It corrodes every ferrous component in the system, and the pipe that let it in is the one thing unaffected.

In a sealed circuit the water is not replenished, so the small amount of oxygen present initially is consumed quickly and corrosion stops. That is why sealed systems work. But if oxygen diffuses continuously through the pipe wall it is constantly resupplied, and corrosion never stops.

The results are a boiler heat exchanger corroding from the inside, steel panel radiators perforating, pump impellers and bodies wearing, and magnetite sludge accumulating throughout - which blocks radiators, clogs heat exchangers, and destroys pumps abrasively.

The diagnosis is often confused because everything fails except the pipework. Black sludge in a system with plastic pipe and repeated component failures is the classic signature of non-barrier pipe used where barrier pipe was required.

EVOH is a copolymer that is almost impermeable to oxygen, applied as a thin layer either on the outside of the tube or sandwiched within the wall.

A very thin layer is sufficient, because its resistance to oxygen diffusion is orders of magnitude greater than polyethylene's. Barrier tubes are tested against a specified maximum oxygen permeation rate at a stated temperature.

EVOH's weakness is moisture - its barrier performance degrades if it absorbs water. So it is normally protected: either co-extruded as a middle layer with polyethylene inside and out, or applied outside with a protective overcoat. A tube with an unprotected external EVOH layer can lose performance if stored wet or buried in damp screed.

Handle and store barrier tube dry, and check the construction if the tube will be embedded in screed or laid in damp conditions.

Required wherever the water recirculates through ferrous components; not required where it does not.

Sealed heating circuits are the clear case - underfloor heating, radiator circuits, and anything connected to a boiler with a steel or cast iron heat exchanger. The water goes round and round past ferrous metal, so continuous oxygen ingress is continuously damaging.

It is not needed on potable hot and cold supply, because the water is drawn off and replaced constantly rather than recirculated, and any dissolved oxygen leaves with it.

It is also unnecessary where the entire circuit is genuinely non-ferrous - all-plastic and copper with a stainless or aluminium heat exchanger - though this is rarer than people assume once pump bodies and valve internals are considered.

Where there is any doubt, specify barrier. The price difference is small and the failure it prevents is expensive and slow to diagnose.

The method used to crosslink the polymer, which affects the degree of crosslinking and some handling properties.

PEX-a is crosslinked with peroxides during extrusion, while the polymer is molten. It achieves the highest and most uniform degree of crosslinking, and it has the most useful practical property: thermal memory, so a kink can be repaired by heating rather than cutting out. It is also the most flexible.

PEX-b uses a silane process completed after extrusion, usually in a warm moist environment. It is widely used and generally the most economical.

PEX-c is crosslinked by electron beam irradiation after extrusion, a clean process with no chemical residues.

All three are suitable for heating when made to the relevant standard and carrying the required barrier. The differences matter most for fitting compatibility - particularly expansion-fit systems, which rely on PEX-a's memory - so check the fitting system against the tube type.

It cannot be fused, and it is joined with mechanical fittings - press, compression, push-fit or crimp.

Crosslinking permanently bonds the polymer chains into a network, which is what gives PEX its temperature and creep resistance. It also means the material is no longer thermoplastic: it cannot be melted and reformed, so butt fusion, socket fusion and electrofusion are all impossible. This is the practical difference from PE-RT, which achieves similar temperature performance without crosslinking and can therefore be fused.

In underfloor heating this rarely matters, because each circuit is laid as one continuous coil from the manifold and back with no joints in the floor at all - which is the correct approach regardless of material.

Where a joint in a screed is unavoidable, use a fitting system approved for burial, pressure test before covering, and record its position on the as-built drawing.