Theorems Flashcards
Dirichletβs Theorem
If k > 0 and gcd(h, k) = 1, then there are infinitely many primes in the arithmetic progression nk + h for n = 1, 2, β¦
Prime Number Theorem
For large x, the ratio π(x) / (x/logx) approaches 1, i.e.
π(x)
βββ- βΆ 1 as x βΆ β
(x/logx)
Properties of divisibility
Let a, b, d, m, n be arbitrary integers unless otherwise indicated. Then
(1) d|m and d|n β d|(am + bn)
(2) ad|an and a β 0 β d|n
(3) d|n and n β 0 β |d| β€ |n|
(4) d|n and n|d β |d| = |n|
(5) d|n and ndβ 0 β (n/d) | n
Common divisor
Given any two integers a and b, there is a common divisor of the form d = ax + by where x, y β Z. Moreover, every common divisor of a and b divides this d.
Euclidβs Lemma
If a|bc and gcd(a, b) = 1, then a|c.
Prime numbers
(1) Every integer n > 1 is a prime number of a product of prime numbers.
(2) There are infinitely many prime numbers.
(3) If p is a prime and p β€ a, then (p, a) = 1
(4) If p is a prime and p|ab, then p|a or p|b. More generally if p|aβ¦a, then p divides at least one of the factors.
Fundamental Theorem of Arithmetic
Every integer n > 1 can be represented as a product of prime factors in only one way, apart from the order of the factors.
β
β 1 / pβ
n=1
Let pβ denote the nth prime. The infinite series β β 1 / pβ n=1 diverges.
Quotient and remainder terms
Given positive integers a and b with b > 0, there exists a unique pair of integers q and r such that a = bq + r with 0 β€ r < b. Moreover, r = 0 if and only if b|a.
Euclidean Algorithm
We are given positive integers and b where bβ€a. Let rβ = a and
rβ = b, and apply the division repeatedly to obtain a set of remainders rβ, β¦, rβββ defined successively by the relations
rβ = rβqβ + rβ 0 < rβ < rβ
rβ = rβqβ + rβ 0 < rβ < rβ
β¦ β¦
rβββ = rβββqβββ + rβ 0 < rβ < rβββ
rβββ = rβqβ + rβββ rβββ = 0
Then rβ = gcd(a, b)
β π(d)
d|n
If n β₯ 1, we have
β π(d) = [1/n] = 1 if n = 1
d|n = 0 if n > 1
β π(d)
d|n
If n β₯ 1, we have
β π(d) = n
d|n
A relation connecting π and π
If n β₯ 1, we have
π(n) = β π(d) n/d
d|n
Product formula for π(n)
For n β₯ 1, we have
π(n) = n β (1 - 1/p)
p|n
Properties of π(n)
(a) π(pα΅
) = pα΅
- pα΅
β»ΒΉ for prime p and a β₯ 1
(b) π(mn) = π(m)π(n)d / π(d), where d = (m, n)
(c) π(mn) = π(m)π(n) if (m, n) = 1
(d) a|b β π(a)|π(b)
(e) π(n) is even for n β₯ 3. Moreover if n has r distinct odd prime factors, then 2 Ν¬ |π(n).
Properties of Dirichlet Convolution
Let f, g, and k be arithmetical functions. Then
a) f β g = g β f
(b) (f β g) β k = f β (g β k
i. e. Dirichlet convolution is commutative and associative.
Identity function on an arithmetical function
Let f be an arithmetical function. Then for all f, we have
f β I = I β f = f
Dirichlet Inverse
If f is an arithmetical function with f(1) β 0, then there exists a unique arithmetical function fβ»ΒΉ such that
f β fβ»ΒΉ = fβ»ΒΉ β f = I
The arithmetical function fβ»ΒΉ is called the Dirichlet inverse of f. Moreover, fβ»ΒΉ is given by the recursion formula
fβ»ΒΉ(1) = 1 / f(1)
fβ»ΒΉ(n) = - (1 / f(1)) β f(n/d) fβ»ΒΉ(d) for n > 1
d|n
d
Set of arithmetical functions
The set of all arithmetical functions f with f(1) β 0 forms an abelian group under the operation β.
Mobius Inversion Formula
The equation f(n) = β g(d) d|n implies g(n) = β f(d) π(n/d) d|n The converse holds.
Von-Mangoldt formula for log
If n β₯ 1, we have
log n = β Ξ(d)
d|n
Von-Mangoldt formula in terms of log and π
If n β₯ 1, we have
Ξ(n) = β π(d) log(n/d) = - β π(d) log d
d|n d|n
When is f(1) = 1?
If f is multiplicative, then f(1) = 1.
