture for if (cid:10)µ∗(t);[Nn−2k](cid:11) = h(f), t ∈ Hn−2k(K(I ,1);Z/2) a where is the generator. The following lemma is proved by an expli it al ulation. 3 The Main Lemma 3 (jointly with P.J.E les (1998)) n−k = n+1+7 n−2k = 15 n ≥ 31 µ : Nn−2k → K(D ,1) Let 2 ( ), , 4 be a y li f h(f,κ,Ξ) = 0 (mod2) stru ture for . Then . Lemma 4. Cy li Stru ture for skew-framed immersions n ≥ 255 f : Mn−k # Rn n−k = For , an arbitrary skew-framed immersion , n+1 +7 2 is regularly homotopi to an immersion with a y li stru ture. Proof The proof is a orollary of Lemma 5 and Proposition 6. De(cid:28)nition (Cy li Stru ture for generi mappings of the standard proje tive spa e) g : RPn−k → Rn n ≥ 3k n = 2l −1 Let Nn−2kbe a generi mappi(nng−, 2k−1) , with double (p∂oNin)tnm−2akn−i1fo⊂ldRPn−k and riti alpoints -dimensionalsubmanifold η : (Nn−2k,∂.N) → (K(D ,1),K(I ,1)) 4 b Let be the stru tured mapping g orresponding to . This stru tured mapping is de(cid:28)ned analogously with the ase of skew-framed immersions. Obviously, the restri tion of the mapping η to the boundary of the douKbl(eIp,o1i)nt⊂s mKa(nDifo,ld1), i.e. to the riti al points b 4 Isub⊂mIanifold, has the target . The standard in lusion d a asthesubgroup oftheindex2iswell-de(cid:28)ned.We shall alla mapping µ : (Nn−2k,∂Nn−2k−1) → (K(I ,1),K(I ,1)) a d g a y li stru ture for , if the following onditions hold: (∗) the homologi al ondition: (cid:10)µ∗(t);[Nn−2k,∂Nn−2k−1](cid:11) = 1 (mod 2), t ∈ Hn−2k(K(I ,1),K(I ,1);Z/2) a d where Hn−2k(K(I ,1)i,sK(tIhe);Zg/e2n)erator in the a d okernel of the homomorphism (remark: this µrel(a[t∂iNven− 2hk−ar1a]) t∈erHisti nu(mKb(eIr,1i;sZ/w2e)l)l-de(cid:28)ned, by expli it al ulations ∗ n−2k−1 d is trivial); (∗∗) the boundary ondition : p ◦η| : ∂Nn−2k−1 → K(I ,1) → K(I ,1), c ∂N b d I = {e,b,a2,ba2} p : K(I ,1) → K(I ,1) b c b d Id,=wh{eer,ea2 ≃ ba2} µis|∂tNhne−s2kt−a1ndard proje tion with the image , oin ides with . 4 Lemma 5. Geometri Control Let (f,κ,Ξ), f : Mn−k # Rn be a skew-framed immersion. Let us assume (see Proposition 6 below) that there exists a generi mapping g : RPn−k → Rn, n ≥ 3k, with a y li stru ture µ : (Nn−2k,∂Nn−2k−1) → (K(I ,1),K(I ,1)). a d (f′,κ,Ξ′) Thenthereexistsaskew-framedimmersion intheregularhomotopy f lass of with a y li stru ture. The idea of the proof of Lemma 5. Take the mapping g ◦κ : Mn−k → RPn−k → Rn. f By [Gr℄, 1.2.2, in the regular homotopy lass of there exists a generi immersion f′ : Mn−k # Rn C0 de(cid:28)ned as a - lose generi regular perturbation ( arbitrary small) of a (singular) mapping g ◦κ. A onstru tion by S.A.Melikhov (2004) J (2l−4+1) = r Z/4 Let usSd7e/niotdeimby(J) =the2lj−o1in+o7f= n−k opiePsLof the standarid : J ⊂- leRnns J spa e , . There is a (cid:21)embedding l ≥ 8 for p.′ : Sn−k → J r S7 → S7/i pˆ :LSent−k/i → J be the joinof opies of thep′stapnd:aRrPdn −okve→r J , be the quotient mappπing: oRfPn−,k → Sn−k/i pbˆe the omposition iof◦thpˆe:sStna−nkd/aird→pJro→je tRionn gˆ : Swn−itkh/i →. TRhne J ompoεsition is well-de(cid:28)ine◦d.pˆLedt: RPn−k → Rn J be an -small generi alternation of the mapping , be ε ε << ε gˆ◦π 1 1 de(cid:28)ned by a -small generi alternation, , of the omposition . 