Home › Blog › Electromechanical & Switchgear › Data Center Cooling & Liquid › Coolant-loop elastomer compatibility
Glycol, Dielectric Fluids, and the Seal: Coolant-Loop Elastomer Compatibility on a Delay
Coolant-loop seal failures are chemistry failures on a delay. An elastomer the fluid attacks does not leak on day one: it swells, softens, or hardens over weeks until the sealing force is gone, long after commissioning signed everything off.
The parent cooling and liquid infrastructure page builds its whole selection logic around that delay, and this article unpacks the three compatibility decisions behind it: why standard silicone sponge fails at glycol-exposed penetrations, how immersion-fluid chemistry picks the tank seal, and why every pairing is confirmed against the fluid maker's elastomer compatibility data rather than assumed from a family name.
The companion mechanical failures, compression set on high-cycle doors and the sealed cabinet that breathes through its own gaskets, finish the picture.
Request a quote for your project →
Why compatibility failures hide from commissioning
A commissioning check verifies the mechanical state of a joint: compressed gasket, closed flange, no leak today. Chemical attack works on a different clock. The fluid migrates into the elastomer network (swell), extracts plasticizers and compounds out of it (hardening and shrinkage), or both, and each mechanism erodes the sealing force gradually. The gasket near a manifold softens and grows where mist and weeps reach it; the immersed lid seal changes durometer over months in the tank.
Neither shows on day one, and both are decided at specification time. The honest predictor is the fluid maker's elastomer compatibility data for the specific fluid and concentration: swelling, hardness change, and extractables against your actual chemistry, which is exactly the data the parent page defers to instead of publishing a green-light chart.
Decision 1: the glycol mist path, silicone hands off to fluorosilicone
Water/glycol loops are the common case, and the subtlety is exposure mode, not just fluid. A dry-side CDU door gasket never wets, so EPDM foam or standard silicone economics win. A coolant manifold penetration is different: glycol mist, drips, and the occasional weep reach the seal for years, and silicone elastomers are a poor match for sustained glycol exposure; the gasket softens and swells where the loop touches it, and the seal force goes with it.
That is why the page maps fluorosilicone sponge (AMS 3323 family) specifically to glycol-exposed penetrations: it keeps silicone's cellular conformance and adds the fuel and chemical resistance, bought at a price premium that belongs only where the exposure actually is. The discipline is a sorting pass: classify every seal location on the cabinet by what can reach it, and let only the mist-path and weep-path joints pay for fluorosilicone.
Decision 2: immersion fluids, where the fluid is the specification
Single-phase immersion cooling inverts the usual sealing question: the gasket is no longer keeping water out of an electrical box, it is living in continuous contact with a dielectric fluid, a synthetic ester, hydrocarbon, or fluorinated chemistry chosen by the tank maker.
Elastomer selection follows that fluid. FKM (75 durometer solid, with Viton sponge where a conformable closed-cell form is needed at low closure force) is the working family for tank lids, sight-glass seals, and feedthrough grommets in the common ester and hydrocarbon fluids. FFKM (GP 70/75D, Steam 75D) is held for the most aggressive chemistries, including some fluorinated fluids and hot spots near heaters, where FKM's compatibility margin is in question.
The honest engineering position, stated plainly on the parent page: compatibility is confirmed against the fluid maker's elastomer compatibility data for the specific fluid, not assumed from the family name, because formulations differ and extractables matter in a loop that bathes electronics.
Decision 3: the mechanical siblings that mimic chemical failure
Two non-chemical mechanisms produce the same symptom (a seal that stopped sealing) and deserve a place in the same review.
First, compression set on high-cycle doors: an RDHx door opens and closes thousands of times, much of it warm, and an organic foam gradually takes a permanent set until exhaust bypasses the coil; silicone foam's flat set behavior across temperature is why BISCO BF / HT-series silicone foam owns that location (set data per ASTM D1056 on the TDS).
Second, the cabinet that breathes: a tightly gasketed CDU thermal-cycles with the loop, swings its internal pressure, inhales humid air past whatever yields first, and condenses moisture inside. The fix is not a tougher gasket but a deliberate pressure path: a die-cut ePTFE vent membrane patch that equalizes the swing while blocking liquid water, sized from the maker's airflow data.
When a "leaking" cabinet shows up, the honest diagnosis sorts these mechanical mechanisms from the chemical ones before any material changes.
A compatibility review you can run from the drawing
- Name the fluid exactly: the coolant brand or chemistry and concentration for water/glycol loops; the specific dielectric fluid for immersion. "Glycol" is not a spec; "the named coolant at the loop concentration" is.
