Twelve Ways a Data Center Glycol Loop Goes Wrong (and How to Tell Which One You Have)
Table of Contents
📋 What You'll Learn
This guide walks you through twelve ways a data center glycol loop goes wrong (and how to tell which one you have) with detailed instructions.
Liquid cooling moved from a niche to the default faster than the maintenance practice around it did. A great many facilities are now running glycol loops at densities nobody was running five years ago, with commissioning records that were never taken and fluid programs inherited from chilled-water practice. When something looks wrong, the first hypothesis is almost always "the coolant went bad" — and it is almost always something else. This guide covers the twelve things that actually go wrong, how each one presents, and the cheapest test that confirms it.
The scale this is happening at. Jane Street's published walkthrough of one of its Texas facilities describes 4,032 GPUs across 56 racks retrofitted for liquid cooling, with roughly 8,000 km of fiber. Retrofits like that are where fluid programs get inherited rather than designed — the loop is new, the maintenance practice around it usually is not, and the commissioning record that would make every reading below interpretable is the thing most likely to be missing.
How to use this guide
Start from what you observed, not from what you suspect. The table below maps a symptom to the modes that plausibly produce it; each mode below has the confirming test. Several symptoms map to several modes, which is the point — the confirming test is what separates them, and it is usually cheaper than the remedy people reach for first.
| What you observed | Modes to consider |
|---|---|
| Freeze point has fallen | 1 (make-up water), 12 (measurement error) |
| pH has fallen | 2 (inhibitor depletion), 9 (oxygen ingress), 12 |
| pH has risen | 8 (incompatible chemistry), 11 (wrong fluid class), 12 |
| Conductivity climbing | 5 (break-in), 1 (untreated water), 2, 7 (corrosion) |
| Fluid looks dark or cloudy | 5 (particulate), 3 (biological), 8 (gel), 12 (observation artifact) |
| Slime, odor, filter blinding | 3 (microbiological growth), 1 (dilution) |
| Approach temperature creeping up | 10 (fouling), 4 (over-concentration), 6 (cold plate) |
| Pump energy up, flow down | 4 (viscosity), 6 (blockage), 8 (gel) |
| Differential pressure rising across cold plates | 6 (cold plate fouling), 8, 5 |
| Metal ions or corrosion products in the panel | 7 (mixed metal), 2, 9 |
First: what the four field numbers actually prove
A field panel is almost always pH, conductivity, TDS and freeze point. They answer genuinely different questions, and the most common analytical error is treating them as four opinions about one question.
| Reading | Measures | Can prove | Cannot prove |
|---|---|---|---|
| pH | Hydrogen-ion activity | Acidic products accumulating faster than the inhibitor buffers them | How much buffer is left — that is reserve alkalinity, a separate titration |
| Conductivity | Total mobile ionic charge | Dissolved ionic material has entered or been generated | Which ions. Inhibitor salts, corrosion products and water hardness read alike |
| TDS | Usually nothing independent — derived from conductivity by a fixed factor | The same thing conductivity proves | It is not a second data point. Citing both as agreement is double-counting one measurement |
| Freeze point | Refractive index → concentration by a printed scale | The glycol-to-water ratio | Anything about fluid health. Thoroughly spent coolant holds its freeze point |
Two rules that apply to every mode below. First, read every number as a delta from the commissioning baseline, never against a generic published window — those windows were written for specific inhibitor chemistries and the wrong one will convict a healthy fluid. Second, a closed circulating loop is one continuously mixed body of fluid. Composition cannot differ from one point in a circuit to another, so any finding that varies by location within a single loop is telling you about solids, hardware or observation conditions, not about chemistry.
1. Make-up water ingress
Water is entering a system engineered to exclude it
Glycol does not preferentially evaporate out of a sealed circuit. If concentration is falling, water is being added — through manual top-ups, a make-up line, an automatic fill valve nobody has audited, or a small leak being quietly replenished. The dilution is rarely the real damage. Untreated site water also brings hardness, chlorides and dissolved oxygen into a loop designed to have none of them, and it dilutes the corrosion inhibitor at the same time it dilutes the freeze protection.
