What Dissolved Oxygen Does Inside a Sealed System

Why this matters

A closed loop is not protected because it is full of water. It is protected because the water in it ran out of oxygen weeks after the fill and never got any more. Once you can put a number on how much oxygen a fill actually contains, a whole class of argument ends: the sludge in a system is a receipt for oxygen that entered, and if there is more sludge than one fill could possibly have produced, the loop has been breathing. That is arithmetic, not opinion, and it settles the conversation with an owner who insists the system has never leaked.

Before you drain a loop to look at this

Read the loop pressure and temperature before opening anything, and do not drain a system above about 140 F. Hot water leaving a drain valve flashes to steam at the opening and the scald reaches further than the hose does. Let the isolated section cool, or drain through a hose run into a container with the discharge end secured so it cannot whip out of the container when flow starts.

Draining a loop in an occupied building takes the heating or cooling with it. Before you commit, confirm nothing on that loop is exposed to freezing while it is empty, and if any part of it serves a coil in an outside air stream, either keep that section charged or blow it out; a drained coil with residual water in an outdoor air stream is a burst waiting for the next cold night.

Do not dry and brush corrosion product. Dried magnetite fines are a respirable dust, and in an older system that product can carry metals picked up from solder and fittings over decades. Weigh it wet and record that it was wet, or dry it in a closed vessel and keep it wetted down when you handle it. Gloves and eye protection to the class named in Section 8 of the loop fluid's safety data sheet, which 29 CFR 1910.1200 requires be available to you.

The drained fluid is treated water. What may go to a sanitary drain, and at what concentration, is the local sewer authority's call, made before the valve opens rather than after the container is full.

Why a sealed system works at all

Steel dissolving in water is an electrical circuit. Iron gives up electrons at the anode, and those electrons have to be accepted somewhere or the whole reaction stalls within moments. In ordinary system water at normal pH, dissolved oxygen is the acceptor. Remove it and the circuit opens; steel in genuinely deaerated water at moderate pH corrodes at a rate that is hard to measure over decades.

That is the entire theory of the closed loop. It is not that the water is treated, though it usually is. It is that the water has nothing left to react with.

Two qualifications belong in the same breath. Hydrogen ion is a second acceptor, so an acidic loop keeps corroding with no oxygen at all, which is why a degraded glycol charge is dangerous in a system that has been sealed for years. And oxygen does not have to be evenly distributed to do damage: a difference in oxygen concentration between two points on the same piece of metal is itself a driving voltage, which is the mechanism behind attack under deposits. The library covers that morphology separately and it is worth reading alongside this.

What one fill actually brings in

Cold water from a main, at atmospheric pressure, carries dissolved oxygen on the order of 10 to 11 mg per litre. That figure is for cold water; solubility falls steeply as temperature rises, to a small fraction of it near boiler temperatures.

That temperature relationship shows up differently at each end of the range and it is worth being able to name both. Fill a heating loop cold and bring it up to temperature and the oxygen comes out of solution as free gas, which is why a newly filled system burps at its vents for the first days and then stops. Fill a chilled water loop and it never gets warm enough to drive the gas out, so the oxygen stays dissolved, stays available, and is spent slowly on the metal instead of leaving through a vent. The heating loop announces its oxygen; the chilled loop quietly uses it.

Now put a mass on it. Iron and oxygen combine in fixed proportions. Going to magnetite, the black oxide you actually recover from a hot closed loop, the reaction consumes about 64 g of oxygen for every 167.6 g of iron, so one gram of oxygen can convert roughly 2.6 g of iron. Going to the red oxides you get in an aerated system the ratio is nearer 2.3. Call it about two and a half grams of iron per gram of oxygen and hold that number.

The consequence is smaller than most techs expect. A 300 gallon loop is about 1,136 litres, so a full fill of cold water carries roughly 11.4 g of oxygen, and that oxygen can convert roughly 30 g of iron. Thirty grams. Spread over the entire wetted steel surface of a building's heating system, that is a film you would struggle to find.

