-1

Which of the numbers $99^{100}$ & $100^{99}$ is the larger? Solve without using logarithms.

J. W. Tanner
  • 60,406
Michael
  • 255
  • 66
    Why do I have to solve it? Even more, why do I have to solve it without using logarithms? – Asaf Karagila Jan 28 '12 at 23:53
  • 6
    Python: 99**100 > 100*99 == True – orlp Jan 29 '12 at 12:02
  • @nightcracker: Was that a typo??! – cardinal Jan 29 '12 at 13:07
  • @cardinal: Actually yes, but it still gives you the answer (it returns False :P) – orlp Jan 29 '12 at 13:37
  • @nightcracker: Except that it's true both with and without the typo, which indicates there's actually a second typo...or Python's doing something a little strange. – cardinal Jan 29 '12 at 13:46
  • 3
    Why all the upvotes? Maybe I should ask a similar (yet abstractly non-duplicate) question... – The Chaz 2.0 Jan 29 '12 at 14:37
  • @TheChaz: This question is currently at the top of the list of "hot questions" on the SE network. – cardinal Jan 29 '12 at 14:45
  • 9
    -1 This question has showed absolutely no effort whatsoever. @TheChaz I don't get the upvotes either. –  Jan 29 '12 at 14:53
  • 5
    @cardinal: I should explain why @nightcracker's Python returns False. Python allows chained comparisons (1 <= x < 9), so it was interpreting 99**100 > 100**99 == True as one of these. True has an integer comparison value of 1, so this is really 99**100 > 10**99 == 1, which is false. 99**100 > 100**99 and (99**100 > 10**99) == True both return True as you'd expect. – DSM Jan 29 '12 at 15:05
  • @cardinal: Thanks for the insight. It's not the end of the world, but sure reinforces the asking of "un-researched" questions. – The Chaz 2.0 Jan 29 '12 at 15:09
  • @DSM: Thanks. I don't use python really. In the interim, I saw that both of your latter examples worked and then tried 10 > 1 == True and 10 > 2 == True. The difference in returned values allowed me to deduce what must be happening. There must be some good reason they decided to allow this in the language specification, but it still strikes me as odd and, as in this case, error-prone. – cardinal Jan 29 '12 at 15:22
  • 7
    No one has explicitly said that the question is simply rude. Michael, using the imperative when asking for a favor will not pay in the long run. –  Jan 29 '12 at 15:30
  • @TheChaz: More likely, it reinforces the asking of questions that seem "amazing" and "hard" to the (even pretty technically oriented) person with an average math background, but that have several fairly simple and elegant answers. – cardinal Jan 29 '12 at 15:31
  • 1
    @cardinal the reason why that's allowed is because chained expressions simplify a lot of things, while explicitly comparing booleans to True or False is something you shouldn't generally need to do anyway as if condition == True: is more verbose than if condition: for no good reason. – Free Monica Cellio Jan 29 '12 at 17:53
  • in irb: 99**100 > 100**99 #=> true – Mas Adit Jan 30 '12 at 12:13
  • Beronulli inequality kills this problem indeed ! :D – Shivang jindal Oct 23 '12 at 11:07
  • I love this question, not because of the question, but the vast variety of answers you give (well you all give the same answer) but many of you have completely different methods, and I just love that about math. – Asimov Sep 09 '14 at 01:10

8 Answers8

68

Note that $$\begin{align} 99^{100} > 100^{99} &\iff 99 \cdot 99^{99} > 100^{99} \\ &\iff 99 > (100/99)^{99} \\ &\iff 99 > \left( 1 + \frac{1}{99}\right)^{99} \end{align}$$

Since $(1 + \frac{1}{n})^n < 3$ for all integers $n$, the above inequalities are all true. Thus, $99^{100} > 100^{99}$. In general, you should expect that $x^y > y^x$, whenever $y > x$.

