Limits at infinity and horizontal asymptotes, the degree rule for rational functions, oblique asymptotes obtained by polynomial division, the end behaviour of polynomials and of the transcendental families, and the complete curve-sketching procedure combining asymptotes with the derivative tests of Section 4.5.
Subject: Calculus I · 72 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 4.6
What a graph does far away, and how to draw all of it
Objectives
Section 4.5 turned the signs of the first and second derivatives into the shape of a graph near its interesting points. That is only half a picture: it says nothing about where the graph goes far to the left and far to the right. This lesson supplies the other half and then puts both together.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 354-380 — learning objectives 4.6.1 to 4.6.5
In Section 4.5 you learned to read a graph's shape from its derivatives: where it rises and falls, where it bends up or down. But every one of those facts is local. None of them tells you what the graph does when x is a million, or minus a million, and a sketch has to show that too.
This lesson adds the missing ingredient, limits at infinity. They give you horizontal asymptotes, the lines a graph settles onto far away, and, when the graph does not settle, oblique asymptotes or a clean statement that it grows without bound.
You will also meet the precise definitions behind the phrase arbitrarily close, learn a quick rule for rational functions, rank the standard families by how fast they grow, and finish by combining everything into a six-step recipe that produces a complete, trustworthy sketch.
Warm-up
\[ f(x) = \frac{1}{x - 3} \]
Discussion prompt
From Section 2.2: what are the two one-sided limits of this function at x = 3, and what do they say about its graph there?
Write down both one-sided limits before you reveal the answer. This is the Section 2.2 idea you are about to turn inside out, so it is worth having it fresh.
Near x equal to 3 the denominator is a tiny number. On the right it is tiny and positive, so one over it is huge and positive; on the left it is tiny and negative, so the fraction is huge and negative. That is what a vertical asymptote is: the input approaches a finite number and the output runs off to infinity.
Today the roles swap. The input runs off to infinity, and you ask what the output does. Sometimes it settles at a finite number, and then you get a horizontal asymptote, the subject of the first part.
Section
Part 1
Concept
Figure (svg): The graph of f of x equals 2 plus 1 over x for x from minus 9 to 9, with a dashed horizontal line at y equals 2. Far to the right the curve sits just above the line; far to the left, just below it. Near x equals 0 it shoots off vertically.
| x | 10 | 100 | 1,000 | 10,000 |
|---|---|---|---|---|
| 2 + 1/x | 2.1 | 2.01 | 2.001 | 2.0001 |
As x gets larger, the one over x shrinks toward zero and the values close in on 2. The same happens for large negative x, from below.
\[ \lim_{x\to\infty}\left(2 + \frac1x\right) = 2, \qquad \lim_{x\to-\infty}\left(2 + \frac1x\right) = 2 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 354-355 — Figure 4.40 and Table 4.2
Look at the right half of the picture first. The curve starts high near the vertical axis and then flattens out, sliding down toward the dashed line at height 2 without ever reaching it. The table puts numbers on that: 2.1, 2.01, 2.001, each extra zero in x buying one more correct decimal place.
The reason is easy to see in the formula. The function is 2 plus one over x, and one over x gets as small as you like once x is large enough. So the values get as close to 2 as you like.
Now look at the left half. For large negative x, one over x is a small negative number, so the values are just below 2: 1.9, 1.99, 1.999. The curve approaches the same line from underneath. Two ends, one limit, approached from opposite sides.
Concept
Limit at infinity — If the values of f become arbitrarily close to a number L once x is large enough, f has limit L at infinity. The same idea for large negative x gives the limit at negative infinity.
\[ \lim_{x\to\infty} f(x) = L \qquad \lim_{x\to-\infty} f(x) = L \]
Horizontal asymptote — If either limit equals a finite number L, the line y equals L is a horizontal asymptote of f.
A function can have a different horizontal asymptote at each end, one at only one end, or none at all.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 355 — the informal definition and the definition of horizontal asymptote
The definition is the same idea as an ordinary limit with one change: instead of x getting close to a number, x gets large. You still ask whether the outputs close in on a single value L.
There are two separate questions, one for each end of the number line, and the answers can differ. That is why the horizontal asymptote definition says or: if either end has a finite limit, you get a horizontal asymptote at that end.
Keep the three possible outcomes in mind. A function can have the same asymptote at both ends, like 2 plus one over x; a different one at each end, like the inverse tangent you will meet shortly; or an asymptote at one end only, like e to the x, which settles at zero on the left and climbs forever on the right.
Notation
Annotate
On: \( \lim_{x\to\infty} f(x) = L \quad\text{or}\quad \lim_{x\to-\infty} f(x) = L \;\Longrightarrow\; y = L \)
Step through the annotations and connect each to the picture of 2 plus one over x. The arrow toward infinity does not mean x reaches infinity; infinity is not a number. It is shorthand for x growing past every bound you could name.
The limit must be a finite number. If the outputs grow without bound, the limit is said to be infinity, but that describes the manner of failing to have a finite limit, and it gives no horizontal line to draw.
The last annotation is the one that saves you from the trap on the next slide but one. A horizontal asymptote is about the far ends only. Near the origin the graph can do anything at all, including cross the line, without affecting the asymptote.
Picture it
Figure (svg): The graph of f of x equals cos x over x plus 1 for x from about 0.6 to 30, oscillating around a dashed line at y equals 1 with shrinking swings; yellow dots mark the first several places the curve crosses the line.
\[ f(x) = \frac{\cos x}{x} + 1, \qquad \lim_{x\to\infty} f(x) = 1 \]
A vertical asymptote cannot be crossed, because the graph has to shoot off to infinity next to it. A horizontal asymptote only describes the far ends, and a graph may cross it as often as it likes.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 355-356 — Figure 4.42
Follow the curve from the left. It starts above the line, dives below it, comes back above, and keeps weaving across the dashed line. The yellow dots mark the crossings, which happen wherever cosine is zero, at every odd multiple of pi over 2, so there are infinitely many of them.
And yet the limit at infinity is 1. The cosine term is divided by x, so the swings shrink: by x equal to 30 they are about a thirtieth in size, and they keep shrinking toward nothing. The curve crosses the line forever while getting ever closer to it.
Contrast a vertical asymptote. There the graph must shoot off toward infinity as x approaches the line, so it cannot wander across it. The two kinds of asymptote are different kinds of statement, and only the vertical kind forbids crossing.
Trap
A common definition from earlier courses:
\[ \text{asymptote} = \text{a line the graph} \]
\[ \text{approaches but never meets} \]
Wrong for horizontal asymptotes.
A horizontal asymptote is defined by a limit at an end, nothing more. The graph of cos x over x plus 1 meets y equals 1 at every odd multiple of pi over 2, and it still has that asymptote.
\[ f\left(\tfrac{\pi}{2}\right) = 0 + 1 = 1 \]
Many people first meet asymptotes through graphs like one over x, where the curve really does approach its lines without touching, and they generalise that into a definition. For vertical asymptotes the habit happens to be harmless. For horizontal ones it is simply false.
The definition you must use is the limit. Ask what the function does as x grows. Whether the graph touches the line at some particular x, or at infinitely many, is a separate question that the definition never asks. If you see a graph cross a line and conclude that the line cannot be an asymptote, check the limit instead.
Worked example
Evaluate both limits at infinity and find the horizontal asymptotes.
\[ f(x) = 5 - \frac{2}{x^2} \]
Split with the limit laws
Why: The limit of a difference is the difference of the limits.
\[ \lim_{x\to\infty}\left(5 - \frac{2}{x^2}\right) = \lim_{x\to\infty} 5 - 2\left(\lim_{x\to\infty}\frac1x\right)\left(\lim_{x\to\infty}\frac1x\right) \]
Use the basic limit
Why: One over x shrinks to zero as x grows.
\[ = 5 - 2 \cdot 0 \cdot 0 = 5 \]
Repeat at the other end
Why: One over x also tends to zero as x goes to negative infinity, and the square makes the sign irrelevant.
\[ \lim_{x\to-\infty}\left(5 - \frac{2}{x^2}\right) = 5 \]
Name the asymptote
Why: The same finite limit at both ends.
\[ y = 5 \text{ at both ends} \]
Check with a large input
Why: At x equal to 100 the value is already within two ten-thousandths of 5, and it sits below 5 because something positive is subtracted.
\[ f(100) = 5 - \frac{2}{10{,}000} = 4.9998 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 356 — Example 4.21a and Figure 4.43
This is the limit laws doing their usual job, now at infinity. The key fact is the basic limit that one over x tends to zero as x grows; the two over x squared is just two times one over x times one over x, so it tends to two times zero times zero.
At negative infinity nothing changes. One over x still tends to zero, and squaring removes the sign, so the limit is again 5. Both ends share the asymptote y equals 5.
The check tells you a little more than the limit does. The subtracted term is always positive, so the values sit just below 5 at both ends, and that is how you would draw the graph approaching its asymptote: from underneath, on both sides.
Worked example
\[ f(x) = \frac{\sin x}{x} \]
Bound the sine
Why: Sine never leaves the interval from minus one to one.
\[ -1 \le \sin x \le 1 \]
Divide by x, for x positive
Why: Dividing by a positive number keeps the inequalities pointing the same way.
\[ -\frac1x \le \frac{\sin x}{x} \le \frac1x \]
Take the limits of the outer functions
Why: Both envelopes shrink to zero.
\[ \lim_{x\to\infty}\left(-\frac1x\right) = 0 = \lim_{x\to\infty}\frac1x \]
Apply the squeeze theorem
Why: Trapped between two functions with limit zero, the middle one has limit zero too.
\[ \lim_{x\to\infty}\frac{\sin x}{x} = 0 \]
Repeat at negative infinity
Why: For negative x the inequalities flip when you divide, but the two envelopes are the same pair, so the squeeze still works.
\[ \lim_{x\to-\infty}\frac{\sin x}{x} = 0 \;\Longrightarrow\; y = 0 \]
Figure (svg): The graph of sine x over x for x from minus 20 to 20, drawn between two dashed envelope curves y equals 1 over x and y equals minus 1 over x; the envelopes pinch toward the x-axis at both ends, and so does the curve.
