Page 21.3: The Knesser Theorem

Time Scales Analysis · LibreTexts

0%

Contents

    Text

    Size
    100%
    Theme
    Typeface
    Line length
    Spacing

    From LibreTexts, under CC BY-NC-SA 12, as the source states it; its equations are written in TeX, which the PDF shows typeset. This page at LibreTexts.

    Page 21.3: The Knesser Theorem

    Suppose that \(\mathbb{T}\) is a time scale unbounded above with forward  jump operator  and delta differentiation operator \(\sigma\) and \(\Delta\), respectively.  For an \(\epsilon>0\) and a \(t_0\in \mathbb{T}\), we denote
    \[
    L_{\epsilon_0}(t_0)=\{t\in \mathbb{T}: 0<t_0-t<\epsilon\}.
    \notag\]

    Lemma

    Let \(m\in \mathbb{N}\) and \(f\) is \(m\)-times differentiable on \(\mathbb{T}\).

    1.  If
    \[
    \liminf_{t\to\infty} f^{\Delta^m}(t)>0,
    \notag\]
    then
    \[
    \lim_{t\to\infty} f^{\Delta^i}(t)=\infty,\quad i\in \{0, \ldots, m-1\}.
    \notag\]

    2.  If
    \[
    \limsup_{t\to\infty} f^{\Delta^m}(t)<0,
    \notag\]
    then
    \[
    \lim_{t\to\infty} f^{\Delta^i}(t)=-\infty,\quad i\in \{0, \ldots, m-1\}.
    \notag\]

    Proof

    1. Let
    \[
    0<\lim_{\delta\to 0}\inf_{t\in L_{\delta}(\infty)}f^{\Delta^m}(t)=\alpha.
    \notag\]
    Suppose that \(\alpha<\infty\).  Then there is an \(\epsilon>0\) such that
    \[
    f^{\Delta^m}(t)\geq \frac{\alpha}{2},\quad t\in L_{\epsilon}(\infty).
    \notag\]
    Thus
    \[
    \int_s^t f^{\Delta^m}(\tau) \Delta \tau\geq \int_s^t \frac{\alpha}{2}\Delta \tau,\quad t, s\in L_{\epsilon}(\infty),\quad s<t,
    \notag\]
    and
    \[
    f^{\Delta^{m-1}}(t)-f^{\Delta^{m-1}}(s)\geq \frac{\alpha}{2} (t-s),\quad t, s\in L_{\epsilon}(\infty).
    \notag\]
    Therefore
    \[
    \lim_{t\to\infty} f^{\Delta^{m-1}}(t)=\infty.
    \notag\]
    Now we suppose that \(\alpha=\infty\). Then for any \(c>0\) there exists an \(\epsilon>0\) such that
    \[
    f^{\Delta^m}(t)\geq \frac{1}{c},\quad t\in L_{\epsilon}(\infty),
    \notag\]
    and
    \[
    \int_s^t f^{\Delta^m}(\tau)\Delta\tau \geq \int_s^t \frac{1}{c}\Delta\tau,\quad t, s\in L_{\epsilon}(\infty),\quad s<t,
    \notag\]
    whereupon
    \[
    f^{\Delta^{m-1}}(t)-f^{\Delta^{m-1}}(s)\geq \frac{1}{c}(t-s),\quad t, s\in L_{\epsilon}(\infty),\quad s<t.
    \notag\]
    Therefore
    \[
    \lim_{t\to\infty} f^{\Delta^{m-1}}(t)=\infty.
    \notag\]
    By induction, we get
    \[
    \lim_{t\to\infty} f^{\Delta^i}(t)=\infty,\quad i\in \{0, \ldots, m-1\}.
    \notag\]