Properties of functions with f(1) = 1
Given an arithmetical function f with f(1) = 1,
(a) f is multiplicative if and only if
f(pβ Ν£ ΒΉ β¦ pβ Ν£ α΅) = f(pβ Ν£ ΒΉ) β¦ f(pβ Ν£ α΅)
for all primes pα΅’ and all integers aα΅’ β₯ 1
(b) If f is multiplicative, then f is completely multiplicative if and
only if
f(p Ν£ ) = f(p) Ν£
for all primes p and all integers a β₯ 1
Dirichlet product of multiplicative functions
If f and g are multiplicative functions, then the Dirichlet product
f β g is also multiplicative.
Note that the Dirichlet product of two completely multiplicative functions is not always completely multiplicative.
Also, if both g and f β g are multiplicative, then f is multiplicative.
Dirichlet inverse of a multiplicative function
If g is multiplicative, then gβ»ΒΉ, the Dirichlet inverse of g, is multiplicative.
Inverse of a completely multiplicative function
Let f be a multiplicative function. Then f is a completely multiplicative function if and only if
fβ»ΒΉ(n) = π(n) f(n)
for all n β₯ 1.
β π(d) f(d)
d|n
If f is multiplicative, then
β π(d) f(d) = β (1 - f(p))
d|n p|n
Product formula for πβ»ΒΉ(n)
πβ»ΒΉ(n) = β (1 - p)
p|n
β Ξ»(d)
d|n
For every n β₯ 1, we have
β Ξ»(d) = 1 if n is a square
d|n 0 otherwise
πββ»ΒΉ(n)
For n β₯ 1, we have that
πββ»ΒΉ(n) = β d Ν£ π(d) π(n/d)
d|n
Let πΌ and π½ be arithmetical functions. Then we have
πΌ β¬ (π½ β¬ F) = (πΌ β π½) β¬ F
where F : (0, β) βΆ C such that F(x) = 0 for 0 < x < 1
(πΌ β F)(m) for integers m
Let F : (0, β) βΆ C such that F(x) = 0 for 0 < x < 1 and let πΌ be an arithmetical function. Also let G : (0, β) βΆ C such that G(x) = 0 for 0 < x < 1 and
G(x) = (πΌ β F)(x) = β πΌ(n) F(x/n)
nβ€x
If F(x) = 0 for all nonintegral x (i.e. x is not an integer), we find that
(πΌ β F)(m) = (πΌ β F)(m)
for all integers m.
Dirichlet convolution identity function on β
The identity function for Dirichlet convolution is also a left identity on the operation β.
Generalised Inversion Formula
Let πΌ be an arithmetical function with Dirichlet inverse πΌβ»ΒΉ. Then the equation G(x) = β πΌ(n) F(x/n) nβ€x implies F(x) = β πΌβ»ΒΉ(n) G(x/n) nβ€x The converse also holds.
Generalised Mobius Inversion Formula
If πΌ is a completely multiplicative function, we have G(x) = β πΌ(n) F(x/n) nβ€x if and only if F(x) = β π(n) πΌ(n) G(x/n) nβ€x
Arithmetic on derivatives of arithmetical functions
If f and g are arithmetical functions, we have
(a) (f + g)β = fβ + gβ
(b) (f β g)β = fβ β g + f β gβ
(c) (fβ»ΒΉ)β = -fβ β (f β f)β»ΒΉ as long as f(1) β 0
Selberg Identity
For n β₯ 1, we have
Ξ(n) log n + β Ξ(n) Ξ(n/d) = β Β΅(d) logΒ²(n/d)
d|n d|n
f(t) = O(g(t)) for t β₯ a
f(t) = O(g(t)) for t β₯ a implies that x x β« f(t) dt = O( β« g(t) dt ) a a for x β₯ a
Euler Summation Formula
If f has a continuous derivative fβ on the interval [y, x], where
0 < y < x, then
x x
β f(n) = β« f(t) dt + β« (t - [t])fβ(t) dt + f(x)([x] - x) - f(y)([y] - y)
y
Euler Summation formula if f has a continuous derivative on
x > 0
Let f have a continuous derivative on x > 0. Then
x x
β f(n) = β« f(t) dt + f(x)([x] - x) + f(1) + β« (t - [t])fβ(t) dt
nβ€x 1 1
Consequences of Eulerβs Summation Formula
If x β₯ 1 then we have:
(a) β 1/n = log x + πΎ + O(1/x)
nβ€x
where πΎ is Eulerβs constant defined by the equation
n
πΎ = lim ( β 1/k - log n )
nββ k=1
(b) β 1/nΛ’ = xΒΉβ»Λ’ / (1-s) + ΞΆ(s) + O(xβ»Λ’) if s > 0, s β 1
nβ€x
(c) β 1/nΛ’ = O(xΒΉβ»Λ’) if s > 1
n>x
(d) β nα΅
= xα΅
βΊΒΉ / (a + 1) + O(xα΅
) if a β₯ 0
nβ€x
Dirichletβs Asymptotic Formula (weak)
For all x β₯ 1, we have
β d(n) = x log x + O(x)
nβ€x
Dirichletβs Asymptotic Formula
For all x β₯ 1, we have
β d(n) = x log x + (2πΎ - 1)x + O(βx)
nβ€x
Average order of π(n)
For all x β₯ 1, we have
β π(n) = 1/2 ΞΆ(2) xΒ² + O(x log x)
nβ€x
Average order of πβ(n) for a > 0
If x β₯ 1 and a > 0, a β 1, we have
β πβ(n) = ( ΞΆ(a + 1) / (a + 1) ) x Ν£ βΊΒΉ + O(x α·© )
nβ€x
where π½ = max{1, a}
Average order of πβ(n) for a < 0
Write a = -π½, so π½ > 0. Let πΏ = max{0, 1 - π½}. Then if x > 1, we have
β πβ(n) = ΞΆ(π½ + 1) x + O(xα΅) if π½ β 1
nβ€x ΞΆ(2) x + O(log x) if π½ = 1
Average order of π(n)
For x > 1, we have
β π(n) = (3/πΒ²) xΒ² + O(x log x)
nβ€x
lim 1/x β π(n)
xββ nβ€x
lim 1/x β π(n) = 0
xββ nβ€x
This is equivalent to the PNT.
