I have to prove the continuity and find the limit $(x,y)\to(1,1)$ of $f:D\to R$, where $$D=\{(x,y)\in R^2:|x|\neq|y|\}$$ and $$ f(x,y) = \frac{1}{y^2-x^2}\int_x^y\ln(e+e^t)dt.$$

I don't know whether my way of thinking is correct and I am a little stuck on the limit.


I have had a few thoughts:

  • The multiplication of continuous functions is continuous, thus if $\int g(x)$ is differentiable and the fraction is continuous, it should be continuous.
  • To be integrable, the integral of an absolute value of the function (the derivative) should have a finite limit.

My steps

$$ f(x,y) = \frac{1}{y+x}\cdot\frac{1}{y-x}\int_x^y\ln(e+e^t)dt$$

Since we can bound the integral by:


Which is convergent on a given interval, thus $g(x)$ is integrable and continuous

Additionally, from the intermediate value theorem we have: $$\frac{1}{y-x}\int_x^y\ln(e+e^t)dt=\ln(e+e^c)~~~~\text{for some}~~c\in(x,y)$$

Therefore: $$ f(x,y)=\frac{1}{y+x}\cdot\ln(e+e^c) $$

Since $|x|\neq|y|$ the function does not take zero value in the denominator, thus this multiplication of functions is continuous.

I am still learning and I would really appreciate if you would point out my mistakes or present a more accurate solution.


I have also tried spheric coordinates, but I can't see nothing helpful in this case:

$$ \frac{1}{y^2-x^2} = \frac{1}{r^2(\sin^2\theta-\cos^2\theta )} $$

One of my thoughts was that I would have to find the limit of the integral and of the fraction or calculate the integral and combine with the fraction, however I decided to test the fraction first.

$$ h(x,y)_{(x,y)\to(1,1)} = \frac{1}{y^2-x^2}$$

I check two sequences: $a_n=(1, \frac{1}{n})$ and $ b_n=(\frac{1}{n},1)$ with $n\to\infty$ and get:

$$ \lim_{n\to -\infty} a_n=\lim_{n\to \infty}\frac{n^2}{n^2-1}= \infty\\ \lim_{n\to \infty} b_n=\lim_{n\to \infty}\frac{n^2}{1-n^2} = -\infty~~~~\text{for big}~~n$$

For this function the limit does not exist since we can show two sequences with different limits.

Again, I am not sure whether my solution presents the correct way of solving. I would like to ask you for some guidance.


There are some problems. First of all, when you use the intermediate value theorem, you find a $c$ for specific $x$ and $y$. So you can't conclude that $f(x,y)=\frac{1}{y^2-x^2}\ln(e + e^c)$ for all $x,y$.

When calculating the limit, you only conclude that $\frac{1}{y^2 -x^2}$ does not have a limit as $(x,y)$ approaches $(1,1)$, which is correct. But this doesn't prove that $f(x,y)$ doesn't have a limit.

To prove the continuity of $f(x,y)$, it is easier to prove continuity of one variable and keeping the other constant for both $x$ and $y$. This will imply continuity of the function of both variables. So let $y$ be given, and define $g(x)=f(x,y)$. On the domain $D$, it is a product of two continuous functions. $\frac{1}{y^2 -x^2}$ is continuous since it is a composition of continuous function. $\int_x^y \ln(e + e^t)dt$ is differentiable as a function of $x$, since $\ln(e + e^t$) is continuous, so in particular it is continuous. Similar arguments proves that the function $h(y)=f(x,y)$ is continuous, so we conclude that $f(x,y)$ is continuous.

The limit at $(1,1)$ is a bit more tricky, let me get back on that when I get the spare time and some paper.

Edit: Now for the limit. This is a little difficult, and I'm not sure you've seen all the techniques used here. Anyway, first notice that

$\ln(e+e^t)=\ln(e(1+e^{t-1})) = 1 + \ln(1+e^{t-1})$

$\ln(1+x)$ has a particularly nice series expansion, which we can now use. First notice that

$\int_x^y 1 + \ln(1+e^{t-1})dt = \int_{x-1}^{y-1} 1 + \ln(1+e^{t})dt$

So let us look at that last integrand. From the series expansion, we get

$1+\ln(1+e^t)=1+e^t-\frac{e^{2t}}{2}+ \frac{e^{3t}}{3}-...$

Integrating this term by term we get

$\int_{x-1}^{y-1} 1 + \ln(1+e^{t})dt=\left[ t + e^t-\frac{e^{2t}}{4}+...\right]_{x-1}^{y-1}$


Notice that $\frac{e^{y-1}-e^{x-1}}{y-x}$ and higher powers tends to $0$ when $(x,y)$ goes to $(1,1)$ (you can use L'hopital to show this). Therefore, then only part that doesn't go to zero when divided by $y-x$ is $y-x$. In total we get

$\lim_{(x,y) \rightarrow (1,1)}\frac{1}{y^2-x^2}\int_x^y \ln(e+e^t)dt = \lim_{(x,y) \rightarrow (1,1)}\frac{1}{y+x}\frac{1}{y-x}(y-x)=\frac{1}{2}$

This got a bit advanced fast, and I'm not 100% if all the calculations are correct, but try to see if you can make heads or tails of it.

  • $\begingroup$ Thank you for such a detailed comment. I would appreciate your way to solve the limit. It is very interesing that the limit can still exist. $\endgroup$
    – Funny
    Oct 21 '20 at 14:32
  • $\begingroup$ Thank you for your answer. I really appreciate, that you cared about my understanding; this proof is clearly explained. However, I am quite confused how you got the division by $y-x$ in the series - quite unexpected for me. Additionally, could you tell me whether I predicted the convergence of an integral correctly? $\endgroup$
    – Funny
    Oct 21 '20 at 19:23
  • $\begingroup$ The reason I am dividing by y-x is because we need to multiply the integral with $\frac{1}{y^2-x^2}$ which you yourself pointed out is equal to $\frac{1}{y+x}\frac{1}{y-x}$. The first part of that product goes to $\frac{1}{2}$ as $(x,y)$ goes to $(1,1)$, so there is no problem, but the other doesn't have a limit. Therefore, I tried to multiply that with the integral (in its series form) to see if I got something convergent. $\endgroup$ Oct 22 '20 at 12:19
  • $\begingroup$ I think you're a little off when you predict the convergence of the integral. Just look at the integrand $\ln(e+e^t)$. This is a continuous function, so by the fundamental theorem of calculus you get that it is integrable on the interval $[x,y]$ (for $y>x$), and that the integral function of $x$ and $y$ is continuous. $\endgroup$ Oct 22 '20 at 12:22

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.