Advanced Mathematics

Analytic Number Theory: Primes, Zeta Functions, and L-Functions

Analytic number theory uses the tools of complex analysis and Fourier analysis to study the distribution of prime numbers and the deep arithmetic structure of the integers. From the Riemann zeta function to sieve methods, this subject sits at the crossroads of analysis and algebra.

Learning Objectives

After studying this guide you will be able to:

  • ✓Define and compute arithmetic functions including the Euler totient, Mobius function, and von Mangoldt function, and apply Mobius inversion.
  • ✓Work with Dirichlet series, identify abscissas of convergence, and derive Euler product representations.
  • ✓State the analytic continuation and functional equation of the Riemann zeta function and locate its trivial and non-trivial zeros.
  • ✓State and sketch the proof of the prime number theorem via contour integration, including the role of zero-free regions.
  • ✓Define Dirichlet characters and L-functions, and outline the proof of Dirichlet's theorem on primes in arithmetic progressions.
  • ✓Apply the Legendre sieve, Brun sieve, and Selberg sieve to bounding prime-counting problems.
  • ✓Compute Gauss sums and Ramanujan sums and understand their role in the circle method.
  • ✓Define modular forms, Hecke operators, and their associated L-functions, and explain their connection to elliptic curves.

1. Arithmetic Functions

An arithmetic function is any function f from the positive integers to the complex numbers. The subject is rich with functions that encode prime factorization information and interact beautifully through Dirichlet convolution.

The Euler Totient Function

The Euler totient phi(n) counts the positive integers up to n that are coprime to n. If n = p1^a1 times p2^a2 times ... times pk^ak is the prime factorization, then:

phi(n) = n times product over primes p dividing n of (1 - 1/p)

  • phi(1) = 1
  • phi(p) = p - 1 for prime p
  • phi(p^k) = p^(k-1) times (p - 1)
  • phi is multiplicative: if gcd(m,n) = 1, phi(mn) = phi(m) times phi(n)
  • Sum over d dividing n of phi(d) = n

The last identity — that the sum of phi(d) over divisors d of n equals n — is a fundamental identity provable by partitioning the integers from 1 to n by their gcd with n. It reflects how the totient function is the Dirichlet inverse of the identity function.

The Mobius Function

The Mobius function mu(n) is defined by:

  • mu(1) = 1
  • mu(n) = (-1)^k if n is a product of k distinct primes
  • mu(n) = 0 if p^2 divides n for some prime p

The Mobius function is the key to Mobius inversion: if g(n) equals the sum over divisors d of n of f(d), then f(n) equals the sum over divisors d of n of mu(n/d) times g(d). This is the number-theoretic analog of the Fundamental Theorem of Calculus.

Key Fact

The sum of mu(d) over all divisors d of n equals 1 if n = 1, and equals 0 if n is greater than 1. This is because the Dirichlet series for mu(n) equals 1/zeta(s), so the Dirichlet convolution of mu with the constant function 1 is the indicator of n = 1.

The von Mangoldt Function

The von Mangoldt function Lambda(n) is defined to be ln(p) when n is a prime power p^k for some prime p and positive integer k, and 0 otherwise. It plays a central role in the explicit formula connecting zeta zeros to prime counts:

psi(x) = sum over n at most x of Lambda(n)

psi(x) = x - sum over zeros rho of x^rho / rho - ln(2 pi) - (1/2) ln(1 - x^(-2))

The Chebyshev psi function counts prime powers weighted by ln(p). The prime number theorem is equivalent to psi(x) ~ x.

Multiplicative Functions and Dirichlet Convolution

A function f is multiplicative if f(mn) = f(m)f(n) whenever gcd(m,n) = 1, and completely multiplicative if this holds for all m, n. The Dirichlet convolution of f and g is defined as:

(f * g)(n) = sum over d dividing n of f(d) times g(n/d)

Dirichlet convolution makes the set of arithmetic functions into a commutative ring. The multiplicative identity is the function epsilon where epsilon(1) = 1 and epsilon(n) = 0 for n greater than 1. The Mobius function is the Dirichlet inverse of the constant function 1. Important examples of multiplicative functions include phi, mu, sigma (sum of divisors), tau (number of divisors), and the Liouville function lambda.

2. Dirichlet Series and Euler Products

A Dirichlet series is a formal series of the form sum from n=1 to infinity of a(n)/n^s, where s is a complex variable. These series are the generating functions of arithmetic functions in analytic number theory, playing the same role that power series play in ordinary analysis.