Check with a number
Why: At x equal to 100 the value is tiny, and smaller in size than one hundredth, as the envelope promises.
\[ \frac{\sin 100}{100} \approx -0.00506, \qquad |-0.00506| < \frac{1}{100} \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 356-357 — Example 4.21b and Figure 4.44
Here the limit laws alone do not work, because sine x has no limit at infinity. You cannot split this into the limit of sine times the limit of one over x. The squeeze theorem gets around that by never asking what sine does, only how big it can be.
Sine is trapped between minus one and one. Divide by a positive x and the quotient is trapped between minus one over x and one over x, the two dashed curves in the figure. Both of those shrink to zero, so whatever is caught between them must shrink to zero too.
The picture is worth reading carefully. The curve crosses the x-axis at every multiple of pi, so this is another function that crosses its horizontal asymptote infinitely often. The check at 100 shows the value is smaller than the envelope, which is exactly what the squeeze promised.
Worked example
\[ f(x) = \tan^{-1} x \]
Recall the tangent's vertical asymptotes
Why: On the interval from minus pi over 2 to pi over 2, the tangent climbs to infinity at the right end and falls to minus infinity at the left end.
\[ \lim_{x\to(\pi/2)^-}\tan x = \infty, \qquad \lim_{x\to(-\pi/2)^+}\tan x = -\infty \]
Reflect in the line y = x
Why: The inverse swaps inputs and outputs, so a vertical asymptote of tangent becomes a horizontal asymptote of its inverse.
\[ \lim_{x\to\infty}\tan^{-1} x = \frac{\pi}{2} \]
The other end
Why: The left-hand vertical asymptote becomes the lower horizontal one.
\[ \lim_{x\to-\infty}\tan^{-1} x = -\frac{\pi}{2} \]
Figure (svg): The graph of the inverse tangent for x from minus 8 to 8, rising through the origin and flattening toward a dashed line at pi over 2 on the right and another at minus pi over 2 on the left.
Check with a large input
Why: At 100 the inverse tangent is within a hundredth of pi over 2, and just below it.
\[ \tan^{-1}(100) \approx 1.5608 < 1.5708 \approx \frac{\pi}{2} \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 357-358 — Example 4.21c and Figures 4.45, 4.46
This limit is read from a graph you already know. On the interval from minus pi over 2 to pi over 2, the tangent climbs from minus infinity to infinity, with vertical asymptotes at both ends. The inverse tangent undoes the tangent, so its graph is the tangent's graph reflected in the line y equals x.
Reflection swaps horizontal and vertical. The vertical asymptote at pi over 2 becomes a horizontal asymptote at height pi over 2 on the right, and the one at minus pi over 2 becomes a horizontal asymptote at minus pi over 2 on the left.
So the inverse tangent has two different horizontal asymptotes, one for each end. The check at 100 lands just below pi over 2, and the gap keeps closing as x grows.
Fill the middle
Fill in the blanks
\lim_3\left(3 + \frac3___\right) = ___, \qquad \lim____\left(3 + \frac______\right) = ___
Why: Four over x shrinks to zero at both ends, so both limits are 3 and the only horizontal asymptote is y = 3. The graph is above it on the right, where 4/x is positive, and below it on the left.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 358 — Checkpoint 4.20
Fill in both limits before you check. The structure is identical to 2 plus one over x from the start of the lesson, only with different constants.
The four over x term tends to zero at both ends, so both limits are 3 and there is one horizontal asymptote, y equals 3. As a bonus, think about the side: on the right four over x is positive, so the graph is above the line, and on the left it is negative, so the graph is below it.
Section
Part 2
Concept
Arbitrarily close and sufficiently large are vague. The formal definition turns them into a challenge and a response: you are given a tolerance epsilon, and you must find a point N past which the function stays within that tolerance.
\[ \forall \varepsilon > 0 \;\exists N > 0: \quad x > N \;\Longrightarrow\; |f(x) - L| < \varepsilon \]
Figure (svg): The curve y equals 2 plus 1 over x for x from 0.8 to 12, with a shaded horizontal band from 1.75 to 2.25 around the line y equals 2. A vertical line at x equals 4 marks N: to the right of it, the curve stays inside the band.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 359-360 — the formal definition and Figure 4.48
Everything so far relied on phrases like arbitrarily close and sufficiently large, which are fine for intuition but too vague to prove anything with. The formal definition makes them into a game with two players.
The first player picks a tolerance epsilon, the half-width of the yellow band around the line y equals L. The second player must answer with a point N on the x-axis such that the whole graph to the right of N lies inside the band. The limit is L exactly when the second player can always win, whatever epsilon is chosen.
In the figure epsilon is a quarter, and N equal to 4 works: past x equal to 4, the curve 2 plus one over x is inside the band. A smaller epsilon would force N further right, but some N always exists, and that is what makes 2 the limit.
Notation
Annotate
On: \( \forall\varepsilon > 0 \;\exists N > 0 : \quad x > N \Rightarrow |f(x) - L| < \varepsilon \)
Read the symbols as a sentence: for every epsilon bigger than zero, there exists an N such that every x beyond N gives an output within epsilon of L. The order matters. Epsilon is chosen first, and N is allowed to depend on it.
The phrase for all x beyond N is what separates a limit from a lucky coincidence. It is not enough for the graph to dip into the band at some point; it has to enter the band and stay there.
Finally, the absolute value is just distance. Saying that f of x minus L is less than epsilon in size is the same as saying the point on the graph is inside the band of half-width epsilon. Once you see the band, the definition becomes a picture.
Worked example
\[ \text{Prove } \lim_{x\to\infty}\left(2 + \frac1x\right) = 2 \]
Write down what must be made small
Why: The distance from f to L simplifies to the size of one over x.
\[ \left|2 + \frac1x - 2\right| = \left|\frac1x\right| \]
Drop the absolute value for positive x
Why: Once x is past a positive N, one over x is positive.
\[ \left|\frac1x\right| = \frac1x \quad (x > 0) \]
Choose N from epsilon
Why: Make one over N equal to epsilon.
\[ \text{Given } \varepsilon > 0, \text{ let } N = \frac{1}{\varepsilon} \]
Show it works for every x past N
Why: A bigger denominator gives a smaller fraction.
\[ x > N \;\Longrightarrow\; \frac1x < \frac1N = \varepsilon \]
Check with a concrete tolerance
Why: For epsilon equal to one hundredth the recipe gives N equal to 100; an input just past it is inside the band.
\[ \varepsilon = 0.01,\; N = 100: \quad \left|f(101) - 2\right| = \frac{1}{101} \approx 0.0099 < 0.01 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 360 — Example 4.22
Every epsilon-N proof has the same shape, and this is the simplest one, so learn the shape here. Start from the distance between f and L and simplify it until you can see how it depends on x. Here it collapses to one over x.
Then work backwards: you need one over x to be less than epsilon, and that happens when x is bigger than one over epsilon. So choose N equal to one over epsilon. The written proof then runs forwards: for x beyond N, one over x is less than one over N, which is epsilon.
The check is not part of the proof, but it is how you catch mistakes in your choice of N. With epsilon a hundredth, N is 100, and an input just past it, 101, gives a distance of about 0.0099, just inside the band.
Worked example
\[ \text{Prove } \lim_{x\to\infty}\left(3 - \frac{1}{x^2}\right) = 3 \]
Simplify the distance
Why: The threes cancel and the square is positive.
\[ \left|3 - \frac{1}{x^2} - 3\right| = \frac{1}{x^2} \]
Solve for the cutoff
Why: One over x squared is below epsilon exactly when x squared exceeds one over epsilon.
\[ \frac{1}{x^2} < \varepsilon \iff x^2 > \frac{1}{\varepsilon} \iff x > \frac{1}{\sqrt{\varepsilon}} \quad (x > 0) \]
Choose N
Why: Take the cutoff itself.
\[ N = \frac{1}{\sqrt{\varepsilon}} \]
Conclude
Why: For every x past N, the chain above runs forward.
\[ x > N \;\Longrightarrow\; \left|f(x) - 3\right| = \frac{1}{x^2} < \frac{1}{N^2} = \varepsilon \]
Check with a concrete tolerance
Why: For epsilon equal to one hundredth, N is 10, and 11 is inside the band.
\[ \varepsilon = 0.01,\; N = 10: \quad \frac{1}{11^2} = \frac{1}{121} \approx 0.0083 < 0.01 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 360 — Checkpoint 4.21
Same shape as the previous example, with one extra step. The distance simplifies to one over x squared, and you want that below epsilon.
Solving the inequality means inverting and then taking a square root: x squared must exceed one over epsilon, so x must exceed one over the square root of epsilon. That becomes your N. Because the square root grows more slowly than its argument, this N is smaller than the one for 2 plus one over x at the same epsilon: the graph enters the band sooner, since one over x squared shrinks faster than one over x.
The numerical check confirms it: with epsilon a hundredth, N is just 10, and at 11 the distance is about 0.0083, safely inside.
Concept
| x | 10 | 20 | 50 | 100 | 1000 |
|---|---|---|---|---|---|
| x³ | 1,000 | 8,000 | 125,000 | 10⁶ | 10⁹ |
Some functions do not settle at all: they grow past every bound. The formal version replaces the band around L with a height M that the graph must clear for good.
\[ \lim_{x\to\infty} f(x) = \infty \iff \forall M > 0 \;\exists N > 0: \; x > N \Rightarrow f(x) > M \]
Figure (svg): The curve y equals x cubed for x from 0 to 6, a horizontal line at height M equals 64, and a vertical line at N equals 4 where the curve crosses that height; to the right of N the curve stays above M.
For minus infinity, M is negative and the graph must stay below it.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 358-361 — Table 4.3, the definitions, and Figure 4.49
The table shows x cubed exploding: a thousand, then eight thousand, then a billion by x equal to 1000. There is no finite number to close in on, so the band around L makes no sense. Instead the challenge is a height M, and the response is a point N past which the graph stays above M.
In the figure M is 64 and N is 4, the cube root of 64. To the right of the vertical line, the whole curve sits in the green region above the bar. Raise the bar and N moves right, but there is always an N.
The phrase stays above matters just as stays inside mattered before. A function that keeps climbing above M and then dropping back below it does not have an infinite limit, and you will meet exactly such a function later in the lesson.
