On Continuous Replicative Functions

Knuth, in The Art of Computer Programming Vol. 1, defines a replicative function as a function $f$ such that $$f(x)+f\left(x+\frac{1}{n}\right)+\cdots +f\left(x+\frac{n-1}{n}\right)=f(nx)$$ whenever $n\in\mathbf{Z}^{+}$. For $f(x)=\lfloor x\rfloor$, this is nothing but Hermite's identity. Knuth gives a few examples of replicative functions including the following.

  • $f(x)=x-\frac{1}{2}$
  • $f(x)=\begin{cases}1 && \text{if }x\in\mathbf{Z} \\ 0 && \text{otherwise}\end{cases}$
  • $f(x)=\begin{cases}1 && \text{if }x\in\mathbf{Z}^{+} \\ 0 && \text{otherwise}\end{cases}$
  • $f(x) = 1$ if there exists $r\in\mathbf{Q}$ and $m\in\mathbf{Z}$ such that $x=r\pi+m$, $f(x) = 0$ otherwise.
  • The three other functions as above with $r$ and/or $m$ restricted to positive values.
  • $f(x) = \log|2\sin(\pi x)|$, if the value $f(x)=-\infty$ is allowed.
  • the sum of any two replicative functions.
  • A constant multiple of a replicative functions.
  • The function $g(x)=f(x-\lfloor x\rfloor)$, where $f$ is replicative.

Knuth then proposes to study the class of continuous replicative functions (the only one listed above is $f(x)=x-\frac{1}{2}$). However, as far as I can see, there seems to be a dearth of literature on this topic; the only piece I am aware of is here, but is not freely accessible.

Knuth also proposes to study the more general class of functions such that $$f(x)+f\left(x+\frac{1}{n}\right)+\cdots +f\left(x+\frac{n-1}{n}\right)=a_nf(nx)+b_n,$$ where $a_n$ and $b_n$ depend on $n$ but not $x$.

Is anyone aware of more information regarding this topic? Any help is greatly appreciated.


The Bernoulli polynomials (the first of which is Knuth's example $f(x) = x-1/2$) satisfy the identity:

$$B_n(mx) = m^{n-1}\sum_{k=0}^{m-1}B_n(x+k/m)$$

This identity is usually interpreted as meaning that the Bernoulli polynomials form a distribution on the additive group $\widehat{\mathbf Z}$ of profinite integers. These distributions play an important role in number theory. The Bernoulli distribution, in a certain sense, is an avatar of the Riemann zeta function.

Another nice example is the Gamma distribution arising from the Gauss multiplication formula:

$$\Gamma(z) \; \Gamma\left(z + \frac{1}{k}\right) \; \Gamma\left(z + \frac{2}{k}\right) \cdots \Gamma\left(z + \frac{k-1}{k}\right) = (2 \pi)^{ \frac{k-1}{2}} \; k^{1/2 - kz} \; \Gamma(kz) \,\!$$

If you'd like to take a look in this direction, you can find some information in Lang's book on cyclotomic fields (where he gives a number of other examples of distributions). Perhaps you can salvage something from the relevant sections.