Inside PG25: The Invisible Chemistry Protecting Every Cold Plate
Put copper, aluminum, and steel in warm, oxygen-bearing liquid and the loop begins a chemistry experiment whether anyone planned one or not.
Inside a cold plate, that experiment is compressed into narrow passages with a large wetted surface.
The fluid must carry heat without encouraging the materials around it to become part of the fluid.
That is why PG25 is more than water plus glycol.
Many data center formulations include Organic Acid Technology, usually shortened to OAT, as part of a carefully matched corrosion inhibitor package.
Glycol helps define the fluid. The inhibitor package helps the hardware survive it.
The 30-second version
- OAT is a technology family, not a complete recipe. The named product still decides the actual chemistry and material compatibility.
- Carboxylate inhibitors can protect active metal sites. Their behavior depends on the formulation, alloy, pH, temperature, oxygen, and contaminants.
- Inorganic and hybrid packages work differently. Broad films, selective protection, and combined strategies create different tradeoffs.
- No single fluid test proves everything. Concentration, pH, reserve alkalinity, inhibitor analysis, and dissolved metals answer different questions.
PG25 is a formulation, not a recipe name
The shorthand starts with roughly 25 percent propylene glycol and 75 percent water by volume.
A finished coolant can also contain a proprietary corrosion inhibitor package and other additives selected by the manufacturer.
Two products can both say PG25 and still use different water quality, inhibitors, antifoam chemistry, dyes, biostability strategies,
and material approvals. Mixing them because the front label looks similar can create a formulation that neither supplier validated.
OAT tells you the inhibitor family. It does not reveal the complete package or make every OAT coolant interchangeable.
What changes at the metal wall
Corrosion is electrochemical. Small anodic and cathodic regions exchange charge while the fluid carries ions between them.
Without suitable protection, metal leaves the surface and the local site can grow.
Carboxylate-based inhibitors can adsorb or chemisorb where protection is needed and interfere with those reactions.
Think of a road crew concentrating on the damaged section instead of resurfacing every lane.
It is selective protection, not a magic coating that appears only after visible damage.
Commercial packages remain proprietary and material-specific. Compatibility data, operating guidance, and fluid analysis matter more than the acronym on its own.
Cold plates make small chemistry problems consequential
Narrow channels create excellent heat transfer and very little tolerance for deposits, incompatible mixtures, or corrosion products.
The first sign may not be a leak. Dissolved metal can rise in a sample. Deposits can change branch resistance.
A surface film can change the boundary where heat enters the coolant. Chemistry, pressure, flow, and temperature may move on different timescales.
The best coolant is therefore not simply the one with the longest advertised life.
It is the one validated for the actual metals, seals, temperatures, heat flux, and maintenance practices in the loop.
Protection has to survive a materials test
A credible inhibitor claim eventually meets metal coupons and a balance.
The Open Compute Project PG fluid guideline describes a modified ASTM D8040 corrosion test and gives maximum corrosion rates for materials including copper and brass.
Known materials are exposed to controlled fluid conditions, then cleaned and weighed carefully to measure material loss.
The idea is simple. The execution is laboratory work.
OCP lists SOLUTHERM OAT PG25 as a ready-to-use fluid with a proprietary OAT inhibitor system for copper, brass, steel, cast iron, and aluminum.
Dow describes DOWFROST LC 25 as a ready-to-use PG25 fluid with a customized inhibitor package for high surface-area copper direct-to-chip systems.
Those are product-specific claims. They are evidence for the named fluids, not permission to treat every PG25 or OAT package as equivalent.
One bottle can answer six different questions
- Appearance can reveal particles, haze, separation, or an obvious color change. It cannot certify concentration or inhibitor health.
- PG concentration checks dilution. It does not prove that the additives remain at the intended level.
- pH shows the current acid-base state. It does not identify the inhibitor or locate corrosion.
- Reserve alkalinity measures acid-neutralizing capacity under a defined method. It is not a universal countdown clock.
- OAT analysis can measure specified organic acids when an approved laboratory method is available. Castrol lists HPLC for its OAT inhibitor analysis.
- Elemental analysis can reveal metals moving relative to baseline, but the result still needs system and sampling context.
Choose a measurement below. Each selection shows what the method reveals and what remains unknown.
There is no universal green band
5, and unadjusted reserve alkalinity above 4 mL for the specification it describes.
It also notes that pH may be lower with OAT inhibitors.
Castrol ON PG25 publishes 24.5 to 29 volume percent glycol and pH from 8.0 to 9.5 for its product. One is an industry guideline and the other is a
product-specific operating range, so the numbers do not need to match.
The correct operating band begins with the approved fluid, project requirements, sample method, sample location, and the clean commissioning baseline.
A lab result becomes useful on a timeline
A bottle captures one moment. The loop has a history.
If dissolved copper rises, the useful context includes temperature, oxygen entry, makeup history, concentration, pH, pressure, flow distribution,
and recent maintenance. If pH moves after an unplanned top-up, the event should already be attached to the trend.
Reliability Engine brings fluid analysis, loop telemetry, maintenance events, and the approved baseline onto one timeline.
That helps teams see whether the chemistry and hardware are still telling the same story and decide what to check next.
Do not ask only whether the coolant is clear or inside one pH band. Ask whether the formulation, the materials, and the operating evidence still agree.
Practical reads on coolant health, GPU thermal margin, and what to check next.