Reading the product before you weigh it

Colour narrows the question before any arithmetic. Black magnetite forms where oxygen is scarce, which is the normal condition of a working closed loop, and a modest quantity of it is what a healthy old system looks like inside. Red brown product means the iron was oxidised further, which takes more oxygen than a starved system has. Recovering red rust from a system described as closed is a finding on its own.

Texture matters too. A fine suspension that clouds the drained water and settles slowly is ordinary. Hard, adherent, layered deposit is a different problem and belongs with the scale and deposit articles rather than here.

One question, two loops that answer it oppositely

The question: does the corrosion product recovered from this loop exceed what a single fill's dissolved oxygen could have produced?

Loop A. Hydronic heating, 300 gallons, twelve years old, opened for a circulator replacement. What came out of the strainer and the low point, collected wet, was about 40 g of black fines.

Magnetite is roughly 72% iron by weight, so 40 g of product carries about 29 g of iron. The fill budget calculated above is about 30 g of iron. The recovered product is very close to a single fill's worth, and the loop has had one fill in twelve years. Nothing here needs explaining. Recharge the inhibitor for the fresh metal the pump work exposed and close it up.

Loop B. Hydronic heating, 400 gallons, ten years old, drained for investigation after repeated pinholes. Recovered product, wet, about 2 kg of black sludge.

Same arithmetic. 400 gallons is about 1,514 litres, so a fill carries roughly 15.1 g of oxygen, which converts roughly 40 g of iron. Two kilograms of magnetite carries about 1,448 g of iron. Divide: about 36 fills.

Thirty six system volumes of oxygenated water have gone through that loop. At 400 gallons each that is roughly 14,600 gallons over ten years, near 1,500 gallons a year, on the order of four gallons a day. Nobody in that building ever saw four gallons a day leave, because an automatic fill valve replaced it silently and the pressure gauge never moved.

The two loops are the same kind of system, the same age bracket, the same fluid. The difference in recovered product is a factor of fifty, and the arithmetic converts that into a number the owner can act on: not "you have a corrosion problem" but "this system has taken in about thirty six times its own volume of fresh water and we are going to find out where."

The test only works in one direction

This is the part that gets misused. Most of the corrosion product in a system is adhered to surfaces and never comes out in a drain, so what you recover is a floor on what is actually there, not a measurement of it.

That makes the finding asymmetric. A large recovered mass proves a large amount of oxygen entered, because the product cannot exist without it. A small recovered mass proves very little, because the product may simply be stuck to the pipe. Loop A's result is consistent with a tight system; it does not certify one. Loop B's result is proof.

Three other assumptions sit under the count and each pushes the same way. Some of the oxygen goes to copper rather than iron. Some is spent oxidising the inhibitor, turning nitrite into nitrate. Some leaves through the vents on first heat up without reacting with anything. Every one of those means the true number of system volumes is higher than your count, not lower, so the count is conservative and you can hand it over as such. Corrosion mechanisms that do not need oxygen, acid attack and the microbiologically driven kind, produce iron product too and push the other way, which is why the count is an order of magnitude tool and not a meter reading.

Checking your own count

Reconcile it against the makeup meter if one exists, because the two are independent estimates of the same thing. Loop B's count predicts on the order of 1,500 gallons a year; if the meter has been in place for a year and reads 40, either the count is wrong or the meter is downstream of the actual water path, and both are worth knowing.

Where there is no meter, fit one at the refill and let the next year settle the argument properly. A count from recovered sludge tells you what has already happened over a decade; a meter tells you what is happening now, and only the second one confirms a repair worked.

State the assumption set with the number whenever you report it. "About 36 system volumes, and that is a floor, because most of the product stays on the pipe" is a defensible sentence. "Thirty six fills" on its own will be repeated back to you as a measurement.

References

• 29 CFR 1910.1200, Hazard Communication, for the safety data sheet whose Section 8 sets glove and eye protection for handling drained treated fluid and corrosion product
• Local sewer authority discharge limits for drained loop water containing inhibitor
• ASHRAE Handbook, Fundamentals, for gas solubility in water as a function of temperature
• See related: Pitting, Crevice and Uniform Corrosion, which owns the differential aeration mechanism referenced here; How Scale Forms and What It Actually Costs