    2.  Let
    \[
    \beta=\lim_{\delta\to 0} \sup _{t\in L_{\delta}(\infty)}f^{\Delta^m}(t)<0.
    \notag\]
    Assume that \(\beta>-\infty\). Then there exists an \(\epsilon>0\) such that
    \[
    f^{\Delta^m}(t)\leq \frac{\beta}{2},\quad t\in L_{\epsilon}(\infty),
    \notag\]
    and
    \[
    \int_s^t f^{\Delta^m}(\tau)\Delta \tau\leq \int_s^t \frac{\beta}{2}\Delta \tau,\quad t, s\in L_{\epsilon}(\infty),\quad s<t.
    \notag\]
    Hence,
    \[
    f^{\Delta^{m-1}}(t)-f^{\Delta^{m-1}}(s)\leq \frac{\beta}{2}(t-s),\quad t, s\in L_{\epsilon}(\infty),\quad s<t,
    \notag\]
    and
    \[
    \lim_{t\to\infty}f^{\Delta^{m-1}}(t)=-\infty.
    \notag\]
    Assume that \(\beta=-\infty\). Then for any \(c_1<0\) there is an \(\epsilon>0\) such that
    \[
    f^{\Delta^m}(t)\leq \frac{1}{c_1},\quad t\in L_{\epsilon}(\infty),
    \notag\]
    and
    \[
    \int_s^t f^{\Delta^m}(\tau)\Delta \tau\leq \int_s^t \frac{1}{c_1}\Delta \tau,\quad t, s\in L_{\epsilon}(\infty),\quad s<t.
    \notag\]
    Hence,
    \[
    f^{\Delta^{m-1}}(t)-f^{\Delta^{m-1}}(s)\leq \frac{1}{c_1}(t-s),\quad t, s\in L_{\epsilon}(\infty),\quad s<t,
    \notag\]
    and
    \[
    \lim_{t\to\infty}f^{\Delta^{m-1}}(t)=-\infty.
    \notag\]
    By induction, we get
    \[
    \lim_{t\to\infty} f^{\Delta^i}(t)=-\infty,\quad i\in \{0, \ldots, m-1\}.
    \notag\]
    This completes the proof.

    Theorem (The Knesser Theorem)

    Let \(n\in \mathbb{N}\) and \(f\) be \(n\) times differentiable on \(\mathbb{T}\). Suppose that there exists an \(\epsilon>0\) such that
    \[
    f(t)>0,\quad \textrm{sgn}\left(f^{\Delta^n}(t)\right)=s\in \{-1, 1\},\quad t\in L_{\epsilon}(\infty),
    \notag\]
    and \(f^{\Delta^n}(t)\ne 0\), \(t\in L_{\delta}(\infty)\), for any \(\delta>0\). Then there exists a \(p\in [0, n]\bigcap \mathbb{N}_0\) such that \(n+p\)  is even for \(s=1\) and odd for \(s=-1\) with
    \[\begin{eqnarray*}
    (-1)^{p+j} f^{\Delta^j}(t)&>& 0,\quad t\in L_{\epsilon}(\infty),\quad j\in [p, n-1]\bigcap \mathbb{N}_0,\\ \\
    f^{\Delta^j}(t)&>& 0,\quad t\in L_{\delta_j}(\infty),\quad \delta_j\in (0, \epsilon),\quad j\in [1, p-1]\bigcap \mathbb{N}_0.
    \end{eqnarray*}\notag\]

