Skip to main content

Section 2.3 Dirichlet product

Subsection 2.3.1 Basic properties of Dirichlet product

Definition 2.3.1.

Let \(f,g\) be arithmetic functions. The Dirichlet product of \(f\) and \(g\text{,}\) denoted \(f * g\text{,}\) is the arithmetic function defined by
\begin{equation} (f * g)(n) = \sum_{d \mid n} f(d) g(n/d) = \sum_{de = n} f(d) g(e) \tag{2.3.1} \end{equation}
where the first sum is over positive divisors of \(n\) and the second sum is over pairs \((d,e)\) of positive integers whose product is \(n\text{.}\)

Example 2.3.2.

\(\phi * 1 = N \text{:}\) Here \(\phi\) is the Euler totient function, \(1\) is the function \(1(n) = 1\) for all \(n\text{,}\) and \(N\) is the function \(N(n) = n\) for all \(n\text{.}\) We have shown that \(\sum_{d \mid n} \phi(d) = n\text{.}\) We can rewrite this as \(\sum_{d \mid n} \phi(d) 1(n/d) = N(n)\text{,}\) where the left hand side is \((\phi * 1)(n)\text{.}\) This shows that \(\phi * 1 = N\) as claimed.
Other notable operations on arithmetic functions include:
  1. The sum: \((f+g)(n) = f(n)+g(n)\)
  2. The pointwise product: \((fg)(n) = f(n) g(n)\)
  3. The composition: \((f \circ g)(n) = f(g(n))\) (assuming \(g\) has values in \(\bbN\text{,}\) the domain of \(f\))
  4. The Cauchy product, defined by \(\sum_{d=0}^n f(d)g(n-d) \text{.}\) (I’m not sure if there is a common notation for the Cauchy product.)

Proof.

For associativity:
\begin{align*} (f*(g*h))(n) \amp = \sum_{d_1 e = n} f(d_1) (g*h)(e) \\ \amp = \sum_{d_1 e = n} f(d_1) \sum_{d_2 d_3 = e} g(d_2) h(d_3) \\ \amp = \sum_{d_1 d_2 d_3 = n} f(d_1) g(d_2) h(d_3) \\ \amp = \sum_{c d_3 = n} \left( \sum_{d_1 d_2 = c} f(d_1) g(d_2) \right) h(d_3) \\ \amp = \sum_{c d_3 = n} (f*g)(c) h(d_3) \\ \amp = ((f*g)*h)(n) \end{align*}
For commutativity:
\begin{align*} (f*g)(n) \amp = \sum_{de = n} f(d) g(e) \\ \amp = \sum_{ed = n} g(e) f(d) \\ \amp = (g*f)(n) \end{align*}
Distributivity is left as an exercise.
Let \(f\) be any arithmetic function. Then
\begin{equation} (f*e)(n) = \sum_{d \mid n} f(d) e(n/d) = f(n) e(n/n) = f(n) \text{,}\tag{2.3.2} \end{equation}
since \(e(n/d) = 0\) for all \(d\) other than \(d=n\text{.}\) So \(f*e = f\text{.}\)

Proof.

If \(f*g = e\) then \(f(1)g(1) = e(1) = 1\text{,}\) so \(f(1) \neq 0\text{.}\) The converse direction is left as an exercise.

Subsection 2.3.2 Dirichlet product of multiplicative functions

Proof.

Suppose \((n_1,n_2) = 1\text{.}\) Observe that if \(de = n_1 n_2\) then there exist factorizations \(d = d_1 d_2\text{,}\) \(e = e_1 e_2\text{,}\) such that \(d_1 e_1 = n_1\) and \(d_2 e_2 = n_2\text{.}\) (Simply, \(d_1\) is the product of prime factors that are in \(d\) and \(n_1\text{,}\) \(d_2\) is the product of prime factors that are in \(d\) and \(n_2\text{,}\) and so on.) Observe that in this factorization, \((d_1,d_2) = (e_1,e_2) = 1\text{.}\)
Now we have
\begin{align*} (f*g)(n_1 n_2) \amp = \sum_{de = n_1 n_2} f(d) g(e) \\ \amp = \sum_{d_1 d_2 e_1 e_2 = n_1 n_2} f(d_1 d_2) g(e_1 e_2) \\ \amp = \sum_{d_1 e_1 = n_1} \sum_{d_2 e_2 = n_2} f(d_1) f(d_2) g(e_1) g(e_2) \\ \amp = \left( \sum_{d_1 e_1 = n_1} f(d_1) g(e_1) \right) \left( \sum_{d_2 e_2 = n_2} f(d_2) g(e_2) \right) \\ \amp = (f*g)(n_1) (f*g)(n_2) \text{.} \end{align*}
This shows that \(f*g\) is multiplicative.