lim 1/x β Ξ(n)
xββ nβ€x
lim 1/x β Ξ(n) = 1
xββ nβ€x
This is equivalent to the PNT.
Partial sums of a Dirichlet product
If h = f β g, let
H(x) = β h(n), F(x) = β f(n), G(x) = β g(n)
nβ€x nβ€x nβ€x
Then
H(x) = β f(n) G(x/n) = β g(n) F(x/n)
nβ€x nβ€x
β F(x/n)
nβ€x
If
F(x) = β f(n)
nβ€x,
we have
β β f(d) = β f(n) [x/n] = β F(x/n)
nβ€x d|n nβ€x nβ€x
β π(n) [x/n]
nβ€x
We have
β π(n) [x/n] = 1
nβ€x
β Ξ(n) [x/n]
nβ€x
We have
β Ξ(n) [x/n] = (log [x])!
nβ€x
β
| β π(n) / n |
n=1
For all x β₯ 1, we have β | β π(n) / n | β€ 1 n=1 with equality holding only when x < 2.
Legendreβs Identity
For every x β₯ 1, we have [x]! = β pα΅ β½α΅βΎ where β πΌ(p) = β [x/pα΅] m=1
(log[x])! when x β₯ 2
If x β₯ 2, we have
(log[x])! = x log x - x + O(log x)
and hence
β Ξ(n) [x/n] = x log x - x + O(log x)
nβ€x
When x β₯ 2,
β [x/p] log p
pβ€x
When x β₯ 2,
β [x/p] log p = x log x + O(x)
pβ€x
Dirichletβs Hyperbola Method
Let
F(x) = β f(n), G(x) = β g(n) and H(x) = β (f β g)(n),
nβ€x nβ€x nβ€x
so that
H(x) = β β f(d) g(n/d) = β f(d) g(q)
nβ€x d|n q, d
qdβ€x
If a, b β R such that ab = x, than
β f(d) g(q) = β f(n) G(x/n) + β g(n) F(x/n) - F(a) G(b)
q, d nβ€a nβ€b
qdβ€x
π(x) / x - ΞΈ(x) / x
For x > 0, we have
0 β€ π(x) / x - ΞΈ(x) / x β€ (log x)Β² / (2 βx log2)
This implies that
lim ( π(x) / x - ΞΈ(x) / x ) = 0
xββ
Hence if either π(x) / x or ΞΈ(x) / x tends to a limit, the so does the other and the limits are equal.
Abelβs Identity
For any arithmetical function a(n), let
A(x) = β a(n)
nβ€x
where A(x) = 0 if x < 1. Assume f has a continuous derivative on the interval [y, x], where 0 < y < x. Then we have
x
β a(n) f(n) = A(x) f(x) - A(y) f(y) - β« A(t) fβ(t) dx
y < n β€ x y
Abelβs Identity where f has a continuous derivative on x > 0
For any arithmetical function a(n), let
A(x) = β a(n)
nβ€x
where A(x) = 0 if x < 1. Assume f has a continuous derivative on
x > 0. Then we have
x
β a(n) f(n) = A(x) f(x) - β« A(t) fβ(t) dt
nβ€x 1
ΞΈ(x) and π(x) in terms of integrals
For x β₯ 2, we have x ΞΈ(x) = π(x) log x - β« π(t) / t dt 2 and x π(x) = ΞΈ(x) / logx + β« ΞΈ(t) / (t logΒ²(t)) dt 2
lim ΞΈ(x) / x
xββ
lim ΞΈ(x) / x = 1
xββ
This is equivalent to the PNT.
lim π(x) / x
xββ
lim π(x) / x = 1
xββ
This is equivalent to the PNT.
