The Incomplete Gamma and Confluent Hypergeometric Functions (Part 3)

In the previous post, I confirmed the curious integral

$\displaystyle \int_0^z t^{a-1} e^{-t} \, dt = \displaystyle e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s}$ ,

where the right-hand side is a special case of the confluent hypergeometric function when $a$ is a positive integer, by differentiating the right-hand side. However, the confirmation psychologically felt very unsatisfactory — we basically guessed the answer and then confirmed that it worked.

A seemingly better way to approach the integral is to use the Taylor series representation of $e^{-t}$ to integrate the left-hand side term-by-term:

$\displaystyle \int_0^z t^{a-1} e^{-t} \, dt = \int_0^z t^{a-1} \sum_{n=0}^\infty \frac{(-t)^n}{n!} \, dt$

$= \displaystyle \sum_{n=0}^\infty \frac{(-1)^n}{n!} \int_0^z t^{a+n-1} \, dt$

$= \displaystyle \sum_{n=0}^\infty \frac{(-1)^n}{n!} \frac{z^{a+n}}{a+n}$ .

Well, that doesn’t look like the right-hand side of the top equation. However, the right-hand side of the top equation also has a $e^{-z}$ in it. Let’s also convert that to its Taylor series expansion and then use the formula for multiplying two infinite series:

$\displaystyle e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} = \left( \sum_{s=0}^\infty \frac{(-z)^s}{s!} \right) \left( \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} \right)$

$= \displaystyle z^a \left( \sum_{s=0}^\infty \frac{(-1)^s z^s}{s!} \right) \left( \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^s \right)$

$= \displaystyle z^a \sum_{n=0}^\infty \sum_{s=0}^n \frac{(-1)^s z^s}{s!} \frac{(a-1)!}{(a+n-s)!} z^{n-s}$

$= \displaystyle \sum_{n=0}^\infty \sum_{s=0}^n \frac{(-1)^s}{s!} \frac{(a-1)!}{(a+n-s)!} z^{a+n}$

Summarizing, apparently the following two infinite series are supposed to be equal:

$\displaystyle \sum_{n=0}^\infty \frac{(-1)^n}{n!} \frac{z^{a+n}}{a+n} = \sum_{n=0}^\infty \sum_{s=0}^n \frac{(-1)^s}{s!} \frac{(a-1)!}{(a+n-s)!} z^{a+n}$ ,

or, matching coefficients of $z^{a+n}$ ,

$\displaystyle \frac{(-1)^n}{n! (a+n)} = \sum_{s=0}^n \frac{(-1)^s (a-1)!}{s! (a+n-s)!}$ .

When I first came to this equality, my immediate reaction was to throw up my hands and assume I made a calculation error someplace — I had a hard time believing that this sum from $s=0$ to $s=n$ was true. However, after using Mathematica to evaluate this sum for about a dozen different values of $n$ and $a$ , I was able to psychologically assure myself that this identity was somehow true.

But why does this awkward summation work? This is no longer a question about integration: it’s a question about a finite sum with factorials. I continue this exploration in the next post.

The Incomplete Gamma and Confluent Hypergeometric Functions (Part 2)

In this series of posts, I confirm this curious integral:

$\displaystyle \int_0^z t^{a-1} e^{-t} \, dt = \displaystyle \frac{z^a e^{-z}}{a} M(1, 1+a, z)$ ,

where the confluent hypergeometric function $M(a,b,z)$ is

$M(a,b,z) = \displaystyle 1+\sum_{s=1}^\infty \frac{a(a+1)\dots(a+s-1)}{b(b+1)\dots (b+s-1)} \frac{z^s}{s!}$ .

This integral can be confirmed — unsatisfactorily confirmed, but confirmed — by differentiating the right-hand side. For the sake of simplicity, I restrict my attention to the case when $a$ is a positive integer. To begin, the right-hand side is

$\displaystyle \frac{z^a e^{-z}}{a} M(1, 1+a, z) = \displaystyle \frac{z^a e^{-z}}{a} \left[1 + \sum_{s=1}^\infty \frac{1 \cdot 2 \cdot \dots \cdot s}{(a+1)(a+2)\dots (a+s)} \frac{z^s}{s!} \right]$

$= \displaystyle \frac{z^a e^{-z}}{a} \left[1 + \sum_{s=1}^\infty \frac{1}{(a+1)(a+2)\dots (a+s)} z^s \right]$

$= \displaystyle \frac{z^a e^{-z}}{a} \left[1 + \sum_{s=1}^\infty \frac{a!}{(a+s)!} z^s \right]$

$= \displaystyle \frac{z^a e^{-z}}{a} \sum_{s=0}^\infty \frac{a!}{(a+s)!} z^s$

$= \displaystyle e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s}$ .

