Minimum Reflux Ratio (Underwood, Binary)

Distillation design is bracketed by two situations no one would ever build. Fenske's minimum stage count assumes total reflux — everything condensed goes back down, nothing is drawn off, and the column makes no product at all. Underwood's minimum reflux is the opposite corner: run at the lowest possible reflux and the column needs infinitely many stages. Every real column lives between those two, and the design problem is choosing where.

What happens at minimum reflux is worth picturing rather than just calculating. As the reflux ratio falls, the rectifying operating line rotates upward until it touches the equilibrium curve. At the point of contact the driving force for mass transfer goes to zero: the vapour leaving a stage is already in equilibrium with the liquid meeting it, so the stage does nothing. That is the pinch, and stepping off stages there produces infinitely many infinitesimal triangles that never get past the pinch point. For a well-behaved binary the pinch sits at the feed, which is the assumption this equation is built on.

Underwood published the general method in 1948, and it is a genuinely more elaborate thing than the expression on this page: two equations, one solved for a root $\theta$ lying between the volatilities of the light and heavy keys, the other summing over every component to give $R_{min}$. That machinery exists because real feeds have more than two components and because distributed non-key components move the pinch off the feed plate. The binary form here is what those equations collapse to when there are exactly two components, constant relative volatility, and a pinch at the feed.

Two habits keep this number useful. First, the $\alpha$ that goes in should be a geometric mean of the values at the top and bottom of the column, not a single measurement — volatility drifts with temperature and composition, and a top-of-column value will flatter the answer. Second, remember what the number is for: real columns are built at 1.1 to 1.5 times $R_{min}$, with the economic optimum usually near 1.2 to 1.3. Below 1.1 the tower height runs away; above 1.5 you are paying for reboiler duty, condenser duty and column diameter and getting almost no stages back for it.