• Practice
  • About 15 min
  • 5 practice questions

Approximations and errors with differentials

Use dy = f′(x)·Δx to approximate values such as √26 and (2.98)³, and to estimate absolute, relative and percentage errors in areas and volumes caused by small errors in measurement.

If something looks wrong, please tell us.

Before calculators, people estimated values such as √26 or (2.98)³ by hand, and engineers still ask how much a small error in measuring a length affects a computed area or volume. Both questions have the same answer: close to a point, a curve is almost the same as its tangent line. This lesson turns that idea into quick approximation and error formulas, a regular source of short exam questions.

The idea: follow the tangent

If x changes by a small amount Δx, the true change in y is Δy = f(x + Δx) − f(x). The tangent predicts the change dy = f′(x)·Δx, called the differential. For small Δx the two are nearly equal, so f(a + h) ≈ f(a) + h·f′(a). Choose a close to the given number, with f(a) and f′(a) easy to work out.

Δy ≈ dy = f′(x)·Δx, so f(a + h) ≈ f(a) + h·f′(a). Relative error = dy/y; percentage error = 100 × dy/y. If y = kxⁿ, the percentage error in y is about n times the percentage error in x.

Worked example 1: √26

Take f(x) = √x, a = 25 and h = 1. f(25) = 5 and f′(x) = 1/(2√x) = 1/10 at x = 25. So √26 ≈ 5 + 1 × 1/10 = 5.1. The true value is 5.0990…, so the estimate is right to two decimal places.

Worked example 2: (2.98)³

Take f(x) = x³, a = 3 and h = −0.02 (the minus sign matters: 2.98 is below 3). f(3) = 27 and f′(3) = 3 × 9 = 27. So (2.98)³ ≈ 27 + 27 × (−0.02) = 27 − 0.54 = 26.46. The true value is 26.4636 to four decimal places.

Worked example 3: ∛127

Take f(x) = x^(1/3), a = 125 and h = 2. f(125) = 5 and f′(x) = 1/(3x^(2/3)), which is 1/(3 × 25) = 1/75 at x = 125. So ∛127 ≈ 5 + 2/75 ≈ 5.0267, against a true value of 5.0265.

Errors in measurement

If a length x is measured with a small error Δx, the resulting error in y = f(x) is about dy = f′(x)·Δx. The relative error is dy/y and the percentage error is 100 × dy/y.

Worked example 4. The radius of a sphere is measured as 10 cm with a possible error of 0.02 cm. V = (4/3)πr³, so dV = 4πr²·dr = 4π × 100 × 0.02 = 8π cm³ (about 25 cm³). The relative error is dV/V = 3·dr/r = 3 × 0.002 = 0.006, that is 0.6%.

Speed method: percentage errors in powers

If y = kxⁿ, then dy/y = n·dx/x, so the percentage error in y is n times the percentage error in x, whatever the constant k. A 1% error in the side of a square gives about 2% error in its area; in the edge of a cube, about 3% in its volume; in the radius of a sphere, about 2% in its surface area and 3% in its volume. For y = √x the percentage error is halved.

Trap: angles must be in radians

Worked example 5. Estimate sin 31°. Take f(x) = sin x and a = 30°, but h must be in radians: 1° = π/180 ≈ 0.017453. So sin 31° ≈ sin 30° + (π/180)·cos 30° = 0.5 + 0.017453 × 0.866025 ≈ 0.51511 (true value 0.51504). Taking h = 1 would give 0.5 + 0.866 = 1.366, more than the largest value a sine can have.

  • Getting the sign of h wrong when the number is below a.
  • Using degrees instead of radians for h in trigonometric approximations.
  • Forgetting to multiply by n when converting a percentage error.
  • Giving dy (the change) when the question asks for the approximate value f(a) + dy.

Key takeaways: f(a + h) ≈ f(a) + h·f′(a) with a chosen to make f(a) easy. The error in y is about f′(x) times the error in x, and for powers the percentage error multiplies by the exponent. Keep h signed and in radians.

Check your understanding

  1. 1. Using differentials with a = 36, what is the approximate value of √37, to three decimal places?

    1. 6.167
    2. 6.042
    3. 6.100
    4. 6.083
    Show answer

    Answer: D. f(x) = √x, f(36) = 6, f′(36) = 1/(2 × 6) = 1/12. So √37 ≈ 6 + 1/12 ≈ 6.083. (6.167 comes from forgetting the 2 in 1/(2√x).)

  2. 2. What is the approximate value of (1.99)⁵ using differentials?

    1. 32.8
    2. 31.92
    3. 30.4
    4. 31.2
    Show answer

    Answer: D. f(x) = x⁵, a = 2, h = −0.01. f(2) = 32 and f′(2) = 5 × 16 = 80. So (1.99)⁵ ≈ 32 + 80 × (−0.01) = 32 − 0.8 = 31.2 (true value 31.208).

  3. 3. The edge of a cube is measured with an error of 1%. What is the approximate percentage error in its volume?

    1. 3%
    2. 1%
    3. 6%
    4. 0.03%
    Show answer

    Answer: A. V = a³, so dV/V = 3·da/a. A 1% error in the edge gives about 3 × 1% = 3% error in the volume.

  4. 4. The radius of a circle is measured as 7 cm with a possible error of 0.05 cm. What is the approximate error in its area?

    1. 0.7π cm²
    2. 0.35π cm²
    3. 1.4π cm²
    4. 2.45π cm²
    Show answer

    Answer: A. dA = 2πr·dr = 2π × 7 × 0.05 = 0.7π cm² (about 2.2 cm²).

  5. 5. Using differentials, what is the approximate value of tan 46°?

    1. 1 + π/180
    2. 3
    3. 1 + π/90
    4. 1.02
    Show answer

    Answer: C. f(x) = tan x, a = 45°, h = 1° = π/180 radians. f′(x) = sec²x = 2 at 45°. So tan 46° ≈ 1 + 2 × π/180 = 1 + π/90 ≈ 1.0349. (3 comes from using h = 1 in degrees.)

Privacy · Terms · Refunds · Delivery