- Presents asfalling freeze point, often across multiple circuits over the same period
- Confirm byestablishing the make-up source and the volume added since commissioning — the single most useful question on this page
- Telldilution with deionized water lowers conductivity. Glycol concentration falling while conductivity rises means the added water carried dissolved solids, which points at untreated site water
- Remedyfix the ingress path first; correcting concentration without it just resets the clock
| Ethylene glycol, % by volume | Approximate freeze point |
|---|---|
| 30% | −14 °C (7 °F) |
| 40% | −22 °C (−8 °F) |
| 50% | −34 °C (−29 °F) |
Values are approximate and vary by formulation — work from the supplier's table for the product in the loop, and note whether it is stated by volume or by weight. Confusing the two is a reliable source of error in its own right.
2. Inhibitor depletion
The additive package is consumed, and pH is a lagging indicator
This is the genuine fluid-end-of-life mode, and it is far less common than it is diagnosed. Aqueous ethylene glycol degrades by thermal oxidation. Work at the U.S. National Bureau of Standards (now NIST) in the 1980s, using ion chromatography on aged solutions, identified the principal products as glycolic, formic and oxalic acids — and established that formation in the absence of oxygen was negligible.
The products are acids, so real degradation drives pH down. A pH that has risen is not a degradation signature. The companion NBS study found aerated, heated glycol in contact with aluminum or copper produced acidic solutions within 3,360 hours — 140 days — under aggressive laboratory conditions with the metal catalyst present. It is not a three-week phenomenon in a sealed loop.
- Presents aspH falling from baseline, conductivity often rising alongside
- Confirm byreserve alkalinity (ASTM D1121) against the commissioning value — the only measurement that speaks directly to remaining buffering capacity
- Thresholdswithin ±0.5 pH units of baseline is normal drift; 0.5–1.0 down is a watch item; more than 1.0 down is a red flag
- Notepublished windows differ by chemistry — conventional packages are commonly cited around 8.0–11.0, OAT and other extended-life chemistries nearer 7.0–9.5. The window without the baseline is not a diagnosis
3. Microbiological growth and under-deposit corrosion
Dilute loops become hospitable, and the damage happens out of sight
Closed loops that have drifted well below their design glycol concentration are a recognized site for microbiological growth, and the mechanism that matters operationally is the corrosion that follows it. Microbiologically influenced corrosion proceeds beneath biofilm and deposit, where conditions are local and aggressive while the bulk fluid your panel samples still looks unremarkable. That is exactly why an untroubled chemistry report does not rule it out, and why pitting sometimes appears in a loop whose numbers look fine.
The usual path in is mode 1 — concentration drifting down over months as water is added — combined with low-velocity sections where solids settle and biofilm has somewhere undisturbed to establish. The two problems compound: deposits shelter growth, and growth generates more deposit.
- Presents asfilters blinding rapidly, odor, slime on wetted surfaces, cloudy fluid, localized pitting that bulk chemistry does not explain
- Confirm bydip slides or a laboratory microbiological count, run alongside glycol concentration — the two belong on the same report, because the concentration explains the count
- Remedyreturn concentration to the design value and deal with the low-velocity deposits that are sheltering growth. Any microbiological treatment is a separate decision to take with a treatment supplier, using a product registered for the purpose and applied strictly per its label
- Do nottreat glycol concentration itself as a microbiological control strategy. Concentration is set by the freeze and heat transfer duty; microbiological risk is managed by a registered treatment and by keeping deposits out of the loop
4. Over-concentration
More glycol is not more protection — it is a thermal penalty
Topping up with concentrate "to be safe" is common and costly. Glycol is more viscous and a poorer heat transfer medium than water, so every point of concentration above what the freeze duty requires buys nothing and costs something. Published figures put the film coefficient reduction at roughly 15–30% for typical HVAC conditions, with pumping energy at high concentrations (60–70%) running 40–60% above a 40% solution.
