2

Let: $x,y,z \ge 0$ and $xy+yz+zx=1$. Prove that: $$\frac{1}{x+y}+\frac{1}{y+z}+\frac{1}{z+x} \ge \frac{5}{2}$$

Here is the solution:

Square both sides and add $(xy+yz+zx)$ in LHS, we have: $$(xy+yz+zx)(\frac{1}{x+y}+\frac{1}{y+z}+\frac{1}{z+x})^2 \ge \frac{25}{4}$$

which can be rewritten as:$$(\sum_{sym}^{}x^5y-\sum_{sym}^{}x^4y^2)+3(\sum_{sym}^{}x^5y-\sum_{sym}^{}x^3y^3)+(\sum_{sym}^{}x^4yz+14\sum_{sym}^{}x^3y^2z+38x^2y^2z^2)\ge 0$$ which is true by Murihead!

My questions are how can they knew how to square both sides and add $(xy+yz+zx)$ to $LHS$, and in the last line of solution, how can they knew when to use Murihead. I'm so confuse when to use Murihead, when to use Weighted AM-GM or Rearrangement inequality. Thank you for all your helps!

  • 1
    There is a proof which just use Cauchy-Schwarz – TATA box Aug 09 '23 at 15:32
  • 1
    See also here: https://math.stackexchange.com/questions/319432/prove-that-frac-1-xy-frac-1-yz-frac-1-zx-geq-frac-5-2. – TATA box Aug 09 '23 at 15:33
  • 2
    I feel like the actual answer to your question is "practice and experience". – Teepeemm Aug 10 '23 at 01:44
  • :))I don't think so, because when I was in low level class, my teacher always said that but when go higher I know there will be some reasons why they did that – Lục Trường Phát Aug 10 '23 at 03:32
  • 1
    You should use Muirhead if it appeared in the lectures in the week before the homework was given. Same applies to most other theorems, and it also applies to exercises following sections in books. – tobi_s Aug 10 '23 at 05:37

3 Answers3

5

Alternative proof.

By squaring both side of the OP, we'll prove that$$(ab+bc+ca)\left(\sum_{cyc}\frac{1}{(b+c)^2}+2\sum_{cyc}\frac{1}{(b+a)(c+a)}\right)\ge \frac{25}{4}.\tag{*}$$ Notice that $$2(ab+bc+ca)\sum_{cyc}\frac{1}{(b+a)(c+a)}=4+\frac{4abc}{(a+b)(b+c)(c+a)}\ge 4. \tag{1}$$ WLOG, assuming that $c=\min\{a,b,c\}.$ We have $2(a+b)>(a+c)(b+c),$ thus \begin{align*} \sum_{cyc}\frac{1}{(b+c)^2}&=\frac{1}{(b+a)^2}+\frac{2}{(b+c)(a+c)}+\frac{(a-b)^2}{(b+c)^2(a+c)^2} \\&\ge \frac{1}{(b+a)^2}+\frac{2}{(b+c)(a+c)}+\frac{(a-b)^2}{4(a+b)^2}. \end{align*}

Proof for $2(a+b)>(a+c)(b+c).$ Since $c=\min\{a,b,c\},$ thus $1=ab+bc+ca\ge 3c^2$ and by AM-GM $$2(a+b)\ge a+c+b+c\ge 2\sqrt{(a+c)(b+c)}>(a+c)(b+c) \iff 4\ge (a+c)(b+c) \iff c^2<3.$$

Also,\begin{align*} \frac{1}{(b+a)^2}&=\frac{ab+bc+ca}{(b+a)^2}\\&=\frac{ab}{(b+a)^2}+\frac{c}{a+b}\\&=\frac{ab}{(b+a)^2}+\frac{2c^2}{(a+c)(b+c)}+\frac{c(c-a)(c-b)}{(a+b)(a+c)(b+c)}\\&\ge \frac{ab}{(b+a)^2}+\frac{2c^2}{(a+c)(b+c)}. \end{align*} Hence \begin{align*} (ab+bc+ca)\sum_{cyc}\frac{1}{(b+c)^2}&\ge \frac{ab}{(b+a)^2}+\frac{2c^2}{(a+c)(b+c)}+\frac{2(ab+bc+ca)}{(a+c)(b+c)}+\frac{(a-b)^2}{4(a+b)^2}\\&=\frac{9}{4}. \tag{2} \end{align*} From $(1)$ and $(2)$ we proved $(*).$

Hence, the proof is done. Equality holds at $a=b=1;c=0$ and its permutations.

TATA box
  • 1
  • 1
  • 5
  • 29
3

It's just homogenization!

Muirhead works with homogeneous symmetric polynomials and homogenization helps to see a possibility to use Muirhead.

It's not always helps.

Another way.

Let $x+y+z\geq2.$

Thus, $$\sum_{cyc}\frac{1}{x+y}=\sum_{cyc}\frac{xy+xz+yz}{x+y}=x+y+z+\sum_{cyc}\frac{xy}{x+y}\geq$$ $$\geq x+y+z+\sum_{cyc}\frac{xy}{x+y+z}=x+y+z+\frac{1}{x+y+z}\geq\frac{5}{2}.$$

Let $x+y+z\leq2$.

Thus, by C-S $$\sum_{cyc}\frac{1}{x+y}=\frac{1}{x+y+z}\sum_{cyc}\frac{x+y+z}{x+y}=$$ $$=\frac{3}{x+y+z}+\frac{1}{x+y+z}\sum_{cyc}\frac{z}{x+y}\geq$$ $$\geq\frac{3}{x+y+z}+\frac{1}{x+y+z}\cdot\frac{(x+y+z)^2}{\sum\limits_{cyc}(xz+yz)}=$$ $$=\frac{3}{x+y+z}+\frac{x+y+z}{2}\geq\frac{5}{2}.$$

  • 3
    Nice proof! I also gave AM-GM and Cauchy-Schwarz proof which split into 2 cases: https://math.stackexchange.com/questions/319432/prove-that-frac-1-xy-frac-1-yz-frac-1-zx-geq-frac-5-2 – TATA box Aug 09 '23 at 16:42
1

Hint.

Calling $f(x,y) = \frac{1}{x+y}$ we have

$$ F(x,y,z) = f(x,y)+f(y,z)+f(z,x) $$

and also

$$ \cases{ f(x,y) \le F(x,y,z)\\ f(y,z) \le F(x,y,z)\\ f(z,x) \le F(x,y,z)\\ } $$

so the solution is reduced to determine the minimum of the problems

$$ \cases{ \min_{x,y}f(x,y), \ \ \text{s.t.}\ \ \ x y =1, x\ge 0, y\ge 0\\ \min_{y,z}f(y,z), \ \ \text{s.t.}\ \ \ y z =1, y\ge 0, z\ge 0\\ \min_{z,x}f(z,x), \ \ \text{s.t.}\ \ \ z x =1, z\ge 0, x\ge 0\\ } $$

Cesareo
  • 33,252