Michael Bowen's VC Course Pages
Answers
Math V21B Section 7.8
Introduction
Please try this problem on your own before looking at my solution. There may be other ways to solve this; I make no claim that this method is the quickest, prettiest, or most efficient.
Problem 37
The original problem is $$I = \int_{ - 1}^0 {\frac{{{e^{1/x}}}} {{{x^3}}}\:dx} $$
The integrand is undefined at $x = 0$, so we conclude that this is likely to be a vertical asymptote, and that we have a Type II improper integral to solve. Because the region (interval) of integration lies immediately to the left of this asymptote, we need to use only one-sided (in this case, left-sided) limits to evaluate the behavior of the antiderivative near $x = 0$. Let us re-write the integral in the form $$I = \int_{ - 1}^0 {\frac{{{e^{1/x}}}} {1}\frac{1} {x}\frac{1} {{{x^2}}}\:dx} $$ which immediately suggests the $w$-substitution $$w = \frac{1} {x} = {x^{ - 1}};\;\;\;\;\;dw = - {x^{ - 2}}dx = - \frac{1} {{{x^2}}}dx.$$ Re-writing $I$ with this substitution gives $$I = \int_?^? {{e^w} \cdot w \cdot \left( { - dw} \right)} = - \int_?^? {w{e^w}\:dw} $$ where the question marks represent the new limits of integration in terms of $w$ rather than $x$, and the negative sign arises from taking the differential of $x^{ - 1}$. Normally, we would explicitly find numeric values for these limits, but since $w = \frac{1} {x}$ is undefined at $x = 0$, replacing the upper limit would be troublesome. So, as an alternative, we will temporarily employ the question marks as surrogate limits until we find a way to work around this issue.
The latter integral is a classic integration-by-parts problem (or we could "cheat" and look up the result in a table). Using the "LIPTE" mnemonic, we set $u$ equal to the algebraic function $w$, from which the rest of the substitutions immediately follow: $$u = w;\;\;\;\;\;du = dw;\;\;\;\;\;dv = {e^w}\:dw;\;\;\;\;\;v = {e^w}$$ Employing the $uv - \int{v\:du}$ paradigm, and not neglecting to specify limits in all required locations, we obtain $$I = - \left[ {\left. {w{e^w}} \right|_?^? - \int_?^? {{e^w}\:dw} } \right] = - \left[ {w{e^w} - {e^w}} \right]_?^? = - \left[ {\left({e^w}\right)\cdot\left( {w - 1} \right)} \right]_?^? = \left[ {\left({e^w}\right)\cdot\left( {1 - w} \right)} \right]_?^?$$ where the leading negative sign disappears in the last step as a result of exchanging the operands in the parenthesized subtraction.
We can now restore the original variable and the original limits of integration by replacing $w$ with $\frac{1}{x}$: $$I = \left[ {\left({e^{1/x}}\right)\cdot\left( {1 - \frac{1} {x}} \right)} \right]_{ - 1}^0 = \mathop {\lim }\limits_{t \to {0^ - }} \left[ {\left({e^{1/x}}\right)\cdot\left( {1 - \frac{1} {x}} \right)} \right]_{ - 1}^t $$ with the last step coming from the technical definition of the Type II improper integral. Again, note that the "lim" limit is left-sided; by contrast, if the limits of integration had been, say, 0 to 1, we would have used a right-sided limit.
To begin the process of evaluating the "lim" limit, let's plug in the $t$ and the $-1$ separately, then subtract. $$I = \mathop {\lim }\limits_{t \to {0^ - }} \left[ {\left( {{e^{1/t}}} \right) \cdot \left( {1 - \frac{1} {t}} \right) - \left( {{e^{1/( - 1)}}} \right) \cdot \left( {1 - \frac{1} { - 1}} \right)} \right] = \mathop {\lim }\limits_{t \to {0^ - }} \left[ {\left( {{e^{1/t}}} \right) \cdot \left( {1 - \frac{1} {t}} \right)} \right] - \left( {{e^{ - 1}}} \right) \cdot \left( {2} \right)$$ $$I = \mathop {\lim }\limits_{t \to {0^ - }} \left[ {\left( {{e^{1/t}}} \right) \cdot \left( {1 - \frac{1} {t}} \right)} \right] - \left( {\frac{2} {e}} \right)$$
To help us understand the behavior of the "lim" limit, let us select a number near (but to the left side of) $t = 0$. (Because of the restriction to $t$ values less than zero, we must make selections such as $t = -0.001$ or $t = -0.0001$ rather than $t = 0.001$ or $t = 0.0001$.)
At $t = -0.001$, we find that $${e^{1/t}} = {e^{1/ - 0.001}} = {e^{ - 1000}} = \frac{1} {{{e^{1000}}}} = \frac{1} {{{\text{huge}}}} \approx 0$$ whereas $$1 - \frac{1} {t} = 1 - \frac{1}{ - 0.001} = 1 - ( - 1000) = 1001 \approx {\text{huge}}$$ which means that the product inside the rectangular brackets is of the form $\left( {0 \cdot {\text{huge}}} \right)$ or $\left( {0 \cdot \infty } \right)$. This is an indeterminate form!
Indeterminate forms are often susceptible to solution using L'Hôpital's rule. However, this rule requires the limiting function to be in the form of a quotient, not a product (which is what we have). With a few steps of manipulation, we can convert this to a quotient of the form $\frac{\infty } {\infty }$.
First, substitute a variable $z = \frac{1} {t}$ into the limit; as $t$ approaches 0 from the left, $z$ will approach negative infinity. (The reason for the new variable is to make the L'Hôpital's-rule derivatives much easier to compute.) With this new variable, we obtain $$I = \mathop {\lim }\limits_{z \to {- \infty} } \left[ {\left( {e^z} \right) \cdot \left( {1 - z} \right)} \right] - \left( {\frac{2} {e}} \right)$$ Second, re-write $e^z$ as $\frac {1}{e^{-z}}$, giving $$I = \mathop {\lim }\limits_{z \to - \infty } \left[ {\left( {\frac{1}{{{e^{ - z}}}}} \right) \cdot \left( {1 - z} \right)} \right] - \left( {\frac{2}{e}} \right) = \mathop {\lim }\limits_{z \to - \infty } \left( {\frac{{1 - z}}{{{e^{ - z}}}}} \right) - \left( {\frac{2}{e}} \right)$$
This quotient is in the desired L'Hôpital form, so apply the rule by differentiating numerator and denominator separately: $$I = \mathop {\lim }\limits_{z \to - \infty } \left( {\frac{{ - 1}} {{ - {e^{ - z}}}}} \right) - \left( {\frac{2} {e}} \right) = \mathop {\lim }\limits_{z \to - \infty } \left( {\frac{1} {{{e^{ - z}}}}} \right) - \left( {\frac{2} {e}} \right) = \mathop {\lim }\limits_{z \to - \infty } \left( {{e^z}} \right) - \left( {\frac{2} {e}} \right) = 0 - \left( {\frac{2}{e}} \right) = \boxed{\displaystyle{ - \frac{2}{e}}}$$
We conclude that the original integral converges to $- \frac{2}{e}$. The result is negative because the graph of the integrand function lies below the $x$-axis on the interval $[-1,0)$.