Abscissa of Convergence

Every Dirichlet series has an abscissa of convergence sigma_c such that the series converges absolutely for Re(s) greater than sigma_c and diverges for Re(s) less than sigma_c. There is also an abscissa of absolute convergence sigma_a which may be larger. For series with non-negative coefficients, sigma_c = sigma_a. The half-plane of convergence is always a half-plane, unlike power series which converge in disks.

Key Dirichlet Series

  • zeta(s) = sum 1/n^s = product 1/(1 - p^(-s)) [Re(s) > 1]
  • 1/zeta(s) = sum mu(n)/n^s [Re(s) > 1]
  • zeta(s)^2 = sum tau(n)/n^s [Re(s) > 1]
  • zeta(s-1)/zeta(s) = sum phi(n)/n^s [Re(s) > 2]
  • -zeta'(s)/zeta(s) = sum Lambda(n)/n^s [Re(s) > 1]

Euler Products

When f is a multiplicative arithmetic function and the Dirichlet series for f converges absolutely, we obtain an Euler product expansion:

sum f(n)/n^s = product over primes p of (sum over k at least 0 of f(p^k)/p^(ks))

For completely multiplicative f (such as Dirichlet characters), the local factor at p simplifies to 1/(1 - f(p)/p^s). The Euler product is the bridge between Dirichlet series and the primes. The unique factorization of integers into primes is precisely what makes the product formula work: every positive integer n appears exactly once when the local factors are multiplied out.

Perron's Formula

Perron's formula recovers partial sums of an arithmetic function from a Dirichlet series via a Mellin-type integral:

sum over n at most x of a(n) = (1/2 pi i) integral (c-i inf to c+i inf) of F(s) x^s / s ds

where F(s) = sum a(n)/n^s and c is in the half-plane of absolute convergence.

Moving the contour of integration to the left, past poles and zeros of F(s), yields an asymptotic formula for the partial sums. This is the fundamental idea behind the analytic proof of the prime number theorem.

3. The Riemann Zeta Function

The Riemann zeta function is the most important single function in analytic number theory. Riemann's 1859 paper "On the Number of Primes Less Than a Given Magnitude" introduced the study of zeta(s) as a complex function and laid the groundwork for a century of work.

Definition and Basic Properties

For Re(s) greater than 1:

zeta(s) = sum from n=1 to infinity of 1/n^s = product over primes p of 1/(1 - p^(-s))

The Euler product shows that zeta(s) is nonzero for Re(s) greater than 1. The function has a simple pole at s = 1 with residue 1. Near s = 1 we have the Laurent expansion: zeta(s) = 1/(s-1) + gamma + O(s-1), where gamma is the Euler-Mascheroni constant (approximately 0.5772).

Analytic Continuation

The zeta function extends to a meromorphic function on all of C. One approach uses the integral representation via the Gamma function. Another uses the formula:

zeta(s) = 1/(s-1) + sum from n=0 to N of (-1)^n gamma_n s^n / n! + O(s^(N+1))

A more elegant continuation uses the alternating zeta function (Dirichlet eta function):

eta(s) = sum (-1)^(n+1) / n^s = (1 - 2^(1-s)) zeta(s)

The eta series converges for Re(s) greater than 0, which extends zeta(s) to Re(s) greater than 0 away from s = 1. Repeated use of functional equations extends it to all of C.

The Functional Equation

Riemann proved that the completed zeta function xi(s) satisfies a perfect symmetry:

xi(s) = (1/2) s(s-1) pi^(-s/2) Gamma(s/2) zeta(s)

xi(s) = xi(1-s)

This functional equation, equivalent to the Poisson summation formula applied to the theta function, shows that zeta is symmetric about the critical line Re(s) = 1/2. It also locates the trivial zeros: since Gamma(s/2) has poles at s = 0, -2, -4, ... the function zeta(s) must vanish at these negative even integers to cancel the poles.

Trivial and Non-Trivial Zeros

Trivial Zeros

Located at s = -2, -4, -6, ... (negative even integers). These arise from the poles of Gamma(s/2) in the functional equation. They are well understood and present no mystery.

Non-Trivial Zeros

Located in the critical strip 0 less than Re(s) less than 1. The functional equation implies they are symmetric about both the real axis and the critical line Re(s) = 1/2. The Riemann Hypothesis conjectures they all lie on Re(s) = 1/2.

The Riemann Hypothesis

All non-trivial zeros of the Riemann zeta function have real part equal to 1/2. First stated by Riemann in 1859, it remains one of the Millennium Prize Problems. Computationally, the first 10 trillion zeros have been verified to lie on the critical line, but no proof exists. The truth of RH would give the sharpest possible error term in the prime number theorem: pi(x) = Li(x) + O(sqrt(x) ln(x)).