David Mitra
  • 74,748
JavaMan
  • 13,153
  • 3
    it holds when $x>2$ – ratchet freak Jan 28 '12 at 23:53
  • 8
    Mr. Man, sir, can you please cite where $(1 + \frac{1}{n})^n < 3$ comes from, or prove it, or something? I'm Rusty on this sort of thing. – Ed Staub Jan 29 '12 at 02:38
  • 7
    @EdStaub: The inequality comes from the calculation of e. e=2.7(ish), and is defined as the limit of that equasion, as n= infinity. – PearsonArtPhoto Jan 29 '12 at 02:54
  • 4
    @EdStaub: Expanding $(1 + \frac{1}{n})^n$ using the binomial theorem, it is enough to show $\frac{1}{2!} + \frac{1}{3!} + \dots + \frac{1}{n!} < 1$. But this follows immediately since $n! \geq 2^{n-1}$ for positive integers $n$ (which you can prove by induction), and hence $$\sum_{k =2}^n \frac{1}{k!} \leq \sum_{k=2}^n \frac{1}{2^{n-1}} < 1$. Adding $2$ to both sides gives the result. There are many places where you can read more (by googling "limit definition of e"), say for example, here: http://www.physicsforums.com/showthread.php?t=176076 – JavaMan Jan 29 '12 at 03:25
  • $\lim_{x \rightarrow 0} e^x-1 = x$. From there, you can get the inequality that $\lim_{x \rightarrow \infty}\left(1 + \frac{1}{x} \right)^{x} =e$ and since it is monotonically increasing function inside the limit, you have that it is true for $x$. – Jalaj Jan 29 '12 at 04:01
  • 1
    @Jalaj $\lim\limits_{x\to 0}e^x-1=0$, not $x$. – Alex Becker Jan 29 '12 at 04:07
  • Sorry, I meant $\lim_{x \rightarrow 0}\frac{e^x-1}{x}=1$. A sheer example of negligence on my part. – Jalaj Jan 29 '12 at 04:11
  • 1
    @ratchet freak: Assuming "it" is $x^y > y^x$ for $y > x$, that would be true for $x \ge e$, not for $x > 2$. – Robert Israel Jan 29 '12 at 08:55
  • @RobertIsrael if you say $x,y \in \mathbb{N}$ then $x>2$ suffices – ratchet freak Jan 29 '12 at 13:43
  • 7
    @Pearsonartphoto: no it doesn't. What you get from the calculation of $e=2.71828\dots$ is that there is an $N$ so that for $x>N$, $\left(1+\frac1x\right)^x<3$. However, what happens for $x\le N$? – robjohn Jan 29 '12 at 22:39
  • @robjohn: $\left(1+\frac1x\right)^x$ is an increasing function of $x$ so you can take $N=0$ – Henry Feb 07 '12 at 13:46
  • 1
    @Henry: my comment was to point out that the fact that $\lim\limits_{n\to\infty}\left(1+\frac1n\right)^n=e<3$ does not give a clue as to $\max\limits_{n\in\mathbb{N}}\left(1+\frac1n\right)^n$. – robjohn Feb 07 '12 at 14:34
29

$99^{100} - 100^{99}$ is:

3560323412732295049306160265725173861897
1207663892369140595737269931704475072474
8187196543510026950400661569100652843274
7182356968017994158571053544917075742738
9035006098270837114978219916760849490001

Since this number is positive, $99^{100}$ is the bigger number.

Hammerite
  • 1,003
  • 34
    Nice reminder that the world has changed since I first did mathematics. – André Nicolas Jan 29 '12 at 01:57
  • 15
    @AndréNicolas I suppose it doesn't use logarithms... – Matt Jan 29 '12 at 02:06
  • 1
    @Matt...uh, why would it? The first term can be calculated with 7 integer multiplications and the second can be done in 8. – cardinal Jan 29 '12 at 02:16
  • 2
    @cardinal Maybe I misinterpreted what AndréNicolas said, but I meant my comment more as a joke than anything; although this method is certainly not what would be expected and doesn't give any insight into which is bigger, it's still 'correct' and doesn't use logarithms. – Matt Jan 29 '12 at 02:19
  • @Matt: I see. I probably misinterpreted your comment then, which I thought might be an ironic one implying logarithms were almost certainly used in the calculation. Sorry about that. Cheers. – cardinal Jan 29 '12 at 02:22
  • 21
    @Matt, technically correct is the best kind of correct. – Hammerite Jan 29 '12 at 02:24
  • 16
    So, Hammerite, is $999999^{1000000}$ larger than $1000000^{999999}$? – Myself Jan 29 '12 at 02:48
  • 9
    @Myself - Yep, http://www.wolframalpha.com/input/?i=999999%5E1000000+%3E+1000000%5E999999. Might have to give it a few seconds ;) – DMan Jan 29 '12 at 04:18
  • 6
    "...Since this number is positive...": as Orwell could have said, many numbers are positive but some are more positive than others... – Georges Elencwajg Jan 29 '12 at 15:05
17

A purely math solution: Using AM-GM inequality:

$$(x+1)^x\times \frac{x}{2} \times \frac{x}{2} < \left(\frac{x(x+1)+x}{x+2}\right)^{x+2}=x^{x+2}.$$

Therefore

$$(x+1)^x < 4x^x$$

and easily we see that $(x+1)^x< x^{x+1}$ for any $x\ge 4$.

