Blog Archive

Tuesday, January 30, 2024

Differential Ring (2): Arithmetic Function Ring

Arithmetic Function as Differential Ring

Definition. We call the element of HomSet(N+,C) arithmetic function. We will use the notation A.

Example. Euler function φ(n)=|(Z/nZ)|, p-adic valuation vp(n)...

Definition. Dirichlet convolution.

For f,gA, their Dirichlet convolution is defined as follows.

(1)fg(n)=d1d2=nf(d1)g(d2)

Proposition. (A,+,) form a commutative ring.

Proof.

It is not hard to see that (A,+) form an abelian group. Now we check the rest axiom.

Obviously, It is closed under , and fg=gf.

For the associative law,

(2)(fg)h(n)=d1d2=n(fg)(d1)h(d2)=d1d2=nd3d4=d1f(d3)g(d4)h(d2)=ijk=nf(i)g(j)h(k)

Similarly,

(3)f(gh)=ijk=nf(i)g(j)h(k)

Hence

(4)(fg)h=f(gh)

For the distributive law.

(5)f(g+h):=d1d2=nf(d1)(g+h)(d2)=d1d2=nf(d1)(g(d2)+h(d2))=d1d2=nf(d1)g(d2)+d1d2=nf(d1)h(d2)=:fg+fh

The unit of A is ε(n):={1,n=10,n2 , (fε)(n)=f(n). For the constant function 1, we have (f1)(n)=d|nf(d).

Hence (A,+,) is a commutative ring.

Definition. If A satisfies (mn)=(m)+(n), then is additive.

Example. vp(n),log(n)...

Another example is from the last blog, Remember dZ(n)=ni=1nvpipi?

Proposition. D(n):=dz(n)n is additive.

Proof .

We already now that dZ(mn)=dZ(m)n+mdZ(n). Hnece D(mn)=dZ(m)m+dZ(n)n.

Proposition. Der(A).

Remark.The derivation is defined by :AA,(f):=f. i.e. (f+g)=f+g,(fg)=fg+fg.

Proof.

[(fg)](n)=(n)(fg)(n)=[(d1)+(d2)]d1d2=nf(d1)g(d2)=d1d2=n(d1)f(d1)g(d2)+d1d2=n(d2)f(d1)g(d2)

i.e. fg(n)+fg(n). (f+g)=f+g follows from distributive law directly.

I will not repeat the differential ring properties again for (A,). See my blog DR (1) .

Let us consider some interesting things. As we know, the kernel of derivative form a ring.

What is the kernel of vp? What is the solution of the ODE vpf=0?

We know that the solution is nontrivial, since vp is a zero divisor in A. vp(n)=0npN+.

Hence the solution of vpf=0 is Ker(vp)={fA|xpN+,f(x)=0}.

Some special arithmetic function

As we can see, f1=d|nf(d) appear at lots of place. If I know f1, how could I know f?

Well, we need the inverse of 1, i.e. The mobius function μ={μ(1)=1μ(n)=(1)r,n=p1...prμ(n)=0,if n is not square free.

We need to prove that μ1=ε.

Proof. Obviously (μ1)(1)=ε(1). Let n=p1e1...pses

(μ1)(n)=d|nμ(d)=μ(1)+μ(pi)+i,jμ(pipj)+i,j,kμ(pipjpk)+...+μ(p1...ps)=r=0s(1)r(sr)=(11)s=0

Corollary. Mobius inversion. Let g=f1,f=gμ.

Definition. Euler function φ(n)=|(Z/nZ)|. (φ(1)=1)

It is not hard to see that φ(pe)=pepe1. Since in (Z/peZ)=Z/peZ{p,2p,3p,...,pe1p}.

Proposition. If gcd(m,n)=1, then φ(mn)=φ(m)φ(n).

It follows from Chinese Remainder Theorem directly. According to CRT, Z/mnZZ/mZ×Z/nZ.

Hence (Z/mnZ)(Z/mZ)×(Z/nZ). |(Z/mnZ)|=|(Z/mZ)×(Z/nZ)|=|(Z/mZ)|×|(Z/nZ)|

Proposition. φ1=idN+. i.e. d|nφ(d)=n

Proof.

φ1=d|nφ(d). For each d|n, (a,n)=d(a/d,n/d)=1.

Let Ad:={0an|gcd(a,n)=d}, then |Ad|=|{0a1n/d,(a1,n/d)=1}|=φ(n/d).

Each Ad is fiber of gcd(,n)f:Z/mZn. i.e. Ad=f1(d). Here n is the set of divisor of n.

Hence Ad form a partition of Z/mZ.

(6)d|n|Ad|=m=d|nφ(n/d)=d|nφ(d)=φ1(n)=idN+(n)

Corollary. φ=μidN.

i.e. φ(n)=d1d2=nμ(d1)d2=ninpi+i,jnp1p2+...=n(11p1)...(11ps).

Futher Properties

Definition multiplicative function and completely multiplicative function

Let f0.

If gcd(m,n)=1f(mn)=f(m)f(n), then f is multiplicative function.

If m,nN+,f(mn)=f(m)f(n), then f is completely multiplicative function.

Example. 1.

Proposition. f is multiplicative function f(1)=1.

Proof. f(n1)=f(n)f(1)

Proposition. The set of multiplicative function/completely multiplicative function is closed under convolution.

Proof.

(7)fg(m)fg(n)=d1|mf(d1)g(m/d1)d2|nf(d2)g(n/d2)=d1|m,d2|nf(d1d2)g(mn/d1d2)=fg(mn)

Proposition f(1)0f is invertible, and the inverse is unique.

If f is invertible, fg(1)=ε(1)=f(1)g(1)=1

If f(1)0,fg(n)=d1d2=nf(d1)g(d2)=0,f(1)g(n)=1<d1,d2<n,d1d2=nf(d1)g(d2)

(8)g(n)=1<d1,d2<n,d1d2=nf(d1)g(d2)f(1)

Corollary. Multiplicative function is invertible.

Proposition. Let f be a multiplicative function, then the inverse of f is multiplicative function as well.

Proof.

Obviously g(11)=g(1)g(1).

Let (a,b)=1

(9)(fg)(ab)=D|abf(D)g(ab/D)=d|a,δ|bf(dδ)g(ab/dδ)=ε(ab)
(10)(fg)(a)(fg)(b)=d|af(d)g(a/d)δ|bf(δ)g(b/δ)=d|a,δ|bf(dδ)g(a/d)g(b/δ)=ε(ab)

Define the function G(ab/dδ)=g(a/d)g(b/δ). Then fG=fg. By the uniqueness of inverse, G=g.

Hence g(a/d)g(b/δ)=g(ab/dδ). Let d=δ=1, g(a)g(b)=g(ab)

Corollary. f=g1 is multiplicative function iff g is multiplicative function.

Since 1 is multiplicative, hence μ is multiplicative as well.

Corollary. The multiplicative functions form a abelian group.

Corollary. (A,+,) is a local ring. Hence we can view A as stalk at the maximal ideal.

 

 

Popular Posts