4. The Prime Number Theorem

The prime number theorem (PNT) is the central result of analytic number theory. It describes the asymptotic distribution of prime numbers — how they thin out as numbers grow large.

Statement

pi(x) ~ x / ln(x) as x approaches infinity

Equivalently: pi(x) / (x / ln(x)) approaches 1

Even better: pi(x) ~ Li(x) = integral from 2 to x of 1/ln(t) dt

Li(x) is the logarithmic integral. The error |pi(x) - Li(x)| is much smaller than x/ln(x).

Gauss conjectured this density around 1793 based on tables of primes. Legendre made a related conjecture. The theorem was finally proved in 1896 by Hadamard and de la Vallee Poussin independently, using properties of the Riemann zeta function.

Chebyshev's Preliminary Results

Before 1896, Chebyshev proved weaker results using elementary methods. He introduced the functions:

  • theta(x) = sum over primes p at most x of ln(p)
  • psi(x) = sum over prime powers p^k at most x of ln(p) = sum Lambda(n) for n at most x

Chebyshev showed that 0.92 x is less than pi(x) ln(x) is less than 1.11 x for large x, proving the PNT is correct in order of magnitude. He also proved Bertrand's postulate: for every n at least 1, there exists a prime between n and 2n.

Proof Sketch via Contour Integration

The analytic proof proceeds in several steps:

Step 1: Link psi(x) to zeta via Perron's formula

Since -zeta'(s)/zeta(s) = sum Lambda(n)/n^s, Perron's formula gives psi(x) as a contour integral of (-zeta'(s)/zeta(s)) times x^s / s.

Step 2: Identify the main term from the pole at s = 1

The simple pole of -zeta'(s)/zeta(s) at s = 1 contributes a residue of x (the x^s/s factor at s = 1). This is the main term in psi(x) ~ x.

Step 3: Show zeta has no zeros on Re(s) = 1

The key estimate uses the identity 3 + 4cos(theta) + cos(2 theta) at least 0 applied to zeta^3(sigma) times |zeta(sigma + it)|^4 times |zeta(sigma + 2it)| at least 1. If zeta vanished at 1 + it0, the fourth power zero would cancel the triple pole, forcing the product below 1 — a contradiction.

Step 4: Move the contour and bound the error

With zeta nonzero on Re(s) = 1, we can shift the contour slightly left. The integrand is bounded on the new contour, giving psi(x) = x + o(x), which is equivalent to the prime number theorem.

Error Terms and Zero-Free Regions

Sharper zero-free regions yield better error terms. The classical zero-free region is: zeta(s) is nonzero in the region sigma greater than 1 - c/ln(|t| + 2) for an absolute constant c. This gives:

pi(x) = Li(x) + O(x exp(-c sqrt(ln x)))

Under RH: pi(x) = Li(x) + O(sqrt(x) ln(x)). Under GRH: similar improvements hold for primes in arithmetic progressions.

5. Dirichlet Characters and L-Functions

To study primes in arithmetic progressions, Dirichlet generalized the zeta function to a family of L-functions built from multiplicative group characters modulo q.

Dirichlet Characters

A Dirichlet character modulo q is a completely multiplicative function chi from the integers to the complex numbers such that:

  • (i) chi(n + q) = chi(n) for all n (periodicity modulo q)
  • (ii) chi(n) = 0 if gcd(n, q) is greater than 1
  • (iii) chi(mn) = chi(m) chi(n) for all m, n (completely multiplicative)
  • (iv) chi(1) = 1

There are exactly phi(q) characters modulo q, forming a group under pointwise multiplication. The principal character chi_0 is defined by chi_0(n) = 1 when gcd(n,q) = 1, and chi_0(n) = 0 otherwise. Characters satisfy the orthogonality relations:

(1/phi(q)) sum over chi mod q of chi(m) chi_bar(n) = 1 if m = n mod q (with gcd(n,q)=1), else 0

(1/phi(q)) sum over a mod q with gcd(a,q)=1 of chi(a) chi'_bar(a) = 1 if chi = chi', else 0

Dirichlet L-Functions

The Dirichlet L-function associated to a character chi modulo q is:

L(s, chi) = sum from n=1 to infinity of chi(n)/n^s = product over primes p of 1/(1 - chi(p)/p^s)

For Re(s) greater than 1. For chi the principal character, L(s, chi_0) = zeta(s) times product over p dividing q of (1 - p^(-s)).