5 1 Proposition 6 d : RPn−k → Rn The Melikhov map is equipped with a y li stru ture. The rest of the paper on erns the proof of this result. This Proposition with Lemma 5 implies Lemma 4 and Lemma 4 with the Main Lemma 3 implies Lemma 2 and the main Theorem 1. The beginning of the proof of the Proposition 6 Γ RPn−k 0 Let be the delated produ t of the standard proje tive spa e Γ = (RPn−k ×RPn−k \∆ )/T′, 0 diag where the quotient is determined with respe t to the free involution T′ : RPn−k ×RPn−k \∆ → RPn−k ×RPn−k \∆ diag diag ′ T (x,y) = (y,x). The lassifying map η : Γ → K(D ,1) Γ0 0 4 π (Γ ) = D T′ 1 0 4 is well-de(cid:28)ned. (NIote⊂thDat ∆ an⊂d tΓhe involution orresponds c 4 antidiag 0 to the subgroup .) Let be a subspa e, alled the antidiagonal, de(cid:28)ned as ∆ = {(x,y) ∈ (RPn−k ×RPn−k/T′)|T(x) = y}, antidiag where T : RPn−k → RPn−k is the standard involution on the overing RPn−k → Sn−k/i. Γ Let us denote by the spa e Γ \(U(∆ )∪U(∆ )), 0 antidiag diag U(∆ ) ∆ U(∆ antidiag antidiag diag where is a small regular neighborhoods of , is a Γ ∆ 0 diag small regular neighborhood of the end of near the deleted diagonal . 1 The author was developed this proof following onversations with Prof. O.Saeki and Dr. R.R.Sadykov (2006)) 6 ε (The radius of the regular neighborhoods depends on a onstant of an approximation in the Melikhov onstru tion.) Γ The spa e is a manifold with boundary. The involution T : RPn−k → RPn−k indu es the free involution T : Γ → Γ. Γ A polyhedron Σ = {[(x,y)] ∈ Γ ,p(x) = p(y)}, Σ ⊂ Γ 0 0 0 0 of double points of the mapping p : RPn−k → J is alled the singular set or the singular polyhedron. The mapping η : Σ → K(D ,1) Σ0 0 4 is well-de(cid:28)ned as the restri tion of the mapping η | . Γ0 Σ0 The subpolyhedron Σ ⊂ Γ 0 0 de omposes as Σ = Σ ∪K, K ⊂ Γ, 0 antidiag where Σ = Σ ∩U(∆ ). antidiag 0 antidiag η | η : K → K(D ,1) The restri tion Γ0 K will be denoted by K 4 . η K Boundary onditions of K The diagonal and antidiagonal boundary omponents of will be denoted by Q = K ∩∂U(∆ ), Q = K ∩∂U(∆ ). diag diag antidiag antidiag η | : Q → K(D ,1) The restri tion K Qantydiag antidiag 4 is de omposes as i ◦η : Q → K(I ,1) ⊂ K(D ,1). a antidiag antidiag a 4 η | : Q → K(D ,1) The restri tion K Qdiag diag 4 is de omposed as i ◦η : Q → K(I ,1) ⊂ K(D ,1). b diag antidiag b 4 7 RK The resolution spa e . RK K Let us onstru t the spa e alled the resolution spa e for . This spa e is in luded into the diagram K(I ,1) ←φ− RK −p→r K. a pr−1(Q ) RQ pr−1(Q ) RQ diag diag diag antidiag Let us denote by , by . The Q antidiag boundary onditions on are: pr RQ −→ Q antidiag antidiag φ ց ւ η antidiag K(I ,1). a Q diag The boundary onditions on are: pr RQ −→ Q diag diag φ ↓ ↓ η diag K(I ,1) ←pb− K(I ,1). d b The diagrams above are in luded into the following diagram: K(I ,1) a ↑ φ տ RK ←− RQ ∪RQ diag antidiag ↓ ↓ K ⊃ Q ∪Q diag antidiag ↓ η ւ K(D ,1). 