Improper integral convergence from minus to positive infinity

Quote from Essential Calculus: Early Transcendentals, by James Stewart:

If $f$ is continuous, then $$\int_{-\infty}^\infty f(x)dx=\lim_{t \to \infty}\int_{-t}^tf(x)dx$$

I thought this would be true if such a limit exists (aka if the area is convergent from $a \to \infty$ and from $-\infty \to a$), but the book answer-sheet marks it as false. Could anyone explain to me why it is false?

Thanks!

• You need to quantify over $f$. – Git Gud Mar 13 '14 at 16:35
• @GitGud mind to elaborate? (I forgot to state that $f$ is continuous) – Aegis Mar 13 '14 at 16:36

Take for example $f(x)=x$. Then $$\int_{-t}^t f(x)\,dx=0,$$ since $f$ is an odd function, and hence the limit $\lim_{t\to\infty}\int_{-t}^t f(x)\,dx$ exists and it is equal to zero. However, the function $f(x)=x$ is NOT integrable on the whole of $\mathbb R$.
Another example is $$\lim_{t\to\infty}\int_{-t}^t \frac{\sin x}{x}\,dx=\pi,$$ although $\dfrac{\sin x}{x}$ is also NOT integrable on the whole of $\mathbb R$.
• Thanks! That made things much clearer. One last thing: would the statement be true if you modified it by saying "if $f$ is integrable on the whole of R then ..." ? – Aegis Mar 13 '14 at 16:41