Relating the PNT to the asymptotic value of the nth prime
Let pβ denote the nth prime. Then the following asymptotic relations are logically equivalent.
lim (π(x) log x) / x = 1
xββ
lim (π(x) log π(x)) / x = 1
xββ
lim pβ / (n log n) = 1
nββ
Order of magnitude of π(n)
For every integer n β₯ 2, we have
1/6) (n / log n) < π(n) < 6 (n / log n
Order of magnitude of pβ
For n β₯ 1, the nth prime satisfies the inequalities
1/6 n log n < pβ < 12 (n log n + n log(12/e) )
Shapiroβs Tauberian Theorem
Let {a(n)} be a non-negative sequence such that β a(n) [x/n] = x log x + O(x) nβ€x for all x β₯ 1. Then (a) For x β₯ 1, we have β a(n) / n = log x + O(1) nβ€x (b) There exists a constant B > 0 such that β a(n) β€ Bx nβ€x for all x β₯ 1. (c) There exists a constant A > 0 and an xβ > 0 such that β a(n) β₯ A(x) nβ€x for all x β₯ xβ
Shapiroβs Theorem on Ξ(n)
For all x β₯ 1, we have β Ξ(n) / n = log x + O(1) nβ€x Also, there exist positive constants cβ and cβ such that β Ξ(n) = π(x) β€ cβx nβ€x for all x β₯ 1 and β Ξ(n) = π(x) β₯ cβx nβ€x for all sufficiently large x.
Shapiroβs Theorem on ΞΈ(n)
For all x β₯ 1, we have β log p / p = log x + O(1) pβ€x Also, there exist positive constants cβ and cβ such that β Ξβ(n) = ΞΈ(x) β€ cβx nβ€x for all x β₯ 1 and β Ξβ(n) = ΞΈ(x) β₯ cβx nβ€x for all sufficiently large x. Here, Ξβ(n) = log p if n is a prime p 0 otherwise
β π(x/n) and β ΞΈ(x/n)
nβ€x nβ€x
We have β π(x/n) = x log x - x + O(log x) nβ€x and β ΞΈ(x/n) = x log x + O(x) nβ€x
An asymptotic formula for the partial sums
β 1/p
pβ€x
There is a constant A such that
β 1/p = log log x + A + O(1 / log x)
pβ€x
for all x β₯ 2.
Relating M(x) to the PNT
We have lim ( M(x) / x - H(x) / (x log x) ) = 0 xββ Hence the PNT implies lim ( M(x) / x ) = 0 xββ Also, M(x) = o(x) as x βΆ β implies π(x) βΌ x as x βΆ β and so lim ( M(x) / x ) = 0 xββ also implies the PNT.
Relating
β π(n) / n
nβ€x
to the PNT
If A(x) = β π(n) / n nβ€x the relation A(x) = o(1) as x βΆ β implies the PNT. In other words, the PNT is a consequence of the statement that β β π(n) / n nβ€x converges and has sum 0.
Selbergβs Asymptotic Formula
For x > 0, we have
π(x) log x + β Ξ(n) π(x/n) = 2x log x + O(x)
nβ€x
Deducing Selbergβs Asymptotic Formula
Let F be a real or complex valued function defined on (0, +β), and let
G(x) = log x β F(x/n)
nβ€x
Then
F(x) log x + β F(x/n) Ξ(n) = β π(d) G(x/d)
nβ€x dβ€x
If N β₯ 1 and π β₯ c > πβ, what is
β
| β f(n) / nβ»Λ’ |
n=N
If N β₯ 1 and π β₯ c > πβ, we have
β β
| β f(n) / nβ»Λ’ | β€ N^-(π - c) β |f(n)| / nαΆ
n=N n=N
For a Dirichlet series F(n), what is
lim F(π + it)
πβ+β
Let β F(s) = β f(n) / nΛ’ n=1 Then lim F(π + it) = f(1) πβ+β uniformly for -β < t < β
Uniqueness Theorem
Given 2 Dirichlet series
β β
F(s) = β f(n) / nΛ’ and G(s) = β g(n) / nΛ’
n=1 n=1
both absolutely convergent for π > πβ, if F(s) = G(s) for each s is an infinite sequence {sβ} such that πβ βΆ β as k βΆ β, then
f(n) = g(n) for all n.
Half-planes with F(s) never zero
Let F(s) = β f(n) / nΛ’ and assume that F(s) β 0 for some s with π > πβ. Then there is a half-plane π β₯ c > πβ in which F(s) is never zero.
Dirichlet convolution in half-plane of convergence
Given two functions F(s) and G(s) represented by Dirichlet series
β β
F(s) = β f(n) / nΛ’ for π>a and G(s) = β g(n) / nΛ’ for π>b
n=1 n=1
In the half-plane where both series converge absolutely, we have
β
F(s) G(s) = β h(n) / nΛ’
n=1
where h = f β g, the Dirichlet convolution of f and g:
h(n) = β f(d) g(n/d)
d|n
Conversely, if F(s)G(s) = β πΌ(n)/nΛ’ for all s in a sequence {sβ} with πβ βΆ β as k βΆ β, then πΌ = f β g.
