Take $V=C[0,1]$ with the usual $L^2$ norm, i.e., $\|f\|_2^2 = \int_{0}^{1}|f(\tau)|^2d\tau$. Is it complete?

Consider the following sequence of functions:

$$f_n(t) = \begin{cases} 0 & t < \frac{1}{2} - \frac{1}{n+1} \\ [t-(\frac{1}{2}- \frac{1}{n+1})](n+1) & \frac{1}{2} - \frac{1}{n+1}\leq t \leq \frac{1}{2} \\ 1 & t \geq \frac{1}{2} \end{cases}$$ The sequence is Cauchy since $\|f_m-f_n\|_2^2 = \frac{1}{m+1}$, but obviously $f_0 = \lim_{n \rightarrow \infty}f_n \notin C[0,1]$.

Now do the same with the usual $L^\infty$ norm, i.e., $\|f\|_\infty = sup_{t \in [0,1]}|f(t)|$. I know that this space is complete, therefore I expect that the sequence $f_n$ is not a Cauchy sequence with the norm $\|\cdot\|_\infty$.

Let $m \leq n$, then:

$$\|f_n-f_m\|_\infty = sup_{t \in [\frac{1}{2} - \frac{1}{m+1},\frac{1}{2}]}\big(\frac{1}{2}-t\big)\big(n-m\big).$$

The question is: ''Is this a Cauchy sequence?''

I think no, because using the $\epsilon$ definition one can argue by contradiction such as: Assume $\forall \epsilon > 0 \; \exists \; N_\epsilon > 0$ s.t. $\|f_n-f_m\|_\infty < \epsilon$. Fix $m = \hat{m} > N_\epsilon$, then $sup_{t \in [\frac{1}{2} - \frac{1}{m+1},\frac{1}{2}]}\big(\frac{1}{2}-t\big)\big(n-m\big)$ depends only on $n$ and call $\delta = \big(\frac{1}{2} -t\big)$. Now take $n = \hat{n} > \frac{\delta\hat{m}+\epsilon}{\delta} > \hat{m} > N_{\epsilon}$, it follows that: $$\|f_{\hat{n}}-f_{\hat{m}}\|_\infty > \epsilon, \; \hat{m}, \hat{n} > N_{\epsilon}$$

Is the reasoning correct?


Your Answer

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

Browse other questions tagged or ask your own question.