Tangent Line Approximation and Linearization
Use the tangent line to estimate values like √4.1 or sin(0.1): L(x) = f(a) + f'(a)(x - a). Worked examples, error checks and over/under estimates.
Tangent Line Approximation: Stop Guessing, Start Estimating
A common wrong belief is that a tangent line touches a curve at exactly one point and never crosses it. That is false. At an inflection point, the tangent line crosses the curve. The honest definition: a tangent line is the limit of secant lines as the second point approaches the first. Its slope equals the derivative at that point. The tangent line approximation uses that line to estimate function values near the point of tangency. It is an AP Calculus staple and a tool you will use repeatedly.
The Linearization Formula: L(x) = f(a) + f'(a)(x - a)
The linearization formula L(x) = f(a) + f'(a)(x - a) (OpenStax Calculus Vol. 1 section 4.2) is the tangent line written as a function. You compute it in three steps: find f(a), find f'(a), and plug into the formula. The result L(x) approximates f(x) for x near a. The smaller the distance |x - a|, the better the approximation.
Choosing a Good Point a: The Anchor Matters
Pick a where you know both f(a) and f'(a) exactly. For square roots, choose a perfect square. For exponentials, choose a value where the function is simple. The point a should be as close as possible to the input you want to approximate. If you need √4.1, pick a = 4 because √4 = 2 and f'(4) is easy. If you pick a = 9, the distance is 4.9 and the error will be large.
Example: Estimate √4.1
Let f(x) = √x. Choose a = 4. f(4) = 2. f'(x) = 1/(2√x), so f'(4) = 1/4. The linearization L(x) = 2 + (1/4)(x - 4). To approximate √4.1, set x = 4.1: L(4.1) = 2 + (1/4)(0.1) = 2 + 0.025 = 2.025. The exact value is √4.1 ≈ 2.02485.That is a good approximation.
Example: Estimate (1.02)^10 or e^0.1
For (1.02)^10, use f(x) = (1 + x)^10. Choose a = 0. f(0) = 1. f'(x) = 10(1 + x)^9, so f'(0) = 10. L(x) = 1 + 10(x - 0) = 1 + 10x. Approximate: x = 0.02 gives L(0.02) = 1 + 0.2 = 1.2. The exact value is (1.02)^10 ≈ 1.21899. Error is about 0.01899.
For e^0.1, use f(x) = e^x. Choose a = 0. f(0) = 1. f'(x) = e^x, so f'(0) = 1. L(x) = 1 + 1(x - 0) = 1 + x. Approximate: x = 0.1 gives L(0.1) = 1.1. The exact e^0.1 ≈ 1.10517. Error is about 0.00517.
Over or Under Estimate? Concavity Tells You
If f is concave up (f''(x) > 0) near a, the tangent line lies below the curve, so the linearization underestimates the true value. If f is concave down (f''(x) < 0), the line lies above the curve, giving an overestimate. For √x at x=4, f''(x) = -1/(4x^(3/2)), negative, so the line overestimates. Our approximation 2.025 was above the true 2.02485. For e^x, f''(x) = e^x, positive, so the line underestimates. The approximation 1.1 was below the true 1.10517.
How Big Is the Error? Compare With the Exact Value
The error formula from the Lagrange remainder is |error| ≤ (M/2)(x - a)^2, where M is the maximum of |f''(c)| on the interval between a and x (OpenStax Calculus Vol. 1 section 4.2). For √4.1, f''(x) = -1/(4x^(3/2)). On [4, 4.1], |f''(x)| is at most about 0.03125. The bound gives |error| ≤ (0.03125/2)(0.1)^2 = 0.00015625. The actual error was 0.00015, within the bound.
| Function | Point a | Approximation | Exact Value | Error | Concavity |
|---|---|---|---|---|---|
| √4.1 | 4 | 2.025 | 2.02485 | 0.00015 | Concave down (overestimate) |
| (1.02)^10 | 0 | 1.2 | 1.21899 | 0.01899 | Concave up (underestimate) |
| e^0.1 | 0 | 1.1 | 1.10517 | 0.00517 | Concave up (underestimate) |
Differentials: dy = f'(x) dx
The differential dy is defined as dy = f'(x) dx (OpenStax Calculus Vol. 1 section 4.2). It represents the change in the linear approximation, not the actual change Δy = f(x + dx) - f(x). For small dx, dy ≈ Δy. In the √4.1 example, with a=4 and dx=0.1, dy = (1/4)(0.1) = 0.025, matching the change in L(x). The actual change Δy = √4.1 - 2 ≈ 0.02485. The difference dy - Δy = 0.00015 is the same error we saw.
To use differentials for approximation: compute dy = f'(a) dx, then f(a + dx) ≈ f(a) + dy. This is exactly the linearization formula in different notation. The College Board AP Calculus AB/BC Course and Exam Description expects you to understand local linearity and use it to approximate function values.
When the Tangent Line Approximation Fails
The biggest failure case is picking a too far from the point you want. If you approximate √50 using a = 0, you get nonsense. Another failure: assuming the function is differentiable. At a corner, no unique tangent line exists; the calculator's nDeriv( function may still give a value (e.g., 0 for |x| at x=0), which is wrong. Always check differentiability first.
Common Questions
What is the difference between linearization and the tangent line equation?
They are the same line. Linearization L(x) = f(a) + f'(a)(x - a) emphasizes the approximation role. The tangent line equation is usually written in point-slope form y - f(a) = f'(a)(x - a). Both are the same mathematical object.
How do I know if my linearization is an overestimate or underestimate?
Check the second derivative at a. If f''(a) > 0, the function is concave up and the linearization underestimates. If f''(a) < 0, it overestimates. For the approximation √4.1, f''(4) = -1/32, negative, so the approximation was too high.
Can I use a tangent line to approximate a value far from a?
No. The error grows with distance squared. The Lagrange error bound shows |error| ≤ (M/2)(x - a)^2. If |x - a| is large, the error can become huge. Use a closer a or a different method.
What does it mean when the tangent line crosses the curve?
At an inflection point, the second derivative changes sign, so the curve goes from concave up to concave down (or vice versa) relative to the tangent line. The line crosses the curve at that point. This does not contradict the definition of a tangent line.
How does the College Board test linearization on the AP exam?
Free-response questions often ask you to use local linearity to approximate a function value. You must write L(x) = f(a) + f'(a)(x - a), show the computation, and sometimes state whether the approximation is an overestimate or underestimate using concavity.