Worked example
\[ \text{Prove } \lim_{x\to\infty} x^3 = \infty \]
Solve the target inequality
Why: The cube is increasing, so it exceeds M exactly past the cube root of M.
\[ x^3 > M \iff x > \sqrt[3]{M} \]
Choose N from M
Why: Take the cube root itself.
\[ \text{Given } M > 0, \text{ let } N = \sqrt[3]{M} \]
Show it works
Why: Cubing preserves the order of positive numbers.
\[ x > N \;\Longrightarrow\; x^3 > N^3 = \left(\sqrt[3]{M}\right)^3 = M \]
Check with a concrete height
Why: For M equal to 1000 the recipe gives N equal to 10, and 11 cubed clears the bar.
\[ M = 1000,\; N = 10: \quad 11^3 = 1331 > 1000 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 361 — Example 4.23
The strategy is the same as before: find the condition on x that forces the output past the target, then use it as N. Because the cube function is increasing, x cubed is bigger than M exactly when x is bigger than the cube root of M.
So given any height M, choose N equal to the cube root of M. For every x beyond N, cubing both sides of the inequality preserves it, because cubing is increasing, and the cube of the cube root of M is M itself.
The check with M equal to 1000 is concrete: N is 10, and at 11 the cube is 1331, over the bar. Try a bigger M yourself, say a million: N becomes 100, and every x past 100 has a cube over a million.
Fill the middle
\[ \text{Prove } \lim_{x\to\infty} 3x^2 = \infty: \quad 3x^2 > M \iff x > \sqrt{\frac{M}{3}} = N \]
Fill in the blanks
M = 300 \;\Longrightarrow\; N = \sqrt10363} = ___, \qquad 3(11)^2 = ___ > 300
Why: In general N is the square root of M over 3. For M = 300 that is the square root of 100, which is 10, and the next whole number, 11, gives 3 times 121 = 363, which is indeed above 300.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 361 — Checkpoint 4.22
Fill in both numbers before checking. The general recipe is already on the slide: three x squared exceeds M exactly when x exceeds the square root of M over 3, for positive x.
With M equal to 300, that is the square root of 100, which is 10. The second blank is the check: an input just past N should clear the bar, and 3 times 11 squared is 363, which does. A complete proof would say: given M, let N be the square root of M over 3; then for x beyond N, three x squared is bigger than three N squared, which equals M.
Section
Part 3
Concept
Figure (svg): Three small panels. Left: a curve flattening onto a dashed horizontal line. Middle: a curve climbing without bound. Right: a sine wave that keeps oscillating between minus 1 and 1.
The behaviour of a function as x goes to infinity or negative infinity is called its end behaviour. At each end, independently, one of three things happens.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 361 — End Behavior
Every end of every function falls into one of the three pictures. Settling onto a horizontal line is the first. Growing without bound, upward or downward, is the second. The third is everything else, most often an oscillation like sine that never commits to anything.
The two ends are independent. The exponential settles on the left and grows on the right; the inverse tangent settles at different heights on the two sides. So end behaviour is always two answers, one for each direction.
The next few slides go through the families of functions you know and sort their ends into these three boxes, starting with the simplest, the powers of x.
Concept
Figure (svg): Two panels. Left: y equals x squared, x to the fourth and x to the sixth, all U-shaped, rising at both ends. Right: y equals x, x cubed and x to the fifth, all rising on the right and falling on the left.
\[ \lim_{x\to\infty} x^n = \infty, \qquad \lim_{x\to-\infty} x^n = \begin{cases} \infty & n \text{ even} \\ -\infty & n \text{ odd} \end{cases} \]
A coefficient in front changes the height, and a negative coefficient flips both ends over.
\[ \lim_{x\to\pm\infty} cx^n = \begin{cases} \lim x^n & c > 0 \\ -\lim x^n & c < 0 \end{cases} \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 361-362 — Figures 4.50 and 4.51
Look at the left panel: x squared, x to the fourth and x to the sixth are all U-shaped. A large negative number raised to an even power is positive and large, so even powers rise at both ends. Higher powers are flatter near the origin and steeper further out, but the ends all point up.
In the right panel the odd powers keep the sign of x. Large positive x gives large positive values, large negative x gives large negative values, so the graph falls on the left and rises on the right.
A constant in front only rescales, unless it is negative. A negative coefficient reflects the graph in the x-axis and turns both ends over. Those two facts, the parity of the power and the sign of the coefficient, decide the ends of any power function.
Worked example
\[ \text{(a) } f(x) = -5x^3 \qquad \text{(b) } f(x) = 2x^4 \]
Read the power and sign in (a)
Why: Odd power, negative coefficient: the ends of x cubed, flipped.
\[ \lim_{x\to\infty}(-5x^3) = -\infty, \qquad \lim_{x\to-\infty}(-5x^3) = \infty \]
Read the power and sign in (b)
Why: Even power, positive coefficient: up at both ends.
\[ \lim_{x\to\infty}2x^4 = \infty, \qquad \lim_{x\to-\infty}2x^4 = \infty \]
Checkpoint 4.23 the same way
Why: Even power, negative coefficient: down at both ends.
\[ \lim_{x\to\infty}(-3x^4) = -\infty \]
Check with inputs of 10 and minus 10
Why: The signs of the values match every limit claimed above.
\[ -5(10)^3 = -5000, \quad -5(-10)^3 = 5000, \quad -3(10)^4 = -30{,}000 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 363 — Example 4.24 and Checkpoint 4.23
Read off two things from each function: is the power even or odd, and is the coefficient positive or negative? Minus 5 x cubed has an odd power and a negative coefficient, so it is x cubed flipped: down on the right, up on the left.
Two x to the fourth has an even power and a positive coefficient, so both ends go up. Checkpoint 4.23, minus three x to the fourth, is the same even power with a negative coefficient, so both ends go down.
The check with inputs of 10 and minus 10 is a habit worth keeping. Plugging in one large value on each side takes seconds and catches the most common slip, a lost minus sign.
Concept
Factor the leading term out of any polynomial of degree n.
\[ f(x) = a_n x^n + a_{n-1}x^{n-1} + \cdots + a_1 x + a_0 \]
\[ = a_n x^n \left(1 + \frac{a_{n-1}}{a_n}\cdot\frac1x + \cdots + \frac{a_0}{a_n}\cdot\frac{1}{x^n}\right) \]
Every term inside the bracket except the 1 has a power of x in its denominator, so the bracket tends to 1 at both ends.
\[ \lim_{x\to\pm\infty} f(x) = \lim_{x\to\pm\infty} a_n x^n \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 363 — the end behaviour of a polynomial
This short derivation is why you can ignore everything except the leading term of a polynomial when you ask about its ends. Factor out the leading term and look at what is left inside the bracket.
Every term in the bracket except the first 1 has a power of x in its denominator, so each one tends to zero as x grows in either direction. The bracket therefore tends to 1, and the polynomial behaves like its leading term times something close to 1.
So the end behaviour of any polynomial is the end behaviour of its leading term, which you just learned to read off from the parity of the power and the sign of the coefficient. The lower terms shape the middle of the graph, its bumps and wiggles, but not its ends.
Picture it
Figure (svg): Two panels. Left, a close view from x equals minus 2 to 2: the curves f of x equals 5x cubed minus 3x squared plus 4 and g of x equals 5x cubed are clearly different. Right, a far view from x equals minus 100 to 100: the two curves lie on top of each other.
| x | 10 | 100 | 1000 |
|---|---|---|---|
| f(x) = 5x³ − 3x² + 4 | 4,704 | 4,970,004 | 4,997,000,004 |
| g(x) = 5x³ | 5,000 | 5,000,000 | 5,000,000,000 |
| f/g | 0.941 | 0.994 | 0.9994 |
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 363-364 — Figure 4.52 and Table 4.4
On the left, in a close view, the two graphs are clearly different: the extra terms lift f above g near the origin and give it a flat spot. On the right, zoomed out by a factor of fifty, they are indistinguishable at the scale of millions.
The table measures this. At x equal to 10 the values differ by about 6 percent. At 100 the ratio is 0.994, and at 1000 it is 0.9994. The ratio is heading to 1, which is precisely what the factoring argument said.
Notice that the difference between f and g does not shrink; it actually grows, since three x squared grows. What shrinks is the difference relative to the size of the values, and that is what the end behaviour of the graph depends on.
Sorting
Sort into buckets
Sort each polynomial by its end behaviour, reading only its leading term.
Sort each polynomial by finding its leading term, the one with the highest power, and reading off parity and sign. Ignore the size of the coefficients and the other terms completely.
Two items are there to catch you. In 7 x to the sixth minus x to the seventh, the leading term is written second, and it is minus x to the seventh, odd and negative. In the product of x minus 1 squared and x plus 2, you have to multiply the leading parts, x squared times x, to see an x cubed with coefficient one.
The first item is also a warning. With a tiny coefficient of a hundredth on x to the fourth and a huge minus 1000 x, the graph dips far below zero near the origin, yet both ends still go up. The leading term always wins eventually.
Section
Part 4
Concept
For a quotient, both the top and the bottom usually grow without bound, and infinity over infinity says nothing. The fix is one move: divide the numerator and the denominator by the highest power of x that appears in the denominator.
\[ \frac{3x - 1}{2x + 5} = \frac{(3x - 1)/x}{(2x + 5)/x} = \frac{3 - \frac1x}{2 + \frac5x} \]
Now every leftover term is a constant or a constant over a power of x, and the limit laws apply term by term.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 364 — End Behavior for Algebraic Functions
For a quotient of polynomials, the limit laws cannot be applied directly, because both the numerator and the denominator usually grow without bound, and infinity over infinity has no fixed value. You have to rewrite first.
The one move to remember is: divide the top and the bottom by the highest power of x that appears in the denominator. This does not change the value of the fraction, since you divide both by the same thing, but it turns the denominator into a constant plus terms that vanish.
After the division, every term is a constant or a constant over a power of x, and each of the second kind tends to zero. The limit is then a matter of reading off constants, as the three parts of Example 4.25 show.
Worked example
\[ f(x) = \frac{3x - 1}{2x + 5} \]
Divide top and bottom by x
Why: The highest power in the denominator is x itself.
\[ f(x) = \frac{3 - 1/x}{2 + 5/x} \]
Apply the limit laws
Why: Each fraction with x underneath tends to zero, at either end.
\[ \lim_{x\to\pm\infty} f(x) = \frac{3 - 0}{2 + 0} = \frac32 \]
Read the asymptotes
Why: The same finite limit at both ends; the denominator vanishes at minus five halves.
\[ \text{HA } y = \tfrac32, \qquad \text{VA } x = -\tfrac52 \]
Figure (svg): The graph of 3x minus 1 over 2x plus 5 for x from minus 16 to 12: a dashed horizontal line at three halves and a dashed vertical line at x equals minus 5 halves. The right branch rises toward three halves from below; the left branch falls toward it from above.
