Differential Calculus
Limits at Infinity and Asymptotes
Limits at infinity describe the end behaviour of a graph: what does as runs off to the far right or the far left. Infinite limits describe the opposite phenomenon, blow-up at a finite point. Together they produce the horizontal, vertical and slant asymptotes that are the first things you draw when sketching a curve, and they are the language used later for improper integrals, series and growth comparisons. Almost every computation here reduces to two moves: spot the dominant term, then make the algebra legitimate by dividing by the right power of .
1. Limits at infinity: definition and basic facts
Definition. Let be defined on some interval . We write if for every there is a number such that . Similarly requires defined on some and .
Definition (horizontal asymptote). The line is a horizontal asymptote of if or . A function therefore has at most two horizontal asymptotes, one per direction. A graph is allowed to cross its horizontal asymptote, even infinitely often: crosses at for every nonzero integer .
Definition (infinite limit at infinity). means: for every there is with . This is not a limit that exists; it is a precise description of how it fails.
Core fact. For every real , . The same holds as whenever is defined for negative (safe for a positive integer).
The limit laws (sum, difference, product, quotient with nonzero denominator limit) hold verbatim for provided each individual limit exists as a finite number. They say nothing about the indeterminate forms , , , , which must be resolved algebraically first.
2. Infinite limits and vertical asymptotes
Definition. means: for every there is with . Replace the last inequality by for ; restrict to or for one-sided versions.
Definition (vertical asymptote). is a vertical asymptote of if at least one of , equals or .
Recipe for a rational function : factor and cancel all common factors first.
- If, after cancelling, and , then is a vertical asymptote. The form is , so the magnitude is infinite; get the sign from a sign table just to the left and just to the right of .
- If every copy of in the denominator cancels — that is, the numerator contains at least as many factors of as the denominator does — the point is a hole (removable discontinuity), not an asymptote, and the limit there is finite. Partial cancellation is not enough: in a factor of does cancel, yet is still a vertical asymptote.
- The form never tells you the answer by itself; it only tells you to simplify.
3. End behaviour of rational functions
Write with , . The dominant-term heuristic says for large ; the rigorous version is to divide numerator and denominator by , the highest power in the denominator, and then apply the limit laws. The quotient law actually finishes the job only when , so that the new numerator has a finite limit; when the new numerator is unbounded, and you argue instead that a quantity tending to divided by one tending to a nonzero finite limit tends to .
| Case | Asymptote | |
|---|---|---|
| horizontal | ||
| horizontal | ||
| or | slant from long division | |
| or | no line; a polynomial asymptote of degree |
Whether an infinite limit is or depends on the sign of and, as , on the parity of .
4. Square roots: , not
The single most common error in this topic. Always use so that for , pulling out of a root flips the sign: .
Worked Example 1. Find both horizontal asymptotes of .
As divide top and bottom by , writing downstairs (legal since ):
As we have , so and dividing by introduces a minus sign inside the root:
So there are two horizontal asymptotes: and .
5. Slant (oblique) asymptotes
Definition. The line is an asymptote of as if . Equivalently, compute and then ; the line exists exactly when both limits exist and are finite. The same definition applies as , possibly with a different line.
For a rational function with , polynomial long division gives with , so the remainder term tends to and is the slant asymptote. The sign of that remainder, , tells you whether the curve sits above the line (positive) or below it (negative).
Worked Example 2. Analyse .
Long division: , so Since as , the line is a slant asymptote in both directions; the curve lies above it for and below it for . The numerator at is , so is a vertical asymptote, with and . There is no horizontal asymptote (a function cannot have a horizontal and a slant asymptote in the same direction).
6. Growth-rate hierarchy
| Slowest to fastest | Representative | Fact (all limits as ) |
|---|---|---|
| logarithms | for all , all | |
| powers and roots | for all , | |
| exponentials | with | when |
| super-exponential | beats every |
Reading the table: in a sum, only the fastest-growing term matters; in a quotient, divide by the fastest-growing term present.
Squeeze Theorem (statement with hypotheses). Suppose for every in some interval (or, for a finite point , for every in some interval around ), and suppose with the same finite value . Then exists and equals . Two hypotheses do the work: the inequality need only hold eventually, but the two outer limits must agree — if they differ, the theorem says nothing. This is the standard tool for a bounded oscillating numerator, e.g. for forces as ; l'Hôpital is useless there because has no limit.
l'Hôpital's Rule (statement with hypotheses). Suppose and are differentiable on an open interval containing , except possibly at (for , on some interval or ), with there. Suppose further that either , or . If exists as a finite number or equals , then exists and equals it. If has no limit (it oscillates, say), the rule gives no information and you must argue another way.
