1
$\begingroup$

Given that: $$\lim_{x \to c}f(x)=L_1$$which means(1): $$\forall\epsilon_1>0, \exists \delta_1>0:|x-c|<\delta_1\implies|f(x)-L_1|<\epsilon_1 $$ and $$\lim_{x \to c}g(x)=L_2, \space \text{where}\space g(x), \space L_2\ne0 $$ which means(2): $$\forall\epsilon_2>0, \exists \delta_2>0:|x-c|<\delta_2\implies|g(x)-L_2|<\epsilon_2 $$

I want to prove: $$\lim_{x \to c}\frac{f(x)}{g(x)}=\frac{\lim_{x \to c}f(x)}{\lim_{x \to c}g(x)}=\frac{L_1}{L_2}$$

which means showing: $$\forall \epsilon>0,\exists \delta>0:|x-c|<\delta\implies|\frac{f(x)}{g(x)}-\frac{L_1}{L_2}|<\epsilon$$

First, let: $$\delta=\min(\delta_1,\delta_2),\space \epsilon=2a, \space \epsilon_1=ak,\space \epsilon_2=an$$where $$ \frac{1}{k}>\frac{1}{|L_2|} \space \text{and} \space \frac{1}{n}>|\frac{f(x)}{g(x)L_2}|$$ $k,n>0$ from how we defined them.

Substituting $\epsilon_1$ in (1) gives us:$$|f(x)-L_1|<\epsilon_1=ak$$ which implies(3): $$\frac{1}{|L_2|}\cdot |f(x)-L_1|<a$$

The same procedure is done for (2), which gives: $$|g(x)-L_2|<\epsilon_2=an$$ which implies(4): $$|\frac{f(x)}{g(x)L_2}|\cdot|g(x)-L_2|<a$$

Adding (3) and (4) gives:$$\frac{1}{|L_2|}\cdot |f(x)-L_1|+|\frac{f(x)}{g(x)L_2}|\cdot|g(x)-L_2|=\frac{1}{|L_2|}\cdot |f(x)-L_1|+|\frac{f(x)}{g(x)L_2}|\cdot|L_2-g(x)|\\=|\frac{f(x)-L_1}{L_2}|+|\frac{f(x)}{g(x)}-\frac{f(x)}{L_2}|\ge |\frac{f(x)-L_1}{L_2}+\frac{f(x)}{g(x)}-\frac{f(x)}{L_2}|$$

Finally, $$|\frac{f(x)-L_1}{L_2}+\frac{f(x)}{g(x)}-\frac{f(x)}{L_2}|=|-\frac{L_1}{L_2}+\frac{f(x)}{g(x)}|=|\frac{f(x)}{g(x)}-\frac{L_1}{L_2}|<2a=\epsilon$$ and $|\frac{f(x)}{g(x)}-\frac{L_1}{L_2}|$ is indeed less than $\epsilon$. We have proved what we wanted.

$\endgroup$
7
  • $\begingroup$ You should add a little more explanation about choosing $n$ (essentially this means that $f/g$ is bounded and this is where the fact that $L_{2}\neq 0$ is needed). $\endgroup$
    – Paramanand Singh
    Oct 22 '17 at 8:35
  • $\begingroup$ The reason why I chose $\frac{1}{n}>|\frac{f(x)}{g(x)L_2}|$ is because it's more "succinct" than $n<|\frac{g(x)L_2}{f(x)}|$ which is what I originally chose during my derivation of the proof. If I had gone with the latter choice, I would need to make $ f(x) \ne 0$. $\endgroup$ Oct 22 '17 at 8:54
  • $\begingroup$ My point was that you should explain why such an $n$ exists. This requires you to show that $|f(x) |<|L_1|+1$ and $|g(x) |>|L_2|/2$. $\endgroup$
    – Paramanand Singh
    Oct 22 '17 at 9:02
  • $\begingroup$ I'm clueless on how did you arrive to that conclusion. I interpreted that as finding a value for $n$ such that it satisfies $\frac{1}{n}>|\frac{f(x)}{g(x)L_2}|$ It took me some time to figure out but I'm lost. How should I start? $\endgroup$ Oct 22 '17 at 9:54
  • 1
    $\begingroup$ $n$ should be constant and not depend on $x$. $\endgroup$
    – Paramanand Singh
    Oct 22 '17 at 10:22
2
$\begingroup$

Your proof is fine. If you have the limit theorem for products, you can instead apply that to $f(x)$ and $1/g(x)$ rather than having to go through all of that (but you would have to also prove $\lim 1/g(x) = 1/\lim g(x)$ which is essentially the difficulty).

$\endgroup$
1
$\begingroup$

The proof is correct except for a minor step. You propose that there is a number $n>0$ such that $|f(x) /g(x) L_{2}|<1/n$. This is true but not obvious / self-evident. Note that by choosing $\epsilon=1$ in the limit definition for $f$ we have a $\delta_{3}>0$ such that $$0<|x-a|<\delta_{3}\implies |f(x) - L_{1}|<1\implies |f(x) |<|L_{1}|+1$$ and choosing $\epsilon=|L_{2}|/2>0$ in the limit definition of $g$ we have a $\delta_{4}>0$ such that $$0<|x-a|<\delta_{4}\implies|g(x)-L_{2}|<|L_{2}|/2\implies |g(x) |>|L_{2}|/2$$ Therefore $$0<|x-a|<\min(\delta_{3},\delta_{4})\implies \left|\frac{f(x)} {g(x) L_{2}}\right|<\frac{2(|L_{1}|+1)}{|L_{2}|^{2}}=\frac{1}{n}\text{ (say)} $$ Now on the basis of $\epsilon, n, k$ choose your $\delta_{1},\delta_{2}$ and keep $\delta$ as minimum of all the deltas.

$\endgroup$
3
  • $\begingroup$ What led you to find the values of $\epsilon=1$ and $\epsilon=\frac{|L_2|}{2}$? Is it just trial and error? Genuinely interested in the process;All my approaches so far to find a suitable $\epsilon$ value involves me going backwards from the result I want to prove. $\endgroup$ Oct 22 '17 at 10:41
  • 1
    $\begingroup$ @DanielMcFluffy: as I said earlier the idea is to bound the ratio $f/g$ and this requires an upper bound on $f$ and lower bound on $g$. The choice of epsilon in both cases is almost arbitrary. But we need a constant as bound so I chose specific real number $1$ (you can choose any number like $0.001$ or $\sqrt{2}$ as long as it is positive. For $g$ we need $\epsilon$ to be less than $|L_{2}|$. Understand that limit definition is not an exercise in symbolic manipulation and formal logic, rather it is the art of bounding expressions suitably. Cont'd. $\endgroup$
    – Paramanand Singh
    Oct 22 '17 at 11:07
  • 1
    $\begingroup$ @DanielMcFluffy: to bound $f/g$ note that $f$ has a limit so $f$ is bounded and $g$ has a non-zero limit so it is bounded away from zero (ie we can ensure that values of $g$ are near to $L_{2}$ and farther from $0$). $\endgroup$
    – Paramanand Singh
    Oct 22 '17 at 11:09

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.