Arctic Char
  • 16,007
hiro
  • 291
  • 11
    "purely math solution" as opposed to? – Najib Idrissi Jan 29 '12 at 07:42
  • 2
    In this case, it's no computation. In general, it's just personal sense.

    Since the question is pretty easy with calculus, I expected that an "elementary" solution (secondary-school) is a best fit.

    – hiro Jan 29 '12 at 13:48
  • 1
    @GeorgesElencwajg I don't see how using a hand calculator, as say I did in my answer, doesn't come as something check-able. Or for that matter if we calculate $99^{100}$ by multiplying 99s out on paper, it only involves 100 multiplications. Sure, that's several sheets of paper and probably takes a few hours to do, but we could all do that before we die if we had the persistance and desire to do so. – Doug Spoonwood Jan 29 '12 at 17:24
  • 3
    @Doug: But why would we want to do such a thing, since (a) this is most boring, (b) using our brain a few seconds would solve the question and would furthermore suggest easy generalizations? – Did Jan 29 '12 at 17:40
  • @Doug: As mentioned in a comment hidden under the fold of another answer, the explicit calculation can be done in under ten multiplications for each term. So, if you're masochistic but still not that masochistic... :) – cardinal Jan 29 '12 at 17:51
  • @DidierPiau If we didn't trust our calculating machines, and we wanted to know how much greater one of the numbers is than the other, doing such a calculation by hand might make sense. – Doug Spoonwood Jan 30 '12 at 02:29
  • 1
    @Doug: If I didn't trust my calculating machines, and I wanted to know how much greater one of the numbers is than the other, I would much prefer to know that $(x+1)^x=\varrho(x)x^x$, where the function $x\mapsto \varrho(x)$ is increasing on $x\geqslant1$ from $\varrho(1)=2$ to $\varrho(+\infty)=\mathrm e$. The proof is easier to check and the result is more informative. – Did Jan 30 '12 at 06:13
  • @DidierPiau Sorry, I don't follow. How does that method tell us how much different the numbers are? – Doug Spoonwood Jan 30 '12 at 13:22
  • 1
    @Doug: The ratio between $(x+1)^x$ and $x^{x+1}$ is $\varrho(x)/x$, hence, between $2/x$ and $\mathrm e/x$ for every $x\geqslant1$, and even between $9/(4x)$ and $\mathrm e/x$ for every $x\geqslant2$. For example, this proves that the ratio $100^{99}/99^{100}$ is between $0.02272727$ and $0.02745739$ without knowing the values of $100^{99}$ and $99^{100}$. This even suggests that this ratio is much closer to the latter value than to the former (to wit, the actual ratio is $0.02731999$). – Did Jan 30 '12 at 16:19
  • @ArcticChar What is the point of your very minor editions on a 9 y.o. post ? – zwim Nov 26 '21 at 23:21
14

$x^{x+1}=x x^x$ while for large $x$, $(x+1)^x\sim e x^x$. Since $99>e$, I would say that $99^{100}>100^{99}$.

More Detail:

To show that $(x+1)^x=\left(1+\frac1x\right)^xx^x<ex^x$, without just saying so and without using logarithms, consider the binomial expansion $$ \left(1+\frac1x\right)^x=1+1+\frac12\frac{x-1}{x}+\frac16\frac{(x-1)(x-2)}{x^2}+\frac{1}{24}\frac{(x-1)(x-2)(x-3)}{x^3}+\dots $$ and note that, at least for $x\in\mathbb{N}$, each term is monotonically increasing. Thus, $\left(1+\frac1x\right)^x$ monotonically increases to $e=\sum\limits_{k=0}^\infty\frac{1}{k!}$.

robjohn
  • 345,667
12

Proof that $x^y > y^x$ for all $y > x > e$: Raising both sides to the ${1 \over xy}$ power, this is equivalent to $x^{1 \over x} > y^{1 \over y}$. The derivative of $x^{1 \over x}$ with respect to $x$ is ${\displaystyle {1 - \ln(x) \over x^2} x^{1 \over x}}$, which is negative whenever $\ln(x) > 1$ i.e. when $x > e$. Thus $x^{1 \over x}$ is a decreasing function of $x$ for $x > e$.