In a high-density liquid-cooled hall, where the whole design intent is moving heat out of a very small area, that penalty lands directly on approach temperature and on the pumps.
- Presents asapproach temperature creeping up, pump energy up, flow down, with a fluid panel that looks perfectly healthy
- Confirm bycomparing measured concentration against the actual freeze duty of the lowest-temperature part of the circuit — not against a habit
- Remedycorrect concentration to design. Establish the required freeze point from the coldest exposed component, typically an outdoor dry cooler, not from the hall
5. Break-in particulate and discoloration
A new loop is supposed to get dirtier before it gets cleaner
Every new piping system arrives carrying material that was never meant to stay in it: mill scale and iron oxides from the pipe, welding slag, jointing compound, cutting and drawing lubricants, flux residue from brazed joints, and construction debris. Pre-commissioning chemical cleaning and velocity flushing exist precisely to remove it, typically a nitrite or molybdate cleaner circulated around 24 hours to lift oil, grease and flux and disperse iron oxide.
No flush is perfect. What remains keeps releasing for weeks to months. Some dissolves and reads as conductivity; some stays suspended and reads as color. Meanwhile the inhibitor package is passivating fresh, reactive metal, which is itself an ionic process. Read in isolation, all of this looks like deterioration. Read alongside pH and reserve alkalinity, it is a system bedding in.
- Presents asconductivity climbing, fluid darkening, filters loading, in a recently commissioned system
- Confirm bythe 5 µm pad test — pass a sample through a filter pad. Color that stays on the pad is suspended solids and filtration is the remedy. Color that passes through is dissolved and warrants a full panel
- Separates from mode 2 byconductivity rising while pH holds and reserve alkalinity stays high is break-in; conductivity rising while pH falls and reserve alkalinity drops is depletion
- Keep the filter element.It has been integrating the loop's particulate load continuously and answers directly what fluid tests answer only by inference
Deposit color is itself diagnostic: reddish-brown indicates iron oxide from mill scale release or ferrous corrosion; blue-green indicates copper; dark grey and gritty indicates construction debris; soft, dark and greasy indicates cutting oil or flux residue that survived the flush.
Color is also an additive. Dye is deliberately added to many coolants for leak detection and product identification. Before treating a color as a symptom, establish what color the fluid was at fill — from the commissioning record, not from memory.
6. Microchannel cold plate fouling
The geometry that makes direct-to-chip work is the geometry that clogs
Microchannel cold plates achieve their heat transfer coefficients through very high surface-area-to-volume ratios and very narrow flow paths. The same features accelerate fouling: sharp corners, sudden expansions and header regions become preferential deposition sites wherever local velocity falls below the threshold that would keep them swept. Thermal gradients add their own drivers, with hot spots promoting localized deposition.
A widely used design rule of thumb is that particles should be smaller than about one third of the channel dimension to avoid bridging. Practical filtration guidance for direct-to-chip service commonly runs to 50 µm or finer, with high-purity loops filtered at 1–5 µm and general cooling loops at 10–50 µm — confirm the requirement against your cold plate vendor’s own specification rather than a general figure.
- Presents asrising differential pressure across cold plates, individual nodes running hot, flow maldistribution between parallel paths
- Confirm bydifferential pressure trending per branch, filter differential, and particle count on a drawn sample
- Remedyfiltration sized to the channel geometry — and note that straining is not filtration; a strainer sized for pump protection will pass everything that matters here
7. Mixed-metal and galvanic corrosion
Copper, aluminum and steel in one wetted circuit
Liquid cooling loops are routinely mixed-metal by construction: copper cold plates and tubing, aluminum components, steel piping, brazed and soldered joints, and a variety of alloys in pumps and valves. Dissimilar metals in a shared electrolyte set up galvanic couples, and the inhibitor package is the thing standing between that arrangement and steady material loss. This is precisely why inhibitor chemistry is specified against the metallurgy rather than chosen by habit.
