Half-life in practice, not just in textbooks
When I first had to actually compute a half-life from real data — not the clean textbook numbers — I learned quickly that the formula alone won't save you. You need to understand what you're measuring and how your equipment is behaving. The math is straightforward, but the messy part is getting reliable numbers out of the lab. Here's how I figure half life when I'm sitting at a bench with a Geiger counter or a spectrophotometer and a set of readings that don't quite line up.
The core equation you actually need
Radioactive decay follows first-order kinetics. That means the rate of decay is proportional to how much stuff is still there. The standard formula is: N(t) = N × e^(-t) Where N(t) is the quantity remaining at time t, N is the starting quantity, and is the decay constant. The half-life is simply the time it takes for N(t) to equal N/2. Solving for that gives you:
t/ = ln(2) / 0.693 / That's the whole thing. Everything else is just extracting from your data.
Get the Full Details
Getting from actual measurements
In practice you don't get to measure N perfectly. You get a sequence of counts or absorbance readings over time. The trick is linearizing the exponential. Take the natural log of your measurements and plot them against time. The slope of that line is -. So the procedure is: collect your data points, take ln of each one, do a linear regression of ln(N) versus t, and the half-life is minus 0.693 divided by the slope. If your slope is -0.023 per hour, your half-life is about 30.1 hours. I once spent two days chasing a half-life measurement for a tracer isotope and kept getting inconsistent results. The problem turned out to be that my detector had a dead time of about 150 microseconds, and at high count rates I was losing roughly 8 percent of my counts to pile-up. Once I applied the dead-time correction — True count = Observed count / (1 - Observed count × dead time) — my half-life stabilized within 2 percent of the accepted value. Without that correction, my calculated half-life was drifting by several hours depending on how concentrated my sample was. That's the kind of thing textbooks never mention.
A common mistake people make
The biggest error I see is using the wrong base for the logarithm. Some people switch to log when doing the linearization and then plug that slope directly into the half-life formula. That won't work. The derivation assumes natural logarithms. If you use log, you need to multiply your slope by ln(10) 2.303 first to convert it back. Otherwise your half-life will be off by a factor of about 2.3. Another mistake is treating the first data point as N. Your first reading is already some time into the decay. If you're working with a short-lived isotope and your first measurement comes 10 minutes after you start, N could be significantly higher than your first reading. It's better to treat N as a free parameter in your regression, or at least account for the elapsed time before your first measurement.
When the simple model breaks down
Half-life calculations assume a closed system with no external influences. That works fine for radioactive decay in isolation. But if you're dealing with a pharmacokinetic half-life — how long a drug stays in the body — things get messier. The body isn't a static container. Absorption, metabolism, and excretion all happen simultaneously, and the apparent half-life can change depending on dose, route of administration, and the patient's kidney function. I've seen people try to apply a single half-life value to a drug dosing schedule for a patient with renal impairment. The standard elimination half-life was 4 hours, but with reduced clearance the effective half-life stretched to 12 hours or more. That's not a calculation error — it's a different system entirely. The exponential decay model still applies, but the value you're using is wrong because the physiological parameters have changed. Similarly, if you're measuring a decay curve and there's background radiation or a contaminant contributing counts at a different rate, your curve won't be a clean exponential. You'll need to subtract the background first, or fit a two-component model. I once had a sample that looked like it had a half-life of about 6 hours, but after subtracting the ambient background (which I measured separately and found to be around 45 counts per minute), the corrected curve revealed a much shorter half-life of roughly 2.3 hours. The uncorrected analysis would have been off by a factor of nearly three.
Quick reference for common isotopes
Carbon-14: 5,730 years. Useful for dating organic material up to about 50,000 years. Iodine-131: 8.02 days. Used in thyroid therapy and diagnostics. Tritium (H-3): 12.32 years. Common tracer in biological studies.
Sodium-24: 15 hours. Used in industrial leak detection. Technetium-99m: 6.01 hours. The workhorse of nuclear medicine imaging.
What you need to actually do the calculation
You need at least 5 to 7 data points spread across a time range of about 2 to 3 half-lives. Fewer points and the regression gets noisy. More than that and you're usually just confirming what you already know. A spreadsheet or any basic data analysis tool will do — Excel, Google Sheets, even a hand calculator with a regression function. Plot your raw counts or concentrations, take the natural log, run the linear fit, and extract from the slope. If you're working with very short half-lives — seconds or less — you'll need fast electronics and a way to start the timer precisely at the moment of production. If they're very long — thousands of years — you can't wait for the decay to happen. Instead you measure the activity (counts per unit time) and calculate from the specific activity and the known atomic mass. The relationship is = Activity / (N × mass × atomic_fraction), where N is Avogadro's number. The method doesn't change. Only your tools and your patience do.
