Calculus › Section D · Applications of derivatives › Topic D.5

Limits at infinity and horizontal asymptotes

Where does a process end up in the long run? Limits at infinity answer that question exactly, and the answers are the horizontal asymptotes you see on so many rational graphs.

The long-run question

So far limits have asked what happens as \(x\) closes in on a finite point. Now flip the vantage: what happens to \(f(x)\) as \(x\) itself grows beyond all bounds? A drug's blood concentration after days, the average cost of a product as production scales, a population approaching its environment's capacity: all are questions about \(x \to \infty\). This is the complement of B.5, where the output blew up near a point. Here the input blows up, and the interesting story is usually that the output settles down. The settling value, when there is one, is drawn on the graph as a horizontal asymptote.

Limits at infinity and horizontal asymptotes

We write \(\lim_{x \to \infty} f(x) = L\) when the outputs can be held as close as we like to \(L\) by taking \(x\) large enough, and similarly for \(x \to -\infty\). When \(L\) is a finite number, the line \(y = L\) is a horizontal asymptote: the graph gets as close as you like to it far out on the right or the left. A function has at most one horizontal asymptote in each direction, so at most two in all, and the two directions can differ; a card below shows a function that settles at \(\frac{3}{2}\) on the right and \(-\frac{3}{2}\) on the left. A graph may also cross its horizontal asymptote: \(\frac{x}{x^2 + 1}\) crosses \(y = 0\) at the origin. The line only describes the far ends.

The one fact everything stands on

Almost the entire technique reduces to this: for any \(r > 0\), \(\frac{1}{x^r} \to 0\) as \(x \to \infty\), because dividing \(1\) by an ever-larger number gives an ever-tinier one. So \(\frac{1}{x}\), \(\frac{1}{x^2}\), \(\frac{1}{\sqrt{x}}\) all vanish at infinity. As \(x \to -\infty\), \(\frac{1}{x^n} \to 0\) for every positive whole number \(n\) too. Any term \(x^k\) divided by a higher power \(x^n\) becomes \(\frac{1}{x^{n-k}}\), one of these vanishing pieces, which is the whole trick. Square roots need care on the negative side, and a card below shows the classic trap.

The divide-through technique

For any rational expression, a ratio of polynomials, the method is: divide every term, top and bottom, by the highest power of \(x\) that appears in the denominator. Every piece you create then looks like a constant or a constant over a power of \(x\), and the vanishing rule from the last card kills the latter one by one. The \(\frac{\infty}{\infty}\) appearance of the original was only cosmetic: the division exposes which parts genuinely matter in the long run. What survives the culling is a constant answer, or a leftover growing term that tells you the limit fails to exist through unbounded growth.

Worked example: matching degrees

Find \(\lim_{x \to \infty} \dfrac{5x^2 - 3x}{x^2 + 2}\). Both sides look like they race to infinity, so divide every term by \(x^2\), the denominator's highest power:

\frac{5x^2 - 3x}{x^2 + 2} = \frac{5 - \frac{3}{x}}{1 + \frac{2}{x^2}}

As \(x \to \infty\), both \(\frac{3}{x}\) and \(\frac{2}{x^2}\) vanish, leaving \(\frac{5}{1} = 5\). The limit is \(5\), and the graph levels at the horizontal asymptote \(y = 5\). A numerical echo: at \(x = 1000\) the ratio is \(\frac{5\,000\,000 - 3\,000}{1\,000\,002} \approx 4.997\), already settling.

The three-degree verdict

After dividing through, a rational function's fate falls into one of three cases, readable directly from the degrees of the top and bottom. Bottom heavier, the denominator's degree is higher: everything on top dies, the limit is \(0\). Degrees equal: the limit is the ratio of the leading coefficients, as the worked example just showed. Top heavier: a positive power of \(x\) survives on top, so the outputs grow without bound in size (towards \(\infty\) or \(-\infty\)), and there is no finite limit and no horizontal asymptote. Comparing degrees is the fastest first pass over any rational long-run question, and the division technique is the honest working behind it.

