Scribd
Upload
160 views18 pages

Milnor Thom

Uploaded by

Son Hoang
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
160 views18 pages

Milnor Thom

Uploaded by

Son Hoang
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 18
On a Theorem of Milnor and Thom Nolan R. Wallach . 1. Introduction In [M1],[T], Milnor and Thom (independently) proved an estimate on the sum of the Betti numbers (relative to an arbitrary field of coefficients) of the set of zeros of polynomials of degree at most k > 0 in R”, essentially C,k” (Milnor’s estimate is k(2k — 1)"~!, Thom’s is essentially twice Milnor’s). In particular, this result gives a quantitative version of Whitney’s (earlier) theorem [W], that says that the number of connected components is finite. Most applications of the Theorem of Milnor and Thom are to the im- plied estimate on the number of connected components. For example, in [B] the estimate was used to determine lower bounds for the complexity of certain algebraic computation trees. One purpose of this article is to give a proof (following Milnor’s methods) of the estimate on the number of con- nected components that uses only advanced calculus, elementary topology and Sard’s theorem (the special cases of Sard’s theorem that are used will also be sketched in this article) that should be accessible to mathematicians and computer scientists who are not experts in algebraic topology. Another is that the proof of Lemma 1 in [M1], left quite a bit to the reader. The first two sections of this article are devoted to an an elementary proof of this lemma (see Theorem 3.4). We also give a less elemenatary proof in section 7 that gives a quantitative upper bound for the number of irreducible com- ponents of a variety over an algebraically closed field in terms of the degrees of a defining set of equations. This result may be of independent interest. Sections 7,8,9 involve more algebraic geometry and constitute whatever is new in this paper. In [M1], Milnor indicates that he has no examples where Cy # 1. This suggests the problem of proving (or disproving) the contention that we can take C,, = 1. In section 8 we prove that ifn = 2 the answer (for the number of connected components) is affirmative. In section 9 we give an affirmative answer for non-singular hypersurfaces for the sum of the Betti numbers. ‘This result gives a sharper upper bound for the sum of the Betti numbers of a set of the form R" — X where X is the zero set of a of polynomials Research partially supported by an NSF Summer Grant. with real coefficients. This article is an outgowth of lectures that the author gave on real algebraic geometry during a three quarter course in algebraic geometry at the University of California, San Diego. Beside graduate students in mathematics there were also regular participants from computer science and economics. We would like to thank R. Paturi for suggesting the Milnor- Thom theorem as a topic in the course and for his lectures on complexity related to [B]. Finally, this article is dedicated to the memory of my friend Joe D’Atri, His untimely death has left a void in the differential geometry community, His interests in and out of mathematics enriched all of our lives. 2. Generic finite varieties Let k denote an algebraically closed field and let V denote an n—dimensional vector space over k. We use the notation P'(V) for the space of polynomial functions on V that are homogeneous of degree r. We set Wrns,.mn = P™(V) x -- x P™(V). We look upon Wrny,....m, 88 an N = 3) (""*"=1) dimensional vector space over k. We fix m; > 0, i = 1,...,n and set W = Wrar,.ima» fg €W then we look upon g as both a polynomial map of k" to k” and an ordered set of polynomials. Set P(V) equal to the algebra of polynomials on V. If 9 €W then set I, equal to the ideal generated by the entries of g. We note that J, is a graded subspace of P(V). Thus Ry = P(V)/Ip inherits a natural grade. Set R} equal to the j-th homogeneous component and hg(t) = > jt? dim R} (thought of as a formal power series). We set AQ) = TJ +t+.. 42-4), a1 The following result is no doubt well known. Proposition 2.1. The set, %m,,...m,; of all_g € W such that hy(t) = h(t) is non-empty and Zariski open in W. ‘We will need some notation before we give our (elementary) proof. Let Z be an n-dimensional vector space over k and let 21,..., 2, be a basis of 2. We grade Z by setting deg(z;) = mj. Then Z = @Z? with Z? the span of the z; with deg(zi) = p. We grade P(V) ® Z by setting (P(V) @ Z) = DPI-+(V) @ Z. We define (9): P(V) ® Z — PIV) by A(g)(F ® x) = fi (here 9 = (91,...,9n)). Then O(g)(P(V) @ Z} Cc PI(V) and A(g)(P(V) ® Z)= Define h; € Z by Ait) = Oh yt. ij Set d= mi +...+ma—n. Set pj hy. It is easy to see that pj > 0. For each 1 < j 0 let &;;(g) be an enumeration of the p; x pj minors of the restriction of 8(g) to (P(V) ® Z)5. We set O° = {9 € Wlif pj > 0 there exists i sj4(g) # 0}. It is clear that Q° is a Zariski open subset. We also note that g = (27, ..., 2M") € 2%, So 2 is non-empty. We note that I, is a graded subspace of P(V). Thus Ry = P(V)/Ig inherits a natural grade. Set Fj equal to the j-th homogeneous component and h(t) = Dy, # dim Rj (thought of as a formal power series). Lemma 2.2. If g € 9° then hy =h. Proof. Fix g € 2°. Since hay1 = 0, O(g)(P(V) @ Z)**! = P(V) AH). Also, if 0 < j < d then dim Ri < dim P4(V) — pj = hj. Let U be a graded subspace of P(V) such that U @ J, = P(V). It is easy to see that dimUN P3(V) = dim Rj and that if us, ..., up is a basis of U then J kg, ..., gnlus = P(V). Let w1,...,wn be indeterminates and grade k{w1, ..., Wa] by setting deg w; = mj. Then we have a graded surjection, V, of k[wi,..., Wn] @U to PV) given by flwi,.,Wn] @Ur flgis-Gnlu. Let Y = ker. Then ¥ is graded and we have an identity of formal power series

You might also like