12

How can I prove that $7^{31}$ is bigger than $8^{29}$?

I tried to write exponents as multiplication, $2\cdot 15 + 1$, and $2\cdot 14+1$, then to write this inequality as $7^{2\cdot 15}\cdot 7 > 8^{2\cdot 14}\cdot 8$. I also tried to write the right hand side as $\frac{8^{31}}{8^2}$.

Michael Lugo
  • 22,354
mrJoe
  • 379
  • 2
    This inequality is equivalent to $31<29 \log_7 8$ or $\frac{31}{29}>\log_7 8$, so if you can find the logarithm and then multiply it by $29$ then you've got it. How to find the logarithm numerically is a substantial question in its own right. ${}\qquad{}$ – Michael Hardy Jul 07 '15 at 14:25
  • This fails:$$\left( 1 + \frac{1}{7} \right)^{7\cdot 4 \frac{1}{7}} \leq e^{4 \frac{1}{7}} \geq 7^2$$ Any better ideas? – abnry Jul 07 '15 at 14:41
  • Maybe you could try to use the binomial theorem. Note that $7=8-1$ and $8=7+1$. – Surb Jul 07 '15 at 16:52
  • I'd love to see a proof using the binomial theorem, but it seems that the two values are two close (relatively) for simple bounds to work. – lhf Jul 07 '15 at 16:52
  • @lhf maybe comparing pairwise the terms in the sums or something like that could help... I'll think about it – Surb Jul 07 '15 at 16:54
  • 2
    I think maybe this problem is from my old problem when Christian Blatter solution a step http://math.stackexchange.com/questions/431461/prove-that-sqrt7-sqrt8-sqrt8-sqrt7?lq=1 – math110 Jul 07 '15 at 16:42

8 Answers8

17

The following (not particularly elegant) proof uses reasonably basic multiplication and division.

We need to show that $7^{31} > 8^{29}$, i.e. that $\dfrac{7^{31}}{8^{29}}>1$.

We have: $\dfrac{7^{31}}{8^{29}}=\dfrac{7^{2}\cdot7^{29}}{8^{29}}=\dfrac{7^{3}}{8}\Big(\dfrac{7}{8}\Big)^{28}=\dfrac{7^{3}}{8}\Big(\dfrac{7^4}{8^4}\Big)^{7}=\dfrac{7^{3}}{8}\Big(\dfrac{2401}{4096}\Big)^{7} > \dfrac{7^{3}}{8}\Big(\dfrac{2400}{4100}\Big)^{7}=\dfrac{7^{3}}{8}\Big(\dfrac{24}{41}\Big)^{7}=\dfrac{7^{3}}{8}\dfrac{24}{41}\Big(\dfrac{24}{41}\Big)^{6}=\dfrac{3 \cdot 7^{3}}{41}\Big(\dfrac{24^2}{41^2}\Big)^{3}=\dfrac{3 \cdot 7^{3}}{41}\Big(\dfrac{576}{1681}\Big)^{3}>\dfrac{3 \cdot 7^{3}}{41}\Big(\dfrac{576}{1683}\Big)^{3}=\dfrac{3 \cdot 7^{3}}{41}\Big(\dfrac{9 \cdot 64} {9\cdot 187}\Big)^{3}=\dfrac{3 \cdot 7^{3}}{41}\Big(\dfrac{64} {187}\Big)^{3}=\dfrac{2^{18} \cdot 3 \cdot 7^3}{11^3 \cdot 17^3 \cdot 41}=\dfrac{2^{18} \cdot 3 \cdot 7^3}{1331 \cdot 17^3 \cdot 41}>\dfrac{2^{18} \cdot 3 \cdot 7^3}{1332 \cdot 17^3 \cdot 41}=\dfrac{2^{16} \cdot 7^3}{111 \cdot 17^3 \cdot 41}=\dfrac{2^{16} \cdot 7^3}{4551 \cdot 17^3}>\dfrac{2^{16} \cdot 7^3}{4557 \cdot 17^3}=\dfrac{2^{16} \cdot 7}{93 \cdot 17^3}=\dfrac{2^{16} \cdot 7}{1581 \cdot 17^2}>\dfrac{2^{16} \cdot 7}{1582 \cdot 17^2}=\dfrac{2^{15}}{113 \cdot 17^2}=\dfrac{2^{15}}{1921 \cdot 17}>\dfrac{2^{15}}{1924 \cdot 17}=\dfrac{2^{13}}{481 \cdot 17}=\dfrac{8192}{ 8177}>1. \quad\square$