We now differentiate, first by using the Product Rule and then differentiating the series term-by-term (blatantly ignoring the need to confirm that term-by-term differentiation applies to this series):

$\displaystyle \frac{d}{dz} \left[\frac{z^a e^{-z}}{a} M(1, 1+a, z) \right] = \displaystyle \frac{d}{dz} \left[ e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} \right]$

$= -e^{-z} \displaystyle \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} + e^{-z} \frac{d}{dz} \left[ \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} \right]$

$= -e^{-z} \displaystyle \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} + e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} \frac{d}{dz} z^{a+s}$

$= -e^{-z} \displaystyle \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} + e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} (a+s) z^{a+s-1}$

$= -e^{-z} \displaystyle \sum_{s=0}^\infty \frac{(a-1)!}{(a+s)!} z^{a+s} + e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s-1)!} z^{a+s-1}$ .

We now shift the index of the first series:

$\displaystyle \frac{d}{dz} \left[\frac{z^a e^{-z}}{a} M(1, 1+a, z) \right] =-e^{-z} \sum_{s=1}^\infty \frac{(a-1)!}{(a+s-1)!} z^{a+s-1} + e^{-z} \sum_{s=0}^\infty \frac{(a-1)!}{(a+s-1)!} z^{a+s-1}$ .

By separating the $s=0$ term of the second series, the right-hand side becomes:

$\displaystyle -e^{-z} \sum_{s=1}^\infty \frac{(a-1)!}{(a+s-1)!} z^{a+s-1} + e^{-z} \frac{(a-1)!}{(a-1)!} z^{a-1} + e^{-z} \sum_{s=1}^\infty \frac{(a-1)!}{(a+s-1)!} z^{a+s-1}$ = e^{-z} z^{a-1}$

since the two infinite series cancel. We have thus shown that

$\displaystyle \frac{d}{dz} \left[\frac{z^a e^{-z}}{a} M(1, 1+a, z) \right] = \frac{e^{-z} z^{a-1}}{a}$ .

Therefore, we may integrate the right-hand side:

$\displaystyle \int_0^z t^{a-1} e^{-t} \, dt = \left[\frac{t^a e^{-t}}{a} M(1, 1+a, t) \right]_0^z$

$\displaystyle = \frac{z^a e^{-z}}{a} M(1, 1+a, z) - \frac{0^a e^{0}}{a} M(1, 1+a, 0)$

$\displaystyle = \frac{z^a e^{-z}}{a} M(1, 1+a, z)$ .

While this confirms the equality, this derivation still feels very unsatisfactory — we basically guessed the answer and then confirmed that it worked. In the next few posts, I’ll consider the direct verification of this series.

The Incomplete Gamma and Confluent Hypergeometric Functions (Part 1)

Yes, the title of this post is a mouthful.

While working on a research project, a trail of citations led me to this curious equality in the Digital Library of Mathematical Functions:

$\gamma(a,z) = \displaystyle \frac{z^a e^{-z}}{a} M(1, 1+a, z)$ ,

where the incomplete gamma function $\gamma(a,z)$ is

$\gamma(a,z) = \displaystyle \int_0^z t^{a-1} e^{-t} \, dt$

and the confluent hypergeometric function $M(a,b,z)$ is

$M(a,b,z) = \displaystyle 1+\sum_{s=1}^\infty \frac{a(a+1)\dots(a+s-1)}{b(b+1)\dots (b+s-1)} \frac{z^s}{s!}$ .

While I didn’t doubt that this was true — I don’t doubt this has been long established — I had an annoying problem: I didn’t really believe it. The gamma function

$\Gamma(a) = \displaystyle \int_0^\infty t^{a-1} e^{-t} \, dt$

is a well-known function with the famous property that

$\Gamma(n+1) = n!$

for non-negative integers $n$ ; this is often seen in calculus textbooks as an advanced challenge using integration by parts. The incomplete gamma function $\gamma(a,z)$ has the same look as $\Gamma(a)$ , except that the range of integration is from $0$ to $z$ (and not $\infty$ ). The gamma function appears all over the place in mathematics courses.

