Let $x_1,\ldots,x_n$ be real numbers such that $|x_1|\geq \ldots\geq |x_n|$ and $\displaystyle \sum_{i=k+1}^n |x_i| \leq \alpha \sum_{i=1}^k |x_i|$ where $1\leq k \leq n-1$ and $\alpha >0$.

Prove that $$\sum_{i=k+1}^n x_i^2 \leq \alpha \sum_{i=1}^k x_i^2$$

I have managed to prove a looser inequality: $$\begin{align} \alpha \sum_{i=1}^k x_i^2 &\geq \frac{\alpha}{k}\left(\sum_{i=1}^k |x_i| \right)^2 \quad \quad \text{Cauchy-Schwarz}\\ &\geq \frac 1k \sum_{i=k+1}^n |x_i| \sum_{i=1}^k |x_i|\\ &\geq \frac 1k \left(\sum_{i=k+1}^n |x_i| \right)^2\\ &\geq \frac 1k \sum_{i=k+1}^n x_i^2 \quad \quad \text{since } \|z\|_1 \geq \|z\|_2 \end{align}$$

Since $\frac 1k$ is not $\geq 1$, I'm stuck...


It is quite simple: Note that $|x_i| \le |x_k|$ for any $i \in \{k+1,\ldots,n\}$. Thus $$\tag{1}\sum_{i=k+1}^n x_i^2 \le |x_k| \sum_{i=k+1}^n |x_i| \le \alpha |x_k| \sum_{i=1}^k |x_i|,$$ where the given condition is used in the last step. Now $|x_k x_i| \le x_i^2$ for an $i=1,\ldots,k$, because $|x_k| \le |x_i|$ for all $i=1,\ldots,k$. So proceeding in (1) gives the claimed estimate $$\sum_{i=k+1}^n x_i^2\le\alpha \sum_{i=1}^k x_i^2.$$

  • $\begingroup$ Nice, thank you. This line of thought did not occur to me. $\endgroup$ – Issou Chankla Nov 21 '18 at 11:17
  • $\begingroup$ In general you lose information, if you apply Cauchy-Schwarz. The constants between $\|\cdot\|_1$ and $\|\cdot\|_2$ are also strict. In fact, the key point here is the condition $|x_1| \ge \ldots \ge |x_n|$. (The statement is wrong without this condition!) $\endgroup$ – p4sch Nov 21 '18 at 12:04

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.