Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that \[ 1 < \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}\leq\frac {3\sqrt {2}}{2} \]
Problem
Source: CWMO 2004, problem 8
Tags: Cauchy Inequality, Inequality, three variable inequality, inequalities, China
02.10.2004 16:13
The left part is easy. The term is larger than $ \frac{a}{\sqrt{a^2+b^2}}+\frac{c}{\sqrt{c^2+a^2}}$. We can assume $ c\geq b$. then it is larger than $ \frac{a}{\sqrt{a^2+c^2}}+\frac{c}{\sqrt{c^2+a^2}} > \frac{a}{a+c}+\frac{c}{a+c} = 1$. Misha
02.10.2004 18:11
How about if $b>c$?
02.10.2004 18:50
The inequality is cyclic so we can assume that c is the largest of a,b, c. In case of b>c you only have to replace all a's by c's all b's by a's and all c's by b's in my proof. Misha
02.10.2004 18:52
Simpler solution for the left part: $ S > \frac{a}{\sqrt{a^2+b^2+c^2}} + \frac{b}{\sqrt{a^2+b^2+c^2}} + \frac{c}{\sqrt{a^2+b^2+c^2}}$.
03.10.2004 15:53
Here is another proof of the left side (remember Engel ;-) ): We have $\sqrt{a^{2}+b^{2}}<\sqrt{a^{2}+b^{2}+c^{2}+2bc+2ca+2ab}=a+b+c$, and thus $\frac{a}{\sqrt{a^{2}+b^{2}}}>\frac{a}{a+b+c}$. Similarly, $\frac{b}{\sqrt{b^{2}+c^{2}}}>\frac{b}{a+b+c}$ and $\frac{c}{\sqrt{c^{2}+a^{2}}}>\frac{c}{a+b+c}$. Thus, $\frac{a}{\sqrt{a^{2}+b^{2}}}+\frac{b}{\sqrt{b^{2}+c^{2}}}+\frac{c}{\sqrt{c^{2}+a^{2}}}>\frac{a}{a+b+c}+\frac{b}{a+b+c}+\frac{c}{a+b+c}=1$. As for the right side, after the substitution $A=a^2$, $B=b^2$, $C=c^2$, it becomes $\frac{\sqrt{A}}{\sqrt{A+B}}+\frac{\sqrt{B}}{\sqrt{B+C}}+\frac{\sqrt{C}}{\sqrt{C+A}}\leq \frac{3\sqrt{2}}{2}$, or, equivalently, $\sqrt{\frac{A}{A+B}}+\sqrt{\frac{B}{B+C}}+\sqrt{\frac{C}{C+A}}\leq \frac{3\sqrt{2}}{2}$, what can be rewritten as $\sqrt{\frac{2A}{A+B}}+\sqrt{\frac{2B}{B+C}}+\sqrt{\frac{2C}{C+A}}\leq 3$. And this is exactly Vasc's bloody inequality 113 in Titu Andreescu, Vasile Cîrtoaje, Gabriel Dospinescu, Mircea Lascu, Old and New Inequalities, GIL, Zalau 2004. Bloody since I have been trying to solve it for an hour until I realized that it's in that book with two solutions, each one page long. Darij
03.10.2004 16:09
I've seen quite a number of inequalities now in the form of \[ \sum_{i=1}^{n} \frac{1}{\sqrt{1+x_i}} \leq S \] where $x_1 x_2 \cdots x_n$ is a constant (often 1, as in this case). If there a general (and neat) way to tackle them?