The confluent hypergeometric function, on the other hand, typically arises in mathematical physics as the solution of the differential equation

$z f''(z) + (b-z) f'(z) - af(z) = 0$ .

As I’m not a mathematical physicist, I won’t presume to state why this particular differential equation is important — except that it appears to be a niche equation that arises in very specialized applications.

So I had a hard time psychologically accepting that these two functions were in any way related.

While ultimately unimportant for advancing mathematics, this series will be about the journey I took to directly confirm the above equality.

Solving Problems Submitted to MAA Journals: Index

I’m doing something that I should have done a long time ago: collecting a series of posts into one single post. The links below show my series on solving problems submitted to the journals of the Mathematical Association of America.

Part 1: Introduction

Part 2a: Suppose that $X$ and $Y$ are independent, uniform random variables over $[0,1]$ . Now define the random variable $Z$ by

$Z = (Y-X) {\bf 1}(Y \ge X) + (1-X+Y) {\bf 1}(Y<X)$ .

Prove that $Z$ is uniform over $[0,1]$ . Here, ${\bf 1}[S]$ is the indicator function that is equal to 1 if $S$ is true and 0 otherwise.

Part 2b: Suppose that $X$ and $Y$ are independent, uniform random variables over $[0,1]$ . Define $U_X$ , $V_X$ , $B_X$ , and $W_X$ as follows:

$U_X$ is uniform over $[0,X]$ ,

$V_X$ is uniform over $[X,1]$ ,

$B_X \in \{0,1\}$ with $P(B_X=1) = X$ and $P(B_X=0)=1-X$ , and

$W_X = B_X \cdot U_X + (1-B_X) \cdot V_X$ .

Prove that $W_X$ is uniform over $[0,1]$ .

Part 3: Define, for every non-negative integer $n$ , the $n$ th Catalan number by

$C_n := \displaystyle \frac{1}{n+1} {2n \choose n}$ .

Consider the sequence of complex polynomials in $z$ defined by $z_k := z_{k-1}^2 + z$ for every non-negative integer $k$ , where $z_0 := z$ . It is clear that $z_k$ has degree $2^k$ and thus has the representation

$z_k =\displaystyle \sum_{n=1}^{2^k} M_{n,k} z^n$ ,

where each $M_{n,k}$ is a positive integer. Prove that $M_{n,k} = C_{n-1}$ for $1 \le n \le k+1$ .

Part 4: Let $A_1, \dots, A_n$ be arbitrary events in a probability field. Denote by $B_k$ the event that at least $k$ of $A_1, \dots A_n$ occur. Prove that $\displaystyle \sum_{k=1}^n P(B_k) = \sum_{k=1}^n P(A_k)$ .

Parts 5a, 5b, 5c, 5d, and 5e: Evaluate the following sums in closed form:

$\displaystyle \sum_{n=0}^\infty \left( \cos x - 1 + \frac{x^2}{2!} - \frac{x^4}{4!} \dots + (-1)^{n-1} \frac{x^{2n}}{(2n)!} \right)$

and

$\displaystyle \sum_{n=0}^\infty \left( \sin x - x + \frac{x^3}{3!} - \frac{x^5}{5!} \dots + (-1)^{n-1} \frac{x^{2n+1}}{(2n+1)!} \right)$ .

Parts 6a, 6b, 6c, 6d, and 6e: Two points $P$ and $Q$ are chosen at random (uniformly) from the interior of a unit circle. What is the probability that the circle whose diameter is segment $\overline{PQ}$ lies entirely in the interior of the unit circle?

Parts 7a, 7b, 7c, 7d, 7e, 7f, 7g, 7h, and 7i: Let $X$ and $Y$ be independent normally distributed random variables, each with its own mean and variance. Show that the variance of $X$ conditioned on the event $X>Y$ is smaller than the variance of $X$ alone.

Integration Using Schwinger Parametrization

I recently read the terrific article Integration Using Schwinger Parametrization, by David M. Bradley, Albert Natian, and Sean M. Stewart in the American Mathematical Monthly. I won’t reproduce the entire article here, but I’ll hit a couple of early highlights.

