A thermosiphon moves fluid using nothing but the density difference heating creates. No pump, no motor, no seal, no power. Where the layout allows it, that is a remarkable amount of equipment removed from a process and a remarkable amount of reliability gained.
A thermosiphon works on one principle: heat a fluid and it becomes less dense, and less dense fluid rises. Arrange an exchanger so that the heated leg is vertical and connects back to a reservoir, and the density difference between the two-phase mixture leaving and the liquid entering establishes a continuous circulation. The circulation rate settles wherever the driving head balances the friction loss, which means the system self-regulates to a considerable degree.
Why this is attractive. A pump on a hot process duty is a maintenance item, a power consumer, a seal that can leak and a single point of failure. Removing it removes all of that. Thermosiphon reboilers are consequently very common on distillation columns, particularly where the bottoms product is hazardous, hot, or would be difficult to seal. Nothing rotates and nothing needs electricity, so the circulation continues as long as heat is being applied.
Vertical and horizontal arrangements. In a vertical thermosiphon the process fluid boils inside vertical tubes with the heating medium in the shell, and the two-phase mixture rises back to the column. This gives a compact plot footprint and short piping, and it is the more common arrangement. In a horizontal thermosiphon the process fluid boils on the shell side with the heating medium in the tubes. That suits fouling service, because the shell side of a horizontal bundle is easier to reach, and duties where the available static head is limited. For shell-side boiling, a divided flow J shell arranged with one inlet and two outlets is the conventional choice, because splitting the outgoing two-phase flow keeps pressure drop low, and pressure drop is the thing that limits circulation.
The relationship to the condenser. A thermosiphon reboiler and a column overhead condenser are coupled through the column, and thermosiphon reboilers are more sensitive to column pressure than pumped alternatives. Circulation depends on the balance between static driving head and friction, and boiling point depends on pressure. If the overhead condenser loses duty and column pressure rises, bottoms boiling temperature rises with it, the vapor fraction in the riser changes, and circulation rate shifts. Push far enough and circulation can become unstable or stall. This is a real operating failure mode and it originates at the condenser, not at the reboiler.
Thermosiphon principles in condensing service. The same density-driven circulation appears on the condensing side of processes. Gravity condensate return systems, closed-loop cooling circuits that circulate without a pump, and reflux arrangements where condensate returns to a vessel under its own head all rely on it. The design discipline is the same: account honestly for the available static head, keep friction losses low, and make sure the system still circulates at minimum load rather than only at design conditions.
Related: kettle style for the pool boiling alternative, and distillation condensers for the other end of the column.
| PRODUCT | STYLE | SHELL TYPE | CIRCULATION |
| Reboiler | Thermosiphon | E or J divided flow | Natural |
In a pumped circuit, if the pressure drop turns out higher than expected you can usually turn the pump up. In a thermosiphon there is nothing to turn up. The driving force is fixed by geometry and fluid density, so every foot of elevation and every fitting in the riser matters. This makes thermosiphon design unusual among heat exchanger problems: the hydraulics have to be settled alongside the plot plan and the piping layout, not afterward.
Get it right and the system is beautifully robust. Get it wrong and the symptoms are unpleasant — sluggish circulation, poor heat transfer, and in bad cases flow instability or complete stall. Retrofitting a fix is difficult, because the remedy is usually elevation or larger piping and neither is easy to add later.
A thermosiphon depends on generating vapor to create the density difference that drives flow. At reduced heat input less vapor forms, the driving head falls, circulation slows, and the exchanger moves toward the condition where it cannot sustain flow. Well-designed systems have adequate margin at minimum expected load; marginal ones work at design rate and misbehave whenever the plant turns down.
The instability worth knowing about is periodic. In some geometries the system can enter a cycle of vapor generation, flow surge, temporary loss of driving head and flow collapse, repeating at intervals. It shows up as oscillating column bottoms level or temperature and it is genuinely difficult to diagnose without knowing to look for it. Adequate static head, correctly sized risers, and avoiding excessive vaporization per pass are the standard defenses.
Vertical thermosiphons boil the process fluid inside the tubes with the heating medium on the shell side. They are compact, need little plot space and use short piping runs, which helps the hydraulics. The tube side is straightforward to clean mechanically. Because the boiling happens inside relatively narrow tubes, vaporization per pass has to be kept moderate to avoid dryout at the tube exit.