03.10.2004 16:36
Darij, could you please post one of the two solutions you mention because I don't have the book (or just mention which inequalities were used in the solution). Misha
06.10.2004 18:40
I have just proved this inequality in excellent way, in my opinion. We can write \[ \sqrt\frac{a}{a+b}+\sqrt\frac{b}{b+c}+\sqrt\frac{c}{c+a}= \] \[ =\frac{a+c}{2(a+b+c)}\cdot\sqrt\frac{4a(a+b+c)^2}{(a+b)(a+c)^2}+ \frac{b+a}{2(a+b+c)}\cdot\sqrt\frac{4b(a+b+c)^2}{(b+c)(b+a)^2}+ \frac{c+b}{2(a+b+c)}\cdot\sqrt\frac{4b(a+b+c)^2}{(c+a)(c+b)^2}=S. \] Now we apply Jensen's inequality for $\sqrt{x}$ and obtain: \[ S\leq \sqrt{\frac{2a(a+b+c)}{(a+b)(a+c)}+\frac{2b(a+b+c)}{(b+c)(b+a)}+\frac{2c(a+b+c)}{(c+a)(c+b)}}=\sqrt{S'}. \] So we are to show that $S'\leq \frac{9}{2}$. You can easily check this inequality, because it is equivalent to $\sum\limits_{\textrm{cyclic}}a^2b\geq 6abc$. Is this solution distinct from Vasc's ones? P.S. Are there copyrights for solutions? Each author merited to be mentioned if his solution is used somewhere.
06.10.2004 20:11
Myth wrote: Is this solution distinct from Vasc's ones? Not only distinct; it is also two times shorter than any of the solutions in the book, and much more elegant than any of them. Myth wrote: P.S. Are there copyrights for solutions? I hope not, but in any case, when I will refer to your solution, I will call it the Myth solution! Darij
06.10.2004 20:22
darij grinberg wrote: Myth wrote: Is this solution distinct from Vasc's ones? Not only distinct; it is also two times shorter than any of the solutions in the book, and much more elegant than any of them. Jensen forever! darij grinberg wrote: Myth wrote: P.S. Are there copyrights for solutions? I hope not, but in any case, when I will refer to your solution, I will call it the Myth solution! I hope all people will respect solution's authors as well as you respect...
06.10.2004 22:06
Myth wrote: darij grinberg wrote: Myth wrote: Is this solution distinct from Vasc's ones? Not only distinct; it is also two times shorter than any of the solutions in the book, and much more elegant than any of them. Jensen forever! More precisely, weighted Jensen forever! I myself had learnt it, as a method, from Peter Scholze in Athens (the inequality itself was known to me before, but I didn't really know what do with it ), and it was immediately clear to me that it should be much more powerful that plain Jensen and it should have a lot of interesting applications. But until now I have seen only very few ones, including your ingenious proof of Vasc's inequality. I guess that most applications of the weighted Jensen inequality remain undiscovered just because people too seldomly try to apply it. And don't forget the weighted Karamata inequality (I believe it's also called Fuchs inequality) - I can't believe that it has no use! Darij
06.10.2004 22:14
A very elegant and neat solution. Congratulations, Myth.
07.10.2004 06:17
darij grinberg wrote: Myth wrote: darij grinberg wrote: Myth wrote: Is this solution distinct from Vasc's ones? Not only distinct; it is also two times shorter than any of the solutions in the book, and much more elegant than any of them. Jensen forever! darij grinberg wrote: More precisely, weighted Jensen forever! I myself had learnt it, as a method, from Peter Scholze in Athens (the inequality itself was known to me before, but I didn't really know what do with it ), and it was immediately clear to me that it should be much more powerful that plain Jensen and it should have a lot of interesting applications. But until now I have seen only very few ones, including your ingenious proof of Vasc's inequality. I guess that most applications of the weighted Jensen inequality remain undiscovered just because people too seldomly try to apply it. And don't forget the weighted Karamata inequality (I believe it's also called Fuchs inequality) - I can't believe that it has no use! I very often use Jensen's inequality (at least, I try such approch in almost each inequalities).
26.10.2004 00:19
What is the "weighted Karamata inequality "??
26.10.2004 01:21
Truly beautiful and impecable! Congratulations, Myth! And it is really sincere!