Proof.

Let \(g(n) = n^s\text{,}\) so \(g\) is multiplicative (in fact, completely multiplicative). Then \(\sigma_s = g * 1\) where \(1\) denotes the arithmetic function \(1(n) = 1\) for all \(n\text{.}\) Here \(1\) is completely multiplicative. Since \(g\) and \(1\) are multiplicative, so is \(\sigma_s\text{.}\)

Remark 2.3.7.

Even if both \(f\) and \(g\) are completely multiplicative, in general \(f*g\) is only multiplicative, not necessarily completely multiplicative. For example let \(f = g = 1\text{,}\) then \((f*g)(n) = \sum_{d \mid n} 1 = d(n) = \sigma_0(n)\text{,}\) the number of positive divisors of \(n\text{.}\) This is multiplicative but not completely multiplicative, for example \(d(4) = 3 \neq d(2)^2\text{.}\)

Proof.

The left hand side is \((f * 1)(n)\text{.}\) Since \(f\) and \(1\) are multiplicative, then so is \(f * 1\text{.}\) Let \(n\) have prime factorization \(n = p_1^{a_1} \dotsb p_k^{a_k}\text{.}\) It follows that
\begin{align*} \sum_{d \mid n} f(d) \amp = (f * 1)(n) \\ \amp = \prod_{j=1}^k (f * 1)(p_j^{a_j}) \\ \amp = \prod_{j=1}^k \sum_{d \mid p_j^{a_j}} f(d) \\ \amp = \prod_{j=1}^k (f(1) + f(p_j) + f(p_j^2) + \dotsb + f(p_j^{a_j})) \text{.} \end{align*}
Finally note \(f(1) = 1\text{.}\)

Proof.

First of all we have \((f*g)(1) = 1\) and \(g(1) = 1\) since \(f*g\) and \(g\)are multiplicative. By definition of Dirichlet product, \((f*g)(1) = f(1)g(1)\text{.}\) Therefore \(f(1) = 1\text{.}\) We have \(f(1 \cdot 1) = 1\text{,}\) that is, \(f(mn) = 1\) for \(m=n=1\text{.}\)
We will prove the following statement by induction on \(N\text{:}\) \(f(mn) = f(m)f(n)\) for all \(m,n\) such that \((m,n)=1\) and \(mn \leq N\text{.}\) We have seen that this holds for \(N=1\text{.}\) Suppose inductively that it holds for \(N-1\text{,}\) and suppose \(m,n\) are positive integers such that \((m,n)=1\) and \(mn=N\text{.}\) Since \(f*g\) is multiplicative we have
\begin{align*} (f*g)(mn) \amp = \sum_{d \mid mn} f(d) g(\frac{mn}{d}) \\ \amp = \sum_{d \mid m, e \mid n} f(de) g(\frac{mn}{de}) \\ \amp = \sum_{d \mid m, e \mid n, de \lt mn} f(de) g(\frac{mn}{de}) + f(mn)g(1) \\ \amp = \sum_{d \mid m, e \mid n, de \lt mn} f(d)f(e)g(\frac{m}{d})g(\frac{n}{d}) + f(mn) \\ \amp = \left( \sum_{d \mid m} f(d)g(\frac{m}{d}) \right) \left( \sum_{e \mid n} f(e)g(\frac{n}{e}) \right) - f(m)f(n) + f(mn) \\ \amp = (f*g)(m) (f*g)(n) - f(m)f(n) + f(mn) \\ \amp = (f*g)(mn) - f(m)f(n) + f(mn) \text{.} \end{align*}
Here we use that \(f(de) = f(d)f(e)\) for \(de \lt mn\text{,}\) where \((d,e)=1\) since \(d \mid m\) and \(e \mid n\text{;}\) we use that \((m/d,n/e) = 1\) and that \(g\) is multiplicative; at the end we use that \(f*g\) is multiplicative.
It follows that \(f(mn) = f(m)f(n)\text{.}\)

Proof.

Both \(f\) and \(f*f^{-1} = e\) are multiplicative.

Subsection 2.3.3 Explanation of Dirichlet product

Definition 2.3.12.