Euler Product
Let f be a multiplicative arithmetical function such that the series βf(n) is absolutely convergent. Then the sum of the series can be expressed as an absolutely convergent infinite product
β
β f(n) = β (1 + f(p) + f(pΒ²) + β¦)
n=1 p
extended over all primes. If f is completely multiplicative, then
β
β f(n) = β 1 / (1 - f(p))
n=1 p
Product formulae for Dirichlet series in half plane of convergence
Assume β f(n)/nΛ’ converges absolutely for π > πβ. If f is multiplicative, we have
β
β f(n) / nΛ’ = β (1 + f(p)/pΛ’ + f(pΒ²)/pΒ²Λ’ + β¦) if π > πβ
n=1 p
and if f is completely multiplicative, we have
β
β f(n) / nΛ’ = β (1 - f(p)/p)β»ΒΉ if π > πβ
n=1
Modulus of Dirichlet series with bounded partial sums
Let sβ = πβ + itβ and assume that the Dirichlet series β f(n)/nΛ’β° has bounded partial sums, say
| β f(n) / nΛ’β° | β€ M
nβ€x
for all x β₯ 1. Then for each π > πβ, we have
| β f(n) / nΛ’ | β€ 2Ma^(πβ - π) (1 + (|s - sβ|) / (π - πβ) )
a
Convergent/ divergent Dirichlet series for s = πβ + itβ
If the series β f(n)/nΛ’ converges for s = πβ + itβ, then it also converges for all s with π > πβ. If it diverges for s = πβ + itβ, then it diverges for all s with π < πβ.
Bounds on πβ and π_c
For any Dirichlet series with π_c finite, we have
0 β€ πβ - π_c β€ 1
πβ and π_c on
β
β (-1)βΏ / nΛ’
n=1
The series β β (-1)βΏ / nΛ’ n=1 converges if π > 0 but only converges absolutely for πβ = 1. Therefore π_c = 0 and πβ = 1.
Analytic functions
Let {fβ} be a sequence of functions analytic on an open subset S of the complex plane, and assume that {fβ} converges uniformly on every compact subset of S to a limit function f. Then f is analytic on S and the sequence of derivatives {fββ} converges uniformly on every compact subset of S to the derivative of fβ.
Dirichlet series on compact rectangles
A Dirichlet series converges uniformly on every compact rectangle lying interior to the half-plane of convergence
π > π_c.
Derivative of Dirichlet series
The sum function β F(s) = β f(n) / nΛ’ n=1 of a Dirichlet series is analytic in its half-plane of convergence π > π_c and its derivative F'(s) is represented in this half-plane by the Dirichlet series β F'(s) = - β (f(n) log n) / nΛ’ n=1 obtained by differentiating term by term
Landauβs Theorem
Let F(s) be represented in the half-plane π > c by the Dirichlet series
β
F(s) = β f(n) / nΛ’
n=1
where c is finite and assume f(n) β₯ 0 for all n β₯ nβ. If F(s) is analytic in some disc about the point s = c, then the Dirichlet series converges in the half-plane π > c - π for some π > 0. Consequently, if the Dirichlet series has a finite abscissa of convergence π_c, then F(s) has a singularity on the real axis at the point s = π_c.
Half-plane of a Dirichlet series as an exponential
Let F(s) = β f(n)/nΛ’ be absolutely convergent for π > πβ and assume that f(1) β 0. If F(s) β 0 for π > πβ > πβ, then for π > πβ we have
F(s) = exp( G(s) )
with
β
G(s) = log(f(1)) + β (fβ β fβ»ΒΉ)(n) / ( (log n)nΛ’ )
n=1
where fβ»ΒΉ is the Dirichlet inverse of f and fβ(n) = f(n) log n.
