The wall at 96%
Distillation works for one reason: the vapour above a boiling mixture is richer in whatever boils easier. Catch it, condense it, boil it again, and it gets richer still. A column is that trick stacked a few dozen times, and the graph above is the whole of it. The black curve is what the vapour is; the dashed line is what the liquid was. The vertical distance between them is everything a stage is worth.
Ethanol and water run out of that distance. At 0.894 mole fraction, which is 95.6% by weight, the curve touches the diagonal: the vapour and the liquid are the same mix. A stage there changes nothing, and so does a hundred of them. That mixture boils at 78.15 °C, below pure ethanol at 78.30 °C, which is the part that surprises people: adding a little water makes ethanol boil sooner. It is an azeotrope, and it is a wall you cannot climb by adding height.
Out now
Try Reflux Student free on Aspen Plus
Reflux Student drives your own Aspen Plus V14 from plain English. You type what you want, it opens the case, makes the change in Aspen, runs it, reads the result back and tells you what it verified. Aspen does the math. You keep the judgment. Windows, your own licence, and free to start is a trial with a limit on it, not a free product.
Enter your email on the next page and the download link lands in your inbox, so you can open it on the Windows machine Aspen lives on. Windows will say it does not recognise the app the first time: click More info, then Run anyway. The certificate is new and Windows trusts it by reputation, which takes downloads to build. Nothing is wrong with the file.
Not on Aspen Plus? Tell me which simulator you use → and you hear the day your build exists.
Read the staircase, not the tower
Each step on the graph is one theoretical stage. Go up from the diagonal to the curve and you have the vapour that leaves that stage; go across to the diagonal and you have condensed it and handed it to the stage above. Step height is the enrichment you bought. Because the curve closes on the diagonal, every step is shorter than the one under it.
Here is what that costs, stepped off the curve above, starting from a 22.1 wt% feed and running at total reflux, which no real column can beat:
| Target x | wt% ethanol | Stages | α there |
|---|---|---|---|
| 0.500 | 71.9 | 2 | 1.888 |
| 0.700 | 85.6 | 3 | 1.300 |
| 0.800 | 91.1 | 6 | 1.128 |
| 0.850 | 93.5 | 10 | 1.058 |
| 0.880 | 94.9 | 17 | 1.019 |
| 0.891 | 95.4 | 27 | 1.004 |
| 0.894 | 95.6 | 45 | 1.000 |
6 stages get you to 91.1%. Another 21 get you to 95.4%. The last 0.13 of a percentage point costs 18 more, and then it stops. That is not a column that is badly designed. That is the shape of the curve, and no amount of steel changes the shape of the curve.
The number that decides whether a column is worth building
Relative volatility, α, is the ratio of how much the vapour favours ethanol to how much the liquid holds it. It is the vertical gap between those two lines expressed as one number, and it is the first thing to look at before anyone draws a tower:
| x | wt% | α | What that means |
|---|---|---|---|
| 0.05 | 11.9 | 8.931 | easy |
| 0.20 | 39.0 | 4.478 | easy |
| 0.40 | 63.0 | 2.382 | easy |
| 0.60 | 79.3 | 1.536 | easy |
| 0.75 | 88.5 | 1.205 | ordinary |
| 0.85 | 93.5 | 1.058 | expensive |
| 0.88 | 94.9 | 1.019 | do not build it |
| 0.89 | 95.4 | 1.006 | do not build it |
Below about α = 1.05 the stage count runs away and the reflux ratio runs away with it, so you are paying for reboiler duty rather than separation. Ethanol and water cross that line at x = 0.856, which is 93.8% wt, and reach α = 1 exactly at the azeotrope. Past it α drops below one: water becomes the lighter component, which is why the curve never comes back.
The three ways round it, and the one plants actually pick
Fuel ethanol has to be dry, because water in a gasoline blend causes phase separation. So the industry does not climb the wall. It goes round it, three ways, and only one of them is still distillation:
- 01An entrainer. Add a third component that forms a new, lower-boiling azeotrope with the water and carries it overhead. Benzene did this for decades and is now avoided on toxicity grounds; cyclohexane is the usual modern choice. It works, and it means running and recovering a solvent you did not previously have in the plant.
- 02A molecular sieve. Stop distilling. Send the vapour through a bed of 3A zeolite whose pores admit water and exclude ethanol, then swing the pressure to drive the water back off and regenerate the bed. It is adsorption, not distillation, which is exactly why the azeotrope does not apply to it. This is what a modern fuel-ethanol plant is built with.
- 03Extractive distillation. Add a heavy solvent such as ethylene glycol that holds the water down in the liquid, pull dry ethanol off the top, then recover the solvent in a second column. Still the preferred route at large scale in some plants, and it trades a solvent-recovery column for the sieve beds.
What is not on the list is a taller column, and neither is a pressure swing. Ethanol and water is the textbook case where pressure swing fails: the azeotrope moves so little with pressure that the two columns would be almost the same column.
What this looks like in a simulator
A simulator finds that crossing point in a second, and it finds it in the wrong place just as fast if the property method is wrong. Ethanol and water is strongly non-ideal, so an activity-coefficient model is not optional: NRTL or UNIQUAC with fitted binary parameters puts the azeotrope where the data says it is. An equation of state picked out of habit will hand you a curve that never touches the diagonal, and a column that separates past 95.6% on paper and not in steel.
The check is one line: run a flash, ask for the azeotrope, and confirm you get 0.894 mole fraction at 78.15 °C. If you do not, nothing downstream of it is worth reading.
Sources: vapour–liquid equilibrium data for ethanol/water at 101.3 kPa, Carey & Lewis (1932), reproduced in Perry’s Chemical Engineers’ Handbook 9e, sec. 13, and in the DECHEMA vapour–liquid equilibrium collection. Azeotrope composition and boiling point: Azeotrope. Why extra stages do nothing past it: LibreTexts, fractional distillation of non-ideal mixtures. Entrainer practice and the move off benzene: Fuel Processing Technology. Vapour-phase 3A zeolite pressure-swing adsorption: 3A zeolite PSA operating guide. Extractive distillation at scale, and why pressure swing is a poor fit here: Engineering Reports (2025), Processes 13(11), 2025.
Finding the wall is the easy part. Getting the model to agree with the plant is not.
The curve above is one binary at one pressure. A real column has a feed with a dozen components in it, a property method somebody picked three years ago, and a convergence failure at two in the morning. Reflux drives the simulator you already have from plain English, takes the setup grind, and reads back every change it makes before it reports it. That is the part I am shortcutting.
The trial is small and we are not precious about which software you run. Whatever you use, we’ll build for it.
Nathan Ruberto · Co‑Founder, CEO
Back to reflux.sh ->