26.10.2004 10:30
anothor solution: let x=b/a, y=c/b, z=a/c, xyz=1,ln x+ln y+ln z=0 let w(t)=(1+e^(2t))^(-1/2) w"(t)>0 Jensen : (1+e^(2ln x))^(-1/2)+(1+e^(2ln y))^(-1/2)+(1+e^(2ln z))^(-1/2) <=3(1+e^(2/3*(ln x+ln y+ln z)))^(-1/2) =3*2(-0.5) (1+e^(2ln x))^(-1/2)=(1+ x^2)^(-0.5) ... (1+ x^2)^(-0.5) +(1+ y^2)^(-0.5) +(1+ z^2)^(-0.5) <=3*2(-0.5) 7615
26.10.2004 11:38
zxingtan wrote: anothor solution: let x=b/a, y=c/b, z=a/c, xyz=1,ln x+ln y+ln z=0 let w(t)=(1+e^(2t))^(-1/2) w"(t)>0 Jensen : (1+e^(2ln x))^(-1/2)+(1+e^(2ln y))^(-1/2)+(1+e^(2ln z))^(-1/2) <=3(1+e^(2/3*(ln x+ln y+ln z)))^(-1/2) =3*2(-0.5) (1+e^(2ln x))^(-1/2)=(1+ x^2)^(-0.5) ... (1+ x^2)^(-0.5) +(1+ y^2)^(-0.5) +(1+ z^2)^(-0.5) <=3*2(-0.5) You are wrong, because your solution requires $w(t)$ is convex for all $t\in\mathbb{R}$, but it is not true.
26.10.2004 12:11
Another nice solution is to consider the incircle of a triangle with sides $a^2+b^2$, $b^2+c^2$ and $c^2+a^2$.
27.10.2004 09:58
I think this is more difficult. If $a,b,c$ are non-negative numbers, then $\sqrt{\frac{a}{4a+5b}}+\sqrt{\frac{b}{4b+5c}}+\sqrt{\frac{c}{4c+5a}}\leq 1$.
23.04.2018 03:11
Ji Chen wrote: Vasc wrote: If $ a,b,c$ are positive numbers, then $ \sqrt {\frac {a}{4a + 5b}} + \sqrt {\frac {b}{4b + 5c}} + \sqrt {\frac {c}{4c + 5a}}\leq 1$.
shalex wrote: For all $ x,y,z\in\mathbb{R_{ + }}$ such that $ xyz = 1$ and $ 0\leqslant\lambda\leqslant\frac {5}{4}$ there is $ \frac {1}{\sqrt {1 + \lambda x}} + \frac {1}{\sqrt {1 + \lambda y}} + \frac {1}{\sqrt {1 + \lambda z}}\leqslant\frac {3}{\sqrt {1 + \lambda}}.$ The solution is nearly the same as Arwen's to the original one in which $ \lambda = 1$. $ \lambda > \frac {5}{4}\Longrightarrow \frac {1}{\sqrt {1 + \lambda x}} + \frac {1}{\sqrt {1 + \lambda y}} + \frac {1}{\sqrt {1 + \lambda z}} < 2.$ holds for all positive numbers $ x,y,z$ with $ xyz = 1.$
See also here : http://www.mathlinks.ro/Forum/viewtopic.php?t=106970 Let $a,b,c$ be nonnegative reals. Prove $$\sqrt{\frac a {4a+5b}}+\sqrt{\frac b {4b+5c}}+\sqrt{\frac c {4c+5a}} \ge \frac 1 2$$If $a, b, c>0.$ Prove that \[\sqrt{\frac{a}{4b+5c}}+\sqrt{\frac{b}{4c+5a}}+\sqrt{\frac{c}{4a+5b}}\ge\frac{2}{3}+\frac{ab+bc+ca}{(a+b+c)^2}\]If $a,b,c$ are non-negative numbers such that $ab+bc+ca>0.$ Find the minimum value of $\sqrt{\frac{a}{4b+9c}}+\sqrt{\frac{b}{4c+9a}}+\sqrt{\frac{c }{4a+9b}} .$ here
28.04.2018 17:49