Check with large inputs
Why: The values at plus and minus 1000 straddle three halves, one from each side, as the figure shows.
\[ f(1000) \approx 1.4958, \qquad f(-1000) \approx 1.5043 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 364-365 — Example 4.25a and Figure 4.53
The highest power in the denominator is x, so divide every term by x. The fraction becomes 3 minus one over x, over 2 plus five over x, and both small fractions vanish at either end, leaving three halves.
That number is just the ratio of the leading coefficients, 3 over 2. For equal degrees this always happens, and you will see it stated as a rule shortly.
The figure shows the full picture: a vertical asymptote where the denominator is zero, at minus five halves, and the same horizontal asymptote at both ends. The check at plus and minus 1000 shows the two ends approaching three halves from opposite sides, below on the right and above on the left.
Worked example
\[ f(x) = \frac{3x^2 + 2x}{4x^3 - 5x + 7} \]
Divide top and bottom by x cubed
Why: That is the highest power in the denominator.
\[ f(x) = \frac{3/x + 2/x^2}{4 - 5/x^2 + 7/x^3} \]
Take the limits term by term
Why: Every term on top has x in its denominator; the bottom keeps the 4.
\[ \lim_{x\to\pm\infty} f(x) = \frac{0 + 0}{4 - 0 + 0} = 0 \]
Read the asymptote
Why: The x-axis is the horizontal asymptote at both ends.
\[ y = 0 \]
Check with large inputs
Why: The values are small and close to three over 4x, the ratio of the leading terms.
\[ f(100) \approx 0.00755 \approx \frac{3}{400} = 0.0075 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 365 — Example 4.25b and Figure 4.54
Now the denominator has degree 3, so divide everything by x cubed. Every term on top acquires a power of x underneath and vanishes, while the bottom keeps its leading coefficient 4. The limit is zero over four, which is zero.
The general lesson: when the denominator has higher degree, it outgrows the numerator and the fraction shrinks to zero at both ends. The horizontal asymptote is the x-axis.
The check shows more than the limit. At x equal to 100 the value is very close to three over four x, the ratio of the leading terms, so for large x this function behaves like three quarters of one over x. That tells you the graph approaches the axis from above on the right and from below on the left.
Worked example
\[ f(x) = \frac{3x^2 + 4x}{x + 2} \]
Divide top and bottom by x
Why: The highest power in the denominator.
\[ f(x) = \frac{3x + 4}{1 + 2/x} \]
Look at the denominator
Why: It tends to one at both ends.
\[ 1 + \frac2x \to 1 \]
Look at the numerator
Why: It follows the sign of x.
\[ 3x + 4 \to \infty \;(x \to \infty), \qquad 3x + 4 \to -\infty \;(x \to -\infty) \]
Conclude
Why: No horizontal asymptote at either end.
\[ \lim_{x\to\infty} f(x) = \infty, \qquad \lim_{x\to-\infty} f(x) = -\infty \]
Check with large inputs
Why: Both values are huge, with the signs claimed.
\[ f(100) \approx 298.04, \qquad f(-100) \approx -302.04 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 365-366 — Example 4.25c and Figure 4.55
Here the numerator has the higher degree. Divide by x, the highest power in the denominator: the bottom becomes 1 plus two over x, which tends to 1, while the top is still 3x plus 4, which does not settle at all.
So the whole fraction behaves like 3x plus 4 for large x, heading to infinity on the right and minus infinity on the left. There is no horizontal asymptote at either end.
But look closely at the check values: 298.04 and minus 302.04. Those are suspiciously close to 3x minus 2 evaluated at 100 and minus 100, which gives 298 and minus 302. That is not a coincidence, and the next slide explains it.
Concept
The last check hid a pattern: 298.04 and minus 302.04 are almost exactly 3x minus 2 at x equal to 100 and minus 100. Long division explains why.
\[ \frac{3x^2 + 4x}{x + 2} = 3x - 2 + \frac{4}{x + 2} \]
\[ \lim_{x\to\pm\infty}\Big(f(x) - (3x - 2)\Big) = \lim_{x\to\pm\infty}\frac{4}{x + 2} = 0 \]
Oblique asymptote — If the vertical distance between the graph of f and a non-horizontal line y = mx + b tends to zero as x goes to infinity or negative infinity, that line is an oblique (slant) asymptote.
Figure (svg): The graph of 3x squared plus 4x over x plus 2 for x from minus 12 to 10, with a dashed slanted line y equals 3x minus 2 and a dashed vertical line at x equals minus 2. Both branches run alongside the slanted line far from the origin.
Check the division by multiplying back
Why: Quotient times divisor plus remainder must return the numerator.
\[ (x + 2)(3x - 2) + 4 = 3x^2 + 4x - 4 + 4 = 3x^2 + 4x \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 366-367 — the oblique asymptote and Figure 4.56
Long division of polynomials splits the function into a polynomial part and a remainder fraction. Here the polynomial part is 3x minus 2, and the remainder fraction is four over x plus 2, which is bottom-heavy and so shrinks to zero at both ends.
That means the vertical distance between the graph and the line y equals 3x minus 2 tends to zero. The graph hugs a slanted line, exactly as a graph with a horizontal asymptote hugs a horizontal one. Such a line is called an oblique, or slant, asymptote.
Look at the figure: far to the right and far to the left, the curve and the dashed line are nearly on top of each other, even though the function is not linear. The last line checks the division the reliable way, by multiplying the quotient by the divisor and adding back the remainder.
Notation
Annotate
On: \( f(x) = \frac{a_n x^n + \cdots + a_0}{b_m x^m + \cdots + b_0}, \quad a_n, b_m \ne 0 \)
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 367 — the summary of Example 4.25
The three parts of Example 4.25 generalise into one rule about the degrees n on top and m on the bottom. Step through the annotations and match each case to its example.
Equal degrees give the ratio of the leading coefficients. A bigger bottom degree gives zero. A bigger top degree gives no horizontal asymptote at all, and the signs of the leading terms decide whether each end goes to plus or minus infinity.
The last annotation is the special case worth remembering: when the top degree is exactly one more than the bottom, long division produces a linear quotient, and that line is an oblique asymptote. If the difference is two or more, the quotient is a curve such as a parabola, and there is no oblique asymptote.
Matching
Match the pairs
Why: (a) is Checkpoint 4.24: equal degrees, leading coefficients 3 and 5. (b) is bottom-heavy. (c) has one degree more on top, and x² − 2x + 5 = (x + 2)(x − 4) + 13. In (d) the leading term of the denominator is −7x⁴, written last: always find the highest power, not the first term.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 366, 380 — Checkpoint 4.24 and Exercises 263, 265, 292
Make each match before checking, and for each one compare the degrees first. The first function is Checkpoint 4.24: equal degrees, leading coefficients 3 and 5, so the asymptote is y equals three fifths at both ends.
The second is bottom-heavy, degree 1 over degree 2, so it goes to zero. The third has one more degree on top, so it has an oblique asymptote, and the division gives x minus 4 with a remainder of 13.
The fourth is the one that catches people. Its denominator is written with the constant first, and the leading term, minus 7 x to the fourth, comes last. Equal degrees, leading coefficients 1 and minus 7, so the asymptote is minus one seventh. Always hunt for the highest power; never assume it is written first.
Fill the middle
\[ f(x) = \frac{3x^3 - 2x + 1}{2x^2 - 4} \]
Fill in the blanks
\frac\frac{3}{2}4 = ___\,x + \frac___\,x + 1}___
Why: Divide the leading terms: 3x³ over 2x² is (3/2)x. Multiply back: (3/2)x times (2x² − 4) = 3x³ − 6x. Subtract from the numerator: −2x + 1 − (−6x) = 4x + 1. The remainder has lower degree than the divisor, so the oblique asymptote is y = (3/2)x.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 377 — Checkpoint 4.29
Fill in both blanks before checking. The degree on top is 3 and the degree on the bottom is 2, one more, so an oblique asymptote is expected.
The first step of long division divides the leading terms: three x cubed over two x squared gives three halves x. Multiply that back through the divisor, subtract, and what is left is 4x plus 1, which has lower degree than the divisor, so the division stops there.
The remainder fraction, 4x plus 1 over 2x squared minus 4, is bottom-heavy and shrinks to zero, so the oblique asymptote is the line y equals three halves x. It passes through the origin because the quotient has no constant term.
Worked example
Find every asymptote of this function from the section's exercise set.
\[ f(x) = \frac{x^3}{4 - x^2} \]
Compare degrees
Why: Three on top, two below: one more on top, so expect an oblique asymptote.
\[ n = 3 = m + 1 \]
Divide
Why: The leading terms give x cubed over minus x squared, which is minus x.
\[ \frac{x^3}{4 - x^2} = -x + \frac{4x}{4 - x^2} \]
Read the oblique asymptote
Why: The remainder fraction is bottom-heavy, so it tends to zero.
\[ \frac{4x}{4 - x^2} \to 0 \;\Longrightarrow\; y = -x \]
Find the vertical asymptotes
Why: The denominator is zero at plus and minus 2, where the numerator is 8 and minus 8, not zero.
\[ 4 - x^2 = 0 \iff x = \pm 2 \]
Figure (svg): The graph of x cubed over 4 minus x squared for x from minus 8 to 8, with dashed vertical lines at x equals minus 2 and 2 and a dashed slanted line y equals minus x. The outer branches run along the slanted line.
Check the division by multiplying back
Why: Quotient times divisor plus remainder returns the numerator.
\[ (4 - x^2)(-x) + 4x = -4x + x^3 + 4x = x^3 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 380 — Exercise 273
This exercise puts the whole asymptote hunt into one function. Start with the degrees: 3 over 2, one more on top, so look for an oblique asymptote rather than a horizontal one.
Dividing x cubed by minus x squared plus 4 gives a quotient of minus x, and the remainder is 4x. The remainder fraction is bottom-heavy, so it vanishes at both ends and the oblique asymptote is y equals minus x, a line that falls from left to right.