- Classify each seal location by exposure: dry-side, mist/weep path, or immersed. The three classes map to three family tracks (economy foams, fluorosilicone, fluoroelastomers).
- Pull the fluid maker's elastomer table for the named fluid and check swelling, hardness change, and extractables for the candidate family.
- Add the non-chemical lines: open-close cycles and service temperature (compression set track), and the vent plan for any tightly sealed, thermal-cycling enclosure.
- Put it all on the drawing: fluid, concentration, exposure mode per seal, and the flame-class callout where the enclosure specification applies one (kSil V-0 and RS-series carry UL 94 V-0 listings per their TDSs).
Frequently asked questions
Why did the manifold gasket fail when the same material seals the door perfectly?
Exposure mode. The door gasket lives dry; the manifold gasket lives in the mist and weep path of the loop, and sustained glycol exposure softens and swells standard silicone until the seal force is gone. Same material, different fluid environment, different outcome. Classify seal locations by what can reach them, and move the mist-path joints to fluorosilicone sponge while the dry-side stays on the economical grades.
Can H-O certify that a seal is compatible with our coolant?
No, and the honest version is better: compatibility is confirmed against the fluid maker's elastomer compatibility data for your specific fluid and concentration, reviewed during quoting, with the seal grade's own TDS carrying its material properties. H-O converts the chosen material and supplies its documentation; the fluid pairing rests on the fluid maker's published data, which is the authority a review will actually accept.
Our loop is changing coolant chemistry. Do the seals need to be re-reviewed?
Yes, and it is a cheap review compared to the alternative. A seal qualified against one chemistry carries no automatic standing in another; ester, hydrocarbon, and fluorinated fluids interact differently with every elastomer family, and concentration changes matter even within one chemistry. Send the new fluid name and concentration with the existing seal drawings; the families may hold, the grades may not.
Does an immersion tank seal need a flame rating too?
Only where the enclosure or facility specification applies one; the fire-protection framing for IT spaces sits in standards like NFPA 75, mapped to equipment level by your facility engineer. When a UL 94 class is called on seal materials, the listing is per grade and per thickness on the TDS, and it rides alongside the compatibility requirement rather than replacing it. State both on the drawing and let the review reconcile them.
Reviewing seals against a coolant or immersion fluid?
Send the fluid name and concentration, the seal locations by exposure mode, and the drawings. Engineering runs the pairing against the fluid maker's data and the grade TDSs and returns a family direction with a quote.
Samples typically ship in 3 to 5 business days for common die-cut configurations; standard production runs about two weeks. MOQ varies by material and part.
Family-owned in Winsted, Connecticut since 1971 · ISO 9001:2015 certified · OEMs worldwide send us a drawing — and get back their part.
Related reading
ApplicationData Center Cooling & Liquid InfrastructureRead →
GuideEPDM, Silicone Sponge, or Fluorosilicone: Choosing Gaskets and Insulation for CDU and RDHx HardwareRead →
GuideHow to Spec Liquid-Cooling Seals and Pipe Insulation: What to Send H-O for a QuoteRead →
IndustryElectromechanical & Switchgear hubRead →
Request a quote for your data-center liquid cooling part
Send H-O your drawing or describe the part—engineering reviews every RFQ and converts samples on your geometry.
Last updated · Reviewed by H-O Products engineering (ISO 9001:2015 certified converter, Winsted, CT).
References
- ASTM D1056 (compression-deflection and set classes on the foam and sponge TDSs); UL 94 (V-0 listings per grade and thickness); AMS 3323 (fluorosilicone sponge designation).
- ASTM D2240 / D412 (durometer and tensile on the fluoroelastomer TDSs); ASHRAE TC 9.9 datacom liquid-cooling guidance (facility-water classes, cited qualitatively); NFPA 75 (fire protection of IT equipment, by designation).
- Fluid makers' elastomer compatibility data for the specific coolant or dielectric fluid (the governing reference for every pairing in this article); maker TDSs for the EPDM, silicone foam and sponge, fluorosilicone, FKM, FFKM, and ePTFE membrane grades.
This article explains coolant-loop compatibility mechanisms qualitatively for specification purposes. No compatibility rating, swell value, or service life is stated for any specific grade or fluid; pairings are confirmed against the fluid maker's elastomer compatibility data and the grade's current technical data sheet during drawing review. Flame listings belong to the listed grades; facility-level requirements belong to the facility engineering.
All parts are custom manufactured to order; minimum production run quantities vary by material and part. H-O is a die-cut converter, not a raw-material producer.