Copper deserves particular attention. NBS work specifically characterized thermal degradation products arising from glycol solutions in contact with copper metal, and the companion study found aerated glycol in contact with copper or aluminum turned acidic within 140 days under test conditions. A mixed-metal loop is a different chemical problem from an all-steel one.
- Presents asmetal ions in a laboratory panel, blue-green or reddish deposits, localized pitting, conductivity rising without an obvious ingress path
- Confirm bylaboratory elemental analysis — copper, iron, aluminum and zinc in solution are the direct evidence
- Remedyverify the inhibitor package is specified for the actual metallurgy present, and confirm reserve alkalinity is being maintained rather than assumed
8. Incompatible inhibitor chemistries mixed
Silicate and carboxylate chemistries do not co-exist
Inhibitor families are not interchangeable, and mixing them is worse than running either. When silicate-containing chemistries are combined with organic acid technology under heat, the silicates can drop out of solution — the reported result is a white-to-grey gelatinous silicic acid precipitate that circulates and accumulates in the narrowest passages available to it.
In an automotive context that means a heater core. In a data center it means the cold plates, which have far less margin. A gel that would take months to make itself known in a 3–5 mm passage behaves very differently in a microchannel.
- Presents asgel or sludge on filters, rising differential pressure, pump seal wear, thermal performance falling across many nodes at once
- Confirm byinspecting filter media and low-flow sections for gelatinous deposit, and auditing every top-up product ever added against the commissioning fluid
- Prevent bydocumenting the fluid at commissioning and controlling top-up stock. Color is not an identifier — dye color varies by manufacturer and tells you nothing reliable about the inhibitor family
Never identify a coolant by its color. Dye is a marketing and leak-detection choice, not a chemistry standard. Two fluids of the same color can be chemically incompatible, and two compatible fluids can look nothing alike. Identify by product documentation and the commissioning record.
9. Oxygen ingress
The degradation reaction needs oxygen, so keeping it out is the control
The NBS work is unambiguous that degradation product formation in the absence of oxygen was negligible. A properly sealed, deaerated closed loop is chemically a different environment from one drawing air. That makes air ingress the upstream cause of a good share of apparent fluid deterioration — and it is a mechanical problem with a mechanical fix.
Common paths: a failing pump seal, an expansion vessel with a lost charge or failed bladder, air separation that is not doing its job, a system operating under negative pressure at high points, or repeated draining and refilling that reintroduces dissolved oxygen every time.
- Presents aspH falling faster than service hours justify, repeated inhibitor depletion after correction, air noise, unstable pressures, entrained air at high points
- Confirm bychecking expansion vessel charge, static pressure at high points, and air vent function before touching the fluid
- Notedissolved carbon dioxide from air contact forms carbonic acid and depresses pH — which also makes it a sampling artifact if samples sit open before testing. Cap samples and run them promptly
10. Fouling-driven heat transfer loss
Approach temperature moves before anything else does
Deposits do not need to block anything to cost real money. Published figures put even a 0.25 mm iron oxide film at a 10% or greater reduction in heat transfer efficiency in chilled-water service, which shows up as a rising approach temperature and a chiller working harder for the same duty.
This is the mode most likely to be running for a long time before anyone opens a ticket, because nothing has failed — the plant is simply less efficient than its design, and there is no alarm for that unless someone is trending it.
- Presents asapproach temperature creeping up, kW per ton rising, duty achieved at higher flow or lower supply temperature than design
- Confirm bytrending approach against commissioning values, and by inspecting heat exchanger surfaces at the next available opportunity
- Remedyaddress the source of the deposit as well as the deposit; cleaning without fixing the mechanism buys time only
11. Wrong fluid class for the specification
An inhibited coolant and a low-conductivity coolant are different products
Conductivity requirements vary by orders of magnitude across cooling applications. Ultra-low-conductivity fluids for fuel cell and certain direct-contact electronics duties are specified in the single-digit µS/cm range, and deionized water to ASTM Type II sits below 1 µS/cm. A conventionally inhibited glycol contains dissolved inhibitor salts by design and reads far higher — because that is what the inhibitor is.