L-functions extend to entire functions for non-principal chi, and satisfy functional equations analogous to that of zeta(s). They have an analytic continuation and non-trivial zeros in the critical strip. The Generalized Riemann Hypothesis (GRH) conjectures all non-trivial zeros of all L(s, chi) lie on Re(s) = 1/2.

Non-Vanishing at s = 1 and Dirichlet's Theorem

The critical step in proving Dirichlet's theorem is showing L(1, chi) is nonzero for all non-principal characters chi. For complex characters this is relatively straightforward from the product formula. For real (quadratic) characters, Dirichlet proved a class number formula:

L(1, chi) = 2 pi h(D) / (w sqrt(|D|)) for fundamental discriminant D

where h(D) is the class number of Q(sqrt(D)) and w is the number of roots of unity. Since h(D) at least 1, L(1, chi) is nonzero.

With L(1, chi) nonzero for all non-principal chi, and using the orthogonality of characters to isolate the arithmetic progression a mod q, one deduces:

sum over primes p = a mod q, p at most x, of 1 ~ x / (phi(q) ln(x))

Primes are asymptotically equidistributed among all phi(q) reduced residue classes modulo q. This is Dirichlet's theorem with the quantitative form given by the prime number theorem for arithmetic progressions.

6. Sieve Methods

Sieve methods are combinatorial and analytic techniques for counting integers in a set that avoid divisibility by primes from a specified set P. They are indispensable when analytic methods alone are insufficient — particularly for questions like twin primes where the zeta function approach breaks down.

The Sieve of Eratosthenes

The oldest sieve. To find all primes up to x: start with all integers from 2 to x. For each prime p up to sqrt(x), cross out all multiples of p. The remaining numbers are prime.

As an analytic counting tool, the sieve of Eratosthenes gives: pi(x) - pi(sqrt(x)) + 1 = number of integers up to x not divisible by any prime up to sqrt(x). By inclusion-exclusion this leads to the Legendre sieve.

The Legendre Sieve

Let S(A, P, z) be the count of elements of a finite set A not divisible by any prime p less than z in P. By inclusion-exclusion:

S(A, P, z) = sum over d squarefree, p|d implies p<z, of mu(d) |A_d|

where A_d is the set of elements of A divisible by d.

The problem with the Legendre sieve is the error term: if |A_d| = |A|/d + r_d, the error sum over d of |r_d| can be as large as 2^z, which overwhelms the main term unless z is very small. This is the "parity problem" — pure inclusion-exclusion cannot distinguish numbers with an even versus odd number of prime factors.

Brun's Theorem and Brun's Constant

Viggo Brun (1919) introduced upper and lower bound sieves by truncating the inclusion-exclusion sum. He proved:

The number of twin prime pairs (p, p+2) up to x is O(x (ln ln x)^2 / (ln x)^2)

A striking consequence: the sum 1/3 + 1/5 + 1/5 + 1/7 + 1/11 + 1/13 + ... over all twin prime pairs converges. This is Brun's constant B2 (approximately 1.9021...). Compare this to the divergence of the sum of reciprocals of all primes (by Euler), which is how we know there are infinitely many primes. The convergence of Brun's constant is consistent with both infinitely many and finitely many twin primes — the sieve cannot resolve the question.

The Selberg Sieve

Atle Selberg (1947) introduced a more powerful upper bound sieve by choosing sieve weights optimally. Instead of mu(d), he uses weights lambda_d chosen to minimize the upper bound:

S(A, P, z) at most sum over a in A of (sum over d | a, d < z of lambda_d)^2

The optimal lambda_d are chosen by a quadratic optimization, yielding the Selberg upper bound sieve.

The Selberg sieve gives bounds of the form: the count of primes in an interval [x, x+y] is at most C times y/ln(y) for an explicit constant C. Combined with the lower bound from the PNT, this shows the sieve gives the right order of magnitude. The Selberg sieve was used by Goldfeld, Bombieri, Iwaniec and others to prove spectacular results, including that there are infinitely many integers n for which n(n+2) has at most a bounded number of prime factors.

7. Exponential Sums and the Circle Method

Exponential sums are sums of the form sum of e^(2 pi i f(n)) where f is a real-valued function. They appear throughout analytic number theory as the Fourier side of counting problems. The circle method uses exponential sums to derive asymptotic formulas for additive number theory problems.

