Analytic Number Theory

Analytic Number Theory

A course in Analytic Number Theory Taught by Barry Mazur Spring 2012 Last updated: January 19, 2014 1 Contents 1. January 244 2. January 268 3. January 31 11 4. February 2 15 5. February 7 18 6. February 9 22 7. February 14 26 8. February 16 30 9. February 21 33 10. February 23 38 11. March 1 42 12. March 6 45 13. March 8 49 14. March 20 52 15. March 22 56 16. March 27 60 17. March 29 63 18. April 3 65 19. April 5 68 20. April 10 71 21. April 12 74 22. April 17 77 Math 229 Barry Mazur Lecture 1 1. January 24 1.1. Arithmetic functions and first example. An arithmetic function is a func- tion N ! C, which we will denote n 7! c(n). For example, let rk;d(n) be the number of ways n can be expressed as a sum of d kth powers. Waring's problem asks what this looks like asymptotically. For example, we know r3;2(1729) = 2 (Ramanujan). Our aim is to derive statistics for (interesting) arithmetic functions via a study of the analytic properties of generating functions that package them. Here is a generating function: 1 X n Fc(q) = c(n)q n=1 (You can think of this as a Fourier series, where q = e2πiz.) You can retrieve c(n) as a Cauchy residue: 1 I F (q) dq c(n) = c · 2πi qn q This is the Hardy-Littlewood method. We can understand Waring's problem by defining 1 X nk fk(q) = q n=1 (the generating function for the sequence that tells you if n is a perfect kth power). Then d fk (q) is the generating function of the thing we want: d X nk d X n q = fk = rk;d(n)q Take the case when k = 1; you get a geometric series. If k = 2, then Y 2m 2m−1 2m+1 f2(q) = (1 − q )(1 + q )(1 − q ) by a theorem of Jacobi. Another example is the partition function: p(n) is the number of ways that n can be expressed as a sum of positive integers: 1 1 X Y 1 p(n)qn = 1 − qm n=1 m=1 A third example is Y X !(q) = q · (1 − qm)2(1 − q11m)2 = a(n)qn 2 3 2 If n 6= 11 is a prime, then a(p) = 1 + p − jE(Fp)j where E = y + y = x − x (here jE(Fp)j 2πiz is the number of points on the curve with values in Fp). When you replace q = e −1 there's a symmetry where you can replace z 7! 11z . 4 Math 229 Barry Mazur Lecture 1 1.2. Formal Dirichlet Series. 1 X c(n) D (s) = c ns n=1 Suppose we have two arithmetic functions n 7! a(n) and n 7! b(n). Define a product a ∗ b = c where X n c(n) = a(d)b d djn (Divisors are assumed always to be positive.) Why do we care? If you have Dirichlet series P a(n) P b(n) Da(s) = ns and Db(s) = ns then X a ∗ b(n) D = = D (s)D (s) a∗b ns a b The star product is commutative, associative, and has the identity " where ( 1 n = 1 "(n) = 0 else This is not the identity function 1 : n 7! 1. The identity function is invertible, and the inverse is the Moebius function ( 0 if n is not square-free µ(n) = (−1)# of prime factors of n else Exercise: check that X µ(d) · 1 = " djn so that X µ(n) · ζ(s) = 1 : ns Definition 1.1. An arithmetic function is called multiplicative if c(u · v) = c(u) · c(v) if (u; v) = 1. So v Y ki c(n) = c(pi ) i=1 Qv ki if n = i=1 pi . Suppose n 7! c(n) is multiplicative, and let p be prime. Let 1 X c(pm) D (s) = c;p pms m=0 X c(n) D (s) = c ns Formally Y (1.1) Dc(s) = Dc;p(s) p 5 Math 229 Barry Mazur Lecture 1 BUT, even if Dc(s) and Dp(s) converge, you may not have the formula, because the P1 −s product may not converge. If n=1 jc(n)n j converges, then everything works. 1 X X n−iy X e−log(n)iy c(n) · n−s = c(n) = c(n) nx nx n=1 If it converges at a point, then it converges at the entire vertical line through that point. 1.3. Riemann ζ function. Using the arithmetic function n 7! 