Coolant and material compatibility.
Compatibility failures do not announce themselves. A seal that swells, a brass fitting facing a stainless body, a coolant topped up with water until the inhibitor is diluted away — each passes a commissioning check and shows up months later as a weep or a temperature trend. This brief covers what a 25% propylene glycol mix does to the wetted path, which material pairs are safe, and the six rules that keep a mixed-metal loop intact.
DL-007 · Published 2026-09-06 · 8 min read
A liquid cooling loop is a chemistry experiment that runs for twenty years. The coolant is not water — it is water, glycol, and an inhibitor package doing three separate jobs — and every material it touches, from the cold plate to the last O-ring in the manifold, is part of the reaction. Compatibility is therefore not a datasheet checkbox; it is a decision about which materials are allowed into the wetted path at all.
The stakes are asymmetric. A coupling that fails mechanically fails loudly and quickly. A coupling that fails chemically fails quietly: the leak rate is fine on day one, the joint looks clean at commissioning, and the first symptom is a dried residue trail under a manifold six or nine months into service.
What a 25% propylene glycol mix actually changes.
PG25 — 25% propylene glycol by volume, the balance water plus inhibitors — is the common indoor default. It is a different fluid from water in four ways that matter to hardware.
| Property at 25 vol% propylene glycol | Value | Why a hardware buyer cares |
|---|---|---|
| Freeze protection | roughly −10 °C | enough for indoor plant and unheated aisles, not for outdoor or uninsulated rooftop runs |
| Specific gravity | ≈ 1.02 at 20 °C | adds about 2% to every pressure drop you calculated for water |
| Heat capacity | ≈ 3.9 kJ/kg·K against 4.18 for water | about 7% more flow for the same heat load, or a slightly larger ΔT at the same flow |
| Thermal conductivity | about 15% lower than water | cold plate performance at the metal surface is slightly worse at the same flow |
| Viscosity | roughly twice the 20 °C value at 0 °C | cold-start pressure drop is higher than the warm number the loop was balanced at |
Values are typical for a 25 vol% propylene glycol mix with an inhibitor package; exact figures follow the coolant maker's data sheet for the product you buy.
The planning basis on the rack side does not change: 1.2 L/min per kW remains a sound first pass, and an 80 kW rack still moves on the order of 96 L/min. What changes is the confidence interval around it. The mix costs you a few percent of heat capacity and a few percent of pumping, and the inhibitor package is doing chemistry the whole time — buffering pH, scavenging dissolved oxygen and maintaining a protective film on the anodic metals in the loop. That film is the reason a mixed-metal loop survives at all, and it is consumable.
What the OCP standard already settles.
The specification does not leave compatibility open. It names the coolant families the wetted materials are qualified against, and it references the safety clause for coolant-contact materials.
- Qualified coolant families — glycol coolants such as DOWFROST LC 25 and JEFFCOOL ISF-25 form the compatibility baseline for wetted materials.
- Safety reference — IEC 62368-1 Annex G.15 frames coolant-contact materials for equipment safety.
- Fluid envelope — the UQD operating window is 17 to +65 °C coolant, with transport and storage from −40 to +75 °C; that window is set by the seal compounds shared across the ecosystem.
- End connections — ORB threads per ISO 11926-3 on plug ends, barbed socket ends sized for 1/4" to 5/8" reinforced EPDM hose.
- Bodies — 316L stainless, brass or aluminium, machined to a ±0.01 mm sealing-face tolerance in our own cells.
Note what that list does and does not do. It fixes the coolant family and the temperature window; it does not tell you which seal compound to order, and it does not police what the rest of your loop is made of. Those two decisions are yours, and they are where compatibility actually gets won or lost.
Seals fail first, and they fail dimensionally.
The elastomer is the only wear part in a dry-break coupling. Everything else in the assembly is designed to outlast the loop; the seal is designed to be replaced.
| Compound | Service window | Water-glycol behaviour | Where it belongs |
|---|---|---|---|
| EPDM | −40 to +120 °C | the reference compound for water-glycol service | default for cold-plate and manifold loops |
| FKM | −15 to +200 °C | built for heat; hot water-glycol is not its strongest service | warm-process loops and glycol-free circuits |
| FVMQ | −55 to +155 °C | keeps flexibility where cold starts dominate | cold-aisle and cold-start designs |
| NBR | −40 to +80 °C | dislikes glycol inhibitor packages and tops out early | cost-driven, moderate-temperature, glycol-free builds |
| HNBR | −40 to +150 °C | good where heat and abrasion arrive together | high-cycle service points with mechanical wear |
Two failure mechanisms explain almost every premature seal failure. The first is swelling: the compound absorbs fluid, changes dimension, and changes the squeeze that makes the seal work. Swelling is not visible on inspection — the O-ring looks like an O-ring — but the set can pass a cold leak test and weep once the loop is at temperature. The second is compression set: the elastomer takes a permanent deformation and stops following the valve face as it moves, so a seal that holds at 20 psi lets go at 60.
