I want to gain a strong validity for formal definition of divergence compare to my version. I'll explain my current understanding below. I hope someone can help me rectifying error I made.

I was guessing out definition of divergence at a point with epsilon-delta proof (more specifically for case: $\lim_{x\to a}$ f(x)=+$\infty$) Given formal definition is $$\forall M\gt0, ( \exists\delta\gt0:0\lt\vert x-a\vert\lt\delta \Rightarrow f(x)\gt M)$$

However, the things that come up with my first trial was $$\forall\delta\gt0, (\exists\epsilon\gt0: \epsilon\lt f(x)\Rightarrow0\lt\vert x-a\vert\lt\delta)$$

Which was I think just x-y axis symmetry of convergence definition, $$lim_{x\to +\infty}f(x)=L \iff \forall\epsilon (\exists\delta\gt0: \delta\lt x\Rightarrow 0\lt\vert f(x)-L\vert\lt\epsilon)$$ I'm vaguely grasping the way to correct my fallacy such as ~ Since $lim_{x\to+\infty}f(x)=L$ is $(x\to +\infty)\to f(x)=L$, So we are not safe to say backward always works (p to q logic)~ or ~ The way how the function is defined conventionally $x\to f(x)$ so in all occasion we should argue about existence of boundary of domain not of range ~

Extra question: Is it work? $$\lim_{x\to c}f(x)=L \iff \forall \delta(\exists\epsilon\gt0:0\lt\vert f(x)-L\vert\lt\epsilon \Rightarrow 0\lt\vert x-c\vert\lt\delta)$$


No, it does not work. Consider the function $f$ given by $f(x)=1$. Then by your definition for all $r\in\mathbb{R}$ we have $\lim_{x\rightarrow 0}f(x)=r$.

To show this let $\delta>0$ randomly given and note that for $\varepsilon=2+|r|$ we have $$|f(x)-r|\leq 1+|r|<\varepsilon$$ for all $|x|<\delta$. You will have similar problems with your proposed definition for convergence.

  • $\begingroup$ I appreciate your disproof. It helped me to evoke extra disproof case. $\endgroup$ – WienAudience Feb 21 '19 at 13:16

Your question is quite hard to understand but I'll try my best to answer it.

You are trying to find a definition for $ \lim_{x\to a} f(x) = + \infty$. i.e. a good definition for " if $x$ approaches $a$ then the function $f$ grows arbitrarily. The correct definition is

$$ \forall M > 0, \exists \delta > 0 : \vert x-a \vert < \delta \Rightarrow f(x) > M.$$

And you came up with $$\forall\delta\gt0, (\exists\epsilon\gt0: \epsilon\lt f(x)\Rightarrow0\lt\vert x-a\vert\lt\delta).$$

Why does this definition not work ? Think about what it's saying ! It says the following

  1. take any positive number $\delta$
  2. Consider the interval around $a$ : $(a - \delta, a + \delta)$
  3. There exists another positive number $\epsilon$ such that if $f(x)$ is greater than $\epsilon$ then $x \in (a-\delta, a+ \delta)$.

There are problems with this definition

Consider a function such that $\lim_{x \to c} f(x) = + \infty$ and $\lim_{x \to c'} f(x) = +\infty$ (in the usual sense) for example $f(x) = \frac{1}{(x+3)^2 (x-3)^2}$. According to your definition $\lim_{x \to 3} f(x) \neq +\infty$.

To see this draw the graph and notice that the function is completely symmetric so if for $\delta$ you find an $\epsilon$ such that if $f(x)$ is greater than $\epsilon$ then $x \in (a-\delta, a+ \delta)$. Then $f(x) = f(-x) > \epsilon$ so $-x \in (a- \delta, a+ \delta)$. Clearly if you take $\delta$ very small you can't have both $x \in (a-\delta, a+ \delta)$ and $-x \in (a- \delta, a+ \delta)$.

Remember that $p \Rightarrow q $ is equivalent to $ \lnot q \Rightarrow \lnot p$ so you're definition is identical to

$$ \forall \delta > 0, \exists \epsilon > 0 : x \not \in [a-\delta,a+\delta] \Rightarrow f(x) \leq \epsilon $$

Therefore any function $f$ which has $\lim_{x \to c} f(x) = + \infty$ and $\lim_{x \to c'} f(x) = +\infty$ (in the usual sense) for $c \neq c'$ will not go to infinity at all according to your definition.

  • $\begingroup$ Thank you for your considerate answer. It truly clarified a lot in my mind. I feel i'm well convinced about theorem now. $\endgroup$ – WienAudience Feb 21 '19 at 13:13

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.