Polynomials Maps and Even Dimensional Spheres Francisco-Javier TURIEL Geometr'ia y Topolog'ia, Facultad de Ciencias Ap. 59, 29080 M'alaga, Spain email: turiel@agt.cie.uma.es Abstract. We construct, for every even dimensional sphere Sn, n 2, and ev* *ery odd integer k, a homogeneous polynomial map f : Sn ! Sn of Brouwer degree k and alg* *ebraic degree 2 | k | -1. ____________- A polynomial map from X Rm to Y Rr is the restriction to X of a polyno* *mial map F : Rm ! Rr such that F (X) Y . When each component of F is homogeneous of degree k, we will say that the polynomial map from X Rm to Y Rr is homogene* *ous of degree k. As usual Sn is the sphere on Rn+1 defined by the equation x21+ ... + * *x2n+1= 1, in short || x ||2= 1, whereas Snr, 1 r n, will be the differentiable manifold,* * diffeomorphic to Sn , defined by the equation (x21+ ... + x2r)2 + x2r+1+ ... + x2n+1= 1. In t* *his work we show: Theorem 1. Suppose n even and 2. Let k be an integer. Then: (a) If k is odd, there exists a homogeneous polynomial map from Sn to Sn of Bro* *uwer degree k and algebraic degree 2 | k | -1. (b) If k is even there exists, for each 2 2r n, a polynomial map from Sn2rt* *o Sn of Brouwer degree k. Representing elements of ssn(Sn) by polynomial maps is an old question [1] * *which was affirmatively solved by Wood, in 1968, provided that n is odd (theorem 1 of* * [3], see[4] as well for the complex sphere). Nevertheless, as far as I know, this pr* *oblem is still 1 open for n even; our theorem settles it when the Brouwer degree is odd. In bot* *h cases the polynomial maps constructed are homogeneous; therefore the problem of repre* *senting elements of ssn(Sn) by homogeneous polynomial maps is solved now, since only ze* *ro and the odd topological degrees may be represented in this way when n is even [2]. The proof of theorem 1 of [3] makes use of a natural polynomial map of topo* *logical degree 2 (lemmas 4 and 5). Nothing similar is known for n even; however the pol* *ynomial map x 2 R3 ! (x21- x22, 2x1x2, x3) 2 R3 send S22into S2 with topological degree* * 2. Part (b) of our theorem generalizes this map. For proving theorem 1 we start constructing a family of real polynomials in* * one vari- X` able. Let '` = ajtj be the Taylor expansion of ' = (1 - t)-1=2, at zero, up * *to order j=0 `; that is to say aj = (2j_-_1)(2j_-_3)_._.2.j1..j!Since the radius of converge* *nce of the power X1 series ajtj is 1 and each aj > 0 , we have a0 = 1 '` (1 - t)-1=2, t 2 [0* *, 1), whence j=0 (t - 1)'2`(t) + 1 0 and '`(t) 1 when t 0 (both inequalities are obvious i* *f t 1). 1X On the other hand if we set ' = t`+1R + '` then tj = (1 - t)-1 = '2 = t`* *+1R"+ '2` j=0 X` on (-1, 1); Therefore '2`= t`+1S + tj where S is a polynomial in t. It follo* *ws, from j=0 that, the existence of a polynomial ~` of degree ` such that (t - 1)'2`(t) + 1 * *= t`+1~`. Lemma 1. For every ` one has (t - 1)'2 + 1 = t`+1~` where ~` is a polynomia* *l of degree `. Moreover ~`(t) 0 and '(t) > 0 for each t 2 R if ` is even, and for * *any t 0 if ` is odd. Proof. It will suffice to show that ~`(t) 0 and '`(t) > 0 if ` 2 is eve* *n and t < 0. First we will prove, by induction on `, the existence of a ffi` > 0 such that '* *` is strictly decreasing on (-1, -1 + ffi`). Note that '` = a`t` + a`-1t`-1 + '`-2 = a((2` -* * 1)t` + 2`t`-1) + '`-2 where a > 0. By induction hypothesis or because '0 = 1, the polynomial '`-2 is decreasin* *g on 2 (-1, -1 + ffi`-2), or on