9

Others may bristle at this "proof," but:

$$7^{31} = 157,775,382,034,845,806,615,042,743 \\ 8^{29} = 154,742,504,910,672,534,362,390,528$$

If all else fails, just calculating the expressions and comparing them will work. This particular problem is only mildly tedious to attack this way if you have pen/paper.

John
  • 26,319
  • 5
    No need for scare quotes. That's a perfectly valid proof. – Eric M. Schmidt Jul 07 '15 at 16:26
  • 1
    This is in a sense a valid proof, but I can't see how you got the numbers or why I should believe you got them right. This is a proof, but not an easily human-verifiable one. – Joonas Ilmavirta Jul 07 '15 at 16:38
  • 3
    Slightly more readable: $7^{31} =101202171144551276401762715267_8$. –  Jul 07 '15 at 16:54
  • 1
    Not for me... similar situation http://math.stackexchange.com/questions/562538/which-is-larger-20-or-240/562543#562543 – JP McCarthy Jul 07 '15 at 17:06
  • 4
    And for those of us who once memorized a few logarithms: $\log 7\approx0.84510$, and $\log 8\approx0.90309$. By an easy hand computation $31\cdot0.84510=26.1981$, and $28\cdot0.90309=26.18961$. The difference between $26.1981$ and $26.18961$ is too large to be the result of roundoff error. – Brian M. Scott Jul 07 '15 at 22:48
  • 1
    @JpMcCarthy: It clearly is a valid proof, and it’s easily (if tediously) checked; claiming otherwise is simply denying the facts. – Brian M. Scott Jul 07 '15 at 22:56
  • @BrianM.Scott I know it is a proof... and to be honest I am not sure if there is a proof that doesn't utilise significant computation... where significant is not well-defined in this context. – JP McCarthy Jul 07 '15 at 23:45
2

Here's an idea for a proof that uses some, but hopefully not too much, arithmetic. If you know the power series

$$-\ln(1-x)=x+{1\over2}x^2+{1\over3}x^3+\cdots$$

then you can get started by finagling the desired inequality as follows:

$$\begin{align} 7^{31}\gt8^{29}&\iff31\ln7\gt29\ln8\\ &\iff31\ln(8-1)\gt29\ln8\\ &\iff31\left(\ln8+\ln\left(1-{1\over8} \right) \right)\gt29\ln8\\ &\iff2\ln8\gt-31\ln\left(1-{1\over8} \right)\\ &\iff6\ln\left(1-{1\over2} \right)\gt-31\ln\left(1-{1\over8} \right)\\ &\iff6\left({1\over2}+{1\over8}+{1\over48}+{1\over64}+\cdots \right)\gt31\left({1\over8}+{1\over128}+{1\over1536}+\cdots \right) \end{align}$$

The final ingredient is to use the inequality

$${1\over n}x^n+{1\over n+1}x^{n+1}+\cdots\lt{1\over n}\left(x^n+x^{n+1}+\cdots \right)={x^n\over n(1-x)}$$

in truncating the infinite sum on the right. It may take a couple of attempts to find truncations that work.

Added later (after seeing math110's answer): I had quite forgotten my own answer (from two years ago) to the problem of proving $\sqrt7^\sqrt8\gt\sqrt8^\sqrt7$. In it, I showed all the steps necessary to establish

$$-\ln\left(1-{1\over8} \right)\lt{137\over1024}\quad\text{and}\quad6\ln2\gt{1063\over256}$$

So all that remains here is to note that

$$4\cdot1063=4252\gt4247=31\cdot137$$

Whew!