https://mp.weixin.qq.com/s?__biz=MzU0NDI5MjczNw==&mid=2247484665&idx=1&sn=9865704acf40177b35fa1eeceb341c51&chksm=fb7f2e17cc08a701c94da2c23cc1584bf53c67a029af8f0a3be098afb756c3ae357736c94cbd&mpshare=1&scene=1&srcid=04284qiU9DgVSHmU5DKkB7c8&pass_ticket=PddpMA1lGncTsH%2FKvyp0%2BXrxQc6%2B3VJ%2Fgv9onzl3RWY%3D#rd
28.08.2018 11:38
Suppose that $ a$, $ b$, $ c$ are positive real numbers. Prove that $$ \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}+\frac{\sqrt{2}}{16}\left(\frac{b^4}{a^4}+\frac{c^4}{b^4}+\frac{a^4}{c^4}\right)\geq\frac {27\sqrt {2}}{16}.$$IMO2001 2008 Jiangxi Higher Education Entrance Examination High-School Mathematics(China Tianjin) No.8(2007):
Attachments:

01.12.2018 06:25
Myth wrote: I have just proved this inequality in excellent way, in my opinion. We can write \[ \sqrt\frac{a}{a+b}+\sqrt\frac{b}{b+c}+\sqrt\frac{c}{c+a}= \]\[ =\frac{a+c}{2(a+b+c)}\cdot\sqrt\frac{4a(a+b+c)^2}{(a+b)(a+c)^2}+ \frac{b+a}{2(a+b+c)}\cdot\sqrt\frac{4b(a+b+c)^2}{(b+c)(b+a)^2}+ \frac{c+b}{2(a+b+c)}\cdot\sqrt\frac{4b(a+b+c)^2}{(c+a)(c+b)^2}=S. \]Now we apply Jensen's inequality for $\sqrt{x}$ and obtain: \[ S\leq \sqrt{\frac{2a(a+b+c)}{(a+b)(a+c)}+\frac{2b(a+b+c)}{(b+c)(b+a)}+\frac{2c(a+b+c)}{(c+a)(c+b)}}=\sqrt{S'}. \]So we are to show that $S'\leq \frac{9}{2}$. You can easily check this inequality, because it is equivalent to $\sum\limits_{\textrm{cyclic}}a^2b\geq 6abc$. Is this solution distinct from Vasc's ones? P.S. Are there copyrights for solutions? Each author merited to be mentioned if his solution is used somewhere. indeed very nice!
05.12.2018 16:53
Suppose that $ a$, $ b$, $ c$ , $ d$ are positive real numbers, prove that \[ \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + d^{2}}}+ \frac {d}{\sqrt {d^{2} + a^{2}}}\leq3.\]
10.05.2019 17:59
Vasc wrote: I think this is more difficult. If $a,b,c$ are non-negative numbers, then $\sqrt{\frac{a}{4a+5b}}+\sqrt{\frac{b}{4b+5c}}+\sqrt{\frac{c}{4c+5a}}\leq 1$. Let $\frac{a}{4a+5b}=\frac{p^2}{9}$, $\frac{b}{4b+5c}=\frac{q^2}{9}$ and $\frac{c}{4c+5a}=\frac{r^2}{9}$ Hence, we need to prove that $p+q+r\leq3$ and since $\frac{5b}{a}=(-4+\frac{9}{p^2})$, we obtain $\{p,q,r\}\in(0;\frac{3}{2})$ Let $p+q+r>3$, $p=kx$, $q=ky$ and $r=kz$, where $k>0$ and $x+y+z=3$. Hence, $k>1$, $\{x,y,z\}\in\{0;\frac{3}{2}\}$ and $125=\prod (-4+\frac{9}{k^2x^2})<\prod (-4+\frac{9}{x^2})$, which is contradiction because we'll prove now that: $125\geq \prod (-4+\frac{9}{x^2})$ Let $3u=x+y+z$, $3v^2=xy+yz+zx$ and $w^3=xyz$ Since, we need to prove $f(w^3)\geq0$, where $f(w^3)$ is increasing function. Hence, $f(w^3)$ reaches minimum value for minimum value of $w^3$, which happens when two of variables are equal. $x=y$ $27(x-1)^2(7x^2+14x+9)\geq0$