The basic premise of the article is that a complicated integral can become tractable by changing it into an apparently more complicated double integral. The idea stems from the gamma integral

$\Gamma(p) = \displaystyle \int_0^\infty t^{p-1} e^{-t} \, dt$ ,

where $\Gamma(p) = (p-1)!$ if $p$ is a positive integer. If we perform the substitution $t = \phi u$ in the above integral, where $\phi$ is a quantity independent of $t$ , we obtain

$\Gamma(p) = \displaystyle \int_0^\infty (\phi u)^{p-1} e^{-\phi t} \phi \, du = \displaystyle \int_0^\infty \phi^p u^{p-1} e^{-\phi u} \, du$ ,

which may be rewritten as

$\displaystyle \frac{1}{\phi^p} = \displaystyle \frac{1}{\Gamma(p)} \int_0^\infty t^{p-1} e^{-\phi t} \, dt$

after changing the dummy variable back to $t$ .

A simple (!) application of this method is the famous Dirichlet integral

$I = \displaystyle \int_0^\infty \frac{\sin x}{x} \, dx$

which is pretty much unsolvable using techniques from freshman calculus. However, by substituting $\phi = x$ and $p=1$ in the above gamma equation, and using the fact that $\Gamma(1) = 0! = 1$ , we obtain

$I = \displaystyle \int_0^\infty \sin x \int_0^\infty e^{-xt} \, dt \, dx$

$= \displaystyle \int_0^\infty \int_0^\infty e^{-xt} \sin x \, dx \, dt$

after interchanging the order of integration. The inner integral can be found by integration by parts and is often included in tables of integrals:

$I = \displaystyle \int_0^\infty -\left[ \frac{e^{-xt} (\cos x + t \sin x)}{1+t^2} \right]_{x=0}^{x=\infty} \, dt$

$= \displaystyle \int_0^\infty \left[0 +\frac{e^{0} (\cos 0 + t \sin 0)}{1+t^2} \right] \, dt$

$= \displaystyle \int_0^\infty \frac{1}{1+t^2} \, dt$ .

At this point, the integral is now a standard one from freshman calculus:

$I = \displaystyle \left[ \tan^{-1} t \right]_0^\infty = \displaystyle \frac{\pi}{2} - 0 = \displaystyle \frac{\pi}{2}$ .

In the article, the authors give many more applications of this method to other integrals, thus illustrating the famous quote, “An idea which can be used only once is a trick. If one can use it more than once it becomes a method.” The authors also add, “We present some examples to illustrate the utility of this technique in the hope that by doing so we may convince the reader that it makes a valuable addition to one’s integration toolkit.” I’m sold.

A vivid illustration of a discontinuous function

The essay Singular Limits in the May 2002 issue of Physics Today has a vivid illustration of a discontinuous function $F(x)$ which measures the ickiness one feels after eating an apple but observing that proportion $x$ of a maggot is still inside the apple. For this function, $\displaystyle \lim_{x \to 0^+} F(x) \ne F(0)$ .

Biting into an apple and finding a maggot is unpleasant enough, but finding half a maggot is worse. Discovering one-third of a maggot would be more distressing still: The less you find, the more you might have eaten. Extrapolating to the limit, an encounter with no maggot at all should be the ultimate bad-apple experience. This remorseless logic fails, however, because the limit is singular: A very small maggot fraction ( $f \ll 1$ ) is qualitatively different from no maggot ( $f=0$ ).

Higher derivatives in ordinary speech

Just about every calculus student is taught that the first derivative is useful for finding the slope of a curve and finding velocity from position, and that the second derivative is useful for finding the concavity of a curve and finding acceleration from position.

I recently came across a couple of quotes that, taken literally, are statements about third and fifth derivatives.

Per Wikipedia, President Nixon announced in 1972 that the rate of increase of inflation was decreasing. Taken literally, this claims that “the second derivative of inflation is negative, and so the third derivative of purchasing power [since inflation is the derivative of purchasing power] is negative.” As dryly stated in the Notices of the American Mathematical Society, “[t]his was the first time a sitting president used the third derivative to advance his case for reelection”; the article then ponders the implications of the abuse of mathematics.

More recently, the popular blog Math With Bad Drawings had some fun analyzing a clause that appeared in a 2013 op-ed piece: “As the rate of acceleration of innovation increases…” Taken literally, the words rate, innovation and increases all refer to a first derivative (innovation would be the rate at which technology changes), while the word acceleration refers to a second derivative. Therefore, taken literally and not rhetorically (which was clearly the authors’ intent), this brief clause is a claim that the fifth derivative of technology is positive.

