KALTEC

BlogApplication notes

Why is my coolant splitting with oil floating on top?

Type
Application notes
Published
20 November 2026
Prepared by
Technical team
Product group
Metalworking Fluids (MWF)

Skimming the oil and adding concentrate is the usual response to a split coolant, but if whatever weakened the emulsifiers stays, the splitting comes back. What actually breaks the emulsion can be leaking machine oil, the wrong temperature or mixing, falling pH, ions building up in the water, or microbes eating the emulsifiers, and two or three often act at once.

Look into the sump in the morning before the machines start, and a yellowish oil film or a cream-like layer sits on what used to be a milky coolant. Stirring seems to mix it back for a while, then it rises again, and sticky residue is left on parts and machines. Even a fresh charge can look the same within a few weeks.

A split coolant cools and lubricates unevenly, so tools wear faster, and rust protection weakens, so parts and machines rust. Floating oil keeps oxygen from the air out of the fluid, so microbes that live without oxygen grow and odour follows (McGuire, 2026; Passman, 2009). The UK Health and Safety Executive, HSE, treats fluid quality as the key to controlling risk and notes that coolant mist can cause occupational asthma, hypersensitivity pneumonitis and dermatitis (HSE, 2021; NIOSH, 1998). Once a coolant has failed badly, sump-side additives rarely bring it back (Burke, 2023), so finding the cause early is the surest way to cut the cost of changeouts.

Emulsion and semi-synthetic coolants are water with oil broken into very small droplets and spread through it. The droplets stay apart because emulsifiers, molecules with one end that likes water and one that likes oil, coat their surfaces, and most products use anionic components such as soaps and fatty acids to give each droplet a negative charge, so like charges push them apart (Brinksmeier et al., 2015; Burke, 2023). When this balance fails, it goes in two steps. First the droplets grow or cluster and rise, which is creaming, and stirring mixes them back in. Further on, droplets merge into a separate oil layer, which is splitting, and that does not come back. Droplet size grows before any separation is visible, so a test that tracks particle size catches instability earlier than a separation test (Deluhery & Rajagopalan, 2005).

In such cases the technical team sets up the five hypotheses below and rules them out one by one. Whatever the case, the first job is the same: draw a sump sample into a clear container, let it stand for a day and compare the floating layer with the concentrate and the machine oils, and on the same day measure concentration, pH, fluid temperature and make-up water hardness against the supplier's recommended range.

Hypothesis 1: leaking machine oil

It happens where the sump serves machines with worn hydraulic seals, automatic slideway lubrication, or oil weeping around gearboxes and spindles. Oil that gets in this way is called tramp oil, and it is not one oil: hydraulic, way, gear and spindle oils arrive through seals, lines and slideways (Burke, 2023). Say the sump has stood without circulation over the weekend and is opened on Monday morning. A brown oil film, darker and tackier than the concentrate, covers the surface, and the machine's hydraulic top-up log has been climbing since last month.

Incoming oil breaks the emulsion by several routes. It too needs coating with emulsifier, so it takes a share of the coolant's emulsifiers and upsets the balance between water-loving and oil-loving character, the HLB (Burke, 2023; McGuire, 2026). Hydraulic and gear oils often contain demulsifiers, made to separate oil from water, and dispersants, which work directly against the coolant's emulsifiers (Burke, 2023; McGuire, 2026). The zinc-based anti-wear additive ZDDP releases divalent zinc ions, which tie up anionic emulsifiers the same way the hardness ions of hypothesis 4 do (Burke, 2023). Oil-soluble coolant additives are drawn out into the tramp oil, lowering their level in the fluid, and the floating layer blocks oxygen and favours anaerobic microbes (McGuire, 2026; Passman, 2009). Once the pump's hard stirring, or shear, and heat have emulsified it, the oil becomes part of the coolant and is hard to take out (Burke, 2023; McGuire, 2026).

A graduated clear container is the handiest tool. Let a sample stand in it for a day, record the share of volume taken by the top layer, and compare the colour, viscosity and smell of the floating oil with the concentrate and the machine oils. A match with a machine oil, together with higher-than-usual lube top-up on that machine, is a strong sign. HSE advises checking tramp oil weekly and keeping it to a minimum, below 2%, with the method agreed with the supplier (HSE, 2021).