R if ` = 2. But the derivative ((2` - 1)t` + 2`t`-1)0* * = ((2` - 1)`t`-1 + 2`(` - 1)t`-2) < 0 on (-1, -1], so '` is strictly decreasing on some * *interval (-1, -1 + ffi`). We show now that '`(t) > (1-t)-1=2 > 0 if t < 0. As (1-t)-1=2 is strictly i* *ncreasing, * *X1 it is enough to prove our result on (-1, 0). On this interval lim`!1 {'`(t)} = * * ajtj = * *j=0 X1 (1 - t)-1=2. But the series ajtj is alternating and the sequence {aj | t |j}* *j2N, whose j=0 limit is zero, strictly decreasing; then '`(t) > (1 - t)-1=2 > 0 for ` even. Finally, if '`(t) > (1 - t)-1=2 > 0 for any t < 0, a straightforward calcul* *ation shows that (t - 1)'2`(t) + 1 < 0, whence ~`(t) 0 since t`+1 < 0. Recall that any polynomial ~ which do not takes negative values has even de* *gree and can be write ~ = ~21+ ~22, where ~1 and ~2 are polynomials of degree half of * *degree of ~. Therefore by setting k = ` + 1, ff = '`, ~` = ~, fi1 = ~1 and fi2 = ~2 one h* *as: Corollary 1. For any odd natural number k there exist three polynomials ff,* * fi1, fi2, the first one of degree k - 1 and the other two with degree k_-_1_2, such tha* *t ff(t) > 0 and (1 - t)ff2(t) + tk(fi21(t) + fi22(t)) = 1 anywhere. Let us proof part (a) of theorem 1. Since topological degrees 1 may be rep* *resented by linear maps, we can assume k 1. On C x Rn-1 = Rn+1, endowed with coordinat* *es (z, y) = (z, y1, ..., yn-1) for which Sn = {(z, y); | z |2 +y21+ ... + y2n-1= 1* *}, we define F (z, y) = ((fi1(| z |2) + ifi2(| z |2))zk, ff(| z |2)y) where ff, fi1 and fi2 are as in corollary 1. Then F (Sn) Sn. Set S1 = {(z, 0); | z |2= 1} Sn. As ff(t) > 0 for each t 2 R, F -1(S1) = * *S1 and F preserves the orientation transversely to S1. Hence the maps F|S1and F|Sn have * *the same topological degree, that is to say k. By construction all the monomials of F have odd degree 2k-1. Multiplying * *everyone 3 of them by a suitable power of | z |2 +y21+ ... + y2n-1the map F becomes homoge* *neous of algebraic degree 2k - 1, whereas F|Sn do not change. For proving (b), first we set "~`(t) = ~`(t2) and "'`(t) = '`(t2). By lemma* * 1 we have (t2 - 1)'"2`(t) + 1 = t2`+2"~`(t), "'`(t) > 0 and "~`(t) 0 for any t 2 R. Thi* *s allows us to find out, for every natural number "k 1, three polynomials "ff, "fi1, "fi2such* * that "ff(t) > 0 and (1 - t2)"ff2(t) + t2"k(f"i21(t) + "fi22(t)) = 1 anywhere. Consider on Rn+1 = R2rx Rn-2r+1 coordinates (x, y) = (x1, ..., x2r, y1, ...* *, yn-2r+1). Let f : R2r ! R2r be a homogeneous polynomial map of algebraic degree 2"k, send* *ing S2r-1 into S2r-1 with topological degree k = 2"k, which always exists (see [3]* *) and J : R2r ! R2r the isomorphism given by Jx = (-x2, x1, ..., -x2r, x2r-1), that i* *s to say the canonical complex structure of R2r. One defines (if "k= 0 just consider a const* *ant map): F (z, y) = (f"i1(k x k2)f(x) + "fi2(k x k2)Jf(x), "ff(k x k2)y) Then F (Sn2r) Sn and the same argument as in part (a), applied to S2r-1 =* * {(x, 0); k x k2= 1} Sn2r, shows that the topological degree of F : Sn2r! Sn equals k. References 1. Baum, P.F.: Quadratics maps and stable homotopy groups of spheres, Ill* *inois J.Math. 11 (1967), 586-595. 2. Golasi'nski, M. and G'omez Ruiz. F.: Polynomial and regular maps into Gr* *assman- nians, K-Theory 26 (2002), 51-68. 3. Wood, R.: Polynomial maps from spheres to spheres, Invent. Math. 5 (* *1968), 163-168. 4. Wood, R.: Polynomial maps of affine quadrics, Bull. London Math. Soc* *. 25 (1993), 491-497. 4