AP Calculus AB and BC

Linearization: Tangent Line Estimates, Over or Under

The linearization of f at x = a is the tangent line L(x) = f(a) + f'(a)(x - a). For x near a, L(x) approximates f(x). If the graph of f is concave up between a and x, the tangent line lies below the curve and the estimate is an underestimate. If it is concave down, the estimate is an overestimate.

The tangent line as an approximation

Zoom in far enough on a differentiable function and its graph looks like a straight line. That line is the tangent line, and near the point of tangency its yy-values are close to the function's yy-values. Using the tangent line to estimate the function is called linearization or tangent line approximation. It is CED Topic 4.6, Approximating Values of a Function Using Local Linearity and Linearization.

The linearization of ff at x=ax = a is the tangent line written in point-slope form and solved for yy:

L(x)=f(a)+f′(a)(x−a)L(x) = f(a) + f'(a)(x - a)

It uses two numbers from the function, the height f(a)f(a) and the slope f′(a)f'(a). The estimate is f(x)≈L(x)f(x) \approx L(x) for xx close to aa. The idea is to trade a value that is hard to compute, such as 4.1\sqrt{4.1}, for arithmetic with a nearby value that is easy, such as 4\sqrt{4}.

How to build and use a linearization

  1. Identify the function ff and choose the center aa: a number close to the target xx where f(a)f(a) and f′(a)f'(a) are known or easy to compute.
  2. Compute the point f(a)f(a).
  3. Compute the slope f′(a)f'(a).
  4. Write L(x)=f(a)+f′(a)(x−a)L(x) = f(a) + f'(a)(x - a).
  5. Substitute the target xx into LL.

Keep the sign of (x−a)(x - a). When the target is to the left of the center, x−ax - a is negative, and a positive slope then pulls the estimate below f(a)f(a). Many problems give f(a)f(a) and f′(a)f'(a) directly, in a sentence or a table, with no formula for ff at all. The method is the same: those two numbers are all the linearization needs.

The approximation gets worse as xx moves away from aa, because the curve bends away from the line. How fast that happens depends on how sharply the graph curves, which is what the second derivative measures.

Overestimate or underestimate

Concavity tells you which side of the curve the tangent line is on, and that settles whether the estimate is too big or too small.

Between aa and xxTangent line positionL(x)L(x) compared with f(x)f(x)
f′′>0f'' > 0 (concave up)Below the graphUnderestimate: L(x)<f(x)L(x) < f(x)
f′′<0f'' < 0 (concave down)Above the graphOverestimate: L(x)>f(x)L(x) > f(x)

The result does not depend on whether xx is to the left or right of aa, and it does not depend on whether ff is increasing or decreasing. A concave up curve bends upward away from its tangent line in both directions, so the line is below it on both sides.

Check the whole interval

The concavity has to hold on the entire interval between aa and xx. In the justification, cite the sign of f′′f'' on that interval, since the sign at the single point aa says nothing about the rest of it.

Writing the justification

When a question asks whether the approximation is an overestimate or an underestimate, the reason is the concavity of ff, stated through the sign of f′′f'' on the interval between the center and the target.

  • Underestimate: the approximation is an underestimate because f′′(x)>0f''(x) > 0 on the interval, so the graph of ff is concave up and the tangent line lies below the graph.
  • Overestimate: the approximation is an overestimate because f′′(x)<0f''(x) < 0 on the interval, so the graph of ff is concave down and the tangent line lies above the graph.

If you are given a graph or table of f′f' instead of f′′f'', use the equivalent fact: ff is concave up where f′f' is increasing and concave down where f′f' is decreasing.

Where students lose points

  • Deciding over or under from whether ff is increasing or decreasing. Direction is irrelevant; only concavity matters.
  • Deciding from whether the target is left or right of the center. That is irrelevant too.
  • Evaluating the slope at the target instead of the center. The slope in L(x)L(x) is f′(a)f'(a).
  • Dropping the sign of x−ax - a when the target is to the left of aa.
  • Reversing the conclusion. Concave up curves sit above their tangent lines, so the tangent line value is too small.
  • Reporting the equation of the tangent line when the question asks for the approximate value, or the reverse.