Comparing an inhibited loop against a DI-water specification produces instant and completely meaningless alarm. The reverse error is worse: topping up a loop that was commissioned for low-conductivity service with an inhibited product silently violates the design basis.
- Presents asa conductivity reading that looks catastrophic against the wrong reference, or a low-conductivity loop that will not hold its specification
- Confirm byestablishing which fluid class the system was commissioned on before interpreting any number against it
- Remedymatch the top-up product to the design basis, and record it where the next person will find it
12. Measurement error masquerading as a fault
Instrument and observation problems produce confident, wrong numbers
A surprising share of coolant investigations are chasing an artifact. These cost nothing to rule out and should precede any remedy.
- Refractometer scaleanalog glycol refractometers commonly carry separate ethylene and propylene glycol scales in one eyepiece. The columns do not agree, because refractive index relates to concentration differently for the two glycols. Reading the wrong column yields a wrong freeze point that errs consistently — which makes it look like a trend. Digital instruments are sold specifically to remove this ambiguity
- Uncalibrated pH penwill invent half a unit without complaint. A sealed loop that reads a full pH unit away from its neighbors on one date and returns on the next has an instrument problem; fluid in a closed circuit does not behave that way
- TDS double-countingmost handhelds derive TDS from conductivity by a fixed factor. Reporting both as corroboration is one measurement quoted twice
- Observation conditionstwo sight glasses under different lighting will not match, and staining on the inside of a glass reads as fluid color. Sample into identical clear containers and compare in the same light before treating any visual difference as real
- Single snapshotone set of readings cannot establish a rate, and rate is what separates break-in from active degradation
- Wrong referencecomparing an in-service loop against currently available product introduces batch variation as a false signal. Compare against the loop's own fill baseline
The commissioning baseline is the cheapest thing on this page
Nearly every mode above is diagnosed by comparison, and most sites cannot make the comparison because nobody recorded the starting point. Ten minutes at fill converts every future panel from an argument into arithmetic.
Record at commissioning, from the loop itself once circulated: pH, conductivity, glycol concentration, reserve alkalinity, the product name and inhibitor family, the batch or lot number, the make-up water source, and a photograph of the fluid in a clear container. Store it where the next person will find it — the value of this record is realized by somebody who does not work there yet.
Where several circuits were filled from the same product at the same time and run comparable duty, they also form a built-in reference set. A parameter that has moved across all of them points to a shared cause — the fluid, the water source, or a site-wide practice. A parameter that has moved on one points to that circuit's own hardware. This comparison costs nothing and is available to most sites already.
Field protocol before anyone drains a loop
Draining is expensive, glycol disposal is regulated, and refilling destroys the evidence needed to identify the cause. If the source is hardware or practice rather than fluid, a fresh charge presents the same way on the same timeline. Ordered by information gained per hour spent:
| # | Action | Rules in or out |
|---|---|---|
| 1 | Verify instruments — refractometer scale, pH calibration | Mode 12, before it becomes a finding |
| 2 | Retrieve the commissioning baseline, fluid identity and batch | Makes every other reading interpretable |
| 3 | Retain and photograph the filter element being changed | Modes 5, 6, 7, 8 — the most direct physical evidence available |
| 4 | Audit every top-up product and volume added since fill | Modes 1, 8, 11 |
| 5 | Establish the make-up water source | Modes 1, 3 |
| 6 | Compare against sibling circuits on the same fill | Separates fluid or site-wide causes from single-circuit hardware |
| 7 | 5 µm pad test on a drawn sample | Mode 5 — separates suspended from dissolved |
| 8 | Check expansion vessel charge and air vent function | Mode 9 |
| 9 | Trend differential pressure per branch and approach temperature | Modes 6, 10 |
| 10 | Full laboratory panel — reserve alkalinity (ASTM D1121), elemental, microbiological | Modes 2, 3, 7 — the decision-grade data |
Which ASTM method is which
Standards in this area are routinely miscited, including in vendor literature, so precision is worth the paragraph. The measurement worth paying a laboratory for is reserve alkalinity, because it is the only one that speaks to remaining inhibitor capacity rather than to the consequences of losing it.
