14
$\begingroup$

A few months ago, while at school, my classmate asked me this curious question: What does $\frac{x^{n}-y^{n}}{n}$ tend to as $n$ tends to $0$? I thought for a few minutes, became impatient, and asked "What?" His reply, log$\frac{x}{y}$, was surprising, but his purported 'proof' was more surprising:

Consider $\lim\limits_{n \to 0}\,\int_y^x t^{n-1}\, dt$. "Pushing the limit into the definite integral", we have $$\int_y^x \lim\limits_{n \to 0}\,t^{n-1}\, dt \implies \int_y^x \frac{1}{t}\, dt \implies \mathsf{log} \frac{x}{y}$$

Leaving the fact that he had the inspiration to pull this integral out of thin air aside, is the limit allowed to pass into the definite integral? We hadn't learned Real Analysis (we were just taking a basic high school, hand-wavy single-variable calculus course), and I remember feeling very uneasy about the sorcery. I still am, hence, this question.

I've since thought about approaching it using $\mathsf{L'Hospital}$, but I still feel uneasy, since it involves differentiating with respect to different variables, which is a little bit confusing. I'd also appreciate your help in this regard.

If you have a better proof, I'll truly appreciate it.

$\endgroup$
  • 1
    $\begingroup$ Actually, you don't need to differentiate with respect to different variables since $x$ and $y$ are constants here (unless you mean not differentiating with respect to $x$ as is most commonly done). Alternatively, you may recognize the limit as the derivative of the function $f(t) = x^t-y^t$ at $t=0$. $\endgroup$ – Joey Zou Jan 24 '16 at 7:25
  • $\begingroup$ Right; now it remains to prove that the integral 'sorcery' is valid. $\endgroup$ – Maxis Jaisi Jan 24 '16 at 7:29
8
$\begingroup$

See here: Can a limit of an integral be moved inside the integral?

In general, passing the limit under the integral is not valid. In this circumstance, however, it is valid, as the functions converge uniformly on a finite interval. Uniform convergence means the following: if we let $M_n = \max\limits_{y\le t\le x}{|t^{n-1}-t^{-1}|}$ be the maximum difference of $t^{n-1}$ from $t^{-1}$, then $\lim\limits_{n\rightarrow 0}{M_n} = 0$. You can check that uniform convergence does hold here, so by a standard (and fairly easy to prove) theorem of analysis, passing the limit under the integral is justified.

$\endgroup$
15
$\begingroup$

By using L'Hospital rule we have \begin{align*} \lim_{n\to 0}\frac{x^n-y^n}{n}&=\lim_{n\to 0}\frac{x^n\log x-y^n\log y}{1}\\ &=\log x-\log y\\&=\log\frac{x}{y} \end{align*}

$\endgroup$
4
$\begingroup$

Let $t = x/y$ and notice that \begin{align*} \lim_{n \to 0} \frac{x^n - y^n}{n} &= \lim_{n \to 0} y^n \frac{t^n - 1}{n} \\ &= \left(\lim_{n \to 0} y^n\right) \left(\lim_{n \to 0} \frac{t^n - 1}{n}\right) \\ &= 1 \cdot \frac{d}{dn}\Big|_{n = 0} t^n \\ &= \ln t = \ln \frac x y \end{align*}

as desired.

$\endgroup$
4
$\begingroup$

Hint: $$ \begin{align} \lim_{n\to0}\frac{x^n-1}{n} &=\log(x)\lim_{n\to0}\frac{e^{n\log(x)}-1}{n\log(x)}\\ &=\log(x)\lim_{u\to0}\frac{e^u-1}u\\[6pt] &=\log(x) \end{align} $$ Subtract

$\endgroup$
2
$\begingroup$

\begin{align} \lim_{n\to0}\frac{x^n-y^n}{n}&=\lim_{n\to0}\frac{e^{n\cdot \ln x}-e^{0\cdot \ln x}-(e^{n\cdot \ln y}-e^{0\cdot \ln y})}{n}\\ &=\left(e^{n\cdot \ln x}\right)'|_{n=0}-\left(e^{n\cdot \ln y}\right)'|_{n=0}\\ &=\ln x - \ln y \end{align}

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.