Fixing the leaks is the most effective measure (Burke, 2023). Aim to remove incoming oil before it emulsifies, roughly within 48 hours (Burke, 2023). Skimmers work only on free-floating oil and need time for it to rise, so place them where flow is slow; when much of the oil is already emulsified, a coalescer, which gathers small droplets into large ones that rise, or a centrifuge is needed (HSE, 2021; McGuire, 2026). Deliberately emulsifying tramp oil into the coolant is not advised, as it brings in sulfur, phosphorus and tackifiers (Burke, 2023).

Hypothesis 2: heat, cold or mixing

Suspect this when fluid temperature climbs, in high summer or during heavy grinding, on machines running a small sump with no chiller, when a fresh charge is made from concentrate kept in an outdoor store in winter, and in sumps topped up with water alone. A winter morning when a drum is brought in from the outdoor store and mixed straight into cold water is one example. A hazy layer sits at the bottom of the drum, and soon after the sump is filled a cream-like layer forms on the surface.

Heat and cold weaken the emulsion by different routes. Nonionic emulsifiers change which they prefer, water or oil, with temperature; the temperature at which they switch is the phase inversion temperature, or PIT, and emulsion stability changes sharply near it. So when fluid temperature climbs well above the design point, the emulsifier leans towards the oil and holds the droplets less firmly (Shinoda & Saito, 1969). One industry specialist lists temperatures above about 71°C, or 160°F, as a cause of coolant failure, and heat also speeds evaporation, so the ions of hypothesis 4 build faster (Burke, 2023).

On the cold side, poor mixing in cold water, a concentrate separated by freezing and thawing, or uneven initial mixing make an emulsion that is weak from the start (Burke, 2023). Such a charge forms a cream layer soon after filling or gives way easily to the other causes. The mix ratio does the same: topping up evaporation losses with water alone thins the mix until there is too little emulsifier to coat the droplets (Burke, 2023; HSE, 2021).

A thermometer and a refractometer give the first answers. Measure sump temperature weekly and consider cooling when it runs well above room temperature, and check concentration weekly with a refractometer against the supplier's range (HSE, 2021). If splitting clusters in summer or after heavy grinding, look at temperature; if it appears only in fresh charges made in winter, look at how the concentrate is stored and mixed. Layers or haze at the bottom of a concentrate drum suggest it has frozen.

When fluid temperature stays high, fit a chiller or a larger sump to spread the heat (HSE, 2021). Store concentrate where it cannot freeze, mix in the order and at the water temperature the manufacturer specifies, and where possible use a proportioning mixer for a steady concentration. Top up not with water alone but with a mix below the recommended concentration (HSE, 2021). The American standard ASTM E1497 can serve as the reference for selection, storage, dispensing and maintenance as a whole (ASTM International, 2023).

Hypothesis 3: falling pH

It happens in fluid whose pH has gone uncorrected for a long time, in systems where amines evaporate away in dry conditions, and on lines machining steel strip coated with rust-preventive oil. Acids released by microbes pull pH down too, and those are taken up under hypothesis 5 (Burke, 2023). Take a line that has machined oiled steel strip for months. The pH on the log edges down week by week, and from one week a cream layer settles on the surface and splits again soon after stirring.

Most emulsion and semi-synthetic coolants are designed for their emulsifiers to work at pH 9.0 to 9.5 (Burke, 2023), and amines are the main components that hold this alkalinity (Brutto, 2013). Fatty acid soaps, the anionic emulsifiers, sit on the droplet surface carrying a negative charge in alkaline conditions, but as pH falls they take up hydrogen ions and turn back into uncharged fatty acids. Without charge they dissolve into the oil, the force that keeps droplets apart weakens, and the droplets merge and grow. Not every small drop does harm, though. According to industry guidance, a fall from 9.3 to about 8.9 does little damage, but below 8.0 the emulsifier package may be damaged beyond recovery (Burke, 2023).