That is why the soak test exists. Sampled sets sit in water-glycol for 72 hours at the −40 to +120 °C class endpoints before release, and the compound identity travels with the batch. If a seal lot has no traceable compound record, it does not go into an assembly — a rule that sounds bureaucratic until a loop is nine months old and someone is trying to work out which elastomer is in it.
Galvanic corrosion is a pairing problem.
Any water-glycol mix conducts well enough to carry a corrosion current. When two different metals share that fluid, one of them becomes the anode and gives up material — permanently and quietly.
| Material | Where it appears in a loop | Galvanic position | Practical rule |
|---|---|---|---|
| 316L stainless | coupling bodies, manifold bodies, ORB ends | most noble of the common set | safe default for the wetted path; it is the cathode in most bad pairings |
| Brass | manifold bodies, fittings, threaded adaptors | mid-series, anodic to stainless | fine on its own; in a stainless assembly it becomes the anode, so avoid small brass inserts facing large stainless bodies |
| Aluminium | cold plates, brackets, some manifold bodies | most anodic of the three | keep out of the wetted path unless the coolant is specified for it and the inhibitor is maintained |
| EPDM seals and hose | every joint, every line | non-metallic — no couple | not a corrosion source, but elastomer surfaces are where deposits collect |
The area ratio is what turns a pair into a problem. A small anodic area facing a large cathodic area concentrates the entire corrosion current into a small part — an aluminium fitting threaded into a large stainless manifold is the textbook bad case, and it will pit long before an aluminium manifold with stainless inserts would. The rule that follows is simple: if two metal families must meet, make the anodic part the larger one or break the electrical path between them with a dielectric washer or bushing at the mechanical joint.
Velocity matters at both extremes. High flow strips the protective film faster than the inhibitor can rebuild it, and near-zero flow in a dead leg lets deposits settle, which creates an oxygen-starved pocket under the deposit where pitting starts. That is the hydraulic argument for balancing from DL-005: an even loop has no dead legs, and a loop with no dead legs has fewer corrosion sites.
Plastics and hose: not automatically inert.
Hose is the largest wetted area in most racks, and the OCP annex names reinforced EPDM for socket ends on 1/4" to 5/8" barbs. EPDM is the right answer for water-glycol at loop temperatures, and it holds its compression set across the 5000-cycle endurance floor.
Elsewhere in the loop, the assumption that a polymer is inert causes two distinct failures. The first is extraction: plasticisers, oligomers and process residues leach into the coolant, travel, and then deposit somewhere that matters — a filter, a narrow port, a cold plate microchannel. The second is environmental stress cracking, where a polymer that resists the glycol perfectly still cracks in the presence of a solvent, a cleaning agent or a change in the inhibitor chemistry.
The practical defence is to check the coolant maker's compatibility list for every polymer in the wetted path, not just the seal, and to re-check it whenever the coolant brand changes. Water plus glycol plus inhibitor is three chemistries, and a switch of product can change the third one while the label still says the same concentration. The same logic applies outside the fluid path: an aluminium panel that catches a drip becomes a corrosion site even though it was never a wetted component, which is one more reason the dry-break claim is worth as much as the flow claim.
Six rules for a mixed-metal loop.
None of these are expensive. All of them are cheaper on a drawing than in a rack.
Compatibility checklist
One more habit is worth adding to the list: when a loop is drained, opened or refilled, the compatibility argument restarts. Airborne contamination, a change of hose brand, a new cleaning agent and a different coolant product all alter the chemistry, and the hardware that was qualified last year is now an untested combination. Record what went in, and when.
Three questions this brief answers most.
Is EPDM suitable for a 25% propylene glycol loop?
Which metal corrodes in a mixed stainless and aluminium loop?
Does glycol concentration matter for compatibility?
Tell us the coolant and the temperature endpoints.
Loop profile in, seal compound and body material back within 48 hours — with the compound identity recorded in the batch documentation that ships in the carton.