16

Let $a=\log_5(\log_4 3)$ and $b=\log_6(\log_6 3)$.
How to compare the two numbers if numerical solution is not used?

The following is my thoughts, but it did not work.

From the question, I get $5^a=\log_4 3$ and $6^b=\log_6 3$.

As $1 >\log_4 3 > \log_6 3>0$, I get $a < 0, b < 0$ and $5^a > 6^b$.

Then it holds $a > b \log_5 6$.

But $b \log_5 6 > b$ does not hold as $b<0$. I can not go on with it. Can someone give any new ideas?

Игорь
  • 416
  • 9
Elysia
  • 171
  • 1
    Here $a=-0.14451407573$ while $b=-0.27299995427$ , just fyi ... – Prem Dec 03 '25 at 08:39
  • 6
    Yes , I know that "Numerical Solution" is not wanted , @AnneBauval , I just gave it "fyi == for your information" to confirm what-ever answer we get with "Analytical Solution" for the verification – Prem Dec 03 '25 at 10:14
  • 2
    can't we just use monotonicity? – SBF Dec 04 '25 at 13:57

3 Answers3

15

My solution might be a little complicated.

$$ \begin{aligned} f(x)&=\log_5(\log_4x)-\log_6(\log_6x)\\ &=\frac{\ln\dfrac{\ln x}{\ln4}}{\ln5}-\frac{\ln\dfrac{\ln x}{\ln6}}{\ln6}\\ &=\ln(\ln x)\left(\frac{1}{\ln5}-\frac{1}{\ln6}\right)+\left[\frac{\ln(\ln6)}{\ln6}-\frac{\ln(\ln4)}{\ln5}\right] \end{aligned} $$ is monotonically increasing, so $f(3)>f(\mathrm{e})=\dfrac{\ln(\ln6)}{\ln6}-\dfrac{\ln(\ln4)}{\ln5}$.
$$ \begin{aligned} g(x)&=\frac{\ln(\ln x)}{\ln x}\\ g'(x)&=\frac{\ln x\dfrac{1}{x\ln x}-\ln(\ln x)\dfrac{1}{x}}{(\ln x)^2}\\ &=\frac{1-\ln(\ln x)}{x(\ln x)^2} \end{aligned} $$ When $x<\mathrm{e}^{\mathrm{e}}$, $g'(x)>0$ so $g(x)$ is monotonically increasing and $g(4)<g(6)$, and thus $$ \frac{\ln(\ln6)}{\ln6}>\frac{\ln(\ln4)}{\ln4}>\frac{\ln(\ln4)}{\ln5} $$ $f(3)>f(\mathrm{e})>0$.

JC Q
  • 2,783
  • 6
    @AnneBauval Thanks for your suggestion. I have said many times and I will repeat, that according to the how to answer page, not every not-quite-well-asked question should be avoided to answer. "Lacking details" is a kind of "bad asked", but not "unworthy of an answer" because the problem can be identified clearly enough, not "are unclear or lacking specific details that can uniquely identify the problem". Sometimes, the OP has done his/her best but is just unable to provide more. – JC Q Dec 03 '25 at 07:55
  • 1
    @JCQ Thanks a lot for your kind help! I'm a new user here and not good at decribing a well-asked question. I'll try to get better at it! – Elysia Dec 03 '25 at 08:06
  • 9
    Anyway, I don't quite understand the two downvotes here. This is not a means to criticize users who answer low-quality questions. This answer is not "not useful". – Bowei Tang Dec 03 '25 at 10:06
  • @BoweiTang I am one of the (3 by now) downvoters. Not for answering a low-quality (now improved) question, but because I found the answer too complicated to be useful, especially to an OP who was "just unable to provide more". – Anne Bauval Dec 03 '25 at 13:05
  • 11
    @AnneBauval Since (having deleting your answer) you've now realised that the problem is more complicated than you thought, maybe you should reconsider your downvote. – Especially Lime Dec 03 '25 at 13:16
9

We provide an elementary proof for the readers.


Observe that :

$$ \begin{align}&\frac 23>\log_32>\frac 12\\ \iff &3^4>2^6>3^3\end{align} $$

Using the above inequality, we obtain the sufficient bounds :

$$ \begin{align}\log_5(\log_43)&=-\log_5(\log_34)\\ &=-\log_5\left(2\log_32\right)\\ &>-\log_5\left(2\cdot \frac 23\right)\\ &=-\log_5\left(\frac 43\right)\end{align} $$

and

$$ \begin{align}\log_6(\log_63)&=-\log_6(\log_36)\\ &=-\log_6\left(1+\log_32\right)\\ &<-\log_6\left(1+\frac 12\right)\\ &=-\log_6\left(\frac 32\right)\end{align} $$


Then, we need to prove :

$$\begin{align}\log_6\left(\frac 32\right)>\log_5\left(\frac 43\right)\end{align}$$

Since

$$\begin{align}\log_6\left(\frac 32\right)>\frac 15&\iff \frac {3^5}{2^5}>6\\ &\iff 243>192\end{align}$$

It suffices to show that :

$$\begin{align}5^{\frac 15}>\frac 43&\iff 5>\frac {4^5}{3^5}\\ &\iff 1215>1024\end{align}$$

which completes the proof.

