On some nested floor functions and their jump discontinuities

It is sufficient to prove that \forall k\in\mathbb{Z^{+}}\setminus\{1\} and \forall n\in\mathbb{Z^{+}} \exists\delta_n=\frac{k-1}{(k-2)k^{n-1}+1} such that

The base case is when n=1. We have that:

and this identity is true because k is a positive integer.
Suppose that this identity holds for n and we’ll prove it right for n+1.
Since k^{n}>k^{n-1} \forall n\in\mathbb{Z^{+}}\forall k\in\mathbb{Z^{+}}\setminus\{1\}, consider:

we want to show that:

but since \delta_{n+1}<\delta_n we have that:

Furthermore:

This is true in fact:

So:

And dividing all by \frac{1}{k-1}, we’ll get:

Call n_1 the left hand side of the inequality and n_2 the right hand side. We’ll prove that they are in fact natural numbers in the next section. Note that:

So the number \frac{1}{k-1}\Bigl[(k-2)k^{n}+1\Bigr]x is strictly between n_1,n_2 ,which are positive integers whose distance between each others is equal to 1. So the floor of

must be equal to n_1 (which is the nearest integer less than or equal to
\frac{1}{k-1}\Bigl[(k-2)k^{n}+1\Bigr]x).
Let \delta_n=\frac{k-1}{(k-2)k^{n-1}+1}. We want to show that:

But now this is obvious because:

And finally \forall\varepsilon>0:

And the thesis follows from the limit definition.

Note that:

So the numerator of n_1 is a multiple of (k-1). Since the numerator of n_2 is equal to the numerator of n_2 plus k-1 we conclude that also n_2 is an integer (because its numerator is again a multiple of k-1).

We’ll prove this result by induction on n.
The base case (n=2) has been already proved before. So we know that
P(f_1)\subset P(f_2).
Suppose that P(f_1)\subset P(f_2)\subset \dots \subset P(f_{n-1}). We’ll prove that:

Let d be an element of P_{n-1}; consider the following limits:

From the induction hyphothesis we know that:

So \exists\delta^{-},\delta^{+}>0 such that:

But from the definition of f_n we have that:

So:

But f_{n-1}=L_d^{-},\forall x\in (d-\delta^{-},d) and f_{n-1}=L_d^{+},\forall x\in (d,d+\delta^{+}), so substituing in we will obtein:

This last step is motivated by using the limit definition.
But we know that L_d^{-},L_d^{+}\in\mathbb{Z^{+}} , and since f_n is monotone increasing and L_d^{-}\not= L_d^{+} we can conclude that:

So:

But using \delta=\min\{\delta^{-},\delta^{+}\} we have:

And finally we have that kx in this interval is bigger than kd (which is a positive rational number bigger than or equal to 1). So:

and combining all the inequalities we’ll get:

Since from Theorem 5 we know that:

it’s sufficient to show that \forall n\geq 2, \exists d\in P(k,k+1,f_n)\setminus P(k,k+1,f_{n-1}). In fact we’ll show that:

First we want to prove that:

Or, using the definition of this set:

In fact we’ll prove by induction on n\geq 2 that using \delta=\frac{1}{k^{n}}:

Assuming that true we will have:

which is our thesis.
By the induction hypothesis:

But using the definition of f_{n+1} with \delta=\frac{1}{k^{n}}<\frac{1}{k^{n-1}} we’ll have:

But furthermore:

And finally:

Now we know that k+\frac{1}{k^{n}}\in P(k,k+1,f_{n+1}), and we want to prove that k+\frac{1}{k^{n}}\not\in P(k,k+1,f_{n}). This is true in fact:

In order to prove it, as seen before, we know that:

But then:

\forall k\geq 2, \forall n\geq 1. So we’ll have:

Where P(k,k+1,f_{n+1}) has n terms of the form k+\frac{1}{k^{r}}, where r\in\{1,\dots,n\}.