A pH meter can settle this, but how it is used decides the result. Measure pH weekly on the same day as concentration and compare it with the supplier's range (HSE, 2021); pH paper is hard to read across a narrow band such as 9.0 to 9.3 and loses sensitivity in humid air, so use a meter calibrated at two points, 7 and 10 (Burke, 2023). Watch the trend rather than single readings. A slow fall over weeks points to amine loss or microbes; a sharp drop within days points first to microbial growth or acid getting in.

Agree the acceptable range with the manufacturer, and raise pH with an amine-based adjuster in small doses. Strong bases such as sodium or potassium hydroxide have little buffering and overshoot easily, and too high a pH can irritate skin or cause chemical burns, so do not use them (Burke, 2023). If microbes are behind the drop, correcting pH alone treats only the symptom, so apply the measures under hypothesis 5 as well (Passman, 2009). If pH has fallen below 8.0 and the fluid has already split, consider a changeout rather than a correction (Burke, 2023).

Hypothesis 4: ion build-up

It shows in plants that mix and top up with hard mains or well water, in hot processes such as grinding where evaporation is fast, and in grinding and cast-iron work that make many fine particles (Burke, 2023). A grinder whose fresh charge loses its milky look within days, while sticky grey scum builds up inside the machine and on the filter, shows the picture. A stirred sample left to stand soon forms a cream layer.

Calcium and magnesium in hard water are divalent cations, carrying two positive charges. When they meet anionic emulsifiers such as soaps and fatty acids, they form metal soaps that do not dissolve in water, which settle or stick to chips and machine surfaces, leaving less emulsifier to coat the oil droplets. The reaction takes minutes (Burke, 2023). The higher the concentration and charge of the cations, the larger the droplets grow and the lower the surface charge, the zeta potential, falls; monovalent ions such as sodium have little effect, while divalent and trivalent ions have a large one (Zimmerman, 2003).

The trouble grows over time. Water evaporates but ions stay, so every water top-up adds cations to the sump, and freshly cut metal surfaces are highly reactive and bind anionic components the same way. In water with many dissolved salts, which conducts electricity well, the droplet charge works less well (Burke, 2023).

The checks are made in two places, the make-up water and the sump. Check the make-up water for hardness, pH and bacteria (HSE, 2021); hardness can be measured by EDTA titration under ASTM D1126 (ASTM International, 2025). Logging sump conductivity weekly shows ions building up. How well a product tolerates hard water can be confirmed with the supplier through hard-water stability results such as the German standard DIN 51367 (DIN, 1991).

Where the water is hard, mix and top up with water treated by reverse osmosis or ion exchange. One industry specialist sets a target of total dissolved solids, the total salts dissolved in the water, below 20 mg/L. A salt-regenerated softener only swaps calcium for sodium and does not cut chloride, and a badly set backwash can even raise it. Very pure water loses the defoaming effect of calcium, so foam may rise. Remove chips and fines within 48 hours and pass all of the fluid through the filter (Burke, 2023).

Hypothesis 5: microbial attack

Microbes grow in sumps where circulation often stops, in sumps where an oil layer covers the surface and shuts out oxygen, and in systems with dead ends or stagnant runs in the piping (HSE, 2021; McGuire, 2026). Think of the first start-up after a holiday, when a sweaty-sock smell rises from the sump, fungal clumps come out on the filter, and a thick cream layer settles on the surface.

Water-mix coolant is a good place for bacteria and fungi to grow, and microbes use some of its components as food (Burke, 2023; Saha & Donofrio, 2012). Fatty acid soaps are among their favoured nutrients (Passman, 2009). In a 2025 experiment, a coolant inoculated with microbes lost 34.15% of its tall oil fatty acids and 26.82% of its triethanolamine within nine days; over 28 days the magnitude of the droplet surface charge, the zeta potential, fell from 51.9 mV to 35.4 mV, and fungal threads formed networks that made the oil droplets clump. In the same experiment pH fell from 8.81 to 7.15 (Shen et al., 2025). The organic acids microbes release pull pH down and lead on to the damage of hypothesis 3 (Burke, 2023; Shen et al., 2025). Inside biofilms on machine walls and pipes, the local pH can drop below 7 even while the bulk sump reads 8.5 to 9.0 (Passman, 2009).