Common mistakes
| Wrong | Right |
|---|---|
| , so as | equals when , so , a positive number, as |
| , so | is indeterminate; multiply by the conjugate to get |
| has vertical asymptotes at | Cancel first: for ; asymptote only at , and a hole at |
| The two one-sided limits are and , so the two-sided limit does not exist, even in the extended sense | |
| l'Hôpital on gives , which has no limit, so the original has none | The form is , but the rule's conclusion needs to exist; divide by and squeeze instead to get |
| l'Hôpital applied to as | Not an indeterminate form: just substitute, getting |
| The graph can never touch | Crossing a horizontal asymptote is allowed and common; the asymptote only controls the behaviour in the limit |
| A degree-3 over degree-1 quotient has a slant asymptote | : the asymptote is a parabola, not a line; there is no linear asymptote |
Key terms
- limit at infinity
- end behaviour
- horizontal asymptote
- infinite limit
- vertical asymptote
- one-sided limit
- dominant term
- indeterminate form
- conjugate multiplication
- absolute value identity for the square root
- slant (oblique) asymptote
- polynomial long division
- removable discontinuity (hole)
- growth-rate hierarchy
- squeeze theorem
- l'Hopital's Rule
Practice Problems
Evaluate and state the horizontal asymptote it produces.
Show hint
Do not compare terms by eye: divide the numerator and the denominator by the highest power of appearing in the denominator.
Show answer
The form is , which is indeterminate, so we must rewrite before taking limits.
The highest power in the denominator is . For we may divide top and bottom by without changing the value:
As we have , and . The denominator tends to , so the quotient law applies:
Check. Nothing in the computation used the sign of , so the same work gives as well. Numerically, , consistent with .
Answer: the limit is , and is a horizontal asymptote (in both directions).
Evaluate each limit and say what it means for the graph.
(a) (b)
Show hint
In both parts divide by , the highest power in the denominator, and then look at what the new numerator does.
Show answer
(a) Divide numerator and denominator by : As , , , , so the limit is .
This is the general rule "numerator degree less than denominator degree gives "; the graph has the horizontal asymptote . (Numerical check: at the value is .)
(b) Divide numerator and denominator by : As the denominator tends to while the numerator . A quantity tending to divided by one tending to tends to , so There is no horizontal asymptote in this direction. Since the degrees differ by exactly , there is a slant asymptote; long division gives because and . The remainder term tends to , so the slant asymptote is .
Answers: (a) , horizontal asymptote ; (b) , no horizontal asymptote but the slant asymptote .
For , find every vertical and horizontal asymptote, and compute the one-sided limits at the vertical asymptote.
Show hint
Check whether the numerator also vanishes where the denominator does; then decide the sign of the denominator on each side of that point.
Show answer
Vertical asymptote. The denominator vanishes only at . There the numerator is , so no cancellation is possible and is a genuine vertical asymptote (the form is ).
Signs. Near the numerator is close to .
- If then through positive values, so .
- If then through negative values, so .
So and ; the two-sided limit does not exist.
Horizontal asymptote. Divide by : So is a horizontal asymptote in both directions.
Independent check by division. , hence which shows at a glance that as , that blows up at , and that the sign of the blow-up is the sign of . Both methods agree.
Answer: vertical asymptote with from the right and from the left; horizontal asymptote .
Evaluate . Explain why l'Hôpital's Rule cannot be used to finish this problem.
Show hint
Divide by and deal with the leftover using a bound on .
Show answer
Step 1: divide by . For ,
Step 2: squeeze the sine term. For all we have , so for Both outer bounds tend to as , so by the Squeeze Theorem .
Step 3: combine. The denominator , so the quotient law gives
Why l'Hôpital fails here. The form is genuinely and both functions are differentiable, so the form hypotheses hold. But the rule also requires that exist (finite or ), and here which oscillates forever between and and has no limit. When that final hypothesis fails, l'Hôpital's Rule yields no conclusion at all — it does not say the original limit fails to exist. The squeeze argument shows the limit is .
Answer: the limit is (so is a horizontal asymptote), and l'Hôpital is inapplicable because does not exist.
Find all horizontal asymptotes of , being careful with the direction .
Show hint
Write and remember what equals when is negative.
Show answer
Domain. We need , i.e. or , and . Both tails are in the domain, so both limits make sense.
Factor out of the root. For every in the domain with ,
Case . Here , so dividing numerator and denominator by ,
Case . Here , so
Numerical check. At : , , and , giving , close to . At the value is .
Answer: two horizontal asymptotes, (as ) and (as ).
Evaluate
(a) (b)
Show hint
Each is an form; multiply and divide by the conjugate expression, then divide by remembering its sign.
Show answer
(a) The form is , which is indeterminate. Multiply and divide by the conjugate (positive for large ):
Now divide top and bottom by , using inside the root:
(b) Now (note for , so the expression is defined). Again the form is indeterminate, of type . Multiply and divide by , which is positive and nonzero here: Divide top and bottom by , and this time , so :
Numerical check of (b). At : and adding gives . ✔
Answers: (a) ; (b) . (Geometrically: has the asymptote on the right and on the left.)
Use the growth-rate hierarchy to evaluate
(a) (b)
Show hint
In (a) divide through by the fastest-growing function present. In (b) pull the dominant power out of each logarithm before dividing.