Barry Cipra
  • 79,832
  • 1
    It's too complicated for me, but thanks for your anwser. – mrJoe Jul 07 '15 at 18:39
  • You could use $$-31\ln\left(\frac78\right) = 31\ln\left(\frac87\right) = 31\left(\frac17 - \frac1{98} + \dots\right)$$, which is an alternating series and thus easily bounded. – NovaDenizen Jul 08 '15 at 01:24
1

The expression

$$7^{31}>8^{29}$$

Is equivalent to $$31\ln(7)>28\ln(8)$$ where $\ln$ denotes the natural logarithm. As $7>e$, this is equivalent to $$\frac{31}{28}>\frac{\ln(8)}{\ln(7)}$$

The above relation can then be easily verified by calculator.

Alternatively, along a similar vein

$$7^{31}=\left(7^{\frac{31}{29}}\right)^{29}$$

As $7^{\frac{31}{29}}\approx8.01>8$ (via my pocket calculator), the inequality follows.

Basically I am still showing this through computation, I'm just trying to make the computations a bit nicer.

1

We have to prove $(\frac {7}{8})^{29}>\frac {1}{49}$. Write

$(\frac {7}{8})^{29}=(1-\frac 18)^{29}=[1- (\frac 18)^{29}]+\binom {29} {1}[1-(\frac 18)^{28}]+….+\binom {29} {14}[1-(\frac18)^{14}]$

Just the first term in this sum of positive is already greater than $\frac {1}{49}$. One has $[1- (\frac 18)^{29}]>\frac {1}{49}\iff 8^{29}-1 > \frac {8^{29}}{49}$ which is quite clear.

Piquito
  • 29,594
  • Is this the binomial formula? What's happening in this answer? – G Tony Jacobs Sep 11 '17 at 14:35
  • @G Tony Jacobs: Do you doubt that $1-\dfrac{1}{49}\gt (\dfrac 18)^{29}\large?$ – Piquito Sep 11 '17 at 18:36
  • No, I don't. That has nothing to do with my comment. I'm just asking what formula you used here. – G Tony Jacobs Sep 11 '17 at 18:37
  • 1
    $(a+b)^n=[a^n+b^n]+\binom {n} {1}[a^{n-1}b+ab^{n-1}]+\binom {n} {2}[a^{n-2}b^2+a^2b^{n-2}]….....$ – Piquito Sep 11 '17 at 18:47
  • Oooooooh! I see it now. I'm not used to seeing the terms grouped like that; thank you for clarifying. – G Tony Jacobs Sep 11 '17 at 18:51
  • You are welcome. – Piquito Sep 11 '17 at 18:53
  • 3
    Unfortunately, the sum is pretty much all wrong. Apart from the first term, the subsequent ones are missing powers of $(-1/8)$ in the first addend and the signs between the addends should alternate. The terms in the corrected sum are clearly not all positive - in fact, if they were, wouldn't it be strange that $(7/8)^{29}$ is almost equal to $1$ (the first term in the sum) or even much greater than $1$ (second term)? – Peter Košinár Mar 25 '21 at 18:57
1

First, $7^5=16807$ and $2^{14}=16384$. And we have $\dfrac{7^{31}}{8^{29}}=\left(\dfrac{7^5}{2^{14}}\right)^6\times\dfrac78$.

Now, for $x\ge0$ and integer $n\ge0$, we have $(1+x)^n\ge1+nx$, hence

$$\left(\frac{7^5}{2^{14}}\right)^6\ge1+6\left(\frac{7^5}{2^{14}}-1\right)=1+\frac{6\times423}{2^{14}}=1+\frac{3\times423}{2^{13}}=\frac{9461}{8192}$$