Horizontal thermosiphons boil on the shell side around a bundle carrying the heating medium. The shell side of a horizontal bundle is more accessible, so this suits fouling process fluids, and the arrangement works with less static head available. The trade is a larger plot footprint and longer connecting piping. A divided flow J shell with one inlet and two outlets is standard here, because halving the two-phase outlet path is one of the most effective ways to keep pressure drop within the available driving head.
The driving force is the weight difference between a column of liquid in the downcomer and a lighter two-phase column in the riser, and it is small, often only 1 to 3 psi. Put numbers on it: 10 ft of hydrocarbon liquid at 45 lb/ft3 gives about 3.1 psi of static head, and if the riser mixture averages 15 lb/ft3 the net driving force is roughly 2 psi.
Every elbow, nozzle and foot of pipe in the loop spends part of that budget, which is why circulation rates are usually set so that 10 to 30 percent of the liquid flashes per pass and no more. Circulation starts on its own once heat is applied and liquid level is established; there is nothing to prime and nothing to start.
Because the pump on a hot bottoms duty is the item most likely to shut the column down, and a thermosiphon has no equivalent failure. A mechanical seal running on 350 F hydrocarbon bottoms is a scheduled replacement item, a double seal with a flush plan adds a support system of its own, and a pump trip takes reboiler duty to zero in seconds with the column still loaded. On a hazardous service the seal is also the most probable leak point in the loop.
Against that, a thermosiphon costs elevation, usually a few extra feet of skirt height or a raised column base, and a riser sized for two-phase flow. Where that elevation can be bought once at construction, it is nearly always cheaper than owning a pump for twenty years.
The faults are visible on the control room trend before they are visible at the exchanger, usually as a bottoms temperature that drifts upward while the steam valve opens further. Sluggish circulation leaves heating surface dry at the top of a vertical unit, so duty falls even though the exchanger is clean and the steam supply is adequate.
Starting problems are the second group: a cold loop full of subcooled liquid may not establish flow until the inlet is warmed, and on some columns a startup bypass or a brief kick of steam to the riser is needed. The third is fouling inside vertical tubes at the dryout zone, where heavy ends bake onto the wall and the exchanger loses capacity over a run. Each of these is cheaper to design out than to fix, because the fixes are elevation and pipe diameter.
Yes, and it effects the reboiler through pressure faster than through anything else. A column running at 15 psia whose condenser loses duty to air blanketing might drift up to 17 psia; for a water-like bottoms that lifts boiling temperature roughly 6 F, which changes the vapor fraction in the riser and with it the driving head. On a marginal design that shift is enough to drop circulation below the stable range.
The practical defense is to keep the condenser vented properly and to check the thermosiphon at a pressure somewhat above the normal operating point, not only at design pressure. When reboiler instability appears in warm weather, warm cooling water at the condenser is usually the cause and the reboiler is only where the symptom shows.
Start from the elevation you have, then the fouling, then the size of the exchanger. A vertical unit needs the column liquid level roughly level with the top tubesheet, which on a large column means the exchanger hangs from the column skirt and the tube length is limited to what the skirt height allows, commonly 10 to 16 ft.
A horizontal unit sits at grade on its own saddles and can be much larger, so very high duties above a few tens of million BTU/hr often go horizontal simply because one vertical shell cannot hold the surface. Fouling bottoms that need shell-side access point horizontal as well. If the elevation is there, the duty is moderate and the fluid is clean, vertical is cheaper to pipe and easier to control.
Beyond the usual thermal data, the item we need and rarely receive is a sketch of elevations: the column liquid level, the exchanger inlet and outlet nozzles, and every fitting in the riser and downcomer with its size. Those dimensions fix the driving head, and the driving head fixes whether the design works at all, so a thermal datasheet alone cannot answer the question.
A bubble point curve for the bottoms matters too, because a wide-boiling mixture changes composition as it vaporizes and the riser density is not what a single-component calculation predicts. If the column already exists, the dimensions of the existing nozzles and skirt tell us what geometry we are allowed to use, which is often the real constraint.