I'm trying to understand the $\varepsilon$-$\delta$ proof for $$\lim_{x \to 0} x^2\sin\left(\frac{1}{x}\right)=0$$

We have to find $\varepsilon>0$ and $\delta>0$ such that $$ \left|x^2\sin\left(\frac{1}{x}\right)\right| < \varepsilon \tag{1}\label{eq1} $$ and $$ |x|<\delta \tag{2}\label{eq2} $$

My work:

We know $$ \begin{aligned} \left|\sin\left(\frac{1}{x}\right)\right| &\leq 1 \\ \implies \left|x^2\sin\left(\frac{1}{x}\right)\right| &\leq |x^2| \\ \end{aligned} $$ $$ \therefore \text{ } \left|x^2\sin\left(\frac{1}{x}\right)\right| < \delta^2 \tag{3}\label{eq3} $$

For $0<\delta<1$, we have $$ \begin{aligned} \delta^2 &< \delta \\ \implies \left|x^2\sin\left(\frac{1}{x}\right)\right| &< \delta \end{aligned} $$ and hence we can choose any $\delta$, such that $0 < \delta \leq \varepsilon$. So far so good.

But, for $\delta \geq 1$, we have $$ \delta^2 \geq \delta $$ The range of value for $\delta$ in this case that I can think of such that $\eqref{eq1}$ and $\eqref{eq3}$ could be reconciled is $$ \left|x^2\sin\left(\frac{1}{x}\right)\right| < \delta \leq \varepsilon $$

But when I check on a graphing tool, regardless of whether $\delta$ is $<1$ or $\geq 1$, $0<\delta\leq\varepsilon$ holds.

What am I doing wrong?


Since $0<\delta\leq\varepsilon$ and from $\eqref{eq2}$, we get $$ |x|<\delta \leq\varepsilon \implies |x|^2 < \delta^2 \leq \varepsilon^2 $$

So, since $$ \left|x^2\sin\left(\frac{1}{x}\right)\right| \leq |x^2| \\ \implies \left|x^2\sin\left(\frac{1}{x}\right)\right| < \delta ^2 \leq \varepsilon^2 $$

So, can we now say $\delta \leq \varepsilon$, regardless of values that either take?

  • 5
    $\begingroup$ Recall that $\delta$ MAY depend on $\varepsilon$. So GIVEN $\varepsilon>0$, just take $\delta=\sqrt{\varepsilon}$. $\endgroup$ Jun 26, 2021 at 16:41

2 Answers 2


You're almost there, and there are several ways to finish up. One thing you could do is what @Tito Eliatron suggests in his comment (this might be the most elegant way):

Let $\epsilon > 0$, set $\delta = \sqrt{\epsilon}$. Then, for $|x|<\delta$ we have

\begin{equation} \left|x^2 \sin{\left(\frac{1}{x}\right)}\right| \leq |x|^2 < \delta^2 = \sqrt{\epsilon}^2 = \epsilon. \end{equation}

But there is another way to "fix" your argument. As you have noted, if $\epsilon \leq 1$, you can choose $\delta = \epsilon$ and use the fact that in this case we have $\delta^2 \leq \delta$, and do the same thing as above. Then we deal with the case $\epsilon > 1$ separately. In this case, we can choose $\delta = 1$, and we have for $|x| < \delta:$

\begin{equation} \left|x^2 \sin{\left(\frac{1}{x}\right)}\right| \leq |x|^2 < \delta^2 = 1^2 = 1 < \epsilon. \end{equation}

Thus, we can set $\delta = \min{(\epsilon, 1)}$ and obtain the same result. I think Tito's suggestion is more beautiful since it avoids different cases, but both methods are valid.


In $\DeclareMathOperator{\epsilon}{\varepsilon}\epsilon$-$\delta$ proofs, $\delta$ is allowed to depend on $\epsilon$, but not vice versa. In your example, if $\epsilon\le1$, then we can set $\delta=\epsilon$, and things work out nicely. Often, weird things can happen for large values of $\epsilon$, and so as a kind of 'safety-net' we require that $\delta$ be at most a specific number. Here, we can set $$ \delta=\min\left(\epsilon,1\right) $$ and things work out nicely. It also helps to work backwards. We need $$ \left|x^2\sin\left(\frac{1}{x}\right)\right|<\epsilon $$ or $$ \left|x^2\right|\left|\sin\left(\frac{1}{x}\right)\right|<\epsilon \, . $$ Now, if we can make $|x^2|$ be smaller than $\epsilon$, then it will be certainly be the case that $|x^2|\left|\sin(1/x)\right|$ will be smaller than $\epsilon$, as $|\sin(1/x)|\le1$. Moreover, if $|x|<1$, then $|x^2|<|x|$. So if $|x|<\epsilon$ and $|x|<1$, then $|x^2|<|x|<\epsilon$, which is what we need. So let $\delta=\min(\epsilon,1)$, which guarantees that $|x|<1$ and $|x|<\epsilon$.


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.