Then the vertical asymptotes: the denominator is zero at 2 and minus 2, and the numerator is not zero there, so both are genuine vertical asymptotes. The figure shows the three branches this creates. The check multiplies the division back, which is the fastest way to be sure a long division is right.
Worked example
\[ f(x) = \frac{3x - 2}{\sqrt{4x^2 + 5}} \]
Find the effective power of the denominator
Why: For large x the 5 is negligible, and the root of x squared is the absolute value of x.
\[ \sqrt{4x^2 + 5} \approx \sqrt{4x^2} = 2|x| \]
Divide by the absolute value of x, for x positive
Why: There the absolute value is x, and x equals the root of x squared, which can move inside the root.
\[ \frac{(3x - 2)/x}{\sqrt{4x^2 + 5}/\sqrt{x^2}} = \frac{3 - 2/x}{\sqrt{4 + 5/x^2}} \to \frac{3}{\sqrt4} = \frac32 \]
For x negative, the absolute value is minus x
Why: Dividing the numerator by minus x flips its sign.
\[ \frac{(3x - 2)/(-x)}{\sqrt{4x^2 + 5}/\sqrt{x^2}} = \frac{-3 + 2/x}{\sqrt{4 + 5/x^2}} \to \frac{-3}{2} \]
Read the asymptotes
Why: A different line at each end.
\[ y = \tfrac32 \text{ as } x\to\infty, \qquad y = -\tfrac32 \text{ as } x \to -\infty \]
Figure (svg): The graph of 3x minus 2 over the square root of 4x squared plus 5 for x from minus 8 to 8, with dashed lines at three halves and minus three halves. The curve rises to three halves on the right; on the left it dips below minus three halves, crossing it once near x equals minus 0.6, and returns toward it from below.
Check with large inputs
Why: The value at 1000 is just under three halves, and at minus 1000 just below minus three halves.
\[ f(1000) \approx 1.4990, \qquad f(-1000) \approx -1.5010 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 367-368 — Example 4.26 and Figure 4.57
The same divide-by-the-highest-power idea works for roots, but you have to decide what the highest power is. Under the root, 4x squared dominates, and the square root of 4x squared is 2 times the absolute value of x. So the denominator behaves like a first power, and you divide by the absolute value of x.
Why the absolute value? Because you want to move the divisor inside the square root, and the only thing that equals the square root of x squared for every x is the absolute value of x. For positive x that is x, and the limit comes out as three halves.
For negative x the absolute value of x is minus x. Dividing the numerator by minus x flips its sign, and the limit becomes minus three halves. Different asymptotes at each end, as the figure shows, including one crossing of the lower asymptote near x equal to minus 0.6.
Trap
A tempting simplification at negative infinity:
\[ \frac{3x-2}{\sqrt{4x^2+5}} = \frac{3x - 2}{x\sqrt{4 + 5/x^2}} \]
\[ \to \frac32 \quad (x \to -\infty)? \]
Wrong sign.
Pulling x out of the root used the root of x squared equals x, which is false for negative x. The correct identity has an absolute value.
\[ \sqrt{x^2} = |x| = -x \quad (x < 0) \]
With minus x in its place the limit is minus three halves. The value at minus 1000, about minus 1.501, confirms it.
This slip gives the wrong sign at negative infinity, and it is very easy to miss because the algebra looks clean. Pulling an x out of a square root silently assumes x is positive.
The safe habit: whenever you take x out of a square root, write the absolute value of x, then replace it by x or by minus x depending on which end you are at. A quick numerical check with x equal to minus 1000 would have caught the error here, since the function's value is about minus 1.5, not plus 1.5.
A useful sanity check without any computation: the square root is never negative, so for large negative x the sign of the whole fraction is the sign of the numerator, 3x minus 2, which is negative. A positive limit is impossible.
Estimation
\[ \lim_{x\to\infty}\frac{\sqrt{3x^2 + 4}}{x + 6} \]
Predict first
Estimate the limit first, then decide exactly. Which is it?
Correct: √3 ≈ 1.732
Why: For large x the numerator is about √3 times x and the denominator about x, so the ratio tends to √3. Dividing top and bottom by x gives √(3 + 4/x²) over 1 + 6/x, which tends to √3/1. The approach is slow: at x = 1000 the value is only 1.7217, because the 6 in the denominator still matters.
\[ \frac{\sqrt{3 + 4/x^2}}{1 + 6/x} \to \sqrt3 \approx 1.7321 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 368 — Checkpoint 4.25
Make an estimate before choosing. For large x, the 4 under the root and the 6 in the denominator are negligible, so the fraction is roughly the square root of 3x squared over x, which is the square root of 3.
The exact method agrees: divide the top and the bottom by x, moving it inside the root on top as the square root of x squared. That is fine here because x is heading to positive infinity. What remains tends to the square root of 3 over 1.
The numbers show something subtle. At x equal to 1000 the value is 1.7217, still off in the second decimal place, because 6 compared with 1000 is not all that small. Limits at infinity can be approached slowly, so a single large input is evidence, not proof.
Section
Part 5
Concept
Figure (svg): Two panels. Left: y equals sine x from minus 10 to 10, a wave trapped between minus 1 and 1. Right: y equals tangent x from minus 5 to 5, with dashed vertical asymptotes at odd multiples of pi over 2.
The six trigonometric functions are periodic, so whatever they do on one period they do again forever. Sine and cosine keep oscillating between minus one and one; tangent has infinitely many vertical asymptotes.
\[ \lim_{x\to\pm\infty}\sin x, \quad \lim_{x\to\pm\infty}\tan x \quad\text{do not exist} \]
Neither has a finite limit, and tangent does not tend to infinity either, because it keeps coming back down to negative values.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 368-369 — Figures 4.58 and 4.59
Periodic functions repeat the same pattern forever, so they cannot settle at a limit unless they are constant. Sine keeps swinging between minus one and one no matter how far out you go; the left panel shows three full swings, and there are infinitely many more.
The tangent in the right panel is worse. It has a vertical asymptote at every odd multiple of pi over 2, so it keeps shooting up to infinity and coming back from minus infinity. It is unbounded, but its limit is not infinity, because it keeps returning to negative values.
These are both examples of the third kind of end behaviour: no finite limit and no infinite limit. When a trigonometric function appears inside something else, as in sine x over x, you usually need the squeeze theorem to say anything at all.
Concept
Figure (svg): On equal axes from minus 4 to 6: the curve y equals e to the x, flat near zero on the left and climbing steeply on the right; the curve y equals ln x, plunging near x equals 0 and climbing slowly on the right; and the dashed mirror line y equals x between them.
| x | −5 | −2 | 0 | 2 | 5 |
|---|---|---|---|---|---|
| eˣ | 0.00674 | 0.135 | 1 | 7.389 | 148.413 |
\[ \lim_{x\to-\infty} e^x = 0, \qquad \lim_{x\to\infty} e^x = \infty \]
\[ \lim_{x\to 0^+}\ln x = -\infty, \qquad \lim_{x\to\infty}\ln x = \infty \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 369-370 — Tables 4.5, 4.6 and Figures 4.60, 4.61
The exponential has two very different ends. On the left it flattens toward zero, as the table shows: e to the minus 5 is less than a hundredth. On the right it explodes, already about 148 at x equal to 5. So the exponential has the horizontal asymptote y equals zero on the left only.
The logarithm is the exponential's reflection in the line y equals x, the dashed line in the figure. Reflection turns the exponential's horizontal asymptote into a vertical one: the logarithm plunges to minus infinity as x approaches zero from the right. On the right it grows without bound, but very slowly: the logarithm of 100 is only about 4.6.
These four limits are worth knowing by heart. They are the raw material for the next example, where the exponential appears inside a fraction.
Worked example
\[ f(x) = \frac{2 + 3e^x}{7 - 5e^x} \]
At positive infinity, divide by e to the x
Why: The exponential is the fastest-growing piece, playing the role of the highest power.
\[ f(x) = \frac{2/e^x + 3}{7/e^x - 5} \]
Let x grow
Why: Constants over a growing exponential vanish.
\[ \lim_{x\to\infty} f(x) = \frac{0 + 3}{0 - 5} = -\frac35 \]
At negative infinity, do not divide
Why: Now the exponential itself tends to zero, so substitute its limit directly.
\[ \lim_{x\to-\infty} f(x) = \frac{2 + 3\cdot 0}{7 - 5\cdot 0} = \frac27 \]
Locate the vertical asymptote
Why: The denominator vanishes where e to the x is seven fifths.
\[ 7 - 5e^x = 0 \iff x = \ln\tfrac75 \approx 0.336 \]
Figure (svg): The graph of 2 plus 3e to the x over 7 minus 5e to the x for x from minus 6 to 4, with a dashed horizontal line at 2 sevenths on the left, another at minus 3 fifths on the right, and a dashed vertical line at x equals ln 1.4, about 0.34.
Check with inputs of 10 and minus 10
Why: Both values agree with the limits to four decimal places.
\[ f(10) \approx -0.60006, \qquad f(-10) \approx 0.28574 \approx \tfrac27 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 370 — Example 4.27
At positive infinity the exponential dominates everything else, so it plays the role of the highest power. Divide the top and bottom by e to the x, and the constants divided by a growing exponential vanish, leaving 3 over minus 5.
At negative infinity the situation is reversed. The exponential shrinks to zero there, so you do not divide by it at all. Just replace e to the x by its limit, zero, and the fraction becomes 2 over 7.
So one function has two different horizontal asymptotes, as the figure shows, plus a vertical asymptote where the denominator vanishes, at the natural log of seven fifths. The check with inputs of 10 and minus 10 matches both limits to four decimal places, which is typical of exponentials: they converge fast.
Error analysis
Annotate
On: \( \lim_{x\to-\infty}\frac{2 + 3e^x}{7 - 5e^x} = \lim_{x\to-\infty}\frac{2/e^x + 3}{7/e^x - 5} = -\frac35 \)
Read the argument and decide which line is wrong before revealing the annotations. The algebra itself is valid: dividing the numerator and the denominator by the same nonzero quantity never changes a fraction.
The error is in the limit. At negative infinity the exponential tends to zero, so two over e to the x and seven over e to the x do not vanish; they grow without bound. The rewritten fraction is infinity over infinity again, and reading off 3 over minus 5 is simply wrong.