11.05.2019 05:37
mecrazywong wrote: Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that \[ 1 < \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}\leq\frac {3\sqrt {2}}{2} \] first is my problem
Attachments:

11.05.2019 05:47
xzlbq wrote: Let $x,y,z>0$,prove \[2\,\sqrt {{\frac { \left( x+y+z \right) \left( xy+xz+yz \right) }{ \left( y+z \right) \left( z+x \right) \left( x+y \right) }}}\geq \sqrt {{\frac {x}{x+y}}}+\sqrt {{\frac {y}{y+z}}}+\sqrt {{\frac {z}{z+x}}}\]<=> \[2\,\sqrt { \left( x+y+z \right) \left( xy+xz+yz \right) }\geq \sqrt {x \left( y+z \right) \left( z+x \right) }+\sqrt {y \left( z+x \right) \left( x+y \right) }+\sqrt {z \left( x+y \right) \left( y+z \right) }\] Note this a=a: \[4\,{\frac { \left( x+y+z \right) \left( xy+xz+yz \right) }{ \left( x+ y \right) \left( y+z \right) \left( z+x \right) }}={\it \sum} \left( z+x \right) {\it \sum} \left( {\frac {x}{ \left( z+x \right) \left( x+y \right) }} \right) \] By C-S, is easy
11.05.2019 08:43
xzlbq wrote: mecrazywong wrote: Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that \[ 1 < \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}\leq\frac {3\sqrt {2}}{2} \] first is my problem Let $\frac{x}{x+y}=\frac{p}{2}$, $\frac{y}{y+z}=\frac{q}{2}$ and $\frac{z}{z+x}=\frac{r}{2}$ Hence, we need to prove that $p^2+q^2+r^2\geq3$ and since $\frac{y}{x}=\frac{2}{p}-1$, we obtain $\{p,q,r\}\in\{0;2\}$ Let $p^2+q^2+r^2<3$, $p=ka$, $q=kb$, $r=kc$, where $k>0$ and $a^2+b^2+c^2=3$ Hence, $k<1$ $1=\prod (\frac{2}{p}-1)=\prod (\frac{2}{ka}-1)>\prod (\frac{2}{a}-1)$,which is contradiction because we'll prove now that: $\prod (\frac{2}{a}-1)\geq 1$ or $\prod (-a+2)(-b+2)(-c+2)\geq abc$ Let $3u=a+b+c$, $3v^2=ab+bc+ca$ and $w^3=abc$, where $3u^2-2v^2=1$, then we need to prove $2w^3\leq 6v^2-12u+8=9u^2-12u+5$ Since $2w^3\leq 2(3uv^2-2u^3+2\sqrt{(u^2-v^2)^3})=5u^3-3u^2+\sqrt{2(-u^2+1)^3}$, so enough to prove $5u^3-3u^2+\sqrt{2(-u^2+1)^3}\leq 9u^2-12u+5$ or $\sqrt{2(-u^2+1)^3}\leq -5u^3+12u^3-12u+5=(-u+1)(5u^2-7u+5)$ squaring $2(-u+1)^3(u+1)^3\leq (-u+1)^2(5u^2-7u+5)^2$ or $0\leq (u-1)^2(27u^4-66u^3+99u^2-74u+23)$
25.05.2020 02:34
Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that \[ \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}\leq\frac {3\sqrt {2}}{2} \]$$\iff$$Suppose that $ a$, $ b$, $ c$ are positive real numbers such that $(1-a)(1-b)(1-c)=abc.$ Prove that$$\sqrt{a}+\sqrt{b}+\sqrt{c}\leq \frac{3\sqrt{2}}{2}$$here
