The membrane is the part of an LNG carrier that sounds impossible when described plainly. A metal skin around a millimetre thick holds a hundred thousand cubic metres of liquid at minus 162 degrees, and the steel a few hundred millimetres away is at whatever temperature the sea is.
Two things make that work: insulation to limit how much heat crosses the gap, and a way of dealing with the fact that anything cooled by 180 degrees tries to shrink.
The contraction problem
Steel contracts as it cools, by roughly twelve parts per million per degree. Cool a thirty-metre tank wall by 180 degrees and it wants to be about six centimetres shorter.
The hull does not cool. It stays near ambient, which is the entire point of the insulation. So the membrane must shrink while the structure it is attached to does not, hundreds of times over a ship’s life, without cracking or tearing away.
There are only two answers: use a material that barely contracts, or build in the slack.
NO96 takes the metallurgical route
Invar is an iron-nickel alloy, about 36 per cent nickel, with an anomalously low thermal expansion coefficient — roughly a tenth that of ordinary steel. The name is short for “invariable”, and it was discovered in the 1890s for precision instruments long before anyone needed it for gas carriers.
In the NO96 system the primary and secondary barriers are both Invar sheet around 0.7 millimetres thick, separated and supported by plywood boxes filled with insulating material. Because Invar hardly moves, the membrane can be a flat sheet welded into a continuous skin, with no allowance for expansion built into its shape.
The elegance costs something. Invar is expensive, it is produced for this purpose by very few mills, and welding it is exacting work: kilometres of seam per tank, all of which must be leak-tight and all of which is inspected. It is one of the reasons so few yards can build these ships.
Mark III takes the geometric route
The alternative is to accept that the metal will move and give it somewhere to go.
Mark III uses a corrugated stainless steel membrane, roughly 1.2 millimetres thick, pressed with corrugations running in two perpendicular directions. As the sheet cools it contracts, and the corrugations flex to absorb it, in the same way a bellows absorbs movement along its axis. Behind the membrane sit panels of reinforced polyurethane foam, and behind those a secondary barrier of composite material.
Because the geometry does the work, the metal can be an ordinary cryogenic stainless rather than a specialist alloy. The trade is that the corrugations occupy volume and complicate the surface, and the foam panels are their own manufacturing discipline.
Neither system is simply better. They coexist because they fail and cost differently, and owners have preferences that outlast individual projects.
What the insulation is actually doing
Insulation thickness sets the boil-off rate, and boil-off rate is a headline commercial specification because the charterer pays for the cargo that evaporates.
Thicker insulation means less heat in, less cargo lost, and less cargo carried — because the insulation occupies space inside a hull of fixed size. Somewhere between roughly 270 and 400 millimetres of layered material is where the optimisation has landed for conventional ships, and successive generations of both systems have pushed the daily rate from around 0.15 per cent through 0.10 to below 0.09.
Below zero it will not go. Heat always enters, and as boil-off and reliquefaction explains, that is not a defect: the evaporation is the cooling.
Why there is always a second barrier
Cryogenic liquid reaching the hull would embrittle it, and a brittle hull is how ships break. So the containment is required to have a secondary barrier capable of holding a leak from the primary for a defined period — long enough to detect it and reach a port.
In NO96 the secondary barrier is a second Invar membrane, identical to the first. In Mark III it is a composite laminate bonded within the insulation. In both, the spaces between barriers are filled with inert gas and continuously monitored: a rise in hydrocarbon content in the interbarrier space is the alarm that something has failed.
Free-standing tanks handle this differently, and Moss and SPB designs need only partial secondary barriers because their failure modes are slower and more detectable.
Why the supply chain is thin here
Almost everything above is licensed. Both membrane systems come from a single French firm, the Invar comes from a very small number of producers, and the installation is a specialist trade requiring yard qualification.
That concentration is not an accident of ownership so much as a consequence of the product: decades of accumulated process knowledge, a customer base of a few dozen ships a year, and a failure mode nobody is willing to experiment with. The supplier directory records the firms involved, each with a public source, and it is deliberately short.