A power series centered at \(c\) is an infinite sum
\begin{equation} f(z) = \sum_{k=0}^{\infty} a_k (z-c)^k \text{.}\tag{2.3.5} \end{equation}
The \(a_k\) are the coefficients of the power series. This sum defines a function of \(z\) provided that it converges.
A Dirichlet series is an infinite sum
\begin{equation} f(s) = \sum_{n=1}^{\infty} \frac{a_n}{n^s} \text{.}\tag{2.3.6} \end{equation}
Again, the \(a_n\) are coefficients of the Dirichlet series. The sum defines a function of \(s\) provided that it converges.
Multiplication of power series corresponds to the Cauchy product of coefficients, and multiplication of Dirichlet series corresponds to the Dirichlet product of coefficients.
Let us multiply two power series:
\begin{align*} (a_0 + a_1 z + \amp a_2 z^2 + a_3 z^3 + \dotsb) (b_0 + b_1 z + b_2 z^2 + b_3 z^3 + \dotsb)\\ \amp = (a_0 b_0) + (a_0 b_1 + a_1 b_0) z + (a_0 b_2 + a_1 b_1 + a_2 b_0) z^2\\ \amp \quad + (a_0 b_3 + a_1 b_2 + a_2 b_1 + a_3 b_0) z^3 + \dotsb \end{align*}
That is,
\begin{equation} \left( \sum a_i z^i \right) \left( \sum b_j z^j \right) = \sum_{i,j} a_i b_j z^{i+j} = \sum_k \left( \sum_{i+j = k} a_i b_j \right) z^k\tag{2.3.7} \end{equation}
This shows that the product of power series with sequences of coefficients \((a_i)\) and \((b_j)\) is given by the power series with coefficients given by the Cauchy product of \((a_i)\) and \((b_j)\text{.}\)
Similarly, the Dirichlet product corresponds to multiplication of Dirichlet series. Let us multiply two Dirichlet series:
\begin{align*} \Big( \frac{a_1}{1^s} + \frac{a_2}{2^s} + \amp \frac{a_3}{3^s} + \frac{a_4}{4^s} + \dotsb \Big) \Big( \frac{b_1}{1^s} + \frac{b_2}{2^s} + \frac{b_3}{3^s} + \frac{b_4}{4^s} + \dotsb \Big)\\ \amp = \frac{a_1 b_1}{1^s 1^s} + \frac{a_1 b_2}{1^s 2^s} + \frac{a_1 b_3}{1^s 3^s} + \frac{a_1 b_4}{1^s 4^s} + \dotsb\\ \amp \quad + \frac{a_2 b_1}{2^s 1^s} + \frac{a_2 b_2}{2^s 2^s} + \frac{a_2 b_3}{2^s 3^s} + \frac{a_2 b_4}{2^s 4^s} + \dotsb\\ \amp \quad + \frac{a_3 b_1}{3^s 1^s} + \frac{a_3 b_2}{3^s 2^s} + \frac{a_3 b_3}{3^s 3^s} + \frac{a_3 b_4}{3^s 4^s} + \dotsb\\ \amp = \frac{a_1 b_1}{1^s} + \frac{a_1 b_2}{2^s} + \frac{a_1 b_3}{3^s} + \frac{a_1 b_4}{4^s} + \dotsb\\ \amp \quad + \frac{a_2 b_1}{2^s} + \frac{a_2 b_2}{4^s} + \frac{a_2 b_3}{6^s} + \frac{a_2 b_4}{8^s} + \dotsb\\ \amp \quad + \frac{a_3 b_1}{3^s} + \frac{a_3 b_2}{6^s} + \frac{a_3 b_3}{9^s} + \frac{a_3 b_4}{12^s} + \dotsb\\ \amp = \frac{a_1 b_1}{1^s} + \frac{a_1 b_2 + a_2 b_1}{2^s} + \frac{a_1 b_3 + a_3 b_1}{3^s} + \frac{a_1 b_4 + a_2 b_2 + a_4 b_1}{4^s} + \dotsb \end{align*}
In general,
\begin{equation} \left( \sum \frac{a_m}{m^s} \right) \left( \sum \frac{b_n}{n^s} \right) = \sum_{m,n} \frac{a_m b_n}{(mn)^s} = \sum_k \frac{ \sum_{mn = k} a_m b_n }{ k^s } = \sum_k \frac{(a*b)_k}{k^s}\tag{2.3.8} \end{equation}