Worked examples

Worked example

Estimating a square root

Use the linearization of f(x)=xf(x) = \sqrt{x} at a=4a = 4 to approximate 4.1\sqrt{4.1}. Is the approximation an overestimate or an underestimate?

  1. Point: f(4)=4=2f(4) = \sqrt{4} = 2.
  2. Slope: f′(x)=12xf'(x) = \frac{1}{2\sqrt{x}}, so f′(4)=12⋅2=14f'(4) = \frac{1}{2 \cdot 2} = \frac{1}{4}.
  3. Linearization: L(x)=2+14(x−4)L(x) = 2 + \frac{1}{4}(x - 4).
  4. Estimate: L(4.1)=2+14(0.1)=2.025L(4.1) = 2 + \frac{1}{4}(0.1) = 2.025.
  5. Concavity: f′′(x)=−14x3/2f''(x) = -\frac{1}{4x^{3/2}}, which is negative for all x>0x > 0. The graph is concave down on [4,4.1][4, 4.1], so the tangent line lies above it.

4.1≈2.025\sqrt{4.1} \approx 2.025. This is an overestimate because f′′(x)<0f''(x) < 0 on the interval. (The actual value is about 2.024852.02485.)

Worked example

Working from given values, with the target on the left

A twice-differentiable function ff has f(2)=5f(2) = 5, f′(2)=−3f'(2) = -3, and f′′(x)>0f''(x) > 0 for all xx. Approximate f(1.8)f(1.8) with the tangent line at x=2x = 2, and decide whether the approximation is an overestimate or an underestimate.

  1. Write the linearization at a=2a = 2: L(x)=5+(−3)(x−2)=5−3(x−2)L(x) = 5 + (-3)(x - 2) = 5 - 3(x - 2).
  2. Substitute the target. Here x−a=1.8−2=−0.2x - a = 1.8 - 2 = -0.2, so L(1.8)=5−3(−0.2)=5+0.6=5.6L(1.8) = 5 - 3(-0.2) = 5 + 0.6 = 5.6.
  3. Use concavity. f′′(x)>0f''(x) > 0 on [1.8,2][1.8, 2], so the graph of ff is concave up there and the tangent line lies below the graph.

f(1.8)≈5.6f(1.8) \approx 5.6. This is an underestimate because f′′>0f'' > 0 on the interval, so the graph is concave up and lies above its tangent line. The fact that ff is decreasing at x=2x = 2 plays no part in the decision.

Worked example

An exponential near zero

Find the linearization of f(x)=exf(x) = e^x at a=0a = 0 and use it to approximate e−0.2e^{-0.2}. Is the approximation too large or too small?

  1. Point: f(0)=e0=1f(0) = e^0 = 1.
  2. Slope: f′(x)=exf'(x) = e^x, so f′(0)=1f'(0) = 1.
  3. Linearization: L(x)=1+1(x−0)=1+xL(x) = 1 + 1(x - 0) = 1 + x.
  4. Estimate: L(−0.2)=1−0.2=0.8L(-0.2) = 1 - 0.2 = 0.8.
  5. Concavity: f′′(x)=ex>0f''(x) = e^x > 0 for every xx, so the graph is concave up and the tangent line lies below it.

e−0.2≈0.8e^{-0.2} \approx 0.8, an underestimate because f′′(x)>0f''(x) > 0. (The actual value is about 0.81870.8187.)

Frequently asked questions

Is linearization the same as the tangent line?

Yes. The linearization L(x)=f(a)+f′(a)(x−a)L(x) = f(a) + f'(a)(x - a) is the equation of the tangent line to ff at x=ax = a, solved for yy. Linearization, tangent line approximation, and local linear approximation all name the same thing.

How do I know if a tangent line approximation is an overestimate?

Check the concavity of ff between the center aa and the target xx. If f′′<0f'' < 0 there, the graph is concave down, the tangent line lies above the curve, and the approximation is an overestimate. If f′′>0f'' > 0, it is an underestimate.

What if the concavity changes between a and x?

Then concavity alone does not settle the question. The tangent line starts on one side of the curve and the curve may or may not cross it before reaching xx. You would need more information about ff, such as its actual value at xx.

Where this shows up outside calculus class