Gauss Sums

The Gauss sum associated to a Dirichlet character chi modulo q is:

tau(chi) = sum from a=1 to q of chi(a) e^(2 pi i a / q)

For a primitive character chi modulo q:

  • |tau(chi)|^2 = q (the modulus squared equals q)
  • tau(chi_bar) = chi(-1) tau(chi)_bar
  • chi(n) = (1/tau(chi_bar)) sum tau(chi, a) e^(-2 pi i an/q)

The last identity expresses chi as a Fourier transform and is fundamental for proving the functional equation of L(s, chi).

Ramanujan Sums

The Ramanujan sum c_q(n) is defined as the sum over a coprime to q of e^(2 pi i an/q):

c_q(n) = sum over a=1 to q with gcd(a,q)=1 of e^(2 pi i an/q) = mu(q/gcd(n,q)) phi(q) / phi(q/gcd(n,q))

Ramanujan sums are integers, multiplicative in q for fixed n, and appear in many expansions of arithmetic functions. For example, Lambda(n) = -sum over q at least 1 of (mu(q)/phi(q)) c_q(n) (Ramanujan's explicit formula for the von Mangoldt function).

Weyl Differencing

Weyl differencing is a technique for bounding exponential sums sum of e^(2 pi i f(n)) when f is a polynomial. The key idea: if S = sum e^(2 pi i f(n)), then |S|^2 = sum sum e^(2 pi i (f(n+h) - f(n))). The inner sum is a new exponential sum with a polynomial of one lower degree. Iterating this process reduces the problem to linear exponential sums, which are geometric series.

Weyl's Inequality

For f(n) = alpha n^k with alpha = a/q + theta, |theta| at most 1/q^2, gcd(a,q) = 1, and N at most q at most N^k, the Weyl sum satisfies:

|sum e^(2 pi i f(n))| is much less than N times (1/q + 1/N + q/N^k)^(1/2^(k-1))

The Hardy-Ramanujan-Littlewood Circle Method

The circle method gives asymptotic formulas for additive problems. Consider Waring's problem: let R_k(n) be the number of ways to write n as a sum of s k-th powers. Define the generating function:

f(alpha) = sum from m=1 to N of e^(2 pi i m^k alpha)

R_k,s(n) = integral from 0 to 1 of f(alpha)^s e^(-2 pi i n alpha) d alpha

The circle [0, 1] is divided into major arcs M (small intervals around rationals a/q with q small) and minor arcs m (the rest). On major arcs, f(alpha) has a good approximation involving Gauss sums, giving the singular series. On minor arcs, Weyl's inequality shows f(alpha) is small, and their contribution is an error term.

Goldbach's Problem via the Circle Method

Vinogradov (1937) proved that every sufficiently large odd integer is a sum of three primes, using the circle method with Weyl sums over primes. The generating function for primes involves the von Mangoldt function: f(alpha) = sum Lambda(n) e^(2 pi i n alpha). On major arcs near a/q, this is well-approximated by a linear combination of Ramanujan sums; on minor arcs, the exponential sum is small by Vinogradov's mean value theorem.

8. Distribution of Primes

Beyond the prime number theorem, analytic number theory studies the fine-scale distribution of primes: how large gaps can be, how evenly they are distributed in short intervals, and what patterns exist between consecutive primes.

Chebyshev's Estimates

Chebyshev proved (circa 1850):

  • theta(x) = sum over primes p at most x of ln(p) is asymptotically between 0.92x and 1.11x
  • pi(x) is between (A x / ln x) and (B x / ln x) for constants A < 1 < B
  • Bertrand's postulate: for all n at least 1, there exists a prime p with n < p at most 2n

Bertrand's postulate (also called Bertrand-Chebyshev theorem) can be proved elementarily using Chebyshev's estimates on binomial coefficients. The central binomial coefficient C(2n, n) is at least 4^n/(2n+1), and the prime factors of C(2n, n) in the range (n, 2n] account for most of its size, forcing at least one such prime to exist.

Prime Gaps

Let g_n = p_(n+1) - p_n be the n-th prime gap. Average prime gaps grow like ln(p_n) by the PNT. But individual gaps can be much larger or smaller:

Large Gaps

The sequence n! + 2, n! + 3, ..., n! + n gives a gap of n - 1 consecutive composites, so g_n can be arbitrarily large. Cramer's conjecture: lim sup g_n / (ln p_n)^2 = 1. The best proved bound is g_n = O(p_n^0.525) (Baker, Harman, Pintz 2001).

Small Gaps

Twin prime conjecture: g_n = 2 infinitely often. Zhang (2013) proved lim inf g_n is less than 70 million. The Polymath project reduced this to 246. Under Elliott-Halberstam conjecture, Maynard showed lim inf g_n at most 12.