β
β f(n) g(n) / n^(a+b)
n=1
Given two Dirichlet series F(s) = β f(n)/nΛ’ and G(s) = β g(n)/nΛ’ with abscissae of absolute convergence πβ and πβ respectively. Then for a > πβ and b > πβ we have
T β
lim (1 / 2T) β« F(a + it) G(b - it) dt = β f(n) g(n) / n^(a+b)
Tββ -T n=1
If F(s) = β f(n)/nΛ’ converges absolutely for π > πβ, then for π > πβ we have T β
lim (1 / 2T) β« |F(π + it)|Β² dt = β |f(n)|Β² / n^(2π)
Tββ -T n=1
In particular, if π > 1, we have
(a) T β
lim (1 / 2T) β« |ΞΆ(π + it)|Β² dt = β 1 / n^(2π) = ΞΆ(2π)
Tββ -T n=1
(b) T β
lim (1 / 2T) β« |ΞΆβ½α΅βΎ(π + it)|Β² dt = β logΒ²α΅(n) / n^(2π) = ΞΆβ½Β²α΅βΎ(2π)
Tββ -T n=1
(c) T β
lim (1 / 2T) β« |ΞΆ(π + it)|β»Β² dt = β πΒ²(n) / n^(2π) = ΞΆ(2π) / ΞΆ(4π)
Tββ -T n=1
(d) T β
lim (1 / 2T) β« |ΞΆ(π + it)|β΄ dt = β dΒ²(n) / n^(2π) = ΞΆβ΄(2π) / ΞΆ(4π)
Tββ -T n=1
Integral formula for coefficients of a Dirichlet series
Assume the series F(s) = β f(n)/nΛ’ converges absolutely for π > πβ Then for π > πβ and x > 0 we have
T
lim (1 / 2T) β« F(π + it) x^(π + it) dt = f(n) if x = n
Tββ -T 0 otherwise
What is
c+iβ
1 / 2πi β« aαΆ» / z dz
c-iβ
Let c > 0 and a be any postitive real number. Then we have
c+iβ 1 if a > 1
1 / 2πi β« aαΆ» / z dz = 1/2 if a = 1
c-iβ 0 if 0 < a < 1
Moreover, we have
c+iT
| 1 / 2πi β« aαΆ» / z dz | β€ a Ν¨ / (π T log(1/a)) if 0 < a < 1
c-iT
c+iT
| 1 / 2πi β« 1/z dz - 1/2 | β€ c / πT if a = 1
c-iT
c+iT
| 1 / 2πi β« aαΆ» / z dz - 1 | β€ a Ν¨ / (π T log a) if a > 1
c-iT
Perronβs Formula
Let F(s) = β f(n)/nΛ’ be absolutely convergent for π > πβ, and let
c > 0 and x > 0 be arbitrary. Then if π > πβ - c, we have
c+iβ β
1 / 2πi β« F(s + z) xαΆ» / z dz = β f(n) / nΛ’
c-iβ nβ€x
where β* means that the last term in the sum must be multiplies by 1/2 when x is an integer.
Perronβs Formula for s = 0
If c > πβ Perronβs Formula is valid for s = 0 and we obtain the following integral representation for the partial sum of the coefficients:
c+iβ β
1 / 2πi β« F(z) xαΆ» / z dz = β f(n)
c-iβ nβ€x
Riemann-zeta function for π > 0, x > 0 and s β 1.
Suppose that π > 0, x > 0 and s β 1. Then for N an integer, we have
β
ΞΆ(s) = β 1 / nΛ’ + NΒΉβ»Λ’ / (s - 1) - s β« {t} / tΛ’βΊΒΉ dt
nβ€N N
Poles of the Riemann-zeta function
The Riemann-zeta function has a simple pole at s = 1 with residue 1, but is otherwise analytic in the half-plane π > 0.
Properties of the gamma function
For π > 0, we have that sπͺ(s) = πͺ(s + 1). Moreover πͺ(1) = 1 and
πͺ(n) = (n - 1)! for all n β N.
Poles of the gamma function
The function πͺ(s) has an analytic continuation as a meromorphic function on C, the only singularities being at s = -k β Z β€ 0. These singularities are simple poles with residues (-1)α΅ / k!.
Representations of the gamma function
πͺ(s) = lim (nΛ’ n!) / s(s + 1) β¦ (s + n) for s β 0, -1, -2, β¦
β
1 / πͺ(s) = s exp(πΎs) β (1 + s/n) exp(-s/n) for all s
n=1
Functional equation for the gamma function
πͺ(s) πͺ(1 - s) = π / sin(πs) for all s
Multiplication formula for the gamma equation
For all s and all integers m β₯ 1, we have
πͺ(s) πͺ(s + 1/m) β¦ πͺ(s + (m-1)/m) = (2π)^((m-1)/2) m^(1/2 - ms) πͺ(ms)
Functional equation for the Riemann-zeta function
For all s, we have
ΞΆ(s) = 2 (2π)Λ’β»ΒΉ πͺ(1 - s) sin(πs / 2) ΞΆ(1 - s)
Riemann Hypothesis
If 0 < Re(s) < 1 and ΞΆ(s) = 0, then Re(s) = 1/2
β (x - n) a(n)
nβ€x
For any arithmetical function a(n), let A(x) = β a(n) (nβ€x) where A(x) = 0 if x < 1. Then x
β (x - n) a(n) = β« A(t) dt
nβ€x 1
LβHopitalβs Rule for increasing linear piecewise functions
Let A(x) = β a(n) (nβ€x) and let Aβ(x) = β«βΛ£ A(t) dt. Assume a(n) β₯ 0 for all n. If we have the asymptotic formula
Aβ(x) βΌ LxαΆ as x βΆ β
for some c > 0 and L > 0, then we have
A(x) βΌ cLxαΆβ»ΒΉ as x βΆ β
LβHopitalβs rule on π(x)
We have
πβ(x) = β (x - n) Ξ(n)
nβ€x
and the asymptotic formula
πβ(x) βΌ 1/2 xαΆ as x βΆ β
implies
π(x) βΌ x as x βΆ β
For c > 0, u > 0, for every k β₯ 1, what is
c+iβ
1 / 2πi β« uβ»αΆ» / (z (z + 1) β¦ (z +k)) dz
c-iβ
If c > 0 and u > 0, then for every integer k β₯ 1, we have
c+iβ
1 / 2πi β« uβ»αΆ» / (z (z + 1) β¦ (z +k)) dz = (1 / k!) (1 - u)α΅ if 0 < u β€ 1
c-iβ 0 if u > 1
the integral being absolutely convergent.
