# Correct way of finding $\delta$ for $\lim_{x \to a} \sqrt{x} = \sqrt{a}$

Prove: $\lim_{x \to a} \sqrt{x} = \sqrt{a}$ using $\epsilon$-$\delta$.

My solution:

We have that $0 < |x-a| < \delta$.

Also, $|\sqrt{x} - \sqrt{a}| < \epsilon$

$\therefore -\epsilon < \sqrt{x} - \sqrt{a} < \epsilon$

$\therefore \sqrt{a} -\epsilon < \sqrt{x} < \sqrt{a} + \epsilon$

$\therefore (\sqrt{a} -\epsilon)^2 < x < (\sqrt{a} + \epsilon)^2$

$\therefore (\sqrt{a} -\epsilon)^2 - a < x - a < (\sqrt{a} + \epsilon)^2 - a$

$\therefore \epsilon^2 - 2\sqrt{a}\epsilon < x - a < \epsilon^2 + 2\sqrt{a}\epsilon$

$\delta = \min(|\epsilon^2 - 2\sqrt{a}\epsilon|,|\ \epsilon^2 + 2\sqrt{a}\epsilon|)$

By letting $\delta = \min(|\epsilon^2 - 2\sqrt{a}\epsilon|,|\ \epsilon^2 + 2\sqrt{a}\epsilon|)$, we get that $|\sqrt{x}-\sqrt{a}|<\epsilon$ if $0 < |x-a| < \delta$.

Thus, $\lim_{x \to a} \sqrt{x} = \sqrt{a}$.

Question: is it correct to find a $\delta$ using the method above?

I've seen a bunch of other methods used to find $\delta$ but I'm totally confused what is the "right" way of doing it. If anyone could explain this or give some other examples, I would appreciate it.

Thank you a lot!

• I would say, yes, this is the most common way to find $\delta$. – Stan Tendijck Jul 4 '18 at 20:06
• Note that in a proof you actually don't need to show how to find a good enough $\delta.$ You just need to say "Given $\epsilon>0$ take $\delta = \ldots$ and suppose that $|x-a| < \delta.$ Then $|f(x) - f(a)| \leq \cdots < \epsilon$." – md2perpe Jul 4 '18 at 21:25

That certainly works. There isn't really a "right" way to do these, but some approaches are shorter than others. In some instances it may be important to find the largest possible $\delta$, but to verify the definition of limit you only need to demonstrate one. For instance, if $a > 0$ you can use $x-a = (\sqrt x - \sqrt a)(\sqrt x + \sqrt a)$ to get $$|\sqrt x - \sqrt a| = \frac{|x-a|}{|\sqrt x + \sqrt a|} \le \frac{|x-a|}{\sqrt a}.$$ Thus if $\delta = \epsilon \sqrt a$ then $|x-a| < \delta$ implies $|\sqrt x - \sqrt a| < \epsilon$.

• Sorry about a dumb question, but how did you get $\delta = \epsilon\sqrt{a}$? – nullbyte Jul 4 '18 at 21:13
• It makes the RHS of the equation above smaller than epsilon – user370967 Jul 4 '18 at 21:15
• @nullbyte. We want to find $\delta$ so that $|x-a|/\sqrt{a} < \epsilon$ when $|x-a| < \delta$. This will be satisfied if $\delta/\sqrt{a} \leq \epsilon.$ Thus we can take $\delta = \epsilon \sqrt{a}.$ – md2perpe Jul 4 '18 at 21:22

There's a bit of an issue: if $\epsilon=2\sqrt{a},$ then you would have $\delta=0.$ Now, in the case that $a=0,$ this is impossible, and you'll instead have $\delta=\epsilon^2.$ In the case that $a>0,$ you should start by assuming that $0<\epsilon<2\sqrt{a},$ which will fix the problem. You will have to justify that it's okay to only consider such $\epsilon,$ however.

• Wow! Good catch. I should have thought about it. Looks like only the solution proposed by Umberto works if $a > 0$. – nullbyte Jul 4 '18 at 21:15

The OP's logic can be salvaged. One problem is they assume on their third $\therefore$ that $u \lt v$ implies $u^2 \lt v^2$ without knowing for sure that $u \ge 0$.

Here is a 'logic patch':

If $a = 0$, simple algebra shows that $\delta = \varepsilon^2$ works.

To show continuity when $a \gt 0$, we can take a few easy steps:

$\tag 1 |\sqrt{x} - \sqrt{a}| < \varepsilon$

$\therefore -\varepsilon < \sqrt{x} - \sqrt{a} < \varepsilon$

$\therefore \sqrt{a} -\varepsilon < \sqrt{x} < \sqrt{a} + \varepsilon$

Now to get rid of $\sqrt x$ and replace it with $x$, we would like to run the 'square it' on our expression, but to do that we must assume that $\sqrt{a} -\varepsilon$ in non-negative. i.e. $\varepsilon \le \sqrt{a}$. But that is not a restriction at all, so we proceed:

$\therefore (\sqrt{a} -\varepsilon )^2 < x < (\sqrt{a} + \varepsilon )^2$

$\therefore -2 \sqrt{a} \, \varepsilon + \varepsilon^2 < x - a < 2 \sqrt{a} \, \varepsilon + \varepsilon^2$

We are trying to find our $\delta \gt 0$ 'setup':

$\tag 2 -2 \sqrt{a} \, \varepsilon + \varepsilon^2 \le -\delta < x - a < \delta \le 2 \sqrt{a} \, \varepsilon + \varepsilon^2$

so that $\text{(2)}$ implies $\text{(1)}$ (we can use our developed "$\therefore$-logic-chain" in reverse).

Since $\varepsilon \le \sqrt{a}$ here, we can define our $\delta \gt 0$ with

$\tag 3 \delta = 2 \sqrt{a} \, \varepsilon - \varepsilon^2$

and so by definition the left side of $\text{(2)}$ 'clicks'. For the right side, it is easy to see that $\delta \le 2 \sqrt{a} \, \varepsilon + \varepsilon^2$ since

$\quad (2 \sqrt{a} \, \varepsilon - \varepsilon^2) \le (2 \sqrt{a} \, \varepsilon + \varepsilon^2) \text{ iff } 2 \varepsilon^2 \ge 0$

So we can use the OP's logic to find a $\delta$ given any starting $\varepsilon$, where if necessary we pound $\varepsilon$ down first to $\sqrt a$.