Cramer's Conjecture and Probabilistic Models

Cramer modeled primes as random: integer n is prime with probability 1/ln(n), independently. In this model, the maximum gap up to x is approximately (ln x)^2. Cramer's conjecture is that actual prime gaps satisfy lim sup (p_(n+1) - p_n)/(ln p_n)^2 = 1. Granville has suggested the constant might be 2e^(-gamma) (approximately 1.123) rather than 1, based on more careful probabilistic modeling that accounts for divisibility by small primes.

9. Modular Forms

Modular forms are the crown jewel of the Langlands program. They are holomorphic functions on the upper half-plane with extraordinary symmetry, and their Fourier coefficients encode deep arithmetic data. The proof of Fermat's Last Theorem by Wiles (1995) is fundamentally a theorem about modular forms.

Definition

Let H denote the upper half-plane. The modular group SL_2(Z) acts on H by Mobius transformations: the matrix (a b; c d) sends z to (az + b)/(cz + d). A modular form of weight k is a holomorphic function f: H to C satisfying:

f((az+b)/(cz+d)) = (cz+d)^k f(z) for all (a b; c d) in SL_2(Z)

Plus the condition that f is holomorphic at the cusp at infinity, meaning f has a Fourier expansion f(z) = sum from n=0 to infinity of a(n) q^n where q = e^(2 pi i z).

A cusp form vanishes at the cusp: a(0) = 0. The space M_k of modular forms of weight k is a finite-dimensional vector space. For even k at least 4, dim M_k = floor(k/12) + (0 or 1 depending on k mod 12). Important examples:

  • Eisenstein series E_k: The Fourier coefficients of E_k involve divisor sums sigma_(k-1)(n). For example, E_4(z) = 1 + 240 sum sigma_3(n) q^n.
  • Ramanujan delta function: Delta(z) = q product from n=1 to inf of (1 - q^n)^24 = sum tau(n) q^n, a weight 12 cusp form. Ramanujan conjectured |tau(p)| at most 2 p^(11/2), proved by Deligne (1974) via the Weil conjectures.
  • Theta series: sum over n in Z^k of q^(n_1^2 + ... + n_k^2) is a modular form whose coefficients count representations as sums of k squares.

Hecke Operators

Hecke operators T_n are linear operators on spaces of modular forms that commute with each other. An eigenform of all Hecke operators simultaneously is called a Hecke eigenform, and its Fourier coefficients are multiplicative — if f is a Hecke eigenform with a(1) = 1 (normalized), then:

  • a(mn) = a(m) a(n) if gcd(m,n) = 1
  • a(p^(k+1)) = a(p) a(p^k) - p^(k-1) a(p^(k-1)) for prime p

This is a Hecke recursion; it implies the Euler product for the L-function of f.

L-Functions of Modular Forms and Connection to Elliptic Curves

The L-function of a weight k Hecke eigenform f is:

L(s, f) = sum a(n)/n^s = product over primes p of 1/((1 - a(p)/p^s)(1 - chi(p) p^(k-1-s)))

This L-function satisfies a functional equation and has analytic continuation. The Modularity Theorem (formerly Taniyama-Shimura-Weil conjecture, proved by Wiles, Taylor-Wiles, Breuil-Conrad-Diamond-Taylor) states: every elliptic curve over Q is modular — its L-function equals L(s, f) for a weight-2 cusp form f. This was the key to proving Fermat's Last Theorem.

The Langlands Program

The Modularity Theorem is one instance of the vast Langlands program, which conjectures deep correspondences between automorphic forms (generalizations of modular forms), Galois representations, and L-functions. Fermat's Last Theorem, Sato-Tate conjecture (proved 2011), and the recent proof of the geometric Langlands conjecture (2024) are milestones in this program.

10. Applications of Analytic Number Theory

Analytic number theory, despite its abstract nature, has concrete applications in cryptography, physics, and the foundations of mathematics. Here we survey the most important connections.

RSA Cryptography

The RSA public-key cryptosystem is based directly on the difficulty of factoring large integers. The setup:

Key generation

Choose large primes p and q. Set n = pq. Compute phi(n) = (p-1)(q-1). Choose e coprime to phi(n) and compute d = e^(-1) mod phi(n). Public key: (n, e). Private key: d.

Encryption and decryption

Encrypt message m: c = m^e mod n. Decrypt: m = c^d mod n. Correctness follows from Euler's theorem: m^(phi(n)) is congruent to 1 mod n when gcd(m, n) = 1, so c^d = m^(ed) = m^(1 + k phi(n)) is congruent to m mod n.