1, then define X 1 ζ(s) = D (s) = 1 ns This converges absolutely if Re(s) > 1, by calculus. There is an infinite product expansion 1 X 1 X 1 Y X = = ( p−ms) ns 1 − 1 p prime ps p prime m=0 (This is in the HW.) Theorem 1.2. X µ(n) ζ(s)−1 = = D (s) ns µ Proof. Dµ(s)ζ(s) = Dµ · D1 = D" = 1 This is analytically true when Re(s) > 1. Basic reason for interest in Dc(s) for any c : n 7! c(n): If we know enough, we will P be able to give very good estimates for p≤X c(p) := πc(X). In particular, when c = 1, we get estimates for π(X), the number of primes ≤ X. P 1 Theorem 1.3 (Euler). p≤X p diverges as X ! 1. Q 1 −1 Q P1 −m Proof. Define λ(X) = p≤X (1 − p ) . Write this as p≤X m=0 p . But X 1 ≤ λ(X) n n≤X 6 Math 229 Barry Mazur Lecture 1 1 P 1 because each n is a term in the expansion of λ. But n diverges (which immediately proves that there are infinitely many primes). Take the log of the infinite product: X 1 log λ(X) = − log 1 − p p≤X But 1 1 1 − log 1 − 1 = + + + ··· p p 2p2 3p3 1 1 1 ≤ + + + ··· p p2 p3 1 1 1 = + 2 · 1 p p 1 − p | P {z } n2 ∗ convergesN for some c constant. So we have X 1 lim log λ(X) = + something convergent X!1 p We showed above that λ(X) goes to 1, which means its log does too; if the LHS diverges, then so does the RHS. 1.4. Arithmetic functions related to the ζ-function. a will be an arithmetic P P n function, and let A(n) = djn a(d). Equivalently, a = µ ∗ A so a(n) = µ(d)A d . In terms of Dirichlet series, DA(s) = ζ(s) · D(a) Define A as the series where X k n 7! σk(n) = d djn and a : n 7! nk. Then X σk(n) D (s) = = ζ(s)D (s) = ζ(s) · ζ(s − k) A ns a X nk X 1 D (s) = = = ζ(s − k) a ns ns−k If k = 0 then define σ0(n) = ν(n) which counts the number of divisors of n. Then X ν(n) = ζ(s)2 ns P We call a = Φ, the Euler Phi function; then A(n) = djn Φ(d); you can check that A(n) = n. ζ(s − 1) D (s) = Φ ζ(s) th Recall rk;d(n), the number of ways that n is expressible as a sum of d k powers, and recall that when k = 2 we had a nice formula for this. Let R2;d(n) be the number of ways 7 Math 229 Barry Mazur Lecture 2 that n is expressible as the sum of d squares, counting (+α)2 and (−α)2 as different, and including 0. Also, order matters here! For example, R2;24(2) = 1104 Slightly more generally, 16 R2;24(n) = σ11(n) + e(n) 691 p where e(n) is the error term: up to taking logs, it is < other term. We have a complete description of the error term. 2. January 26 Define 1 Y ∆ = q (1 − qn)24 n=1 and recall X n2 f2 = q : It turns out that ∆ ≡ f2 (mod 2) : RECALL, a Dirichlet series is a formal expression that looks like D(s) = P a(n)n−s. We also had arithmetic functions n 7! a(n) 2 C. L-functions will be a special case. 2.1. Some properties to learn about. • Domain of definition. For example, ζ(s) doesn't make sense as a power series to the left of the line Re(s) = 1. However, there is a meromorphic continuation, with pole only at s = 1. (We have good control to the right of the line, but the most interesting information comes from the other half.) Also, there are symmetries: usually a relationship between L(s) and L(1 − s) (i.e. you can sort of reflect across the line). P • Poles and Zeroes. One way (the only way) of understanding p≤X a(p) is to under- P stand the poles and zeroes. In the case of ζ(s), this gives information about p≤X 1 = π(X), the number of primes ≤ X. The Riemann hypothesis says that the zeroes of the 1 ζ-function are at the negative even integers, on the line Re(s) = 2 , and nowhere else. If you're interested in statistics, consider P w(p)a(p) (i.e. a weighted sum). A good thing P to study is p≤X log(p).

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    80 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us