Smell and slime give the first warning. A rotten-egg or sweaty-sock smell, a slimy film on the sump walls, and fungal mats that block filters and pipes are the typical signs (Burke, 2023). Put numbers on it weekly with dipslides, coated slides dipped in the fluid and incubated to count microbes; 10⁴ cfu/mL, live cells per mL, or more means microbial growth, and 10⁶ or more means poor control (HSE, 2021). Dipslides take a day or more, so for a quick read use the ASTM E2694 method, which measures ATP, an energy molecule found only in living cells. The test takes about five minutes and shows total microbial load without separating bacteria from fungi (ASTM International, 2021).

Biocide is the last resort. Correct concentration, pH, tramp oil, metal contamination, temperature and circulation first, and only if growth continues add biocide at the supplier's recommended dose. At 10⁶ cfu/mL or more, act at once, which usually means draining and cleaning the system. If dead ends or stagnant runs remain in the piping after cleaning, biofilm returns, so remove them or flush them regularly (HSE, 2021).

Testing the hypotheses

For sorting the five hypotheses, a single bottle of settled sample tells the most. Stir the sump, fill a clear container and leave it a day: if the top layer mixes back on stirring it is creaming, and if an oil layer stays apart it is splitting (Deluhery & Rajagopalan, 2005). If the colour, viscosity and smell of the floating oil match a machine oil and lube top-up has risen, deal with hypothesis 1 first (Burke, 2023; HSE, 2021).

Next come the numbers measured that same day. Concentration or fluid temperature out of range points to hypothesis 2, and pH below 8 to hypothesis 3, with hypothesis 5 in view as well if there is odour too (Burke, 2023; HSE, 2021). Both take only minutes with a refractometer, a thermometer and a pH meter.

If every number is in range but splitting continues, move on to the slower measurements. High make-up water hardness, or steadily rising sump conductivity with metal-soap scum, points to hypothesis 4 (ASTM International, 2025; Burke, 2023); odour and biofilm together, with high dipslide or ATP readings, point to hypothesis 5 (ASTM International, 2021; HSE, 2021).

A fresh mix standing beside the sump sample in the same kind of container makes a useful control. If even the fresh mix soon forms a cream layer, look at the concentrate, the water or the mixing; if the fresh mix holds while only the sump fluid splits, look at oil, ions or microbes picked up in service. After a fix, keep measuring concentration, pH and microbes weekly (HSE, 2021), and close the case only once the readings stay in range over several checks.

The hypotheses can overlap. Industry guidance finds that around three failure mechanisms are often present at once on site (Burke, 2023). One chain runs from a tramp oil layer blocking oxygen, to anaerobic microbes making acid, to pH falling and the emulsifiers failing (Burke, 2023; McGuire, 2026; Passman, 2009). In that case deal first with oil removal, then concentration and pH, then microbes, and for a coolant that has already split badly, consider cleaning out and replacing it rather than rescuing it with additives (Burke, 2023; HSE, 2021).

Limits and cautions

Control ranges for concentration, pH and hardness differ by product, so take the manufacturer's technical data and safety data sheet, the SDS, as the reference. Synthetic coolants, which use few or no emulsifiers, are made to reject tramp oil, so oil floating on them is usually machine oil (McGuire, 2026).

Wear gloves and eye protection as the SDS directs when sampling or handling pH adjusters and biocides. Mist from heavily contaminated coolant is linked to respiratory disease, so limit exposure while draining and cleaning (HSE, 2021; NIOSH, 1998). Dispose of spent fluid under the applicable rules. If splitting returns soon after a changeout, or the system is large, such as a central supply, it is safer to leave the root-cause analysis to a specialist.

