Why Half-Life Is Statistical

Introduction

A half-life of ten minutes does not mean each nucleus lasts exactly ten minutes. Some decay soon, some much later, and no measurement of the simple kind used in this course can predict which particular nucleus goes next. Yet a large population falls in a regular pattern. The contrast is not a contradiction: random individual events can have stable statistical averages.

Core explanation

For one nuclide under fixed conditions, each surviving nucleus has a constant probability of decay per unit time. The probability that a nucleus remains undecayed through one half-life is one-half. For a large collection of independent nuclei, about half are therefore expected to remain after that interval. “About” matters. A real count need not be exactly one-half because each nucleus's outcome is uncertain.

Imagine only four radioactive nuclei. After one half-life, the expected remaining number is two, but observing one, three or even zero is possible. A calculation that returns exactly two describes the average over many repetitions of this four-nucleus experiment, not a guarantee for one trial. With a million nuclei, relative fluctuations are much smaller, so the measured fraction is ordinarily close to one-half. That is why half-life tables work so well for macroscopic samples.

The average number of survivors after time t is N₀e^(−λt), where λ is the decay constant. The probability that a particular nucleus survives to time t is e^(−λt). This survival probability does not “remember” how long it has already survived. If a nucleus is still present now, its chance of surviving the next half-life is still one-half under the simple constant-rate model. Surviving past one half-life does not make it overdue.

Activity is likewise statistical. A 100-Bq source averages 100 decays per second over appropriate conditions, but one one-second measurement may contain more or fewer than 100 events. A detector adds further counting randomness and efficiency effects. Longer counting periods often improve the relative precision of an average rate, although a rapidly changing source must be analysed over intervals short enough to capture its decline.

For independent random counts with a roughly constant mean, a Poisson model often gives a standard deviation near the square root of the expected count. If about 10,000 events are counted, square root 10,000 is 100, or about 1% relative fluctuation. If only 100 events are counted, square root 100 is 10, or about 10%. This approximation helps explain why a smooth decay curve emerges from many events while a sparse count plot looks noisy. It is not a promise that every observation lies within one standard deviation.

Statistical unpredictability is different from lack of physical rules. The decay products, conserved quantities and probability distribution obey physical laws. The uncertainty is about which individual nucleus will decay at a given instant. The phrase “random decay” should therefore not be mistaken for “anything can happen”: an alpha-emitting isotope does not arbitrarily change its A and Z by unrelated amounts.

The model's independence assumptions matter. A detector may miss events, electronics can have dead time and a sample may contain multiple isotopes or daughter activities. Those effects can distort observed counts away from a simple one-nuclide exponential. Proper interpretation separates counting variation from changes in equipment, geometry or sample composition.

Step-by-step reasoning

1. Identify whether a question asks about one nucleus or a large population. 2. For one nucleus, state a survival probability, not an exact scheduled decay time. 3. For many nuclei, use the half-life fraction to find the expected remaining count. 4. Recognise that measured counts fluctuate around an expectation. 5. Check whether detector limits, background or multiple nuclides could affect an observed curve.

Visual explanation

Draw two panels. In the first, show four dots at the start and several possible one-half-life outcomes with zero to four dots remaining. In the second, show thousands of dots represented by a smooth shaded bar whose expected height halves. Beside the panels, plot a jagged small-sample count trace and a smoother large-sample average, both following an overall falling trend.

Real-world analogy

Repeated fair coin tosses cannot tell which individual toss will be heads, but many tosses give a fraction near one-half. Half-life has a similar individual-versus-population distinction. The analogy is limited because decay is a continuous-time process and different nuclides have different probabilities rather than one universal coin.

Real-world example

When monitoring a weak radioactive source, a counter might record 18 events during one interval and 25 during the next even if the average rate has barely changed. A scientist compares multiple intervals, background and uncertainty before declaring a real rise or fall. A short-term count fluctuation is not evidence that the isotope's half-life changed.

Why?

Why can a smooth curve describe random decays? The uncertainty of each event is averaged over a very large number of nuclei. Absolute fluctuations grow with count size, but their percentage relative to the total shrinks, making the overall population trend reliably close to its expected exponential curve.

Common misconception

“A nucleus that has survived several half-lives is more likely to decay soon because it is old.” Under the ordinary radioactive-decay model, its future probability depends on the nuclide, not its age. It is not overdue; the remaining population continues to halve over each later half-life.

Worked example

A sample contains 800 identical unstable nuclei and has a 2-hour half-life. After 2 hours, the expected number of undecayed parents is 800/2 = 400. After 4 hours, it is 200. These are expected values, not exact guaranteed counts in one real sample. For a single chosen nucleus, the chance that it survives 4 hours is (1/2)² = 1/4. If that nucleus has in fact survived 4 hours, its chance of surviving the next 2 hours is still 1/2, not 1/4, because the decay probability is memoryless.

Quick check

1. Can a sample of four nuclei have three survivors after one half-life? Answer: Yes. Two survivors is the expected number, but random outcomes can differ in a small sample.

Exam focus

Use “expected” or “on average” when giving half-life counts. Explain that individual decay times cannot be predicted, while large populations show a reliable fractional decline. Avoid claiming that an old surviving nucleus is overdue.

Advanced insight

For independent identical nuclei, the number surviving at a chosen time follows a binomial distribution with survival probability p = e^(−λt). At one half-life p = 1/2; for N₀ initial nuclei, the expected survivors are N₀/2 and the standard deviation is √(N₀/4). The relative spread falls as the initial sample grows, quantitatively explaining the smooth macroscopic law.

Summary

Half-life is a statistical population measure. Individual nuclear decay times are unpredictable, but a large collection has a reliable expected exponential decline. Small samples and short detector intervals fluctuate around that trend, and surviving nuclei are not “overdue” for decay.

Practice questions

1. A single nucleus has survived two half-lives. What is its probability of surviving one more half-life under the simple model? Answer: One-half; its future survival probability is unchanged by its past survival. 2. Why is a calculation of 50 survivors from an initial 100 not a guarantee? Answer: It is an expected count after one half-life; individual nuclei decay randomly, so a real small sample may differ. 3. Which generally has smaller relative counting fluctuation: 100 counted events or 10,000 counted events? Answer: 10,000 events. A square-root estimate gives about 1% relative fluctuation versus about 10% for 100 events.