Show answer
(a) The fastest-growing term in sight is , so divide numerator and denominator by : By the hierarchy, for every , so and . The denominator tends to , so
(b) Extract the dominant power inside each logarithm and use (valid since both factors are positive for ):
Divide top and bottom by (nonzero for ), and note while , so that ratio tends to :
Independent check of (b) by l'Hôpital. The form is , both functions are differentiable for , and the denominator's derivative is nonzero there, so the rule may be tried: and since this limit exists, l'Hôpital's conclusion is valid and confirms the answer.
Answers: (a) ; (b) .
Let . Find all vertical and horizontal asymptotes, identify any removable discontinuity and the value that repairs it, and compute the one-sided limits at each vertical asymptote.
Show hint
Factor both quadratics completely before deciding which zeros of the denominator are asymptotes.
Show answer
Factor. The domain excludes and .
Cancel. For every (and ),
Removable discontinuity at . The factor cancelled completely, so is a hole, not an asymptote: Defining makes continuous there.
Vertical asymptote at . After cancellation the denominator still vanishes at while the numerator equals , so is a vertical asymptote.
- As : numerator , denominator , so .
- As : numerator , denominator , so .
Horizontal asymptote. Using either the original expression or the reduced one, divide by : (The degrees of the original numerator and denominator are equal, with leading coefficients and , giving the same value.) So is a horizontal asymptote in both directions.
Answer: hole at ; vertical asymptote with from the right and from the left; horizontal asymptote .
Find the slant asymptote and the vertical asymptote of , and determine on which side of the slant asymptote the curve lies as and as .
Show hint
The degree of the numerator exceeds that of the denominator by exactly one: do polynomial long division and look at the remainder term.
Show answer
Step 1: long division. Divide by :
Check: . ✔ Hence
Step 2: the slant asymptote. Since as , we have , which is exactly the definition of the line being an asymptote. So the slant asymptote is There is no horizontal asymptote, since as and as .
Step 3: the vertical asymptote. The denominator vanishes at ; the numerator there is , so is a vertical asymptote. From :
- : , so ;
- : , so .
Step 4: position relative to the line. The difference is , whose sign is the sign of . Therefore the curve lies above the line for (in particular as ) and below it for (in particular as ).
Answer: slant asymptote ; vertical asymptote with from the right and from the left; the curve is above the slant asymptote as and below it as .
Give a complete asymptote analysis of : all vertical asymptotes with the four one-sided limits, and the slant asymptote with the side the curve approaches from.
Show hint
Divide the cubic by the quadratic first; the quotient and the remainder answer almost every part of the question at once.
Show answer
Step 1: long division. Divide by :
So the quotient is with remainder : Check: . ✔
Step 2: vertical asymptotes. vanishes at . The numerator at is and at is , so no cancellation occurs and both and are vertical asymptotes.
Signs come from (the polynomial part stays bounded):
- : , , so and .
- : , , so and .
- : , , so and .
- : , , so and .
(Sanity check: for we have and for we have , matching the four signs above.)
Step 3: slant asymptote. Since as , , so is a slant asymptote in both directions. Because once , the difference is positive in both tails: the curve approaches from above as and also from above as .
Step 4: horizontal asymptotes. None: as and as (the degree of the numerator exceeds that of the denominator).
Answer: vertical asymptotes (with on the right, on the left) and (with on the right, on the left); slant asymptote , approached from above in both directions; no horizontal asymptote.
Let . State the domain, and find the oblique asymptotes as and as using and .
Show hint
Compute the slope first by pulling out of the root, and get the intercept with a conjugate multiplication.
Show answer
Domain. exactly when or . Both tails are included, so both asymptotes make sense.
Right-hand asymptote (, so ).
Slope: Intercept (an form, so use the conjugate):
So the right-hand asymptote is .
Left-hand asymptote (, so ).
Slope: Intercept:
(In the third equality we divided numerator and denominator by , which turns into .) So the left-hand asymptote is .
Numerical check. versus ; versus . Both differences are small and shrinking. ✔
Answer: domain or ; oblique asymptotes as and as (the curve lies just below each line). There are no vertical asymptotes: is continuous on its whole domain.
Find the value of the constant for which Explain why the naive answer "the limit is because both roots behave like " is wrong.
Show hint
Both terms grow like , so this is : rationalise with the conjugate and keep track of the linear term that survives.
Show answer
Step 1: rationalise. For large both radicands are positive. Multiply and divide by the conjugate :
Step 2: divide by . For we have , so each root divided by can be written with the pulled inside:
Step 3: take the limit. As , , and , so
Step 4: solve. . (With the radicand is positive for all large , so the expression is defined; the answer is consistent.)
Numerical check. With and : and , whose difference is . ✔
Why the naive answer fails. It is true that both roots are asymptotic to , i.e. each equals , but the constant-order parts need not cancel: while . Subtracting two quantities that both tend to is an indeterminate form precisely because the difference of the lower-order terms survives. Only an algebraic rearrangement (here, the conjugate) makes that surviving term visible.
Answer: .