The lesson generalises. Dividing by the dominant term only helps if you divide by what dominates at that particular end. At the right end that is the exponential; at the left end it is the constants, and the correct answer is two sevenths.
Prediction
\[ f(x) = \frac{3e^x - 4}{5e^x + 2} \]
Predict first
What are the limits as x → ∞ and as x → −∞?
Correct: 3/5 and −2
Why: At +∞ divide by eˣ: (3 − 4/eˣ)/(5 + 2/eˣ) → 3/5. At −∞, eˣ → 0, so substitute: (0 − 4)/(0 + 2) = −2. Numerically f(10) ≈ 0.59995 and f(−10) ≈ −1.99970. Two horizontal asymptotes, y = 3/5 on the right and y = −2 on the left.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 370 — Checkpoint 4.26
Choose an answer before revealing. This is Example 4.27 again with new numbers, so the method is the same: divide by e to the x at the right end, substitute zero for e to the x at the left end.
At the right end the fraction tends to 3 over 5. At the left end it becomes minus 4 over 2, which is minus 2. The numerical values at 10 and minus 10 confirm both.
If you picked three fifths at both ends, you treated the exponential like a polynomial with the same degree top and bottom. That reasoning only works where the exponential is growing, which is at one end only.
Concept
Many limits at infinity are races between two functions that both grow without bound. The winners are ranked: any power of x eventually beats any power of the logarithm, and any exponential with base above one eventually beats any power of x.
\[ \ln x \;\ll\; x^{p} \;\ll\; e^{x} \qquad (p > 0) \]
| x | ln x | √x | x³ | eˣ |
|---|---|---|---|---|
| 10 | 2.303 | 3.16 | 1,000 | 22,026 |
| 100 | 4.605 | 10 | 10⁶ | 2.7 × 10⁴³ |
| 1000 | 6.908 | 31.6 | 10⁹ | about 2 × 10⁴³⁴ |
So the limit of the slower over the faster is zero. Section 4.8 proves these rankings with L'Hôpital's rule; for now they are backed by the table and the next example.
Many limits at infinity are races: both the top and the bottom grow without bound, and you need to know which grows faster. The ladder on this slide ranks the standard families. Every positive power of x beats the logarithm, and every exponential with base bigger than one beats every power.
The table makes the gaps concrete. At x equal to 1000, the logarithm is about 7, the square root about 32, x cubed a billion, and e to the x a number with 435 digits. Once the faster function pulls ahead, the ratio of the slower to the faster collapses toward zero.
These rankings are proved properly in Section 4.8 with L'Hôpital's rule. For now, treat them as facts backed by evidence, and use them to settle races quickly. The next example shows why the evidence must come from far enough out.
Worked example
Decide the limit, and see why a calculator at moderate x points the wrong way.
\[ \lim_{x\to\infty}\frac{x^{100}}{e^x} \]
Compare logarithms instead
Why: The logarithm of the ratio is the difference of the logarithms, and it is much easier to handle.
\[ \ln\frac{x^{100}}{e^x} = 100\ln x - x \]
Evaluate at 100
Why: The power is ahead: the ratio is enormous.
\[ 100\ln 100 - 100 \approx 360.5 \;\Longrightarrow\; \frac{x^{100}}{e^x} \approx 10^{156.6} \]
Evaluate at 1000
Why: The exponential has taken over: the ratio is tiny.
\[ 100\ln 1000 - 1000 \approx -309.2 \;\Longrightarrow\; \frac{x^{100}}{e^x} \approx 10^{-134.3} \]
Figure (svg): The logarithms of x to the hundredth and of e to the x plotted for x from 1 to 900: the curve 100 ln x and the straight line y equals x. The curve is far above the line at first; the line overtakes it at about x equals 647.
Find the crossover
Why: Solving 100 ln x equal to x numerically.
\[ 100\ln x = x \iff x \approx 647.3 \]
Conclude by the ladder
Why: Past the crossover the exponential keeps winning.
\[ \lim_{x\to\infty}\frac{x^{100}}{e^x} = 0 \]
Check that it keeps shrinking
Why: Going from 1000 to 1001 multiplies the ratio by about 0.41, and that factor only falls as x grows.
\[ \frac{f(1001)}{f(1000)} = \left(1.001\right)^{100}e^{-1} \approx 0.41 \]
This is a race where a calculator at moderate x gives exactly the wrong impression. At x equal to 100, the power is so far ahead that the ratio is about ten to the 156. Anyone who stopped there would guess the limit is infinity.
Comparing logarithms makes the race visible. The logarithm of the power is 100 ln x, a slowly growing curve; the logarithm of the exponential is x, a straight line. The figure shows the curve well ahead at first, and the line overtaking it at about 647. From then on the exponential leads, and by 1000 the ratio is ten to the minus 134.
The final check explains why it can never recover. Stepping from 1000 to 1001 multiplies the power by about 1.1 but multiplies the exponential by e, about 2.7, so the ratio falls by more than half at each step, and that factor only gets smaller.
Tweak it
Parameter explorer
The curve is xᵖ e⁻ˣ, a power divided by the exponential. Slide p from 0.5 to 4. Where is the hump, and what happens after it, for every p?
\[ y = x^{{p}}e^{-x} \]
Start with p at 2 and slide it up slowly. The curve rises from zero, reaches a hump, and then falls back toward zero. Watch where the hump sits: it is at x equal to p. A bigger power pushes the hump further right and makes it taller.
Now push p to 4. The hump is higher, almost 5, and the fall comes later, but it still comes. No matter how large you make p, the exponential eventually overwhelms the power and the curve returns to zero.
That is the growth ladder in one picture. A power can delay the exponential's win, as x to the hundredth did until 647, but it cannot prevent it. Slide p down to a half as well: the curve is small throughout and decays almost immediately.
Ranking
Put in order
Order from slowest-growing to fastest-growing as x → ∞.
Why: Logarithms lose to every positive power (ln x over √x is 0.0138 at x = 10⁶). Powers are ordered by exponent. Every exponential beats every power, and between exponentials the larger base wins: 2ˣ/eˣ = (2/e)ˣ → 0 because 2/e < 1.
Drag the six functions into order, slowest first, before checking. Use the ladder: logarithms at the bottom, then powers by exponent, then exponentials by base.
The two places people hesitate are the ends. The logarithm is slower than the square root, even though the square root is itself slow; at a million, their ratio is about one seventieth. And between two exponentials the bigger base wins, because the ratio of 2 to the x over e to the x is two over e to the x, and two over e is less than one, so its powers shrink to zero.
Counterexample
Discussion prompt
Someone claims: if f(x) takes arbitrarily large values as x grows, then the limit of f at infinity is infinity. Give a counterexample and explain exactly which part of the definition fails.
Write your counterexample before revealing. The claim confuses two different ideas: taking arbitrarily large values somewhere, and staying large everywhere past some point.
The function x sin x is the standard counterexample. At the peaks of sine it equals x, so it reaches every height you like. But half a period later it equals minus x, and at every multiple of pi it is zero. It swings between larger and larger positive and negative values forever.
The formal definition shows exactly what fails. An infinite limit requires an N past which every single output is above M. Here, however far out you go, there are inputs where the function is zero, so no N works even for M equal to one. The tangent from earlier fails for the same reason.
Real world
Figure (svg): The curve C of t equals 12 times 1 minus e to the minus 0.3 t for t from 0 to 25 hours, rising steeply from 0 and flattening under a dashed line at 12 milligrams per litre; a dot marks 95 percent of the limit at about t equals 10.
\[ C(t) = 12\left(1 - e^{-0.3t}\right) \text{ mg/L}, \quad t \text{ in hours} \]
Discussion prompt
After a continuous infusion starts, the blood concentration of a drug follows the formula above. What is the long-run level, from which side is it approached, and roughly when is it within 5 percent of that level?
Answer all three parts before revealing. The first is a limit at infinity of the kind you did in Part 1: the exponential term tends to zero as time grows, so the concentration tends to 12 milligrams per litre.
The side of approach comes from the sign of what is subtracted. The exponential is always positive, so the concentration is always 12 times something less than one: below 12, approaching it from underneath, as the curve in the figure does.
The third part is what a clinician actually needs. Setting the concentration to 95 percent of its limit and solving gives about 10 hours. The horizontal asymptote is the steady state; the equation tells you how long it takes to get close to it. This is the most common reason horizontal asymptotes appear outside a mathematics class.
Section
Part 6
Concept
Steps 1 to 4 are this lesson; steps 5 and 6 are Section 4.5. Only after all six do you draw.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 370-371 — Problem-Solving Strategy: Drawing the Graph of a Function
This is the book's strategy, and the order is deliberate. The first four steps use only the function itself and give you the frame: where the graph can exist, where it meets the axes, what happens at the far ends, and where it breaks.
The last two steps use the derivatives from Section 4.5 and fill in the shape inside that frame: where the graph rises and falls, where it bends. Doing the frame first means the derivative information has somewhere to go, and it also gives you checks, since the shape has to join up with the asymptotes.
When you sketch, plot the intercepts, the extrema and the inflection points as dots, draw the asymptotes as dashed lines, and then connect the dots following the sign charts. The examples that follow do exactly this.
Worked example
\[ f(x) = (x - 1)^2(x + 2) \]
Domain and intercepts
Why: A polynomial is defined everywhere; set x to zero, then f to zero.
\[ f(0) = 2, \qquad f(x) = 0 \iff x = 1 \text{ or } x = -2 \]
Ends
Why: Multiply out to find the leading term.
\[ f(x) = x^3 - 3x + 2 \;\Longrightarrow\; f \to \pm\infty \text{ as } x \to \pm\infty \]
Vertical asymptotes
Why: A polynomial has none.
\[ \text{none} \]
First derivative
Why: Differentiate the expanded form and factor.
\[ f'(x) = 3x^2 - 3 = 3(x - 1)(x + 1) \]
Classify the critical points
Why: The sign goes plus, minus, plus across minus 1 and 1.
\[ f(-1) = 4 \text{ (local max)}, \qquad f(1) = 0 \text{ (local min)} \]
Second derivative
Why: Zero only at the origin, negative before it and positive after.
\[ f''(x) = 6x \;\Longrightarrow\; \text{inflection at } (0, 2) \]
Check the second-derivative test agrees
Why: Concave down at the maximum, concave up at the minimum, exactly as the first-derivative signs said.
\[ f''(-1) = -6 < 0, \qquad f''(1) = 6 > 0 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 371-372 — Example 4.28
A polynomial makes the first four steps quick: the domain is everything, there are no vertical asymptotes, and the ends come from the leading term, which you find by multiplying out. It is x cubed, so the graph falls on the left and rises on the right.