Yeah I know, I used logarithms. But someone needed to say this ;)

Zarrax
  • 44,950
4

I cheat and use a basic fact about $e$.

$${99^{100}\over 100^{99}} = 99\left({99\over 100}\right)^{99}\approx {99\over e} > 1.$$

ncmathsadist
  • 49,383
  • Nice cheat, what is the general formula for this fact about $e$ please? Thanks. – NoChance Jan 29 '12 at 15:09
  • 1
    For large $n$, $(1 + \lambda/n)^n \sim e.$$ – ncmathsadist Jan 29 '12 at 15:18
  • 2
    @ncmathsadist Your comment has a small typo. Surely, you meant to write either $(1 + \lambda/n)^n \sim e^{\lambda}$ or $(1 + 1/n)^n \sim e$. – Srivatsan Jan 29 '12 at 17:27
  • Yes, $(1 + \lambda/n)^n \sim e^\lambda$ – ncmathsadist Jan 29 '12 at 17:36
  • 6
    Somebody should probably mention that such asymptotics, per se, can give NO information about the $n$th term of a sequence, even for $n=99$, hence the (true) basic fact used here CANNOT prove the desired inequality. – Did Jan 29 '12 at 19:19
  • That'd be true, but I think the approximation is illuminating. It is not terribly hard to establish that for any $n$, $2\le (1 + 1/n) \le 3$. Massage this a bit and it will make the argument airtight. – ncmathsadist Jan 29 '12 at 20:33
  • 3
    Starting from $2\le1+1/n\le3$, I doubt one can go far. And if you wish to use $(1+1/n)^n\le3$ for every $n$, then mention it... – Did Jan 30 '12 at 16:29
2

$100^{99}$=$(10*10)^{99}$=$(10^{99})(10^{99})$=$10^{198}$ exactly.

$99^{100}$=$(9*11)^{100}$=$(9^{100})(11^{100})$ exactly. My "hand" calculator approximates $9^{100}$ as about $(2.656)(10^{95})$.

11=(2)(2)(2.75).

$2^{100}$ equals about (1.267)($10^{30}$), $2.75^{100}$ equals about (8.575)($10^{43}$). Dropping the coefficients here we can thus approximate ($11^{100}$) by a lower bound of ($10^{30}$)($10^{30}$)($10^{43}$)=$10^{103}$.

Keeping the coefficients on the approximation of $9^{100}$ we have a lower bound for $99^{100}$ as $(2.656)((10^{95}$)($10^{103}$))=(2.656)($10^{198}$) which comes as greater than $10^{198}$.

So, $99^{100}$>$100^{99}$.

Note that if we kept the coefficients in here, we would also have more of an idea as to how much greater $99^{100}$ is than $100^{99}$. Some of the other answers do this, some don't. This doesn't necessarily make this answer better though, since such information might come as extraneous to the problem.

-1

From experimenting with small numbers:

scala> (0 to 5).map (x=> (math.pow (x, x+1) - math.pow (x+1, x))).mkString ("; ")
res18: String = -1.0; -1.0; -1.0; 17.0; 399.0; 7849.0

scala> (0 to 5).map (x=> (math.pow (x, x+1), math.pow (x+1, x))).mkString ("; ")
res19: String = (0.0,1.0); (1.0,2.0); (8.0,9.0); (81.0,64.0); (1024.0,625.0); (15625.0,7776.0)

you can conclude, that the first one is growing faster than the second. Of course this is only an indication.

  • Do you have computer code in this answer? I simply don't know how to interpret "scala> (0 to 5).map" and the like. – Doug Spoonwood Jan 30 '12 at 04:40
  • @DougSpoonwood: Yes, Scala code. (0 to 5) creates a Range, a collection of the numbers (0, 1, ..., 5). The map (x => takes each of them, and puts them, named x, into a function, math.pow (x, x+1) - ... so for the first element it is math.pow (0, 0+1) or 0^(1), then 1^2, 2^3 and so on minus 1^0, then 2^1, 3^2 and so on. mkString is only used for formatting the output a bit. (3^4-4^2) = 81.0-64.0 = 16 – user unknown Jan 30 '12 at 14:50