For ∀n∈N, if an arithmetic functionf that is not f(n)=0 satisfies the following, it is called a multiplicative function.
f(mn)=f(m)f(n),gcd(m,n)=1
If a multiplicative function satisfies the following condition, it is called a completely multiplicative function.
f(mn)=f(m)f(n),m,n∈N
Basic Properties
[1]: If f is multiplicative, then f(1)=1.
[2]: That f is a multiplicative function is equivalent to for all primesp1,⋯,pr and all a1,⋯,ar∈N, f(p1a1⋯prar)=f(p1a1)⋯f(prar).
[3]: If f is multiplicative, for primes p that satisfy p∣n,
d∣n∑μ(d)f(d)=p∣n∏(1−f(p))
[4]: If f is multiplicative, then f being a completely multiplicative function is equivalent to for all primesp and all a∈N, f(pa)=[f(p)]a.
[5]: If f is multiplicative, then f being a completely multiplicative function is equivalent to the inverse of f with respect to Dirichlet convolutionf−1 being represented as follows.
f−1(n)=μ(n)f(n)
[6]: If f is completely multiplicative, then F(n):=∑d∣nf(d) is multiplicative.
Naturally, being completely multiplicative implies being multiplicative.
An arithmetic function being multiplicative is akin to the ‘independent’ condition spoken of across mathematics. Since function values can be thought of in smaller fragments, it naturally harbors rich mathematical properties.
Theorem [1] is derived very simply but implies f(1)=0, which becomes a critical condition for the set of multiplicative functions to form a group.
Proof
[1]
Since f(1)=f(1⋅1)=f(1)f(1), f(1)=1 must hold.
■
[2]
(⇒)
Since f is multiplicative and always gcd(pa,qb)=1 for different primesp=q,
f(p1a1⋯prar)===f(p1a1)f(p2a2⋯prar)f(p1a1)f(p2a2)f(p3a3⋯prar)f(p1a1)⋯f(prar)
(⇐)
If we set m=p1a1⋯prar, n=q1b1⋯qsbs, and gcd(m,n)=1, then
f(mn)====f(p1a1⋯prarq1b1⋯qsbs)f(p1a1)⋯f(prar)f(q1b1)⋯f(qsbs)f(p1a1⋯prar)f(q1b1⋯qsbs)f(m)f(n)
■
[3]
Strategy: The properties of the Möbius function must be used.
Definition of the Möbius function: Given a primep1,⋯,pk, let a natural numbern be represented as n=p1a1⋯pkak. The arithmetic functionμ defined as follows is called the Möbius function.
μ(n):=⎩⎨⎧1(−1)k0,n=1,a1=⋯=ak=1,otherwise
Multiplicativity: For all m,n∈N that satisfy gcd(m,n)=1, μ(mn)=μ(m)μ(n)
g(n):=d∣n∑μ(d)f(d)
And let’s say m satisfies g(m,n)=1. Then, for d=ab that satisfies a∣m and b∣n, gcd(a,b)=1 also holds, so we can consider sigma separately. Also, since μ is multiplicative and f is multiplicative by the premise,
g(mn)=====ab∣mn∑μ(ab)f(ab)a∣mb∣n∑μ(ab)f(ab)a∣mb∣n∑μ(a)μ(b)f(a)f(b)a∣m∑μ(a)f(a)b∣n∑μ(b)f(b)g(m)g(n)
Therefore, g is also multiplicative. Then, according to Theorem [2], g(p1a1⋯prar)=g(p1a1)⋯g(prar), so we only need to calculate g(pa) for a fixed primep. According to Theorem [1], f(1)=1 and μ(1)=1, and for higher orders k≥2, all μ(pk)=0,
g(pa)====d∣pa∑μ(d)f(d)μ(1)f(1)+μ(p)f(p)1⋅1−1⋅f(p)1−f(p)
Summarizing for natural numbersn=p1a1⋯prar results in
g(n)=p∣n∏g(pa)=p∣n∏(1−f(p))
■
[4]
(⇒)
Since f is completely multiplicative,
f(pa)=f(p)(pa−1)=[f(p)]2(pa−2)=⋯=[f(p)]a
(⇐)
Let’s consider that for a primep1,⋯,ps, an integer a1,⋯,as,b1,⋯,bs∈N0, and a natural numberc1,⋯,ct∈N, any m and n are represented as follows.
m=p1a1⋯ptatn=p1b1⋯ptbtmn=p1c1⋯ptct
If m and n do not have primepi as a divisor, then ai=0 and bi=0, respectively. Since f is multiplicative,
f(mn)========f(p1c1⋯ptct)f(p1c1)⋯f(ptct)[f(p1)]c1⋯[f(pt)]ct[f(p1)]a1[f(p1)]b1⋯[f(pt)]at[f(pt)]bt[f(p1)]a1⋯[f(pt)]at[f(p1)]b1⋯[f(pt)]bt[f(p1a1)]⋯[f(ptat)][f(p1b1)]⋯[f(ptbt)]f(p1a1⋯ptat)f(p1b1⋯ptbt)f(m)f(n)
■
[5]
(⇒)
Let’s say g(n):=μ(n)f(n).
Since f is completely multiplicative and d∣n∑μ(d)=I(n),
$$
\begin{align*}
(gf)(n) =& \sum_{d \mid n} \mu (d) f (d) f \left( {{ n } \over { d }} \right)
\\ =& f(n) \sum_{d \mid n} \mu (d)
\\ =& f(n) I(n)
\end{align}
$$
In Theorem [1], since $f(1) = 1$ and the identity functionI is I(n)=0 in n>1, f(n)I(n)=I(n). Summarizing, (g∗f)(n)=I(n), so g=f−1.
According to the definition of the Möbius function, if k≥2 then μ(pk)=0,
μ(1)f(1)f(pa)+μ(p)f(p)f(pa−1)+0+⋯+0=0
Since the premise is that f is a multiplicative function, according to Theorem [1], f(1)=1. Meanwhile, the Möbius function is μ(1)=1 and μ(p)=−1, therefore,
1⋅1⋅f(pa)+(−1)⋅f(p)f(pa−1)=0
Summarizing,
f(pa)=f(p)f(pa−1)
When recursively resolved, f(pa)=f(p)a, therefore, by Theorem [4], f is completely multiplicative.
■
[6]
If we assume gcd(m,n)=1, we can distinguish the divisors of m from those of n. For convenience, let’s say a1,⋯,aM are the divisors of m, and b1,⋯,bN are the divisors of n.
F(mn)====d∣mn∑f(d)f(1)+f(a1)+⋯+f(aM)+f(b1)+⋯+f(bN)+f(a1bN)+⋯+f(aMb1)+⋯+f(a1a2bN)+⋯+f(mn)a∣m∑f(a)b∣n∑f(b)F(m)F(n)
■
Apostol. (1976). Introduction to Analytic Number Theory: p33. ↩︎