At the centre of a liquefaction train is a single object where the gas finally becomes a liquid. It is the most expensive item on the site, it takes years to procure, and there are two ways to build it.
The coil-wound exchanger
Imagine many kilometres of small-diameter aluminium tubing wound in layers around a central mandrel, inside a tall pressure shell. Feed gas flows inside the tubes and condenses; refrigerant sprayed at the top flows down the outside of the bundle and boils, taking the heat away.
On a large train the finished object is several metres across, tens of metres tall and weighs hundreds of tonnes. It is fabricated in a factory and shipped whole, which makes transport a genuine project constraint — heavy-lift vessels, route surveys, and sometimes a jetty built to receive it before anything else on site.
Two things make it the default for large mixed-refrigerant trains.
It tolerates two-phase flow. Refrigerant boiling on the shell side is a mixture of liquid and vapour, distributed over a very large bundle. A wound geometry handles that gracefully, and maldistribution — some tubes seeing more flow than others — degrades performance rather than causing damage.
It tolerates thermal cycling. The winding accommodates expansion and contraction as the unit cools and warms, and it does so for decades of start-ups and shutdowns.
The plate-fin exchanger
The alternative packs plates and corrugated fins into a brazed aluminium block. Fluids run in alternating layers through passages a few millimetres across, and the surface area per unit volume is enormous.
They are compact, efficient, and far cheaper for a given duty than a wound bundle. Several can be arranged in a cold box to do collectively what one wound exchanger does alone.
They are also less forgiving. The narrow passages are sensitive to maldistribution and to anything solid arriving in the stream, and the brazed structure is less tolerant of rapid temperature change than a wound bundle — thermal shock is a real failure mode, so start-up and shutdown rates are controlled carefully.
Where each one lands
Large mixed-refrigerant trains — C3MR, AP-X, DMR — use coil-wound exchangers, because the duty is single, enormous and two-phase.
Cascade processes use plate-fin blocks, because the duty is split across three separate refrigerant loops and each individual step suits a compact exchanger.
Nitrogen expander cycles and small-scale plants use plate-fin, for the same reason plus space.
Floating plants go either way, and the decision turns on weight and motion as much as thermodynamics: a very tall wound exchanger on a hull that moves is a structural problem a compact block does not present.
Why the supply chain is thin here
This is the narrowest point in the whole liquefaction supply chain, and it is worth being precise about why.
The physical article is not conceptually complicated. What is hard is making one to tolerances that survive thirty years of thermal cycling at minus 160 degrees, with tens of thousands of joints that all have to hold, and being able to demonstrate that it will before anyone commits a multi-billion-dollar project to it.
That capability is accumulated over decades, in a small number of factories, against a customer base of a few dozen units a year worldwide. Nobody enters the business speculatively, because the qualification barrier is a track record and a track record cannot be bought.
The result is that the item with the longest lead time on a liquefaction project is frequently this one, and a wave of simultaneous final investment decisions competes for the same factory slots. A project can be permitted, financed and contracted and still wait.
Where it appears on this site
The supplier directory records the firms in this part of the chain, each with a public source, and it is deliberately short — that shortness is the finding, not a gap in the research. The blog piece on three narrow links treats the same concentration alongside the membrane licensors and the shipyards.
Where a specific firm is linked to a specific asset, the link comes from that firm’s own published source and is recorded in the curated data rather than inferred.