If $a,b,c,d >0$, and $a+b+c+d=4$, prove that $$a^{ab}+b^{bc}+c^{cd}+d^{da} \geq \pi.$$

I don't think Jensen's inequality will help here, but I think first determining where equality holds will be useful. Maybe taking the logarithm or exponential of both sides will also be useful, but I want to in the end get rid of the plus signs in order to simplify it.

  • 1
    $\begingroup$ This reminds me the inequality $a^a+b^b>a^b+b^a$ from Wikipédia. Not sure if it help though... $\endgroup$ – Surb May 6 '16 at 20:54
  • 2
    $\begingroup$ May I ask how you came across this problem? Also, do you know whether the constant $\pi$ is optimal? $\endgroup$ – Wojowu May 6 '16 at 20:55
  • 3
    $\begingroup$ The best I've got out of Mathematica is 3.1605859174508652189…, using NMinimize[{(a^2)^(a^2 b^2) + (b^2)^(b^2 c^2) + (c^2)^(c^2 d^2) + \ (d^2)^(d^2 a^2), a != 0 && b != 0 && c != 0 && d != 0 && a^2 + b^2 + c^2 + d^2 == 4}, {a, b, c, d}, WorkingPrecision -> 100] $\endgroup$ – Patrick Stevens May 6 '16 at 21:26
  • 3
    $\begingroup$ Alright, did I fall for a "scam", i.e. someone cooked up a function and figured out a numeric lower bound? I found the original post on artofproblemsolving and it seems the original poster was banned from the site. Is there an actual solution to this problem? $\endgroup$ – Ivan Mar 14 '17 at 2:39
  • 5
    $\begingroup$ An interesting inequality, but even without the numerical evidence for the bound being around $3.16$ there doesn't seem to be any reason for $\pi$ to feature here. Seems arbitrary $\endgroup$ – Yuriy S Apr 25 '18 at 10:04

TL;DR: The inequality has been proven for all cases except the following five:

  • $1 < a < 2$, $b < 1$, $c > 1$, $d < 1$

  • $1 < a < 2$, $b < 1$, $c < 1$, $d < 1$

  • $2 < a < 3$, $b < 0.207$, $c > 1$, $d < 1$

  • $2 < a < 3$, $b < 0.256$, $c < 1$, $d < 1$

  • $3 < a < 4$, $b < 0.129$, $c < 1$, $d < 1$

This partial answer heavily uses the results that for a real number $k>0$,

  • $\min x^{kx}=(\sqrt[e]{e^k})^{-1}$,

  • $k^{kx}>k^{kx_0}$ for $k>1$ and $x>x_0$,

  • $k^{kx}<k^{kx_0}$ for $k<1$ and $x>x_0$.

As the sum $$S=a^{ab}+b^{bc}+c^{cd}+d^{da}$$ is cyclic, we need only concern $a=1$ and $a>1$ to prove the truth of the inequality. Thus there are sixteen cases that we need to consider (note that in most cases the condition $a+b+c+d=4$ is implicitly used).

$1)$ $a=b=c=d=1$

Clearly $S=1+1+1+1>\pi$.

$2)$ $a=b=1$, $c>1$, $d<1$

As $c<2$, $S>1+1+1+(\sqrt[e]{e})^{-1}>\pi$.

$3)$ $a=b=1$, $c<1$, $d>1$

As $d<2$, $S>1+1+(\sqrt[e]{e^2})^{-1}+1>\pi$.

$4)$ $a=1$, $b>1$, $c>1$, $d<1$

We have $S>1+1+1+(\sqrt[e]{e})^{-1}>\pi$.

$5)$ $a=1$, $b>1$, $c<1$, $d>1$

As $d<2$, $S>1+1+(\sqrt[e]{e^2})^{-1}+1>\pi$.

$6)$ $a=1$, $b>1$, $c<1$, $d<1$

We have $S>1+1+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e})^{-1}>\pi$.

$7)$ $a=1$, $b<1$, $c>1$, $d<1$

If $b\ge0.6$, $c\le2.4$ so $S\ge1+0.6^{0.6\cdot2.4}+1+(\sqrt[e]{e})^{-1}>\pi$. If $b<0.6$, $c>1.4$ so $S>1+(\sqrt[e]{e^3})^{-1}+\min\{1.4^{1.4d}+d^d\}>\pi$.

$8)$ $a=1$, $b<1$, $c<1$, $d>1$

If $b\ge0.89$, $d\le2.11$ so $S\ge1+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^{2.11}})^{-1}+1>\pi$. If $b<0.89$, $3>d>1.11$ so $S\ge 1+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^3})^{-1}+1.11^{1.11}>\pi$.

$9)$ $a=1$, $b<1$, $c>1$, $d>1$

As $c<2$, $S>1+(\sqrt[e]{e^2})^{-1}+1+1>\pi$.

$10)$ $a>1$, $b>1$, $c>1$, $d<1$

As $a<2$, $S>1+1+1+(\sqrt[e]{e^2})^{-1}>\pi$.

$11)$ $a>1$, $b>1$, $c<1$, $d>1$

As $d<2$, $S>1+1+(\sqrt[e]{e^2})^{-1}+1>\pi$.