| Standard | What it actually is | Role |
|---|---|---|
| ASTM D1121 | Standard Test Method for Reserve Alkalinity of Engine Coolants and Antirusts — titration with 0.100 N HCl to pH 5.5 | This is the reserve alkalinity method. In-service monitoring |
| ASTM D2619 | Standard Test Method for Hydrolytic Stability of Hydraulic Fluids (Beverage Bottle Method) — copper catalyst, water, 93 °C, 48 h | Formulation qualification. Frequently miscited as a reserve alkalinity method. It is not one |
| ASTM D3306 | Standard Specification for Glycol Base Engine Coolant for Automobile and Light-Duty Service | A product specification, not a field test |
The short version
- Degradation lowers pH. Glycol oxidizes to glycolic, formic and oxalic acids. A pH that has risen is not degradation.
- Every number is a delta from baseline, not a comparison against a generic window written for a different inhibitor chemistry.
- One loop is one chemistry. Anything varying by location within a circuit is solids, hardware or observation conditions.
- Falling concentration means water is getting in — and dilute loops are where microbiological problems start.
- More glycol is a thermal penalty, not extra insurance.
- Never identify a coolant by dye color, and never mix inhibitor families.
- Rule out the instrument before the fluid. It is free and it is often the answer.
- Keep the filter element. It is the best evidence in the system and it is usually in a skip by the time anyone asks.
Sourcing glycol, or trying to work out what your loop is telling you?
We supply ethylene and propylene glycol and inhibited coolant blends from stock, with a Certificate of Analysis on every order — and we are happy to look at a test panel with you before anyone drains anything.
Browse coolants and antifreezeReferences & Authoritative Sources
Degradation chemistry from peer-reviewed work conducted at the U.S. National Bureau of Standards (now NIST). Test method definitions from ASTM International. Chemical identity from the National Institutes of Health PubChem database.
- Rossiter, W. J., Godette, M., Brown, P. W., & Galuk, K. G. — An investigation of the degradation of aqueous ethylene glycol and propylene glycol solutions using ion chromatography. Solar Energy Materials, Vol. 11 (1985). National Bureau of Standards. Identifies glycolic, formic and oxalic acids as principal thermal-oxidative degradation products.
- Clifton, J. R., Rossiter, W. J., & Brown, P. W. — Degraded aqueous glycol solutions: pH values and the effects of common ions on suppressing pH decreases. Solar Energy Materials, Vol. 12, Issue 1 (1985), pp. 77–86. Establishes pH decrease on degradation and the 3,360-hour aerated metal-contact result.
- Rossiter, W. J. et al. — Characterization of potential thermal degradation products from the reactions of aqueous ethylene glycol and propylene glycol solutions with copper metal. Solar Energy Materials (1987).
- ASTM D1121 — Standard Test Method for Reserve Alkalinity of Engine Coolants and Antirusts. ASTM International.
- ASTM D2619 — Standard Test Method for Hydrolytic Stability of Hydraulic Fluids (Beverage Bottle Method). ASTM International.
- ASTM D3306 — Standard Specification for Glycol Base Engine Coolant for Automobile and Light-Duty Service. ASTM International.
- PubChem CID 174: Ethylene Glycol — National Center for Biotechnology Information, U.S. National Library of Medicine. CAS 107-21-1, C2H6O2, MW 62.07, IUPAC name ethane-1,2-diol.
