$\def\scrBp{\mathscr B\boldsymbol.}\def\rD{{\rm D}\kern.4mm}\def\ssp{\kern.4mm}\def\bbR{\mathbb R} $For a certain purpose I invented the Banach space $\scrBp^\infty(\ssp\bbR\ssp)$ defined as follows. The vectors are all smooth functions $x:\bbR\to\bbR$ such that there is $M\in\bbR^+$ with $|\,\rD^kx(t)\,|\le M$ and $\lim_{\,s\to\pm\infty\,}|\,\rD^kx(s)\,|=0$ for all $k\in\mathbb N_0$ and $t\in\bbR$ . The norm of $x$ is the infimum of the set of all these $M$ . Now I am wondering whether this space has any nonzero vectors. Hence the question.

  • 1
    $\begingroup$ I doubt so, since neither $e^{-x^2}$-like functions nor functions from $C_c^\infty$ will be in there (the uniform constant breaks these) as far as I can tell. $\endgroup$ – AlexR Nov 6 '14 at 15:58
  • $\begingroup$ Can a continuous and differentiable function approach some finite limit as $x\to\infty$, without its derivative approaching $0$? $\endgroup$ – Nishant Nov 6 '14 at 16:13
  • 2
    $\begingroup$ @Nishant An example is $t\mapsto t^{-1}\,\sin(t^2)$ . $\endgroup$ – The-unKnowN Nov 6 '14 at 16:18
  • 3
    $\begingroup$ A smooth function with uniformly bounded derivatives is real-analytic, and its Taylor series converges on all of $\mathbb{C}$, so it would be the restriction of an entire function to the real axis. I'm almost sure that means your space is $\{0\}$, but I don't see a proof yet. $\endgroup$ – Daniel Fischer Nov 6 '14 at 16:18
  • $\begingroup$ In view of Daniel Fischer's comment above, I added the "complex-analysis" tag. $\endgroup$ – The-unKnowN Nov 8 '14 at 5:04

It turns out that there are a lot of these functions. To see this, one possibility is to use the theory of the Fourier transform:

Choose some $\varphi\in C_{c}^{\infty}\left(\left(-\frac{1}{2\pi},\frac{1}{2\pi}\right)\right)$ and let $f:=\mathcal{F}\varphi=\widehat{\varphi}$ be the Fourier transform of $\varphi$, i.e. $$ f\left(\xi\right)=\int_{\mathbb{R}}\varphi\left(x\right)\cdot e^{-2\pi ix\xi}\,{\rm d}x. $$ By the usual properties of the Fourier transform, we conclude that $f\in\mathcal{S}\left(\mathbb{R}\right)$ is a Schwartz-function, which means $$ C_{m,n}:=\sup_{x\in\mathbb{R}}\left|x^{m}\cdot\partial^{n}f\left(x\right)\right|<\infty $$ for all $m,n\in\mathbb{N}_{0}$. In particular, this implies (for $\left|x\right|\geq1$): $$ \left|\partial^{n}f\left(x\right)\right|=\left|\frac{x\cdot\partial^{n}f\left(x\right)}{x}\right|\leq\frac{C_{1,n}}{\left|x\right|}\xrightarrow[x\to\pm\infty]{}0 $$ so that $f$ fulfills the second one of your requirements.

By choosing (e.g.) $\varphi\geq0$ and $\varphi\not\equiv0$, we can also ensure that $f\left(0\right)=\int_{\mathbb{R}}\varphi\left(x\right)\,{\rm d}x>0$, i.e. $f\not\equiv0$ (alternatively, we have $\varphi=\mathcal{F}^{-1}f$, so that $f\not\equiv0$ holds as soon as $\varphi\not\equiv0$ is true).

Finally, we can differentiate "under the integral sign" to get \begin{eqnarray*} \left|\partial^{n}f\left(\xi\right)\right| & = & \left|\int_{\mathbb{R}}\varphi\left(x\right)\cdot\frac{{\rm d}^{n}}{{\rm d}\xi^{n}}e^{-2\pi ix\xi}\,{\rm d}x\right|\\ & = & \left|\int_{\mathbb{R}}\varphi\left(x\right)\cdot\left(-2\pi ix\right)^{n}\cdot e^{-2\pi ix\xi}\,{\rm d}x\right|\\ & \leq & \int_{\mathbb{R}}\left|\varphi\left(x\right)\right|\cdot\left|2\pi x\right|^{n}\,{\rm d}x\\ & \overset{{\rm supp}\left(\varphi\right)\subset\left(-\frac{1}{2\pi},\frac{1}{2\pi}\right)}{\leq} & \int_{-\frac{1}{2\pi}}^{\frac{1}{2\pi}}\left|\varphi\left(x\right)\right|\cdot\left|2\pi x\right|^{n}\,{\rm d}x\\ & \leq & \int_{-\frac{1}{2\pi}}^{\frac{1}{2\pi}}\left|\varphi\left(x\right)\right|\,{\rm d}x\leq\left\Vert \varphi\right\Vert _{L^{1}}, \end{eqnarray*} so that the derivatives of $f$ are also uniformly bounded, which is the first of your desired properties.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.