The security of RSA relies on the integer factorization problem being hard. Analytic number theory is used both to generate cryptographically large primes efficiently (using probabilistic primality tests based on Fermat's little theorem and Miller-Rabin) and to analyze the security of factoring algorithms like the number field sieve, whose complexity involves smooth numbers counted by Dickman's rho function.

Equidistribution and Weyl's Theorem

Weyl's equidistribution theorem (1916) states that the sequence (n alpha) mod 1 is equidistributed in [0,1] for every irrational alpha. This means: for any subinterval [a,b],

(1/N) times count of n at most N with (n alpha) mod 1 in [a,b] approaches b - a

The proof uses Weyl's criterion: a sequence is equidistributed if and only if for every nonzero integer h, (1/N) sum e^(2 pi i h n alpha) approaches 0. This reduces to bounding a geometric series and using the irrationality of alpha. Equidistribution results are also proved for primes: the sequence (p_n alpha) mod 1 is equidistributed for irrational alpha (Vinogradov), and more refined results follow from GRH.

Quantum Chaos and the GUE Hypothesis

A remarkable experimental observation: the statistical distribution of normalized spacings between zeros of the Riemann zeta function matches the distribution of eigenvalue spacings in the Gaussian Unitary Ensemble (GUE) of random matrix theory. This was discovered numerically by Montgomery and Dyson around 1972.

Montgomery's Pair Correlation Conjecture

Let F(alpha) = sum over zeros rho = 1/2 + i gamma of w(gamma) x^(i gamma), where w is a weight. Montgomery conjectured that the pair correlation of zeta zeros equals 1 - (sin(pi u) / (pi u))^2, which is exactly the GUE pair correlation. This suggests a deep connection between the Riemann zeta function and quantum systems whose classical limit is chaotic.

Random matrix theory also governs the distribution of the maximum of |zeta(1/2 + it)| over t in [T, 2T] (conjectured: ~ (1/2) ln(T) with GUE fluctuations), and the value distribution of L-functions at the central point s = 1/2 (Katz-Sarnak conjectures).

Practice Problems

Problem 1: Euler Totient Identity

Prove that sum over d dividing n of phi(d) = n using Mobius inversion and the multiplicativity of phi.

Show Solution

Solution:

Partition the set (1, 2, ..., n) by defining A_d = integers k in (1,...,n) with gcd(k, n) = d. Then A_d has exactly phi(n/d) elements (since k/d ranges over integers coprime to n/d). So sum over d dividing n of phi(n/d) = n. Re-indexing d to n/d gives sum over d dividing n of phi(d) = n.

In terms of Dirichlet convolution: phi * 1 = id, where id(n) = n. This is equivalent since phi * 1 evaluated at n equals n.

Problem 2: Dirichlet Series Product

Show that the Dirichlet series for tau(n) (number of divisors) equals zeta(s)^2, and find the Dirichlet series for sigma(n) (sum of divisors).

Show Solution

Solution:

tau(n) = sum over d dividing n of 1, which is the Dirichlet convolution of the constant function 1 with itself: tau = 1 * 1. The Dirichlet series for 1 is zeta(s), so the Dirichlet series for tau is zeta(s)^2.

Similarly, sigma(n) = sum over d dividing n of d = 1 * id (Dirichlet convolution of the constant 1 with the identity function id(n) = n). The series for id(n) is sum n/n^s = sum 1/n^(s-1) = zeta(s-1). Therefore the Dirichlet series for sigma(n) is zeta(s) zeta(s-1), converging for Re(s) greater than 2.

Problem 3: Zero-Free Region Argument

Explain why the identity 3 + 4 cos(theta) + cos(2 theta) at least 0 implies zeta has no zeros on the line Re(s) = 1, and what goes wrong if we try to prove zeta has no zeros on Re(s) = 1/2 by the same method.

Show Solution

Solution:

For sigma greater than 1, taking logarithms in the Euler product: Re(ln zeta(sigma + it)) = sum over primes p of sum over k at least 1 of Re(p^(-k(sigma+it))) / k = sum over p, k of cos(kt ln p) / (k p^(k sigma)).

Using 3 + 4 cos(theta) + cos(2 theta) = 2(1 + cos theta)^2 at least 0 with theta = t ln p gives: 3 Re ln zeta(sigma) + 4 Re ln zeta(sigma+it) + Re ln zeta(sigma+2it) at least 0, which means zeta(sigma)^3 |zeta(sigma+it)|^4 |zeta(sigma+2it)| at least 1.