The derivative factors neatly as 3 times x minus 1 times x plus 1. Testing one point in each interval gives the signs plus, minus, plus, so there is a local maximum at minus 1, where the value is 4, and a local minimum at 1, where the value is 0. That minimum touches the x-axis, because 1 is a double root.
The second derivative is 6x, negative left of the origin and positive right of it, so the inflection point is at 0, 2. The check shows the two derivative tests agreeing at both critical points, which is a good sign that nothing was miscalculated.
Picture it
Figure (svg): Left: the graph of x minus 1 squared times x plus 2 for x from minus 3 to 2.5, with a local maximum at minus 1, 4, a local minimum at 1, 0, and an inflection point at 0, 2. Right: two sign charts. The first derivative is plus, minus, plus with cuts at minus 1 and 1. The second derivative is minus, plus with a cut at 0.
Each interval of each chart becomes a piece of the curve: plus on the first chart means rising, minus on the second means bending downward.
The charts on the right are the whole of steps 5 and 6 in a form you can read at a glance. Each number line is cut at the points where its derivative is zero, and the sign in each piece comes from one test value.
Now trace the graph from left to right with the charts beside you. First chart plus, second chart minus: rising and bending down, up to the maximum at minus 1, 4. First chart minus: falling, through the inflection point at 0, 2, where the second chart switches to plus. First chart plus again after 1: rising and bending up from the minimum at 1, 0.
The yellow inflection point sits exactly where the curve changes from a cap to a cup. Drawing the sign charts first and the curve second is the most reliable way to avoid a sketch that contradicts its own derivatives.
Step zero
\[ f(x) = (x - 1)^3(x + 2) \]
Discussion prompt
Before any derivative: what do the domain, intercepts and ends already tell you? Then find the extrema and inflection points.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 372 — Checkpoint 4.27
Write down what the first four steps give you before touching a derivative. The intercepts come straight from the factors and from f of zero, which is minus 2. The leading term is x to the fourth, so both ends go up, and a polynomial has no vertical asymptotes.
Then the derivative: the product rule gives three times x minus 1 squared times x plus 2, plus x minus 1 cubed, and factoring out x minus 1 squared leaves 4x plus 5. The critical points are minus five quarters and 1. Only the first is an extremum, a local minimum of about minus 8.54.
The point x equal to 1 is the interesting one. The derivative is zero there, but it is a square times something positive on both sides, so it does not change sign. The graph flattens momentarily and keeps rising, and because the second derivative changes sign there, it is an inflection point, not an extremum.
Worked example
\[ f(x) = \frac{x^2}{1 - x^2} \]
Domain and intercepts
Why: The denominator vanishes at plus and minus 1; the only zero of the numerator is 0.
\[ x \ne \pm 1, \qquad \text{intercept } (0, 0) \]
Ends: divide by x squared
Why: Equal degrees, leading coefficients 1 and minus 1.
\[ \lim_{x\to\pm\infty}\frac{1}{1/x^2 - 1} = -1 \;\Longrightarrow\; y = -1 \]
Vertical asymptotes from one-sided limits
Why: Near 1 the numerator is about 1, and the denominator is a small number whose sign depends on the side.
\[ \lim_{x\to1^+} f = -\infty, \quad \lim_{x\to1^-} f = \infty, \quad \lim_{x\to-1^+} f = \infty, \quad \lim_{x\to-1^-} f = -\infty \]
First derivative
Why: Quotient rule; the numerator simplifies to 2x.
\[ f'(x) = \frac{(1 - x^2)(2x) - x^2(-2x)}{(1 - x^2)^2} = \frac{2x}{(1 - x^2)^2} \]
Read its sign
Why: The denominator is a square, so the sign is the sign of x: decreasing left of 0, increasing right of it.
\[ \text{local min } f(0) = 0, \quad \text{no local max} \]
Second derivative
Why: Differentiate again and simplify.
\[ f''(x) = \frac{6x^2 + 2}{(1 - x^2)^3} \]
Read its sign
Why: The numerator is always positive, so the sign is the sign of 1 minus x squared. (The book prints the middle interval as minus 1 to minus 1; it is minus 1 to 1.)
\[ f'' < 0 \text{ on } (-\infty, -1), \quad f'' > 0 \text{ on } (-1, 1), \quad f'' < 0 \text{ on } (1, \infty) \]
Figure (svg): The graph of x squared over 1 minus x squared for x from minus 4 to 4: dashed vertical lines at x equals minus 1 and 1, a dashed horizontal line at y equals minus 1. The middle branch is a cup with its minimum at the origin, rising to infinity at both vertical asymptotes; the two outer branches come up from minus infinity and flatten toward y equals minus 1 from below.
Check with test values
Why: The values match the tables: concave up at 0, concave down at 2, and just past 1 the function is hugely negative.
\[ f''(0) = 2, \quad f''(2) \approx -0.963, \quad f(1.001) \approx -500.8 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 373-375 — Example 4.29
This is the full strategy on a rational function, and the first four steps now do real work. The domain excludes plus and minus 1, the only intercept is the origin, the degrees are equal so the horizontal asymptote is the ratio of leading coefficients, minus 1, and the one-sided limits at plus and minus 1 tell you which way the graph shoots off on each side of each vertical asymptote.
The first derivative simplifies to 2x over a square, so its sign is the sign of x. The graph falls, then rises, with a local minimum at the origin. The second derivative's numerator is always positive, so the concavity is decided by the sign of 1 minus x squared: concave down on the outside, concave up in the middle.
Note the book's misprint: it writes the middle interval as minus 1 to minus 1, which is an empty interval. It means minus 1 to 1. The checks confirm the signs numerically, including a value of about minus 500 just to the right of 1.
Trap
From the concavity chart of Example 4.29:
\[ f'' < 0 \text{ just left of } 1 \]
\[ f'' > 0 \text{ just right of } 1 \]
\[ \Longrightarrow \text{ inflection at } x = 1? \]
Wrong.
An inflection point is a point on the graph where the concavity changes. At x equal to 1 the function is not even defined, so there is no point there. The concavity changes across the gap, and that is all.
\[ 1 \notin \text{domain} \]
\[ \Longrightarrow \text{ no inflection point} \]
It is natural to scan a concavity chart for sign changes and call every one an inflection point. In Example 4.29 the chart changes sign at minus 1 and at 1, and neither is an inflection point.
An inflection point is a point on the graph, so the function has to be defined there, and the graph has to be continuous through it. At x equal to plus or minus 1 there is no point at all, only a gap with a vertical asymptote. The concavity changes across the gap, which is perfectly fine, but nothing is inflected.
So when you list inflection points, list only the sign changes that happen at points of the domain, and check each one by computing the function value there.
Worked example
\[ f(x) = \frac{x^2}{x - 1} \]
Domain and intercepts
Why: Only x equal to 1 is excluded; the only zero is at the origin.
\[ x \ne 1, \qquad \text{intercept } (0, 0) \]
Ends: divide
Why: One degree more on top, so divide to find the slanted line.
\[ \frac{x^2}{x - 1} = x + 1 + \frac{1}{x - 1} \;\Longrightarrow\; y = x + 1 \]
Vertical asymptote
Why: Near 1 the numerator is about 1 and the denominator changes sign.
\[ \lim_{x\to1^+} f = \infty, \qquad \lim_{x\to1^-} f = -\infty \]
First derivative
Why: Quotient rule, then factor.
\[ f'(x) = \frac{(x - 1)(2x) - x^2}{(x - 1)^2} = \frac{x(x - 2)}{(x - 1)^2} \]
Classify the critical points
Why: Plus, minus, minus, plus across 0, 1 and 2.
\[ f(0) = 0 \text{ (local max)}, \qquad f(2) = 4 \text{ (local min)} \]
Second derivative
Why: It never vanishes; its sign is the sign of x minus 1.
\[ f''(x) = \frac{2}{(x - 1)^3}: \quad \text{concave down on } (-\infty, 1), \text{ up on } (1, \infty) \]
Figure (svg): The graph of x squared over x minus 1 for x from minus 4 to 6, with a dashed vertical line at x equals 1 and a dashed slanted line y equals x plus 1. The left branch has a local maximum at the origin and falls away on both sides; the right branch has a local minimum at 2, 4. Both branches hug the slanted line far out.
Check the asymptote numerically
Why: At 100 the function and the line differ by exactly the remainder, one over 99.
\[ f(100) = \frac{10{,}000}{99} \approx 101.0101 = 101 + \frac{1}{99} \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 375-377 — Example 4.30
Here the degree on top is one more than on the bottom, so step 3 is a long division rather than a limit. The quotient is x plus 1 and the remainder fraction is one over x minus 1, so the oblique asymptote is y equals x plus 1.
The remainder fraction also tells you which side of the asymptote the graph is on: positive to the right of 1, so the right branch is above the line, and negative to the left, so the left branch is below it. The figure shows exactly that.
The derivative has sign pattern plus, minus, minus, plus across 0, 1 and 2, which gives a local maximum at the origin and a local minimum at 2, 4. It may look odd that the maximum value is smaller than the minimum value, but they are on different branches. The check at 100 lands exactly one over 99 above the asymptote, the size of the remainder.
Comparison
Comparison matrix
| function | domain | ends | vertical asymptote | extrema |
|---|---|---|---|---|
| x²/(1 − x²) | x ≠ ±1 | HA y = −1 | x = ±1 | min (0, 0) |
| x²/(x − 1) | x ≠ 1 | oblique y = x + 1 | x = 1 | max (0, 0), min (2, 4) |
| (3x + 5)/(8 + 4x) | x ≠ −2 | HA y = 3/4 | x = −2 | none: f′ = 4/(8 + 4x)² > 0 |
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 375 — Checkpoint 4.28 (third row)
Fill in the blanks before checking. The first two rows are Examples 4.29 and 4.30, so you can check them against the slides you just saw. Comparing them side by side shows how the degree rule changes the ends: equal degrees give a horizontal asymptote, one extra degree on top gives an oblique one.