02.11.2020 06:27
sqing wrote: Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that $$\frac{a}{\sqrt{b^2+c^2}}+\sqrt{2}\left(\frac{b}{a+c}+\frac{c}{a+b}\right) \ge \frac{3\sqrt{2}}{2}$$ $$\frac{a}{b+c}+\frac{b}{c+a}+\frac{c}{\sqrt{2(a^2+b^2)}} \ge \frac{3}{2}$$Suppose that $ a$, $ b$, $ c$ are positive real numbers. Prove that $$\frac{a}{b+c}+\frac{b}{c+a}+\sqrt{\frac{2c}{a+b}} \ge {2}$$here: $$ \frac{a}{b+c}+\frac{b}{c+a}+\frac{4(c+\sqrt{ab})}{a+b}\geq 4$$ $$ \frac{a}{b+c}+\frac{b}{c+a}\ge \frac{a+b}{2}\left(\frac{1}{b+c}+\frac{1}{c+a}\right)$$
Attachments:



04.10.2021 14:23
sqing wrote: High-School Mathematics(China Tianjin) No.8(2007): If $a>0$, $b>0$, $c>0$, and $\sqrt[3]{\!abc\mspace{0.5mu}}\ge8$, then $\frac1{\sqrt{\mspace{-1mu}1+a}\mspace{1mu}}+\frac1{\sqrt{\mspace{-1mu}1+b}\mspace{1mu}}+\frac1{\sqrt{\mspace{-1mu}1+c}\mspace{1mu}}\ge\frac3{\sqrt{\mspace{-2mu}1+\sqrt[3]{\!abc\mspace{0.5mu}}\mspace{1mu}}}$?
30.05.2022 04:49
Can we prove that suppose $x_1,\dots ,x_n>0,n\ge3$ then $\frac {x_1}{\sqrt {{x_2}^{2} + \dots +{x_n}^{2}}}+\frac {x_2}{\sqrt {{x_1}^{2} +{x_3}^{2}+ \dots +{x_n}^{2}}}+\dots +\frac {x_n}{\sqrt {{x_1}^{2} + \dots +{x_{n-1}}^{2}}} \le t$ What's the max value of t?
25.12.2024 07:41
sqing wrote: Suppose that $ a$, $ b$, $ c$ are positive real numbers, prove that \[ \frac {a}{\sqrt {a^{2} + b^{2}}} + \frac {b}{\sqrt {b^{2} + c^{2}}} + \frac {c}{\sqrt {c^{2} + a^{2}}}\leq\frac {3\sqrt {2}}{2} \]$$\iff$$Suppose that $ a$, $ b$, $ c$ are positive real numbers such that $(1-a)(1-b)(1-c)=abc.$ Prove that$$\sqrt{a}+\sqrt{b}+\sqrt{c}\leq \frac{3\sqrt{2}}{2}$$here Our key idea is to make this geometrically, let $O$ be the origin of the complex plane. Consider points $c,-ci, -a,bi$ as $P,Q,R,S$ respectively. Note that our inequality is simply equivalent to proving that $1<\sin \angle OSR+\sin\angle OPS+\sin \angle ORQ\leq \frac{3\sqrt 2}{2}$. However, letting $u=\sin \angle OSR, v=\sin\angle OPS, w=\sin \angle ORQ$, observe that $\tan \angle OSR \tan \angle OPS \tan \angle ORQ=1$. This implies that $uvw=\sqrt{(1-u^2)(1-v^2)(1-w^2)}$ by Pythagorean identity and moving the cosines to the RHS. Now, squaring gives the result sqing wanted us to prove, that is $a,b,c$ are positive real numbers such $(1-a)(1-b)(1-c)=abc$, and we want $1<\sqrt{a}+\sqrt{b}+\sqrt{c}\leq 3\frac{\sqrt 2}{2}$. I used calculus to finish, I might writeup it here later
25.12.2024 08:20
SPLENDID IDEAS!