$12)$ $a>1$, $b>1$, $c<1$, $d<1$

If $d\ge0.675$, $S>1+1+(\sqrt[e]{e^{0.675}})^{-1}+0.675^{0.675}>\pi$. If $a\ge2$, $d<0.675$, then $S>2^2+0+0+0>\pi$. If $a<2$, $d<0.675$, then $S>1+1+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^2})^{-1}>\pi$.

$13)$ $a>1$, $b<1$, $c>1$, $d<1$

If $3>a>2$, $c<2$ so $b\ge0.207$, $S>2^{2\cdot0.207}+(\sqrt[e]{e^2})^{-1}+1+(\sqrt[e]{e^3})^{-1}>\pi$.

$14)$ $a>1$, $b<1$, $c<1$, $d>1$

If $d\ge2$, $S>0+0+0+2^2>\pi$. If $d<2$, $S>1+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^2})^{-1}+1>\pi$.

$15)$ $a>1$, $b<1$, $c>1$, $d>1$

As $c<2$, $S>1+(\sqrt[e]{e^2})^{-1}+1+1>\pi$.

$16)$ $a>1$, $b<1$, $c<1$, $d<1$

If $4>a>3$, $b\ge0.129$, $S>3^{3\cdot0.129}+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^4})^{-1}>\pi$. If $3>a>2$, $b\ge0.256$, $S>2^{2\cdot0.256}+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e})^{-1}+(\sqrt[e]{e^3})^{-1}>\pi$.

  • $\begingroup$ There miss many cases by example when $a<1$ and we don't have the case when the minimum is reached... $\endgroup$ – user635269 Apr 19 at 13:59
  • $\begingroup$ @FatsWallers As I've said in my post, because the sum $a^{ab}+b^{bc}+c^{cd}+d^{da}$ is cyclic, proving the inequality for $a\ge 1$ is sufficient due to the constraint $a+b+c+d=4$. If you try doing it for $a<1$, you get the same conclusions as for say, $b<1$ or $c<1$. $\endgroup$ – TheSimpliFire Apr 19 at 15:41
  • $\begingroup$ Okay sorry I read too fast lol ^^ . $\endgroup$ – user635269 Apr 20 at 12:44

We can easily prove a weaker inequality as follows. It is known that if $x>0$ then $x^x\ge \left(\tfrac 1e\right)^{1/e}=E$. Since the original inequality is cyclic, witout loss of generality we can assume that $d\ge 1$. Then

$$a^{ab}+b^{bc}+c^{cd}+d^{da}\ge E^b+E^c+E^d+1^a\ge 3E^{(b+c+d)/3}+1\ge 3E^{4/3}+1>2.8369.$$


Warning : It's a partial proof but I think it's interesting . We have the following theorem :

Let $a,b,c,d>0$ such that $abcd\geq e^{-2}$ then we have : $$a^{ab}+b^{bc}+c^{cd}+d^{da}\geq 4(e^{-0.5})^{e^{-1}}$$

Proof : We begin to apply AM-GM to get : $$a^{ab}+b^{bc}+c^{cd}+d^{da}\geq 4(a^{ab}b^{bc}c^{cd}d^{da})^{0.25}$$ Hence we want to prove : $$4(a^{ab}b^{bc}c^{cd}d^{da})^{0.25}\geq 4(e^{-0.5})^{e^{-1}}$$ Now we take the logarithm : $$0.25\sum_{cyc}ab\ln(a)\geq -0.5e^{-1}$$

But $x\ln(x)$ is convex so we get : $$0.25\sum_{cyc}ab\ln(a)\geq 0.25(ab+bc+cd+da)\ln(\frac{ab+bc+cd+da}{a+b+c+d})$$

Hence we want to prove :

$$0.25(ab+bc+cd+da)\ln(\frac{ab+bc+cd+da}{a+b+c+d})\geq -0.5e^{-1}$$

Wich is true because : $$\frac{ab+bc+cd+da}{a+b+c+d}\geq e^{-0.5}$$


$$a+b+c+d\geq 4e^{-1}$$

Now remains to add the condition $a+b+c+d=4$ to get :

Let $a,b,c,d>0$ such that $abcd\geq e^{-2}$ and $a+b+c+d=4$ then we have : $$a^{ab}+b^{bc}+c^{cd}+d^{da}> 4(e^{-0.5})^{e^{-1}}$$

An other idea will be to extend the initial theorem (using the same technique) and see what happend .

Finally we conclude with the fact that it completes partialy the result of TheSimpliFire .

  • 1
    $\begingroup$ Why $\frac{ab+bc+cd+da}{a+b+c+d}\geq e^{-0.5}$? Why this and $a+b+c+d\geq 4e^{-1}$ (by the way, since $abcd\ge e^-2$ then AM-GM implies $a+b+c+d\geq 4e^{-1/2}$) imply $0.25(ab+bc+cd+da)\ln(\frac{ab+bc+cd+da}{a+b+c+d})\geq -0.5e^{-1}$? Note, that a priori the logarithm may be negative, so we have to take this into account. $\endgroup$ – Alex Ravsky May 1 at 17:01

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.