4 d : RPn−k → Rn Let us onsider Melikhov's mappiin◦gp◦π : RPn−k → Sn−(kt/hiis→isJa⊂sRmnall geneNri n−a2lkternation of the omposition d ). NLent−2k ⊂ Γbe thedoublepointmanifold(withNbno−u2nkdary)of ,theembedding 0 is well-de(cid:28)ned. The manifold is de omposed into two manifolds (with boundary) along the ommon omponent of the boundary Nn−2k = N ∪N , antidiag d N = Nn−2k ∩U(∆ ), antidiag antidiag N = Nn−2k ∩Γ. d 8 Lemma 7 res : N → RK d There exists a mappµin:gN → K(I ,1) alled the resolution mapping that d a indu es a mapping in luded into the following diagram: K(I ,1) = K(I ,1) a a ↑ φ ↑ µ տ res RK ←− N ⊃ W ∪W d diag antidiag ↓ η ↓ η ւ K(D ,1) = K(D ,1), 4 4 W antidiag with boundary onditions on : µ| = i ◦η : W −→ K(I ,1) −i→a K(D ,1) Wantydiag⊂Nd a antidiag antidiag a 4 , W diag and with boundary onditions on : µ = i ◦p ◦η : W → K(I ,1) → K(I ,1) → K(I ,1). a b diag diag b d a Lemma 8 The mapping µ = η| ∪µ : Nn−2k = N ∪N → K(I ,1) a Nantidiag antidiag d a d determines a y li stru ture for . Proof of Lemma 8 ∗ We have to prove the equality in the De(cid:28)nition of the Cy li Stru ture for generi mappings. T : Γ → Γ Γ/T Γ Γ Let us onsider the fπree(Γi/nTvo)lution E and the quotient . 1 Γ DThe fundamental group c ∈ E,\dDenotce2d=bya2 , is a quadratic extension of 4 4 by means of an element , D .⊂ThEe element of order 4 is 4 ommutes with all elements in the subgroup . The following diagram is well-de(cid:28)ned: 9 N −→ RK −→ K ⊂ Γ d ↓ ↓ ↓ ↓ N /T −→ RK/T −→ K/T ⊂ G/T d ↓ ↓ ↓ ↓ K(I ,1) = K(I ,1) K(E,1) = K(E,1). a a RK → RK/T → K(I ,1) φ c ∈ a The omposition oin ides with (The π (RK/T)\π (RK) π (RK) ⊂ 1 1 1 π (RK/T) ommutes with all eleRmKen/tTs i→n thKe(sIu,b1g)roup 1 a of the index 2 andc the mapping is well-Ide(cid:28)ned. a The image of the element is the generator of the y li group .) The N → N /T → K o(mIp,o1s)ition of the left verti al arroµw:sNin→theKd(Iia,g1ra)m d (Ndn−2Nkd,µ ) a d a a oin ides with t(hNe′nm−a2kp,pµin′)g N′n−2k = .NT′nh−e2pka∪irN′n−2k is obordant to a pair a , where cycl d , and N′n−2k N′n−2k cycl is a losed manifold. The manifold (with boundary) cycl is the N′n−2k/T double overing over the orientedmaniHfold(w(Kith(Ib,o1u)n,dKa(rIy),1)c;yZcl) Nd.The n−2k a d base of the over µre′pr(e[sNen′nt−s2ak, ∂yN l′e])in∈ H (K(I ,1),K(I ,1)) . Therefore therelative y le a,∗ d n−2k a d istrivialand (cid:10)µ∗(t);[Nn−2k,∂Nn−2k−1](cid:11) = (cid:10)µ′∗(t);[N′n−2k](cid:11). a a cycl The last hara teristi number oin ides with (cid:10)κn−k;[N¯n−k](cid:11) = 1. Lemma 8 is proved. K A natural strati(cid:28) ation of the polyhedron . Proof of Lemma 7 J (S7/i) j = 1...,r J j Let be the join of lens spa es , . The spa e admits J(k ,...,k ) 1 s a natural strati(cid:28) ation de(cid:28)ned by the olle tion of subjoins 0 < k < ··· < k < r 1 s generated by lenses with numbers . The preimage p−1(J(k ,...,k )) ⊂ RPn−k, 1 s p : RPn−k → J R(k ,...,k ) 1 s , is denoted by . A point x ∈ R(k ,...,k ) ⊂ RPn−k 1 s is determined by the olle tion of oordinates (x ,...,x ,λ) k1 ks s x ∈ (up to the antipodal transformation of the (cid:28)rst oordinates), where kj S7 λ (s−1) j , and is a bary entri oordinate on the standard -simplex. 10