Error Function in Measurements: A Simple Introduction

The error function, written erf(x), is defined as (2 / square root of pi) times the integral from 0 to x of e to the power (minus t squared) dt. It measures the accumulated area under the bell curve from the center out to x, and it is the key to turning measurement uncertainties into concrete probabilities.

It matters because real measurements always scatter around the true value, usually following the normal (Gaussian) distribution. The error function answers the practical question: what is the chance that my reading lands within a given distance of the true value?

Key Concepts

• Definition: erf(x) = (2 / square root of pi) times the integral from 0 to x of e to the power (minus t squared) dt. The prefactor makes erf(infinity) = 1.
• Basic values: erf(0) = 0, erf(infinity) = 1, and erf is an odd function, so erf(minus x) = minus erf(x).
• Complementary error function: erfc(x) = 1 minus erf(x), the tail area beyond x, useful for rare events.
• Cumulative normal distribution: the probability that a standard normal variable is below x equals (1/2)[1 + erf(x / square root of 2)].

Formulas

• erf(x) = (2 / square root of pi) times the integral from 0 to x of e to the power (minus t squared) dt.
• Normal CDF: Phi(x) = (1/2) [1 + erf(x / square root of 2)].
• Probability within plus or minus k sigma of the mean: P = erf(k / square root of 2).
• Useful values: erf(1) is about 0.8427, erfc(1) is about 0.1573, erf(2) is about 0.9953.

Worked Example

A digital voltmeter has a random error with standard deviation sigma = 0.05 volts. What is the probability that a single reading falls within 0.05 volts and within 0.10 volts of the true value?

1. Within plus or minus 0.05 volts is within plus or minus 1 sigma, since sigma itself is 0.05 volts.
2. The probability is erf(1 / square root of 2) = erf(0.7071), which is about 0.6827, or 68.3 percent.
3. Within plus or minus 0.10 volts is within plus or minus 2 sigma.
4. The probability is erf(2 / square root of 2) = erf(1.4142), which is about 0.9545, or 95.5 percent.
5. So roughly one reading in three misses the plus-or-minus 0.05 volt band, but only about one in twenty-two misses the plus-or-minus 0.10 volt band.

Common Misconceptions

• "The error function gives the error of a measurement." It does not give an error value at all; it gives the probability that the error falls inside a chosen band.
• "erf has a simple closed form." It does not. Like the integral of the Gaussian itself, erf cannot be written with elementary functions and must be evaluated numerically or from tables.
• "68 percent means within plus or minus one unit." The 68 percent rule is about one standard deviation, not one unit; the band width depends on the instrument's own sigma.

Real-World Applications

• Experimental physics: quoting results as value plus-or-minus sigma relies on erf to interpret what the band means.
• Quality control: factories use erf-based bands to decide whether a batch meets tolerance limits.

The error function is the bridge between the bell curve and everyday decisions about measurements. Any time you see a plus-or-minus uncertainty quoted, the error function is quietly defining what that band promises.