2
$\begingroup$

Suppose $a_0\geq\dots \geq a_n \geq 0$ is a sequence of non-negative numbers, where $n+1\leq \sum_{k=0}^{n} a_k \leq n+2$. Then, I want to prove that the following statement is true,

$2\Big[ \sum_{k=0}^{n} (k+1) a_k \Big]^2 -\Big[1+ \sum_{k=0}^{n} a_k \Big]\Big[ \sum_{k=0}^{n} k(k+1)a_k \Big] \geq 0.$

$\endgroup$
7
  • 1
    $\begingroup$ Do not use $N$ where $n$ will do. I have edited accordingly. $\endgroup$ Commented Jan 24 at 3:28
  • 3
    $\begingroup$ @IosifPinelis Why? $\endgroup$ Commented Jan 24 at 3:49
  • 1
    $\begingroup$ @IosifPinelis - you introduced a typo into the title ... $\endgroup$ Commented Jan 24 at 4:06
  • 1
    $\begingroup$ @mathworker21 : (i) There is no reason to use $N$ where $n$ will do. (ii) For each entry of $n$ instead of $N$, one keystroke is saved. $\endgroup$ Commented Jan 24 at 4:09
  • 1
    $\begingroup$ @MichaelEngelhardt : Fixed it. $\endgroup$ Commented Jan 24 at 4:10

1 Answer 1

6
$\begingroup$

A counterexample is given by the following conditions: $n=185$, $$a_k=\sum_{j=k}^n b_j,\quad b_j:=\frac{c_j}{j+1}, \quad c_j:=\frac{1000}7\,1(j=47)+\frac{302}7\,1(j=185).$$


Indeed, then $a_0\ge\cdots\ge a_n\ge0$, $$s_0:=\sum_{k=0}^n a_k=\sum_{k=0}^n \sum_{j=k}^n b_j =\sum_{j=0}^n b_j\sum_{k=0}^j 1 \\ =\sum_{j=0}^n b_j(j+1) =\sum_{j=0}^n c_j=\frac{1000}7+\frac{302}7=186\in[n+1,n+2],$$ $$s_1:=\sum_{k=0}^n(k+1)a_k=\sum_{k=0}^n(k+1)\sum_{j=k}^n b_j =\sum_{j=0}^n b_j\sum_{k=0}^j(k+1) \\ =\sum_{j=0}^n b_j\frac{(j+1)(j+2)}2 =\sum_{j=0}^n c_j\frac{j+2}2=\frac{105474}7,$$ $$s_2:=\sum_{k=0}^n k(k+1)a_k=\sum_{k=0}^n k(k+1)\sum_{j=k}^n b_j =\sum_{j=0}^n b_j\sum_{k=0}^j k(k+1) \\ =\sum_{j=0}^n b_j\frac{j(j+1)(j+2)}3 =\sum_{j=0}^n c_j\frac{j(j+2)}3=\frac{4250230}7,$$ so that the left-hand side of your inequality is $$2s_1^2-(1+s_0)s_2=-\frac{1168732}{49}<0.\quad\Box$$

$\endgroup$
1
  • $\begingroup$ Hey. Thank you so much for the counter example. I was trying to prove convexity of a function and the inequality above comes out as the part of the second derivative. Probably the convexity argument is not true then. I was trying to see if i can establish some more conditions on my actual function that can help me get rid of the counter examples like above. But, probably that is not prossible. $\endgroup$ Commented Jan 29 at 18:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .