Find all positive reals $x,y,z $ such that \[2x-2y+\dfrac1z = \dfrac1{2014},\hspace{0.5em} 2y-2z +\dfrac1x = \dfrac1{2014},\hspace{0.5em}\text{and}\hspace{0.5em} 2z-2x+ \dfrac1y = \dfrac1{2014}.\]
Problem
Source:
Tags: number theory, least common multiple, inequalities, symmetry, Vieta, algebra, system of equations
01.01.2015 09:19
my proof = add all three equations to get $ (1/x) + (1/y) + (1/z) = 3/2014 $ and taking lcm and cross multiplying and than adding we get $ x+y+z = 3(2014) $ now by AM-HM inequality we have $(x+y+z)((1/x) +(1/y)+(1/z)) = 3(2014)(1/2014)3 = 9 \geq 9 $ thus equality occurs. hence $x=y=z = 2014 $
01.01.2015 10:37
aditya21 wrote: and taking lcm and cross multiplying and than adding we get $ x+y+z = 3(2014) $ What lcm? The numbers $x,y,z$ are real, not integer. You mean clearing denominators, then adding. Then there is the small issue that $x,y,z$ are not known to be positive, so applying AM-HM isn't quite justified. For example, denoting $x=2014a$, $y=2014b$, $z=2014c$, we get $a+b+c = 1/a+1/b+1/c = 3$, which also has other (not all positive) real solutions than $a=b=c=1$.
01.01.2015 11:37
but mavpronevma AM-HM or cauchy schwarz inequality can be applied to reals i suppose it doesnt violate the condition of x,y,z being reals can you clarify? also by lcm i wanted to say like $ 2x-2y+(1/z) =1/2014$ can be wriiten as $ 2zx-2yz+1 = (z/2014)$ and similarly for others and than adding all 3 equations.
01.01.2015 11:46
Yes, read my post above. There is for example the solution $a=\dfrac {-1-\sqrt{27/11}}{2}$, $b=\dfrac {-1+\sqrt{27/11}}{2}$, $c=4$. Cauchy-Schwarz can be applied for example in the form $(x^2+y^2+z^2)(r^2/x^2+s^2/y^2+t^2/z^2) \geq (r+s+t)^2$, even for some negatives among $x,y,z,r,s,t$, but not for $(x+y+z)(r^2/x+s^2/y+t^2/z) \geq (r+s+t)^2$. Just the relations for $x+y+z$ and $1/x+1/y+1/z$ are not enough; you need to also refer back to one of the original relations.
01.01.2015 12:12
ohh thanks mavpronevma! so how do we solve this system of equations? also one of my friend wrote that since the equation is symmetric in $x,y,z$ than $x=y=z$ which is wrong according to me can you clarify?
01.01.2015 12:17
Of course the symmetry argument is wrong; the system $x+y+z=6$, $xy+yz+zx=11$, $xyz=6$ is symmetric, but its solutions are $\{1,2,3\}$.
01.01.2015 12:20
do you have a solution to this question mavropnevma?
01.01.2015 17:55
Just do what you did to get \[\dfrac1x+\dfrac1y+\dfrac1z=\dfrac{3}{2014}\] and \[x+y+z=3(2014)\] Then let $xy+yz+zx=3k$ hence $xyz=2014k$. By Vieta's $x,y,z$ are roots of $f(n)=n^3-6042n^2+3kn-2014k$. Since $x,y,z$ are real, we know $\Delta \geq 0$ aka $-108 k^3+876138336 k^2-1776894406964928 k\geq 0$ which is just $(k-4056196)^2 k\leq 0$. Now the only positive solution is $k=4056196$ which gives the beloved solution $x=y=z=2014$ else $k\leq 0$ and you can use the cubic formula (although there will be negative solutions, if the problem wanted it solved in positive reals, it would've been much easier).
01.01.2015 18:21
utkarshgupta wrote: Find all real $x,y,z $ such that $2x-2y+(1/z) = (1/2014)$ , $ 2y-2z +(1/x) = (1/2014) $ and $ 2z-2x+ (1/y) = (1/2014) $ the question asked to find positive reals so $x=y=z=2014$ is the answer
01.01.2015 18:36
That was also my guess ... judging by the degree of difficulty of the other problems, it must have had this extra, simplifying condition.
04.01.2015 06:37
FWIW : Finding positive real solution(s) is easy but finding all (even not restricted to real) solutions is not hard either. (Assuming solving a cubic equation is not hard ..specially if numerical (approximate) solution is sufficient) $2x-2y+(1/z) = (1/2014)$ , $ 2y-2z +(1/x) = (1/2014) $ and $ 2z-2x+ (1/y) = (1/2014) $ As pointed out, adding all we get $1/x+1/y+1/z=3/2014$ and writing the three as $2z(x-y)+1=z/2014$ ...etc , and add them together we have: $x/2014+y/2014+z/2014=3$ Again for simplicity (or general solution for any k) let $k=2014$, and $x=ka,y=kb,z=kc$ We have $a+b+c=3$ (eq 1) $1/a+1/b+1/c=3$ eq(2) and of course, $2k^2 c(a-b)+1=c$ (eq 3) We can simplify a bit by assuming $a=1+p,b=1+q,c=1+r$ so we get $p+q+r=0$ And $1/(1+p)+1/(1+q) = 3 - 1/(1+r)$ $\frac {2+p+q}{1+p+q+pq} = \frac {3+3r -1} {1+r}$ Substituting $p+q=-r$ to get this equation in pq and r .. we have $\frac {2-r}{1-r+pq} = \frac {2+3r}{1+r}$ which gives us $pq = \frac {2r^2} {2+3r}$ We have the values of p+q (=-r) and pq (in terms of r) and eq 3 gives the value of (p-q) Eq 3 now is $2k^2(1+r)(p-q)=c-1=r$ which gives us $p-q = (1/2k^2)(r/(1+r))$ so $(p+q)^2 = (p-q)^2+4pq$ now yields a nice equation in $r$ : $r^2 = \frac {(1/4k^4)r^2}{(1+r)^2} + \frac {8r^2} {2+3r}$ $r^2=0 $ is obviously a solution, other solutions are given by: $1 = \frac {(1/4k^4)}{(1+r)^2} + \frac {8} {2+3r}$ This simplify nicely to $12k^4 r^3 -(3+36k^4)r-(2+24k^4)$ Or $12k^4 r^3 -3(12k^4+1)r-2(12k^4+1)$ Which again comes out neatly as, $ r^3 - 3nr -2n$ where $n = 1+1/(12*2014^4)$ (Please note that n is very close to 1) The values (since $n$ is pretty close to $1$ ) of $r$ is close to -1,+1,2 (Because when $n=1$ reduces above to $r^3-3r-2 = (r+1)^2(r-2)=0$ This gives value of $c$ close to $0,0,3$ For the case of 2014, we have, The additional solutions for values $c$ are approximately $-.00000004..,+.00000004..,3.000000000000004$ etc So the solutions for $x,y,z$ are: $(2014,2014,2014)$ and close to (0,0,3*2014) More accurately $v$ of the order of $10^{-11}$ and $u$ is of the order of $10^{-4}$ and $(x,y,z)$ are $(-u,(u-v), (3*2014+v)$ where, $v << u <<1 $ (Cyclic combinations)
10.07.2023 01:10
Sorry for reviving this post but I have found out another solution to this question and I want to know whether it is correct or it has some flaw W.L.O.G assume $x \geq y \geq z$, then $\frac{1}{z} \geq \frac{1}{y} => 2x-2y+\frac{1}{z} \geq 2z-2x+\frac{1}{y}$, equality iff $x=y=z$. As equality is given in the question, we get our only solution as $x=y=z=2014$.