Contour integral representation for πβ(x) / xΒ²
If c > 1 and x β₯ 1, we have
c+iβ
πβ(x) / xΒ² = 1 / 2πi β« xΛ’β»ΒΉ / s(s + 1) (- ΞΆβ(s) / ΞΆ(s) ) ds
c-iβ
Contour integral representation for πβ(x) / xΒ² - 1/2 (1 - 1/x)Β²
If c > 1 and x β₯ 1, we have
c+iβ
πβ(x) / xΒ² - 1/2 (1 - 1/x)Β² = 1 / 2πi β« xΛ’β»ΒΉ h(s) ds
c-iβ
where
h(s) = 1 / s(s + 1) (- ΞΆβ(s) / ΞΆ(s) - 1 / (s - 1) )
Upper bounds for |ΞΆ(s)| and |ΞΆβ(s)| near the line π = 1
For every A > 0, there exists a constant M (depending on A) such that
|ΞΆ(s)| β€ M log t and |ΞΆβ(s)| β€ M logΒ² t
for all s with π β₯ 1/2 satisfying
π > 1 - A / log t and t β₯ e
Riemann-zeta formula β₯ 1 for π > 1
If π > 1, we have
ΞΆΒ³(π) |ΞΆ(π + it)|β΄ |ΞΆ(π + 2it)| β₯ 1
Non-vanishing of ΞΆ(1 + it)
We have ΞΆ(1 + it) β 0 for every real t
Inequalities for | 1 / ΞΆ(s) | and | ΞΆβ(s) / ΞΆ(s) |
There is a constant M > 0 such that | 1 / ΞΆ(s) | < M logβ· t and | ΞΆ'(s) / ΞΆ(s) | < M logβΉ t whenever π β₯ 1 and t β₯ e.
First Order Poles of fβ(s) / f(s)
If f(s) has a pole of order k at s = πΌ, then the quotient fβ(s) / f(s) has a first order pole at s = πΌ with residue -k.
Where is F(s) = - ΞΆβ(s) / ΞΆ(s) - 1 / (s - 1) analytic?
The function
F(s) = - ΞΆβ(s) / ΞΆ(s) - 1 / (s - 1)
is analytic at s = 1.
Formula for π(x) βΌ x
For x β₯ 1, we have
β
πβ(x) / xΒ² - 1/2 (1 - 1/x)Β² = 1 / 2π β« h(1 + it) exp(it log x) dt
-β
where the integral
β
β« | h(1 + it) | dt
-β
converges. Therefore, by the Riemann-Lebesgue Lemma, we have
πβ(x) βΌ 1/2 xΒ²
and hence
π(x) βΌ x as x βΆ β
Zero-free regions for ΞΆ(s)
Assume π β₯ 1/2. Then there exist constants A > 0 and C > 0 such that
| ΞΆ(π + it) | > C / logβ· t
whenever
1 - A / logβΉ t < π β€ 1 and t β₯ e
This implies that ΞΆ(π + it) β 0 for such π and t.
Number of Dirichlet characters
There are exactly π(k) distinct Dirichlet characters modulo k.
Number of Dirichlet characters
There are exactly π(k) distinct Dirichlet characters modulo k.
Properties of Dirichlet characters
Let k be a positive integer.
(i) Then we have that π(n) = 0 or π(n) = exp(2πin / π(k)) for some
m β N.
(ii) Let (n, k) = 1. Then the inverse of a Dirichlet character π
modulo k is given by the complex conjugate π bar which is
defined by
_ ___
π(n) = π(n)
Furthermore, π bar is also a Dirichlet character modulo k.
Sums over π(n)
Let k N. Then we have
(i) k
β π(n) = π(k) if π = πβ
n=1, (n,k)=1 0 if π β πβ
(i) β π(n) = π(k) if n β‘ 1 mod k
π mod k 0 otherwise
What is _
β πα΅£(m) πα΅£(n)
Let πβ, β¦, π_(π(k) - 1) denote the Dirichlet characters modulo k, where k β N. Let m, n β Z such that (n, k) = 1. Then we have
π(k)-1 _
β πα΅£(m) πα΅£(n) = π(k) if m β‘ n mod k
r=0 0 otherwise
β π(n) f(n)
Let π be any non-principal character modulo k and let f be a non-negative function which has a continuous negative derivative fβ(x) for all x β₯ xβ. Then if y β₯ x β₯ xβ, we have
β π(n) f(n) = O(f(x))
where we sum over x < n β€ y.