Sources

  1. ASTM International. (2021). ASTM E2694-21, Standard test method for measurement of adenosine triphosphate in water-miscible metalworking fluids. https://store.astm.org/e2694-21.html
  2. ASTM International. (2023). ASTM E1497-23, Standard practice for selection and safe use of water-miscible and straight oil metal removal fluids. https://store.astm.org/e1497-23.html
  3. ASTM International. (2025). ASTM D1126-17(2025), Standard test method for hardness in water. https://store.astm.org/d1126-17r25.html
  4. Brinksmeier, E., Meyer, D., Huesmann-Cordes, A. G., & Herrmann, C. (2015). Metalworking fluids: Mechanisms and performance. CIRP Annals, 64(2), 605–628. https://doi.org/10.1016/j.cirp.2015.05.003
  5. Brutto, P. E. (2013, August). Amines 101 for metalworking fluids. Tribology & Lubrication Technology. https://www.stle.org/files/TLTArchives/2013/08_August/Technical_Analysis1.aspx
  6. Burke, J. M. (2023, April). Understanding and controlling water dilutable metal removal fluid failure. LUBE Magazine, 174, Lube-Tech No. 145. https://www.lube-media.ambrit.com/wp-content/uploads/Lube-Tech-145-Understanding-and-controlling-water-dilutable-metal-removal-fluid-failure.pdf
  7. Deluhery, J., & Rajagopalan, N. (2005). A turbidimetric method for the rapid evaluation of MWF emulsion stability. Colloids and Surfaces A: Physicochemical and Engineering Aspects. https://doi.org/10.1016/j.colsurfa.2004.12.001
  8. Deutsches Institut für Normung [DIN]. (1991). DIN 51367:1991-08, Testing of metal working fluids; determination of the stability of emulsified metal working fluids in hard water. https://www.din.de/en/wdc-beuth:din21:1734393
  9. Health and Safety Executive [HSE]. (2021). MW5, Managing fluid quality: COSHH essentials for machining with metalworking fluids. https://www.hse.gov.uk/pubns/guidance/mw5.pdf
  10. McGuire, N. (2026, March). Can tramp oil go from a bug to a feature? Tribology & Lubrication Technology. https://www.stle.org/files/TLTArchives/2026/03_March/Feature.aspx
  11. National Institute for Occupational Safety and Health [NIOSH]. (1998). Criteria for a recommended standard: Occupational exposure to metalworking fluids (DHHS [NIOSH] Publication No. 98-102). https://www.cdc.gov/niosh/docs/98-102/
  12. Passman, F. J. (2009, August). Microbial problems in metalworking fluids. Tribology & Lubrication Technology. https://www.stle.org/files/TLTArchives/2009/08_August/Cover_Story.aspx
  13. Saha, R., & Donofrio, R. S. (2012). The microbiology of metalworking fluids. Applied Microbiology and Biotechnology, 94, 1119–1130. https://doi.org/10.1007/s00253-012-4055-7
  14. Shen, Y., Zhang, W., Wu, L., Dong, Y., Guo, G., Dong, L., & Guo, Z. (2025). Microbial proliferation deteriorates the corrosion inhibition capability, lubricity, and stability of cutting fluid. Frontiers in Microbiology, 16, 1522265. https://doi.org/10.3389/fmicb.2025.1522265
  15. Shinoda, K., & Saito, H. (1969). The stability of O/W type emulsions as functions of temperature and the HLB of emulsifiers: The emulsification by PIT-method. Journal of Colloid and Interface Science, 30(2), 258–263.
  16. Zimmerman, J. B. (2003). Formulation and evaluation of emulsifier systems for petroleum- and bio-based semi-synthetic metalworking fluids [Doctoral dissertation, University of Michigan]. https://deepblue.lib.umich.edu/handle/2027.42/123762

Product and site consultation

Leave your company, site and the product groups in use. Our engineer will get in touch to go through your site conditions and concerns. Materials, equipment, operating conditions and downstream steps show what should change and what can stay. If nothing needs to change, we will say so.

Contact us
After you contact us
  1. 01
    Site setup

    We confirm process conditions and needs first.

  2. 02
    Spec design

    We set the product spec and selection criteria.

  3. 03
    Comparison

    We compare matching candidates and propose.

  4. 04
    Tailored supply

    We supply on time and manage the stock.

  5. 05
    Site check

    We visit, check conditions of use and record.

  6. 06
    Use management

    We track use and change cycles to catch issues.

  7. 07
    Regular analysis

    We analyze the records and send a yearly report.