0

This limit is proposed to be solved without using the L'Hopital's rule or Taylor series: $$ \lim_{x\to a}\frac{a^x-x^a}{x-a}, $$ where $a>0$. I know how to calculate this limit using the L'Hopital's rule: $$ \lim_{x\to a}\frac{a^x-x^a}{x-a}= \lim_{x\to a}\frac{a^x\ln a-ax^{a-1}}{1}=a^a\ln a- a^a, $$ but I have no idea how to calculate this limit without using derivatives.

  • 3
    Also: https://math.stackexchange.com/q/3444103/42969, https://math.stackexchange.com/q/2536965/42969 – all found with Approach0 – Martin R Nov 19 '21 at 12:46
  • Why do you want to avoid L'Hospital in this case ? It is a useful rule , and here I doubt there is any alternative. – Peter Nov 19 '21 at 12:48
  • @Peter Because such is the task in a book of problems. Problems for L'Hôpital's rule are in another section – mathstudent1 Nov 19 '21 at 13:22
  • @Martin R No, it doesn't. The solutions in the cited answers use derivatives in some way (Taylor series or something else) I need a solution without derivatives – mathstudent1 Nov 19 '21 at 13:25
  • You somehow need to describe the behavior of $x^a$ and $a^x$ for $x$ close to $a$. I wonder how that can be achieved without using anything related to derivatives. – Martin R Nov 19 '21 at 13:52
  • @mathstudent1, if you don't want to use derivatives then ok. Just bring them to the form of limit definition of derivatives and simplify the limit. Check out for proofs of derivatives of different functions using limits online. – 2 is even prime Nov 19 '21 at 13:55
  • 2
  • @Invisible Yes, the last link does answer the question, thanks – mathstudent1 Nov 19 '21 at 16:38
  • @mathstudent1: Note that that solution uses what they call the “remarkable limit”: $\lim_{ x \to 0}\frac{a^{kx}-1}{x}=k;\ln(a)$. How would you prove that without using derivatives? – Martin R Nov 19 '21 at 18:03
  • @Martin R It follows from $\lim_{x \to \infty}\left(1 + \frac{1}{x}\right)^x = e$, which is also proved without using derivatives – mathstudent1 Nov 19 '21 at 18:38

3 Answers3

1

The limit is same thing to write as, $$\lim_{k\to 0}\frac{a^{a+k}-(a+k)^{a}}{k}$$ Now, I will use some non standard Analysis $$st\left(\frac{a^{a+\epsilon}-(a+\epsilon)^{a}}{\epsilon}\right)$$ Where $\epsilon$ denotes an infinitesimal. And $st(...)$ for standard part function. In Euler's famous book, "Indroductio in analysin infnitorum" chapter 7 he derives the formula that $a^{\epsilon}=1+\ln(a)\epsilon$ Putting this in required limit and using the binomial theorem, $$st\left(\frac{1}{\epsilon}\left(a^{a}+a^{a}\epsilon\ln a-a^{a}-\epsilon a^{a}-\binom{a}{2}a^{a-2}(\epsilon)^{2}-...\right)\right)$$ Notice that $a^{a}$ cancels off each other. The infinitesimal in the denominator cancels with epsilon with the numerator with $a^{a}(\ln a-1)$. All other terms will get $\epsilon^{n}, n\in\mathbb{N}$ as their coefficients. So the required answer is, $a^{a}(\ln a-1)$. $$\lim_{x\to a}\frac{a^x-x^a}{x-a}=a^{a}(\ln a-1)$$

  • wow ... I'm afraid my level of knowledge does not allow me to understand your answer. I'll try to come back to it later when my level rises. – mathstudent1 Nov 19 '21 at 16:18
1

$$ L=\lim_{x\to a}\frac{a^x-x^a}{x-a}=\lim_{h\to 0}\frac{a^{a+h}-(a+h)^a}{h}=\lim_{h\to 0}a^a\frac{a^h-(1+h/a)^a}{h}.$$ Use binomial expansion $(1+z)^k=1+kz+k(k-1)z^2/2!+...$ and $\lim_{z\to 0} \frac{a^z-1}{z}=\ln a$ $$\implies L=a^a\lim_{h \to 0}\frac{a^h-(1+h+ h^2a(a-1)/a^2}{h}=a^a[\lim_{h\to 0}\frac{a^h-1}{h}-1]=a^a (\log a-1)$$

Z Ahmed
  • 43,235
  • Yes, this is fine if $a$ is an integer, but the assignment only says that $a>0$ – mathstudent1 Nov 19 '21 at 16:24
  • No, $a$ has not been assumed to be integer here. – Z Ahmed Nov 19 '21 at 17:46
  • @Z Ahmed Then this is not a binomial expansion, but a Taylor expansion, as far as I know – mathstudent1 Nov 19 '21 at 18:35
  • OK, I don't mind if you call it Taylor expansion, but the solution is unfazed. – Z Ahmed Nov 19 '21 at 19:05
  • @Z Ahmed The point is not how it is called, but whether it can be obtained without using derivatives. If it is possible, I will be surprised and accept your answer. – mathstudent1 Nov 19 '21 at 19:59
  • $(1+x)^p=1+px+p(p-1)x^2/2!+p(p-1)(p-2)x^3/3!+....$ is binomial series. If $p$ is a natural number it is a polynomial of degree $p$. If not then it is an infinite series which is vailid only if $|x|<1$. There is no need to make any reference to Taylor series which is based on derivatives where as binomail expansion is purely algebraic it is taught well before derivatives are taught. – Z Ahmed Nov 20 '21 at 02:19
  • @Z Ahmed Could you please clarify how this can be proved algebraically without using derivatives for any $p>0$? I am not aware of such proof and it would be interesting to look at it. – mathstudent1 Nov 20 '21 at 04:09
1

You do not need L'Hopital or Taylor series here. If you rewrite the numerator as $(a^x-a^a)-(x^a-a^a)$ and use the definition of the derivative, you have

$$\lim_{x\to a}{a^x-x^a\over x-a}=\lim_{x\to a}{a^x-a^a\over x-a}-\lim_{x\to a}{x^a-a^a\over x-a}=f'(a)-g'(a)$$

where $f(x)=a^x$ and $g(x)=x^a$ have derivatives $f'(x)=a^x\ln a$ and $g'(x)=ax^{a-1}$, respectively, hence

$$\lim_{x\to a}{a^x-x^a\over x-a}=a^a(\ln a-1)$$

This may look like L'Hopital, but it's not. Indeed, anytime you have a L-Hopital-like problem, $\lim_{x\to a}f(x)/g(x)$ with $g'(a)\not=0$, you don't need to invoke L'Hopital to evaluate the limit; L'Hopital can be a convenience in such settings, but it is really only needed when $\lim_{x\to a}f'(x)$ and $\lim_{x\to a}g'(x)$ are also both $0$.

Barry Cipra
  • 79,832
  • Thanks, but this is not the answer I was looking for as it uses derivatives. The answer I needed was found by @Invisible: https://math.stackexchange.com/q/2648245/721644 – mathstudent1 Nov 19 '21 at 16:41
  • @mathstudent1, the answer there basically does the same thing, it just calls them "remarkable limits" instead of "derivatives." – Barry Cipra Nov 19 '21 at 17:53
  • I disagree with you. Both "remarkable" limits are known to students before the introduction of the concept of a derivative – mathstudent1 Nov 19 '21 at 18:32