Near s = 1, zeta(sigma) ~ 1/(sigma-1). If zeta had a zero of order m at 1+it0, |zeta(sigma+it0)| would be O((sigma-1)^m). The inequality forces (sigma-1)^(-3+4m) to remain bounded below, requiring 4m at most 3, hence m = 0 (no zero) since m must be a positive integer.

On Re(s) = 1/2: zeta(sigma)^3 still blows up but now the pole is weaker relative to the potentially vanishing terms, and the functional equation mixes zeros and poles in a way that prevents the same argument from forcing non-vanishing. The method is fundamentally tied to the presence of the pole at s = 1.

Problem 4: Character Orthogonality

Using orthogonality of Dirichlet characters, show that the number of primes p at most x with p congruent to a mod q (gcd(a,q) = 1) is approximately pi(x)/phi(q).

Show Solution

Solution:

By character orthogonality, the indicator of (n is congruent to a mod q) can be written as: 1_(n = a mod q) = (1/phi(q)) sum over chi mod q of chi_bar(a) chi(n).

Summing over primes p at most x: pi(x; q, a) = sum over primes p at most x of 1_(p = a mod q) = (1/phi(q)) sum over chi mod q of chi_bar(a) (sum over primes p at most x of chi(p)).

For the principal character chi_0, the inner sum is approximately pi(x). For non-principal chi, the sum over chi(p) is o(x/ln x) by the prime number theorem for arithmetic progressions (which uses L(1,chi) nonzero). Therefore pi(x; q, a) ~ pi(x)/phi(q).

Problem 5: The Selberg Sieve Bound

State and apply the Selberg upper bound sieve to prove that the number of twin prime pairs (p, p+2) up to N is O(N / (ln N)^2).

Show Solution

Solution:

Let A = (n(n+2) : n at most N). We want to count n at most N such that both n and n+2 are prime, which requires n(n+2) to have no prime factor at most sqrt(N).

Apply the Selberg sieve with sieve set P = (primes at most z) where z = sqrt(N), and the set A. The key quantity is the density of n mod p such that n(n+2) is divisible by p. For odd prime p, there are 2 residues modulo p where n(n+2) is divisible by p (namely n = 0 and n = -2 = p-2). For p = 2, there is 1 such residue (n = 0).

The Selberg sieve gives an upper bound: the count is at most C times N / (sum over d at most z, squarefree, p|d implies p in P, of mu(d)^2 / g(d))^2, where g(d) is the product of p/(p-2) over prime divisors of d. This product over the sieve gives (ln z)^2 times a convergent product, yielding the bound O(N / (ln N)^2). The twin prime conjecture says the true order is N / (ln N)^2 with an explicit constant involving the twin prime constant C_2 = product over primes p at least 3 of p(p-2)/(p-1)^2.

Exam Tips

Know Your Dirichlet Series

Memorize the standard identities: zeta(s)^2 = sum tau(n)/n^s, 1/zeta(s) = sum mu(n)/n^s, -zeta'(s)/zeta(s) = sum Lambda(n)/n^s. Be ready to derive new identities by multiplying known series.

PNT Proof Outline

Know the four steps: (1) link psi(x) to -zeta'/zeta via Perron, (2) main term from the pole at s=1, (3) nonvanishing of zeta on Re(s)=1 via the 3+4cos+cos(2) trick, (4) contour shift and error estimate.

Character Computations

Practice applying the orthogonality relations to decompose sums over arithmetic progressions. Remember: sum over chi of chi(a)chi_bar(b) equals phi(q) if a = b mod q and gcd(b,q) = 1, else 0.

Functional Equation

Know the functional equation xi(s) = xi(1-s) and what it implies: trivial zeros at negative even integers, symmetry of nontrivial zeros about Re(s) = 1/2, and the location s = 1/2 as the center of symmetry.

Sieve Method Setup

For sieve problems, always identify: the set A being sieved, the primes P doing the sieving, and the density function omega(p) (fraction of residues mod p that are sieved out). The key quantity is the product sum relating the sieve level to the main term.

Modular Forms Checklist

For any result about modular forms, check: the weight k, the level (SL_2(Z) or Gamma_0(N)), whether it is a cusp form, and whether it is a Hecke eigenform. The Hecke eigenform condition is what gives the Euler product for the L-function.

Related Topics

Number Theory

Elementary number theory: divisibility, primes, congruences, quadratic reciprocity — the algebraic foundations for analytic number theory.

Complex Analysis (Advanced)

Contour integration, the residue theorem, entire functions, and the Phragmen-Lindelof principle — the analytic toolkit used throughout this subject.

Real Analysis

Rigorous foundations for limits, series convergence, and integration — essential background for understanding Dirichlet series convergence and Perron's formula.