= Improper integration by truncation
For a nonnegative <function> $f$ on a <compact interval>, value truncation replaces large heights by a cap:
$$
f_r(x)=\min(f(x),r),\qquad r>0.
$$
One elementary convention requires every $f_r$ to be <Riemann integrable> and defines the truncated <integral> as $\lim_{r\to\infty}\int f_r$, allowing $+\infty$. The <integrals> are nondecreasing in $r$, so the <limit> exists in the <extended real numbers>; finiteness is an additional convergence condition. For nonnegative <measurable functions>, the <monotone convergence theorem> identifies this value with the <Lebesgue integral> whenever the capped <Riemann integrals> exist.
For $f(x)=x^{-1/2}$ on $(0,1]$, assign any finite value at zero. For $r\geq1$, splitting at $x=r^{-2}$ gives
$$
\int_0^1 f_r(x)\,dx=r\,r^{-2}+\int_{r^{-2}}^1x^{-1/2}\,dx=2-\frac1r\longrightarrow2.
$$
This removes a height cutoff. The usual <improper integral> instead removes a small endpoint interval, $\lim_{\varepsilon\downarrow0}\int_\varepsilon^1f$; both conventions agree here. Neither convention justifies cancellation of two infinite positive and negative parts.
Back to article page