If in addition f(x) βΆ 0 as x βΆ β, then the infinite series
β
β π(n) f(n)
n=1
converges, and we have, for x β₯ xβ,
β
β π(n) f(n) = β π(n) f(n) + O(f(x))
nβ€x n=1
β π(n) / n
If π is any non-principal character modulo k and if x β₯ 1, we have
β
β π(n) / n = β π(n) / n + O(1 / x)
nβ€x n=1
β (π(n) log n) / n
If π is any non-principal character modulo k and if x β₯ 1, we have
β
β (π(n) log n) / n = β (π(n) log n) / n + O((log x) / x)
nβ€x n=1
β π(n) / βn
If π is any non-principal character modulo k and if x β₯ 1, we have
β
β π(n) / βn = β π(n) / βn + O(1 / βx)
nβ€x n=1
A(n) = β π(d)
d|n
Let π be any real-valued character modulo k and let
A(n) = β π(d)
d|n
Then A(n) β₯ 0 for all n and A(n) β₯ 1 if n is a square.
Properties of B(n) such that
A(n) = β π(d) and B(n) = β A(n) / βn
d|n nβ€x
For any real-valued non-principal character π modulo k, let
A(n) = β π(d) and B(n) = β A(n) / βn
d|n nβ€x
Then
(a) B(x) βΆ β as x βΆ β.
(b) B(x) = 2βx L(1, π) + O(1) for all x β₯ 1
When is L(1, π) nonzero?
For any real-valued non-principal character π modulo k, we have L(1, π) β 0.
Formula for β log p / p pβ€x pβ‘h mod k if k > 0 and (h, k) = 1
If k > 0 and (h, k) = 1, we have for all x > 1,
β log p / p = (1 / π(k)) log x + O(1)
pβ€x
pβ‘h mod k
where the sum is extended over those primes p β€ x which are congruent to h mod k.
Formula for β log p / p pβ€x pβ‘h mod k for all x > 1 in terms of characters
For x > 1, we have
β log p / p = (1 / π(k)) log x
pβ€x π(k)-1 _
pβ‘h mod k + (1 / π(k)) β πα΅£(h) β πα΅£(p) log p / p
r=1 pβ€x
+ O(1)
Formula for
β log p / p
pβ€x
for all x > 1
For each x > 1 and π β πβ, we have
β π(p) log p / p = - Lβ(1, π) β π(n) π(n) / n + O(1)
pβ€x pβ€x
Size of
L(1, π) β π(n) π(n) / n
nβ€x
For x > 1 and π β πβ, we have
L(1, π) β π(n) π(n) / n βͺ 1
nβ€x
Formula for
Lβ(1, π) β π(n) π(n) / n
nβ€x
If π β πβ and L(1, π) = 0, we have
Lβ(1, π) β π(n) π(n) / n = log x + O(1)
nβ€x
Formula for β log p / p pβ€x pβ‘h mod k for all x > 1 in terms of number of characters
Let N(k) denote the number of non-principal characters π modulo k such that L(1, π) = 0. For x > 1, we have
β log p / p = ( (1 - N(k)) / π(k) ) log x + O(1)
pβ€x
pβ‘h mod k
Prime Number Theorem for arithmetic progressions
For (a, k) = 1, we have
πβ(x) βΌ π(x) / π(k) βΌ (1 / π(k)) (x / log x)
as x βΆ β.
Alternative formulation for the PNT for arithmetic progressions
If the relation
πβ(x) βΌ π(x) / π(k) as x βΆ β
holds for every integer a relatively prime to k, then
πβ(x) βΌ π_b(x) as x βΆ β
whenever (a, k) = (b, k) = 1. The converse is also true.
Size of integral of a continuous function on a contour
Consider a contour πΎ and a continuous function f defined on πΎ. Suppose that |f(z)| β€ M for all z on πΎ. Then
| β« f(z) dz | β€ ML
πΎ
where L is the length of πΎ.
Cauchyβs Integral Formula
Let U be a domain. Let πΎ be a positively oriented simple closed contour with its image and interior lying entirely within U. Suppose that a is a point in the interior of πΎ. If f is holomorphic on U, then
f(a) = 1 / 2πi β« f(z) / (z - a) dz
πΎ
Laurentβs Theorem
If f is holomorphic on an annulus A = {z β C : R < |z - a| < S} for
0 β€ R < s. Then there exists a sequence (bβ), n β Z, of complex numbers such that
β
f(z) = β bβ(z - a)βΏ
n=-β
for every z β A. Such a series is referred to as a Laurent series. Moreover for any r with R < r < S and for any n β Z, if πΎ is a circular closed contour with centre a and radius r, then
bβ = 1 / 2πi β« f(w) / (w - a)βΏβΊΒΉ dw
πΎ
Cauchyβs Residue Theorem
If πΎ is a closed contour traversed anticlockwise, if f is a function which is holomorphic on a domain containing the image and interior of πΎ, except for a finite number of isolated singularities on the interior (call them aβ, β¦, aβ), then
k
β« f(z) dz = 2πi β Res(f, a)
πΎ j=1