Frequently Asked Questions
What are the most common causes of data center coolant problems?
Twelve failure modes account for most of them: make-up water ingress, inhibitor depletion, microbiological growth in dilute loops, over-concentration, break-in particulate from new pipework, microchannel cold plate fouling, mixed-metal corrosion, incompatible inhibitor chemistries forming a silicate gel, oxygen ingress, fouling-driven heat transfer loss, running the wrong fluid class for the conductivity specification, and measurement error. Degraded glycol is a real mode but far less common than it is diagnosed.
Does the pH of glycol coolant go up or down as it degrades?
Down. Aqueous ethylene glycol degrades by thermal oxidation into glycolic, formic and oxalic acids, so degradation products are acidic and pH falls as the inhibitor's buffering reserve is consumed neutralizing them. Work at the U.S. National Bureau of Standards established this and also found that degradation product formation is negligible without oxygen. A pH that has risen is not a degradation signature and should be investigated as contamination, an incompatible top-up product, or instrument error.
Why is my glycol coolant dark or cloudy?
Most often suspended particulate rather than a change in the glycol. New piping releases mill scale, iron oxides, weld slag, jointing compound, cutting lubricant and flux residue for weeks to months after commissioning. Pass a sample through a 5 micron filter pad: color that stays on the pad is solids and filtration is the remedy, while color that passes through is dissolved and warrants a full laboratory panel. Cloudiness can also indicate microbiological growth or a gel formed by mixing incompatible inhibitor chemistries. Many coolants also carry added dye, so check what color the fluid was at fill.
Is rising conductivity in a newly commissioned glycol loop a problem?
Not on its own. Break-in debris dissolving from new pipework raises conductivity, and the inhibitor package passivating fresh metal is itself an ionic process. The distinguishing test is what accompanies it: conductivity rising while pH holds steady and reserve alkalinity stays high is a break-in signature, whereas conductivity rising while pH falls and reserve alkalinity drops indicates inhibitor depletion. Note also that TDS on most handheld meters is derived from conductivity by a fixed factor, so it is not an independent second reading.
What does falling glycol concentration across several loops mean?
Water is entering the systems. Glycol does not preferentially evaporate from a sealed circuit, so concentration falling across several circuits over the same period points to water being added through top-ups, a make-up line, an automatic fill valve, or a leak being replenished. Dilution with deionized water lowers conductivity, so if conductivity rises while glycol concentration falls, the added water carried dissolved solids, which indicates untreated site water rather than deionized water.
Can you mix different coolant inhibitor types?
No. Inhibitor families are not interchangeable. Combining silicate-containing chemistries with organic acid technology under heat can drop silicates out of solution as a white-to-grey gelatinous silicic acid precipitate that circulates and accumulates in the narrowest passages available, which in a liquid-cooled data center means the cold plates. Never identify a coolant by dye color either - color is a leak-detection and marketing choice, not a chemistry standard, so two fluids of the same color can be chemically incompatible.
Does a higher glycol concentration give better protection?
No, above what the freeze duty requires it is a thermal penalty. Glycol is more viscous and a poorer heat transfer medium than water. Published figures put the film coefficient reduction at roughly 15 to 30 percent for typical HVAC conditions, with pumping energy at 60 to 70 percent concentrations running 40 to 60 percent above a 40 percent solution. Set concentration from the freeze duty of the coldest exposed component, typically an outdoor dry cooler.
Which ASTM method measures reserve alkalinity in coolant?
ASTM D1121, Standard Test Method for Reserve Alkalinity of Engine Coolants and Antirusts, which titrates the sample with 0.100 N hydrochloric acid to pH 5.5. ASTM D2619 is frequently miscited for this and is not a reserve alkalinity method - it is the Standard Test Method for Hydrolytic Stability of Hydraulic Fluids (Beverage Bottle Method), a formulation qualification test rather than an in-service monitoring test. ASTM D3306 is a product specification, not a field test.
Related Chemical Collections
About the Author