Игорь
  • 416
  • 9
  • Comments have been moved to chat; please do not continue the discussion here. Before posting a comment below this one, please review the purposes of comments. Comments that do not request clarification or suggest improvements usually belong as an answer, on [meta], or in [chat]. Comments continuing discussion may be removed. – Shaun Dec 07 '25 at 17:37
-2

$\log_a x = \frac {\ln x}{\ln a}$

$\log_5(\log_4 3) = \frac {\ln 3 - \ln 4}{\ln 5}$

And

$\log_6(\log_6 3) = \frac {\ln 3 - \ln 6}{\ln 6}$

if $\log_5(\log_4 3) > \log_6(\log_6 3)$

$\frac {\ln(\frac{\ln 3}{\ln 4})}{\ln 5} > \frac {\ln (\frac {\ln 3}{\ln 6})}{\ln 6}$

$\frac {\ln(\ln 3) - \ln(\ln 4)}{\ln 5} > \frac {\ln (\ln 3) - \ln(\ln 6)}{\ln 6}$

$\frac {\ln \ln 3}{\ln 5} - \frac{\ln \ln 4}{\ln 5} > \frac{\ln \ln 3}{\ln 6} - \frac {\ln \ln 6}{\ln 6}$

Comparing terms

$\frac {\ln \ln 3}{\ln 5} > \frac{\ln \ln 3}{\ln 6}$

and

$- \frac{\ln \ln 4}{\ln 5} > - \frac {\ln\ln 6}{\ln 6}$

The last line is true, but it isn't obvious.

user317176
  • 12,546
  • No. If the base is greater than zero, $\log x$ is an increasing function. The result may be negative for $0<x<1$, but it is still increasing. – user317176 Dec 05 '25 at 00:17
  • 2
    I think there's a small misunderstanding. While $\log_a x$ is increasing in $x$ for fixed $a>1$, comparing $\log_b x$ for different bases $b$ depends on whether $x>1$ or $0<x<1$. Here $y=\log_6 3\in(0,1)$, so for $1<5<6$ we have $\log_5 y<\log_6 y$. Hence $\log_5(\log_6 3)<\log_6(\log_6 3)$, so the second inequality in your chain is reversed. – Игорь Dec 05 '25 at 00:34
  • So, the problem isn't that $\log x$ is increasing when $x< 1$ it is that the first line is wrong if $x<1$? – user317176 Dec 05 '25 at 00:38
  • That's right. The general inequality $\log_m x > \log_n x$ for $1<m<n$ only holds when $x>1$. When $0<x<1$, it is reversed: $\log_m x < \log_n x$. Here, $x = \log_6 3 \in (0,1)$, so the first line of your chain should be reversed. – Игорь Dec 05 '25 at 00:47
  • Okay, here is a new proof. It is not as elegant, but at least it isn't wrong. – user317176 Dec 05 '25 at 00:47
  • 3
    There is a small technical issue in your calculation. You wrote

    $$\log_5(\log_4 3) = \frac{\ln 3 - \ln 4}{\ln 5}, \quad \log_6(\log_6 3) = \frac{\ln 3 - \ln 6}{\ln 6},$$

    but actually

    $$\log_4 3 = \frac{\ln 3}{\ln 4} \quad\text{so}\quad \ln(\log_4 3) = \ln\left(\frac{\ln 3}{\ln 4}\right) \neq \ln 3 - \ln 4.$$

    Since the subsequent steps rely on this equality, the comparisons and conclusion are not valid.

    – Игорь Dec 05 '25 at 01:04
  • 3
    "The last line is true, but it isn't obvious." If it isn't obvious, and you don't establish it, then you haven't written an answer. – Gerry Myerson Dec 05 '25 at 03:20
  • 1
    @GerryMyerson This question really deserves more attention. It’s surprisingly subtle and can mislead even careful readers. At first glance it looks completely trivial—I initially thought so myself. For instance, a recent answer falls into the same trap. It’s genuinely a well-crafted problem. – Игорь Dec 05 '25 at 03:59