Finally,

$$\frac{9461}{8192}\times\frac78=\frac{66227}{65536}>1$$

Therefore,

$$\frac{7^{31}}{8^{29}}>1$$

0

Lemma. $5^{15} > 2^{32}\cdot7.$

Proof. The first eight entries in the sixteenth row of Pascal's Triangle are easily computed recursively by hand (or may be read from a table). Using them, the Binomial Theorem gives \begin{gather*} \frac{5^{15}}{2^{30}} =\left(1 + \frac14\right)^{15} > \sum_{i=0}^7\frac{\binom{15}i}{4^i} = 1 + \frac{15}4 + \frac{105}{4^2} + \frac{455}{4^3} + \frac{1365}{4^4} + \frac{3003}{4^5} + \frac{5005}{4^6} + \frac{6435}{4^7} \\ = \frac{6435 + 4(5005 + 4(3003 + 4(1365 + 4(455 + 4(105 + 4(15 + 4))))))}{4^7} \\ = \frac{6435+4(5005+4(3003+4(1365+4\cdot1179)))}{4^7} = \frac{6435+4(5005+4(3003+4\cdot6081))}{4^7} \\ = \frac{6435+4(5005+4\cdot27327)}{4^7} = \frac{6435+4\cdot114313}{4^7} = \frac{463687}{4^7} > 4\cdot7, \end{gather*} because $4^8\cdot7 = 65536\cdot7 = 458752 < 463687.\ \square$

We have $7^4 = 2401 > 2400 = 2^5\cdot3\cdot5^2,$ so $7^{32} > 2^{40}\cdot3^8\cdot5^{16}.$

But $3^8\cdot5 = 6561\cdot5 = 32805 > 32768 = 2^{15},$ so $7^{32} > 2^{55}\cdot5^{15}.$

Now the lemma gives $7^{32} > 2^{87}\cdot7,$ whence $7^{31} > 2^{87} = 8^{29}.\ \square$

  • N.B. The mention of Pascal's Triangle was not intended to suggest (as I've just realised with horror it might) that one has to compute the first fifteen rows before computing the sixteenth! What I had in mind was simply the recursion $$ \binom{n}{i+1} = \binom{n}{i}\cdot\frac{n-i}{i+1} \quad (i = 0, 1, 2. \ldots, n - 1). $$ – Calum Gilhooley Dec 21 '21 at 00:51
-1

I tried just approximate the numbers and I think I found a very ugly way to prove the inequality. Are there any mistakes?

We have $7^4>24\cdot10^2$ so $7^{31}=7^3\cdot(7^4)^7>7^3\cdot(24\cdot10^2)^7=2^{35} \cdot 3^7 \cdot 5^{14} \cdot 7^3$. Also $8^{29}=2^{87}$. So one has to prove the inequalities $2^{52}<3^7 \cdot 5^{14} \cdot 7^3=750141\cdot5^{14}<750142\cdot5^{14}$ or $2^{51}<375071\cdot5^{14}<375072\cdot5^{14}$.

This mean we have to prove that $2^{51}<2^5 \cdot 3 \cdot 3907\cdot5^{14}$ or $2^{46}<3\cdot3907\cdot5^{14}<3\cdot3908\cdot5^{14}=3\cdot2^2\cdot977\cdot5^{14}$ or equivalently $2^{44}<3\cdot977\cdot5^{14}$. But $3\cdot977\cdot5^{14}<3\cdot976\cdot5^{14}=3\cdot2^4\cdot61\cdot5^{14}$ so one has to prove that $2^{40}<3\cdot61\cdot5^{14}$.

But $3\cdot61\cdot5^{14}<3\cdot62\cdot5^{14}$ so one has to prove that $2^{40}<3\cdot62\cdot5^{14}$ or $2^{39}<3\cdot31\cdot5^{14}$. But $3\cdot31\cdot5^{14}<3\cdot2^5\cdot5^{14}$ so one has to prove that $2^{34}<3\cdot5^{14}$.

Now, lets approximate some square roots:

The inequality above is equivalent to $2^{17}<\sqrt{3}\cdot5^7$. But $\sqrt{3}>1.7$ so one has to prove that $2^{17}<1.7\cdot5^7$. This is the same as $\sqrt{2}\cdot2^8<\sqrt{1.7\cdot5}\cdot5^3$ or $256\cdot\sqrt{2}<\sqrt{8.5}\cdot125$. This is the same as $(256/125)^2<8.5/2$ or $256^2\cdot2<8.5\cdot125^2$ or $131072<132812.5$.

  • This proof doesn't work. You proved that $2^{52}<750142\cdot 5^{14}$, which doesn't imply that $2^{52}<750141\cdot 5^{14}$. – Batominovski Jul 07 '15 at 23:48