The third row is Checkpoint 4.28. Equal degrees give the horizontal asymptote three quarters, and the denominator vanishes at minus 2 where the numerator is minus 1, so that is a vertical asymptote. The quotient rule gives a derivative of 4 over the square of 8 plus 4x, which is positive everywhere in the domain, so the function increases on each piece and has no extrema.
Its second derivative is minus 32 over the cube of 8 plus 4x, which makes it concave up to the left of minus 2 and concave down to the right. The intercepts are 0, five eighths and minus five thirds, 0.
Worked example
\[ f(x) = (x - 1)^{2/3} \]
Domain and intercepts
Why: A cube root is defined for every real number, and squaring it is too.
\[ f(0) = 1, \qquad f(x) = 0 \iff x = 1 \]
Ends
Why: A positive power of something growing without bound.
\[ \lim_{x\to\pm\infty}(x - 1)^{2/3} = \infty \]
First derivative
Why: Power rule; it is never zero and fails to exist at 1.
\[ f'(x) = \frac23(x - 1)^{-1/3} = \frac{2}{3(x - 1)^{1/3}} \]
Classify x = 1
Why: Negative to the left, positive to the right: a local minimum, with value zero.
\[ f(1) = 0 \text{ (local min)} \]
Look at the slope near 1
Why: The one-sided limits of the derivative are infinite with opposite signs.
\[ \lim_{x\to1^+} f'(x) = \infty, \qquad \lim_{x\to1^-} f'(x) = -\infty \;\Longrightarrow\; \text{cusp} \]
Second derivative
Why: Negative wherever it exists, because the power of 4/3 makes the denominator positive.
\[ f''(x) = -\frac{2}{9(x - 1)^{4/3}} < 0 \quad (x \ne 1) \]
Figure (svg): The graph of x minus 1 to the two thirds for x from minus 10 to 20: a curve that falls to a sharp point at 1, 0 and rises again, concave down on both sides, with the y-intercept at 0, 1.
Check with values
Why: Two inputs at the same distance from 1 give the same height, and 9 and 28 land on whole numbers.
\[ f(0) = f(2) = 1, \quad f(9) = 8^{2/3} = 4, \quad f(28) = 27^{2/3} = 9 \]
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, pp. 377-378 — Example 4.31
This function is defined everywhere, because a cube root, unlike a square root, accepts negative numbers. Its ends both go to infinity, since it is the square of a cube root and grows without bound in both directions.
The derivative is never zero, but it fails to exist at x equal to 1, which makes 1 a critical point. The sign changes from negative to positive, so there is a local minimum there, with value zero. What is new is the shape of that minimum: the derivative's one-sided limits are minus infinity and plus infinity, so the graph comes in vertically from both sides and forms a sharp point, a cusp.
The second derivative is negative on both sides, so the graph is concave down everywhere except at the cusp. The check uses the symmetry about x equal to 1 and two inputs where the cube roots are whole numbers.
Prediction
\[ f(x) = 5 - x^{2/3} \]
Predict first
Where is the cusp, and what is the end behaviour?
Correct: Cusp at (0, 5); f → −∞ at both ends
Why: f′(x) = −(2/3)x^(−1/3) is undefined at 0, with limit +∞ from the left and −∞ from the right, so the peak at (0, 5) is a sharp point. And x^(2/3) grows without bound at both ends (it is a square of a cube root, so never negative), so 5 − x^(2/3) → −∞ both ways: f(±1000) = 5 − 100 = −95.
OpenStax Calculus Volume 1, §4.6 Limits at Infinity and Asymptotes §4.6, p. 378 — Checkpoint 4.30
Choose an answer before revealing. This is Example 4.31 shifted and flipped: x to the two thirds has a cusp at the origin, and subtracting it from 5 turns the graph upside down and lifts it by 5.
So the cusp is at the origin's image, 0, 5, now pointing upward as a sharp peak. The derivative's one-sided limits there are plus infinity from the left and minus infinity from the right.
For the ends, x to the two thirds grows without bound at both ends and is never negative, so 5 minus it goes to minus infinity at both ends. At plus or minus 1000 the value is already minus 95.
Section
Part 7
Pattern
Figure (svg): A flow diagram. Start at the ends: take the limits as x goes to plus and minus infinity. A finite limit gives a horizontal asymptote. An infinite limit sends you to divide: if the quotient is linear, it is an oblique asymptote. Separately, zeros of the denominator that survive cancelling give vertical asymptotes.
Stewart, Calculus: Early Transcendentals 8e, §4.5 Summary of Curve Sketching §4.5, pp. 315-322 — Stewart's summary of curve sketching
This is the procedure for finding every asymptote of a rational, radical or exponential function. Start with the ends, because they are the new material, and treat each end separately.
If the limit at an end is finite, you have a horizontal asymptote there. If it is infinite and the function is rational, divide; a linear quotient is an oblique asymptote. For square roots, divide by the absolute value of x and watch the sign at negative infinity; for exponentials, divide by the exponential only at the end where it grows.
Then turn to the gaps in the domain. A zero of the denominator gives a vertical asymptote only if it survives cancelling; the one-sided limits there tell you how to draw it. With every asymptote on the page, the derivative steps from Section 4.5 fill in the rest.
Check
Check your understanding
Evaluate the limit as x → −∞ of √(9x² + 1)/(2x − 3).
Answer: B
Why: For x < 0, √(9x² + 1) = |x|√(9 + 1/x²) = −x√(9 + 1/x²). Dividing top and bottom by x gives −√(9 + 1/x²)/(2 − 3/x) → −3/2. Numerically, at x = −1000 the value is about −1.4978.
The question is designed around the most common error of the lesson. For large x the numerator behaves like 3 times the absolute value of x, and the denominator like 2x, so the answer depends on the sign of x.
At negative infinity the absolute value of x is minus x, and the ratio tends to minus three halves. If you chose plus three halves, you pulled x out of the square root as if it were positive; go back to the trap slide and redo the division with the absolute value.
Check
Check your understanding
Which of these functions has an oblique asymptote?
Answer: C
Why: The numerator has degree exactly one more than the denominator. Division gives (2x² + 3x)/(x − 1) = 2x + 5 + 5/(x − 1), so y = 2x + 5 is an oblique asymptote.
Compare degrees for each rational option. An oblique asymptote needs the top degree to be exactly one more than the bottom degree, and only one option qualifies.
The distractors each fail in an instructive way. Equal degrees give a horizontal asymptote. A difference of two degrees gives a parabola-shaped end, not a line. And e to the x over x is not rational at all; the exponential outgrows every line, so no straight line can follow it.
Check
Check your understanding
Which list gives every asymptote of f(x) = (5x² − 3)/(2x² + x)?
Answer: A
Why: Equal degrees give y = 5/2 (f(1000) ≈ 2.4987). The denominator x(2x + 1) is zero at 0 and −1/2, and the numerator is −3 and −1.75 there, not zero, so both are vertical asymptotes.
A complete answer has two parts, ends and gaps, and the options test whether you did both. The ends come from the degree rule: equal degrees, so the ratio of leading coefficients, five halves.
The gaps come from factoring the denominator as x times 2x plus 1, giving zeros at 0 and minus one half. The numerator is not zero at either, so both are vertical asymptotes. If the numerator had been zero at one of them, the factor might have cancelled and left a hole instead.
Explain it to yourself
Discussion prompt
In two or three sentences: why can a graph cross its horizontal asymptote, even infinitely often, but never cross a vertical asymptote?
Write your explanation before revealing. The heart of it is that the two kinds of asymptote are defined by two different kinds of limit.
A vertical asymptote means the outputs blow up as the input approaches a fixed number, so near that line the graph is racing off toward infinity and cannot cross from one side to the other along the way. A horizontal asymptote means only that the outputs approach a number as the input grows. That is a statement about the trend at the far end, and it says nothing about any individual output.
So a graph may cross its horizontal asymptote once, as in Example 4.26, or infinitely often, as with cos x over x plus 1, provided the distance to the line still shrinks to zero.
Exit ticket
\[ f(x) = \frac{x^2}{x^2 - 1} \]
Discussion prompt
Give the symmetry, every asymptote, and the local extrema of this function.
Give all three answers before revealing. Replacing x by minus x leaves the function unchanged, so it is even and its graph is symmetric about the vertical axis; that halves the work.
Equal degrees give the horizontal asymptote y equals 1. The denominator vanishes at plus and minus 1 while the numerator does not, so those are vertical asymptotes. The derivative simplifies to minus 2x over a square, positive for negative x and negative for positive x, so the origin is a local maximum.
Compare this with Example 4.29, which is almost the same function with the denominator's sign reversed. There the origin was a minimum and the asymptote was minus 1. One sign change flips the whole picture, which is why the frame steps are worth doing carefully.
Recap
| question | tool | answer looks like |
|---|---|---|
| limit at ±∞ | limit laws, squeeze, divide by the highest power | L (horizontal asymptote), ±∞, or no limit |
| rational p/q | degree rule | aₙ/bₘ if n = m; 0 if n < m; ±∞ if n > m |
| n = m + 1 | long division | oblique asymptote y = mx + b |
| square roots | divide by |x|, with |x| = −x for x < 0 | possibly different asymptotes at each end |
| races | ln x ≪ xᵖ ≪ eˣ | slower over faster tends to 0 |
| proof | ε–N (finite), M–N (infinite) | an N for every ε or M |
Then the six steps: domain, intercepts, ends, vertical asymptotes, first derivative, second derivative. Section 4.7 turns these shape tools toward optimization.
Stewart, Calculus: Early Transcendentals 8e, §2.6 Limits at Infinity; Horizontal Asymptotes §2.6, pp. 126-139 — the same material in Stewart
A limit at infinity asks what a function does as the input grows without bound. A finite answer gives a horizontal asymptote at that end, which the graph may cross. The formal definitions make that precise with a band of half-width epsilon, or a height M, and a point N beyond which the graph behaves.
For rational functions, divide by the highest power in the denominator, or use the degree rule directly; one extra degree on top means long division and an oblique asymptote. For square roots, divide by the absolute value of x and mind the sign at negative infinity. For races, use the ladder: logarithms lose to powers, and powers lose to exponentials.
Finally, the six steps join this lesson to the last one. The frame comes first, the derivatives fill it in, and the sketch comes last. Section 4.7 uses these same tools to find the best value of a quantity, which is optimization.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.