Horrible False Analogy

I had forgotten the precise assumptions on uniform convergence that guarantees that an infinite series can be differentiated term by term, so that one can safely conclude

$\displaystyle \frac{d}{dx} \sum_{n=1}^\infty f_n(x) = \sum_{n=1}^\infty f_n'(x)$ .

This was part of my studies in real analysis as a student, so I remembered there was a theorem but I had forgotten the details.

So, like just about everyone else on the planet, I went to Google to refresh my memory even though I knew that searching for mathematical results on Google can be iffy at best.

And I was not disappointed. Behold this laughably horrible false analogy (and even worse graphic) that I found on chegg.com:

Suppose Arti has to plan a birthday party and has lots of work to do like arranging stuff for decorations, planning venue for the party, arranging catering for the party, etc. All these tasks can not be done in one go and so need to be planned. Once the order of the tasks is decided, they are executed step by step so that all the arrangements are made in time and the party is a success.

Similarly, in Mathematics when a long expression needs to be differentiated or integrated, the calculation becomes cumbersome if the expression is considered as a whole but if it is broken down into small expressions, both differentiation and the integration become easy.

Pedagogically, I’m all for using whatever technique an instructor might deem necessary to to “sell” abstract mathematical concepts to students. Nevertheless, I’m pretty sure that this particular party-planning analogy has no potency for students who have progressed far enough to rigorously study infinite series.

Solving Problems Submitted to MAA Journals (Part 7i)

The following problem appeared in Volume 131, Issue 9 (2024) of The American Mathematical Monthly.

Let $X$ and $Y$ be independent normally distributed random variables, each with its own mean and variance. Show that the variance of $X$ conditioned on the event $X>Y$ is smaller than the variance of $X$ alone.

In previous posts, we reduced the problem to showing that if $f(x) = 1 + \sqrt{2\pi} x e^{x^2/2} \Phi(x)$ , then $f(x)$ is always positive, where

$\Phi(z) = \displaystyle \frac{1}{\sqrt{2\pi}} \int_{-\infty}^z e^{-z^2/2} \, dz$

is the cumulative distribution function of the standard normal distribution. If we can prove this, then the original problem will be true.

Motivated by the graph of $f(x)$ , I thought of a two-step method for showing $f$ must be positive: show that $f$ is an increasing function, and show that $\displaystyle \lim_{x \to -\infty} f(x) = 0$ . If I could prove both of these claims, then that would prove that $f$ must always be positive.

I was able to show the second step by demonstrating that, if $x<0$ ,

$\displaystyle f(x) = |x| e^{x^2/2} \int_{-\infty}^x \frac{1}{t^2} e^{-t^2/2} \, dt$ .

As discussed in the last post, the limit $\displaystyle \lim_{x \to -\infty} f(x) = 0$ follows from this equality. However, I just couldn’t figure out the first step.

So I kept trying.

And trying.

Until it finally hit me: I’m working too hard! The goal is to show that $f(x)$ is positive. Clearly, clearly, the right-hand side of the last equation is positive! So that’s the entire proof for $x<0$ … there was no need to prove that $f$ is increasing!

For $x \ge 0$ , it’s even easier. If $x$ is non-negative, then

$f(x) = 1 + \sqrt{2\pi} x e^{x^2/2} \Phi(x) \ge 1 + \sqrt{2\pi} \cdot 0 \cdot 1 \cdot \frac{1}{2} = 1 > 0$ .

So, in either case, $f(x)$ must be positive. Following the logical thread in the previous posts, this demonstrates that $\hbox{Var}(Z_1 \mid Z_1 > a+bZ_2) < 1$ , so that $\hbox{Var}(X \mid X <Y) < \hbox{Var}(X)$ , thus concluding the solution.

And I was really annoyed at myself that I stumbled over the last step for so long, when the solution was literally right in front of me.

Solving Problems Submitted to MAA Journals (Part 7h)

The following problem appeared in Volume 131, Issue 9 (2024) of The American Mathematical Monthly.

Let $X$ and $Y$ be independent normally distributed random variables, each with its own mean and variance. Show that the variance of $X$ conditioned on the event $X>Y$ is smaller than the variance of $X$ alone.