    Proof

    1.  Let \(s=-1\).  Assume that \(f^{\Delta^{n-1}}(t_1)\leq 0\) for some \(t_1\in L_{\epsilon}(\infty)\). Then there exists a \(t_2\in L_{\epsilon}(\infty)\), \(t_2>t_1\), such that
        \[
        f^{\Delta^{n-1}}(t)\leq f^{\Delta^{n-1}}(t_2)< f^{\Delta^{n-1}}(t_1),\quad t\in L_{\epsilon}(\infty), \quad t>t_2.
        \notag\]
        Hence,
        \[
        \limsup_{t\to\infty} f^{\Delta^{n-1}}(t)<0
        \notag\]
        and by the above lemma, we conclude that
        \[
        \lim_{t\to\infty} f(t)=-\infty.
        \notag\]
        This ia a contradiction. Therefore
        \[
        f^{\Delta^{n-1}}(t)>0,\quad t\in L_{\epsilon}(\infty).
        \notag\]
        Define
        \[\begin{eqnarray*}
        M&=& \bigg\{ m\in \mathbb{N}_0: (-1)^{m+j}f^{\Delta^j}(t)>0,\quad t\in L_{\epsilon}(\infty),\\ \\
        && j\in (m, n-1]\bigcap \mathbb{N}_0,\quad n+m\quad \textrm{odd}\bigg\}.
        \end{eqnarray*}\notag\]
        Then \(n-1\in M\) and there exists \(p=\min M\). If \(p\in \{0, 1\}\) we are done. Assume that \(p\geq 2\). Suppose that \(f^{\Delta^{p-1}}(t)<0\), \(t\in L_{\epsilon}(\infty)\). Suppose that \(f^{\Delta^{p-2}}(t_1)\leq 0\) for some \(t_1\in L_{\epsilon}(\infty)\). Then there exists a \(t_2\in L_{\epsilon}(\infty)\), \(t_2>t_1\), such that
        \[
        f^{\Delta^{p-2}}(t)\leq f^{\Delta^{p-2}}(t_2)< f^{\Delta^{p-2}}(t_1),\quad t\in L_{\epsilon}(\infty),\quad t>t_2.
        \notag\]
        Hence and the above lemma, we obtain that
        \[
        \lim_{t\to\infty} f(t)=-\infty.
        \notag\]
        This is a contradiction. Therefore \(f^{\Delta^{p-2}}(t)>0\), \(t\in L_{\epsilon}(\infty)\). Also,
        we have that
        \[
        (-1)^{p-2+j} f^{\Delta^j}(t)>0,\quad t\in L_{\epsilon}(\infty),\quad j\in [p-2, n-1]\bigcap \mathbb{N}_0,
        \notag\]
        which is a contradiction by the definition of \(p\). Therefore there exists a \(t_3\in L_{\epsilon}(\infty)\) so that
        \[
        f^{\Delta^{p-1}}(t_3)\geq 0.
        \notag\]
        Because \(f^{\Delta^p}(t)>0\), \(t\in L_{\epsilon}(\infty)\), we conclude that
        \[
        f^{\Delta^{p-1}}(t)>0,\quad t\in L_{\epsilon}(\infty),\quad t>t_3.
        \notag\]
        Hence,
        \[
        \liminf_{t\to\infty}f^{\Delta^{p-1}}(t)>0
        \notag\]
        and by the above lemmaLemma, we obtain
        \[
        \lim_{t\to\infty} f^{\Delta^j}(t)=\infty,\quad 0\leq j\leq p-2.
        \notag\]
       

    2.  Let \(s=1\). Assume that there exists a \(t_4\in L_{\epsilon}(\infty)\) such that \(f^{\Delta^{n-1}}(t_4)\geq 0\). Then there exists a \(t_5\in [t_4, \infty)\), \(t_5>t_4\), such that
            \[
            f^{\Delta^{n-1}}(t)\geq f^{\Delta^{n-1}}(t_5)> f^{\Delta^{n-1}}(t_4),\quad t\in L_{\epsilon}(\infty),\quad t>t_5.
            \notag\]
            Hence,
            \[
            \liminf_{t\to\infty} f^{\Delta^{n-1}}(t)>0
            \notag\]
            and by the above lemma, we conclude that
            \[
            \liminf_{t\to\infty}f^{\Delta^j}(t)=\infty,\quad j\in \{1, \ldots, n-2\}.
            \notag\]
            So, \(p=n\). This completes the proof.
     


    This page titled Page 21.3: The Knesser Theorem was last modified on Mon, 26 Jan 2026 18:49:57 GMT and is shared under a CC BY-NC-SA 12 license and was authored, remixed, and/or curated by Svetlin G. Georgiev.