Worked example: the top wins

Examine \(\lim_{x \to \infty} \dfrac{3x^3 - x}{x^2 + 1}\). Divide by the denominator's top power, \(x^2\): \(\frac{3x - \frac{1}{x}}{1 + \frac{1}{x^2}}\). The denominator now drifts to \(1\), but the top still holds \(3x\), which grows without bound, and no finite limit exists. Notice the finer detail the algebra reveals: for large \(x\) the ratio runs along very close to the line \(y = 3x\), since \(\frac{3x^3 - x}{x^2 + 1} = 3x - \frac{4x}{x^2 + 1}\), and the correction term dies. The graph approaches a slanted line, which D.6 will use as an oblique asymptote in sketching.

The negative-side trap

Squares and square roots forget signs, and that matters on the left tail. The catchphrase: \(\sqrt{x^2} = |x|\), and for negative \(x\) the absolute value is \(-x\), not \(x\). Compute \(\lim_{x \to -\infty} \dfrac{3x - 2}{\sqrt{4x^2 + 1}}\). For \(x < 0\), \(|x| = -x\), so divide top and bottom by \(-x\). The top becomes \(\frac{3x - 2}{-x} = -3 + \frac{2}{x} \to -3\). The bottom becomes \(\frac{\sqrt{4x^2 + 1}}{|x|} = \sqrt{4 + \frac{1}{x^2}} \to 2\). The limit is \(-\frac{3}{2}\), while the same expression on the positive side approaches \(+\frac{3}{2}\). Two directions, two different horizontal asymptotes.

Long-run answers in the world

Average cost is the standard example. Suppose a workshop has fixed costs of 50,000 dollars, and each unit costs a further 40 dollars to make, whatever the volume. Making \(x\) units then costs \(C(x) = 50\,000 + 40x\) dollars, and the average cost per unit is \(\frac{C(x)}{x} = 40 + \frac{50\,000}{x}\). At 1000 units that is 90 dollars; at 100,000 units it is 40.50 dollars. As \(x \to \infty\) it approaches 40: the fixed cost is shared across ever more units, and \(y = 40\) is the horizontal asymptote. A limit at infinity is the mathematical form of the question "what does this eventually settle to?", and comparing degrees answers it in seconds.

Common mistakes to avoid

The recurring slips: answering \(\frac{\infty}{\infty}\) as \(1\) or calling it impossible, when the divide-through technique resolves it; using the ratio of leading coefficients when the degrees differ, when it applies only to equal degrees; dropping the sign on the negative side, computing \(\frac{3x}{\sqrt{4x^2+1}} \to \frac{3}{2}\) for \(x \to -\infty\) when the answer is \(-\frac{3}{2}\), because the root hands you \(|x|\); assuming a graph cannot cross its horizontal asymptote, when it can, since the line only describes the tails; and confusing the two infinite-limit stories, blow-up near a point from B.5 versus settling or growing as \(x\) runs away here.

Key takeaways

\(\lim_{x \to \infty} f(x) = L\) says the outputs settle to \(L\) as inputs grow, and the graph then hugs the horizontal asymptote \(y = L\). The engine is one fact, \(\frac{1}{x^r} \to 0\), deployed by dividing through by the denominator's highest power. Rational functions triage by degree: bottom heavier means \(0\), equal degrees means the ratio of leading coefficients, top heavier means unbounded growth and no horizontal asymptote. On the negative side, \(\sqrt{x^2}\) is \(|x| = -x\), so watch the signs. Left and right tails may settle at different heights. D.6 puts these long-run answers to work in curve sketching.

Flashcards

Flashcard

x → ∞

The long-run question

Where do the outputs head as x grows beyond all bounds?

Test your understanding

Question 1 of 5

What is the limit of (6x² + 5)/(2x² − x) as x approaches infinity?

Question 2 of 5

What is the limit of (7x + 1)/(2x² + 3) as x approaches infinity?

Question 3 of 5

What happens to (2x³ + x)/(3x² − 1) as x approaches infinity?

Question 4 of 5

What is the horizontal asymptote of f(x) = (4x² + x)/(2x² − 5)?

Question 5 of 5

What is the limit of (3x − 2)/√(4x² + 1) as x approaches negative infinity (x → −∞)?

Standards: 1

← Previous topicNext topic →