Mean Residence Time from a Tracer

Space time $\tau = V/v_0$ is a nominal figure computed from a drawing. Mean residence time $\bar{t}$ is what the fluid actually did, measured by injecting a tracer and watching it come out. The two are equal in a vessel that behaves the way the drawing says it does, and comparing them is the single most useful diagnostic a reactor engineer has — it finds problems no amount of inspection will show, in a vessel that is running.

The method is a first moment. Inject a sharp pulse of tracer — a dye, a salt, a radioisotope — and record concentration at the outlet until it returns to baseline. The area under that curve is the zeroth moment $M_0$; the area under a plot of $t\,C$ against $t$ is the first moment $M_1$; and $\bar{t} = M_1/M_0$ is their ratio, which is simply the centre of gravity of the exit-age distribution. The tracer's concentration units cancel in the division, so only the time units have to be consistent — but they have to be consistent, and mixing minutes into one moment and seconds into the other is an easy way to be out by sixty.

Reading the comparison is where the value is. If $\bar{t}$ comes out clearly SHORTER than $\tau$, some of the vessel is not being used: dead zones behind baffles, stagnant corners, a settled sludge bank, or outright channelling from inlet to outlet. The active volume is roughly $V \times \bar{t}/\tau$, and the reaction is being asked to happen in that smaller vessel — which is why a chlorine contact tank or a digester can fail its design duty while every instrument on it reads correctly. If $\bar{t}$ comes out LONGER than $\tau$, suspect the measurement before the vessel: an error in the flow, or tracer adsorbing onto packing and leaching back out.

One practical trap dominates the technique. The first moment weights each measurement by $t$, so the long thin tail of the curve — the part that looks like noise and is tempting to stop recording — carries disproportionate weight in the answer. Truncating a tracer test early biases $\bar{t}$ low and can hide exactly the dead volume the test was run to find. Sample well past the visible end of the curve, and treat the baseline as data. The same dataset, taken further, also gives the variance, and from the variance an equivalent number of tanks in series — which turns a qualitative "this basin mixes badly" into a number a designer can put into a performance calculation.