Given $a>b>2$ both positive integers, which of $a^b$ and $b^a$ is larger?

I tried an induction approach. First I showed that if $b=3$ then any $a \geq4$ satisfied $a^b<b^a$.

Then using that as my base case I tried to show that given any pair of positive integers $a,b$ satisfying $a>b>2$ and $a^b<b^a$, then $(a+1)^{b+1}<(b+1)^{a+1}$ - but that is where I got stuck.

Any help would be appreciated.

  • 6
    $\begingroup$ what if you took root $1/ab$ from both sides and studied $y=x^{1/x}$. $\endgroup$
    – Maesumi
    Jun 4, 2013 at 3:19
  • $\begingroup$ If you let $a=b+k$, then you want to solve $(b+k)^b < b^{(b+k)}$, which is equivalent to $(1+\frac{k}{b})^b < b^k$. If you know the definition that $e^k$ is the limit of the increasing sequence $(1+\frac{k}{n})^n$, then this implies $b>e$, or equivalently that $b\ge3$. $\endgroup$
    – user641
    Jun 4, 2013 at 7:56
  • 1
    $\begingroup$ $a\gt b\gt 2$, does not state everything. If specifically the order with respect to $e$ is said, like $e\gt a\gt b\gt 2$ or $a\gt b\gt e\gt 2$, then it would be more clear. $\endgroup$
    – user249332
    Dec 15, 2015 at 16:55
  • 1
    $\begingroup$ related: math.stackexchange.com/questions/517555/… $\endgroup$
    – Henry
    Jan 2, 2020 at 9:54

5 Answers 5


Suppose for a moment: $$a^b = b^a.$$ Taking natural log: $$ b\ln a = a\ln b, $$ which is $$ \frac{\ln a}{a} = \frac{\ln b}{b} . $$ Now consider the function: $$ f(x) = \frac{\ln x}{x}, $$ where $$ f'(x) = \frac{1 - \ln x}{x^2} < 0 \quad \text{ if }\;x>e. $$ I believe you could take it from here.

  • $\begingroup$ Thanks so much for the help, I had realized that calculus would make pretty short work of this, and I see how your proofs go. However, I'm not supposed to use it. Apparently there is some sort of induction/bounding argument that does it, just for integers. By the way, thanks for letting me finish off your proof, it's much more informative than otherwise. $\endgroup$
    – John Marty
    Jun 4, 2013 at 3:42
  • $\begingroup$ @JohnMarty I see, but I am not an expert in number theory. I will try to think of a number theoretic answer and edit my post then. $\endgroup$
    – Shuhao Cao
    Jun 4, 2013 at 3:43

The result follows easily using calculus. Here's an elementary approach, which uses the fact that $a, b$ are positive integers.

Consider $n \geq 3$. Then $$(n+1)^n=\sum_{i=0}^{n}{\binom{n}{i}n^{n-i}}=1+n^2+\sum_{i=0}^{n-2}{\binom{n}{i}n^{n-i}}<n^n+\sum_{i=0}^{n-2}{n^n}=n^{n+1}$$ since $\binom{n}{i} \leq n^i$, and $1+n^2<n^n$ for $n \geq 3$.

Therefore $n^{\frac{1}{n}}>(n+1)^{\frac{1}{n+1}}$ for $n \geq 3$. This immediately implies that $a^{\frac{1}{a}}<b^{\frac{1}{b}}$, so $a^b<b^a$.



Which is larger, $b\ln a$ or $a\ln b$?

Which is larger, $\dfrac{\ln a}{a}$ or $\dfrac{\ln b}{b}$?

How does $\dfrac{\ln x}{x}$ behave as $x$ increases? Looks like a job for the derivative.

  • $\begingroup$ As I mentioned to Shuhao Cao, I'm not supposed to use calculus for this question. Any hint on how to proceed with a number theory proof would be great. $\endgroup$
    – John Marty
    Jun 4, 2013 at 4:00

Another elementary approach:
First notice that it suffices to prove that $n^{n+1}>(n+1)^n$ for $n\ge 3$. We divide both sides by $n^n$ to turn it into $n>(1+1/n)^n$. As a last step, show by induction that $(1+1/n)^n\le n$ for $n\ge3$.

The case $n=3$ is clear. Now, if $(1+1/k)^k\le k$, then $(1+1/(k+1))^{k+1}\le (1+1/(k))^{k+1}\le k(1+1/(k))\le k+1$.

Inform me if anything needs improvements. Thanks in advance.


Most of the answers submitted are fabulous, but let's endeavor to make this as simple as humanly possible, leveraging intuition primarily.

~Begin by asking yourself: Which is larger, $2^3$ or $3^2$? $~8<9$.

~Continue by asking: Which is larger, $3^4$ or $4^3$? $~81>64$. --> This is a reversal of the inequality because the 'weight' of the exponent operation begins to dominate as numbers becomes larger.

~Conclude by asking: Which is larger, $4^5$ or $5^4$? $~1024>625$. --> The disparity is more extreme in absolute terms and likely grows as n increases to ∞.

---The trend suggests strongly that $a>b>2$. Of course, more rigor is needed to definitively confirm that the trend does not reverse, but intuition leaves little doubt.


You must log in to answer this question.

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