In previous posts, we reduced the problem to showing that if $g(x) = \sqrt{2\pi} x e^{x^2/2} \Phi(x)$ , then $f(x) = 1 + g(x)$ is always positive, where

$\Phi(z) = \displaystyle \frac{1}{\sqrt{2\pi}} \int_{-\infty}^z e^{-z^2/2} \, dz$

is the cumulative distribution function of the standard normal distribution. If we can prove this, then the original problem will be true.

When I was solving this problem for the first time, my progress through the first few steps was hindered by algebra mistakes and the like, but I didn’t doubt that I was progressing toward the answer. At this point in the solution, however, I was genuinely stuck: nothing immediately popped to mind for showing that $g(x)$ must be greater than $-1$ .

So I turned to Mathematica, just to make sure I was on the right track. Based on the graph, the function of $f(x)$ certainly looks positive.

What’s more, the graph suggests attempting to prove a couple of things: $f$ is an increasing function, and $\displaystyle \lim_{x \to -\infty} f(x) = 0$ or, equivalently, $\displaystyle \lim_{x \to -\infty} g(x) = -1$ . If I could prove both of these claims, then that would prove that $f$ must always be positive.

I started by trying to show

$\displaystyle \lim_{x \to -\infty} g(x) = \lim_{x \to \infty} x e^{x^2/2} \int_{-\infty}^x e^{-t^2/2} \, dt = -1$ .

I vaguely remembered something about the asymptotic expansion of the above integral from a course decades ago, and so I consulted that course’s textbook, by Bender and Orszag, to refresh my memory. To derive the behavior of $g(x)$ as $x \to -\infty$ , we integrate by parts. (This is permissible: the integrands below are well-behaved if $x<0$ , so that $0$ is not in the range of integration.)

$g(x) = \displaystyle x e^{x^2/2} \int_{-\infty}^x e^{-t^2/2} \, dt$

$= \displaystyle x e^{x^2/2} \int_{-\infty}^x \frac{1}{t} \frac{d}{dt} \left(-e^{-t^2/2}\right) \, dt$

$= \displaystyle x e^{x^2/2} \left[ -\frac{1}{t} e^{-t^2/2} \right]_{-\infty}^x - x e^{x^2/2} \int_{-\infty}^x \frac{d}{dt} \left(\frac{1}{t} \right) \left( -e^{-t^2/2} \right) \, dt$

$= \displaystyle x e^{x^2/2} \left[ -\frac{1}{x} e^{-x^2/2} - 0 \right] + |x| e^{x^2/2} \int_{-\infty}^x \frac{1}{t^2} e^{-t^2/2} \, dt$

$\displaystyle = - 1 +|x| e^{x^2/2} \int_{-\infty}^x \frac{1}{t^2} e^{-t^2/2} \, dt$

$\displaystyle = -1+ |x| e^{x^2/2} \int_{-\infty}^x \frac{1}{t^2} e^{-t^2/2} \, dt$ .

This is agonizingly close: the leading term is $-1$ as expected. However, I was stuck for the longest time trying to show that the second term goes to zero as $x \to -\infty$ .

So, once again, I consulted Bender and Orszag, which outlined how to show this. We note that

$\left|g(x) + 1\right| = \displaystyle |x| e^{x^2/2} \int_{-\infty}^x \frac{1}{t^2} e^{-t^2/2} \, dt < \displaystyle |x| e^{x^2/2} \int_{-\infty}^x \frac{1}{x^2} e^{-t^2/2} \, dt = \displaystyle \frac{g(x)}{x^2}$ .

Therefore,

$\displaystyle \lim_{x \to -\infty} \left| \frac{g(x)+1}{g(x)} \right| \le \lim_{x \to -\infty} \frac{1}{x^2} = 0$ ,

so that

$\displaystyle \lim_{x \to -\infty} \left| \frac{g(x)+1}{g(x)} \right| =\displaystyle \lim_{x \to -\infty} \left| 1 + \frac{1}{g(x)} \right| = 0$ .

Therefore,

$\displaystyle \lim_{x \to -\infty} \frac{1}{g(x)} = -1$ ,

$\displaystyle \lim_{x \to -\infty} g(x) = -1$ .

So (I thought) I was halfway home with the solution, and all that remained was to show that $f$ was an increasing function.

And I was completely stuck at this point for a long time.

Until I realized — much to my utter embarrassment — that showing $f$ was increasing was completely unnecessary, as discussed in the next post.