Designing a distillation column

McCabe-Thieleminimum refluxnumber of traysUnderwood and Gillilandcolumn design

Designing a distillation column: the material balance split, minimum reflux, the stages that follow, the operating lines and tray efficiency.

Column Material Balance (Distillate and Bottoms Split)

D = F\,\frac{z_F - x_B}{x_D - x_B}

How much of the feed leaves overhead and how much out the bottom, from nothing but the three compositions. Two balances — total flow and light key — solved together, and the answer is a lever rule.

Minimum Reflux Ratio (Underwood, Binary)

R_{min} = \frac{1}{\alpha - 1}\left(\frac{x_D}{z_F} - \frac{\alpha\left(1 - x_D\right)}{1 - z_F}\right)

The least reflux that could ever achieve a given split, at the price of an infinite number of stages. The lower limit of the design, opposite Fenske's upper one, and the number every real reflux ratio is quoted as a multiple of.

Reflux Ratio

R = \frac{L}{D}

The liquid sent back down the column divided by the product drawn off the top. One number that sets the slope of the rectifying line, the height of the tower, and most of the energy bill.

Gilliland Correlation (Actual Stages)

\frac{N - N_{min}}{N + 1} = 1 - \exp\!\left[\left(\frac{1 + 54.4X}{11 + 117.2X}\right)\!\left(\frac{X - 1}{\sqrt{X}}\right)\right],\quad X = \frac{R - R_{min}}{R + 1}

The empirical bridge across the middle of column design: given the two unbuildable limits — Fenske's fewest stages and Underwood's least reflux — it returns the stage count a column actually needs at the reflux you intend to run.

Rectifying Operating Line (McCabe–Thiele)

y = \frac{R}{R + 1}\,x + \frac{x_D}{R + 1}

The straight line above the feed that ties the vapour rising off a stage to the liquid falling onto it. Slope R/(R+1), and it always passes through the point (x_D, x_D) on the 45° line — which is how it gets drawn.

Feed Line (q-Line)

y = \frac{q}{q - 1}\,x - \frac{z_F}{q - 1}

The line the feed's thermal condition draws on a McCabe–Thiele diagram. Both operating lines must cross on it, and where they cross is the optimum feed stage — so q decides where the feed nozzle goes.

Boilup Ratio

V_B = \frac{V}{B}

The vapour the reboiler raises divided by the bottoms drawn off beneath it. The stripping section's answer to the reflux ratio, and the term the steam bill is really written in.

Stripping Operating Line (McCabe–Thiele)

y = \frac{V_B + 1}{V_B}\,x - \frac{x_B}{V_B}

The straight line below the feed, written in terms of the boilup ratio. Slope greater than 1, and it passes through (x_B, x_B) on the 45° line — the mirror of the rectifying line, anchored on the bottoms instead of the distillate.

Overall Column Efficiency

E_o = \frac{N_t}{N_a}

The bridge from the calculation to the hardware: theoretical stages divided by the actual trays needed to do their work. Real trays never reach equilibrium, so there are always more of them than the theory asked for.

How they fit together

Nine formulas, one column, and they resolve in a fixed order because each one closes a degree of freedom the next needs. The material balance split is first and it is not optional bookkeeping — feed rate and the three compositions fix the distillate and bottoms flows entirely, before any thermodynamics enters. Specify a recovery your balance does not permit and every later step will faithfully size a column for a separation that cannot exist.

Then the reflux decision, which is the whole economics of the column in one number. Underwood's minimum reflux is the floor: below Rmin no number of trays achieves the separation, because the operating line touches the equilibrium curve and a pinch forms. It is a limit and never an operating point — at Rmin you would need infinite stages. The opposite bound is total reflux, which needs the fewest stages and produces no product at all. Real columns run between them at an operating reflux ratio of roughly 1.2 to 1.5 times Rmin, and that range is not arbitrary: below it the column grows tall quickly, above it you are boiling money, since reflux is reboiler duty and condenser duty at the same time. Gilliland's correlation then converts the choice into an actual stage count from Rmin and the Fenske minimum stages. It is an empirical fit through scattered data, good to a stage or two, so treat it as a sizing estimate rather than a design — worth stating because it looks as exact as the equations around it.

The three line equations are what you would draw on a McCabe-Thiele diagram, and their job is to locate the feed. The rectifying line is fixed by R and xD, the stripping line by the boilup ratio and xB, and the q-line tells you where they meet. That q is the term to be careful with: it is the feed's thermal condition, 1 for a saturated liquid, 0 for a saturated vapour, above 1 for subcooled and negative for superheated. It rotates the feed line and moves the intersection point, which moves the feed tray, and a feed introduced on the wrong tray degrades a column that is otherwise correctly sized — this is one of the most common findings when a real column underperforms its design. Overall efficiency is the last line and the one that turns theory into steel: theoretical stages divided by actual trays, typically 0.6 to 0.8 for a hydrocarbon system and considerably worse for viscous or foaming ones. Twelve theoretical stages at 70% efficiency is a seventeen-tray column, and forgetting this step is how a column gets built five trays short.