Arithmetic Functions and Multiplicative Functions — Functions on the Integers
We systematically define the principal multiplicative functions (phi, sigma, mu, d), develop the ring structure of Dirichlet convolution, and give a rigorous proof of the Moebius inversion formula. We also clarify the connection to the inclusion-exclusion principle.
A function f:Z>0→C is called an arithmetic function.
Definition 2 (Multiplicative function).
An arithmetic function f is multiplicative if f(1)=1 and f(mn)=f(m)f(n) whenever gcd(m,n)=1. If f(mn)=f(m)f(n) for all positive integers m,n, then f is completely multiplicative.
2 The Principal Arithmetic Functions
Definition 3 (Divisor function and sum-of-divisors function).
For a positive integer n:
d(n)=∑d∣n1: the number of positive divisors of n (also written σ0(n)).
σ(n)=∑d∣nd: the sum of the positive divisors of n (also written σ1(n)).
More generally, σk(n)=∑d∣ndk.
Theorem 4.
Both d and σ are multiplicative. For n=p1e1⋯pkek,
d(n)=i=1∏k(ei+1),σ(n)=i=1∏kpi−1piei+1−1.
Proof.
Multiplicativity follows from the bijection between divisors of mn (when gcd(m,n)=1) and pairs of divisors of m and n. For a prime power, the divisors of pe are 1,p,p2,…,pe, giving d(pe)=e+1 and σ(pe)=1+p+⋯+pe=(pe+1−1)/(p−1).□
Definition 5 (M\"obius function).
The Möbius functionμ:Z>0→{−1,0,1} is defined by
μ(n)=⎩⎨⎧1(−1)k0if n=1,if n=p1p2⋯pk (a product of k distinct primes),if n has a squared prime factor.
Lemma 6.
d∣n∑μ(d)={10if n=1,if n>1.
Proof.
The case n=1 is immediate. For n>1, let p1,…,pk be the distinct prime factors of n. The divisors d of n with μ(d)=0 correspond to products of subsets of {p1,…,pk}, so ∑d∣nμ(d)=∑j=0k(jk)(−1)j=(1−1)k=0.□
3 Dirichlet Convolution
Definition 7 (Dirichlet convolution).
The Dirichlet convolution of arithmetic functions f and g is
(f∗g)(n)=d∣n∑f(d)g(n/d).
Theorem 8.
Dirichlet convolution is commutative and associative, with identity element ε(n)=[n=1] (the function that is 1 at n=1 and 0 elsewhere). Every arithmetic function f with f(1)=0 has a Dirichlet inverse f−1 satisfying f∗f−1=ε.
Proof.
Commutativity follows from the substitution d↔n/d: ∑d∣nf(d)g(n/d)=∑d∣nf(n/d)g(d). Associativity is verified by rewriting the double sum as a triple sum ∑abc=nf(a)g(b)h(c). That f∗ε=f follows from ε(n/d)=[d=n]. The inverse is defined recursively: f−1(1)=1/f(1), and for n>1, f−1(n)=−f(1)1∑d∣n,d<nf(n/d)f−1(d).□
Remark 9.
The Dirichlet convolution of two multiplicative functions is again multiplicative. Denoting by 1 the constant function 1(n)=1, we have μ∗1=ε; that is, μ is the Dirichlet inverse of 1.
4 The Möbius Inversion Formula
Theorem 10 (M\"obius inversion formula).
For arithmetic functions f and g,
g(n)=d∣n∑f(d)for all n⟺f(n)=d∣n∑μ(d)g(n/d)for all n.
Proof.
In the language of Dirichlet convolution, the two conditions read g=f∗1 and f=g∗μ. Since μ∗1=ε, we can pass between them using associativity: if g=f∗1, then g∗μ=f∗1∗μ=f∗ε=f; conversely, if f=g∗μ, then f∗1=g∗μ∗1=g∗ε=g.□
Example 11.
It is known that ∑d∣nφ(d)=n (classify 1,…,n according to gcd(k,n)=d; there are φ(n/d) integers in each class). Applying Möφ(n)=∑d∣nμ(d)⋅n/d=n∑d∣nμ(d)/d, which is equivalent to φ(n)=n∏p∣n(1−1/p) and agrees with the derivation via inclusion-exclusion.