The Dielectric Union and What It Does Not Fix

Why this matters

The dielectric union is the reflex answer to dissimilar metals in a water line, and it gets credited with far more than it does. It is a real device that solves a real and narrow problem: it breaks metallic continuity at one point. Almost everything a shop expects it to prevent, it prevents only when several other conditions happen to be true at the same time.

That gap is why a tech installs one, the corrosion continues, and the customer concludes the shop does not know what it is doing. This card is mostly about the conditions where it does nothing, because that is the part nobody is taught.

What it actually does

Two dissimilar metals corrode galvanically when three conditions are all present: an electrical connection between them, an electrolyte bridging them, and a difference in potential between the two metals. A dielectric union removes the first condition at one location by putting an insulating element between the two threaded halves.

That is the entire mechanism. It does not change the water, it does not change the metals, and it does not reach beyond the fitting. Every limitation below follows from that one sentence.

Condition one: the metals are still connected somewhere else

An insulator in the middle of a circuit does nothing if the circuit closes elsewhere, and in a building it usually does.

Through the electrical bonding system, which is required. Metal water piping in a building is commonly bonded as part of the electrical system, and NFPA 70 (the National Electrical Code) Article 250 governs that bonding. Where a fitting interrupts continuity in bonded piping, the bonding path is restored with a listed bonding jumper across the fitting, which means the two metals remain electrically connected by design.

This is not a defect to be corrected. Removing or omitting a required bonding conductor to make your dielectric fitting work electrically is a serious safety error: that conductor is part of the path that clears a fault and keeps metal piping from becoming energized. If you have to choose between galvanic isolation and required bonding, bonding wins every time, and the corrosion problem is solved another way. If you are working inside an enclosure to inspect a bonding connection, de-energize, lock and tag under 29 CFR 1910.333(b)(2), then prove dead with the live-dead-live sequence in NFPA 70E-2021, 120.5, checking your meter on a known live source before and after.

Through metal that has nothing to do with plumbing. A shared strut or hanger touching both pipes. A pump body with both metals bolted to it. A metallic flexible connector or a braided hose whose braid spans the joint. An equipment cabinet both are fastened to. Any of these closes the circuit around your union, and none of them are visible if you are looking only at the fitting.

Condition two: the water is a path of its own

The insulator sits in the metal, not in the water. Conductive water flows continuously through the union and around the insulating element, so the fluid provides a path in parallel with the one you just broke. How much that matters depends on how conductive the water is and how long the insulating section is: a short insulator in aggressive, high-conductivity water leaves a much better parallel path than a long non-metallic spool in soft water.

There is a second water-borne route that no fitting at the joint touches at all. Dissolved copper released upstream travels with the flow and deposits on more active metal downstream, and every deposit becomes its own tiny cell that pits the base metal. That mechanism does not need the two metals to be joined at any point, so an isolating fitting is irrelevant to it. The controls for it are flow direction, with the more noble metal placed downstream, and water chemistry.

Condition three: it was never galvanic in the first place

This is the most common reason a union changes nothing, and it is a diagnostic failure rather than a hardware failure. Several corrosion mechanisms look similar at a fitting and none of them are addressed by isolation:

• Oxygen attack, from aeration, frequent make-up water or ingress into a closed system. Produces general rust and tuberculation on ferrous components anywhere in the system.
• Chloride pitting, from water chemistry or a chloride source. Deep, localized holes in an otherwise clean surface.
• Low pH or acid attack, including from degraded glycol in a treated loop.
• Erosion-corrosion, from velocity and turbulence, classically just downstream of a restriction, an elbow or a poorly reamed cut. The signature is directional, smooth and undercut rather than crusty.
• Microbially influenced corrosion, under deposits in low-flow or dead-leg sections.

Before specifying isolation, name the mechanism. Metal loss concentrated on one side of a dissimilar joint with the other side clean supports a galvanic story. Loss distributed through the system, or on both metals, or with a directional wash pattern, does not, and an isolating fitting installed against those mechanisms is a fitting bought for nothing.

Condition four: the union becomes its own problem

The fitting introduces failure modes the plain connection did not have.

Its insulating element and gasket are materials in the system with their own temperature and chemical limits, and they age. A degraded liner can fail as a leak or can lose its isolating function while still holding pressure, which means the fitting is now doing nothing and looks fine. The internal profile also collects scale and debris, so on hard water or a dirty system it becomes a restriction and a deposit site, and deposits under a restriction are where localized attack starts.

It can also be defeated at install. A conductive joint compound run across both halves, a metal filing bridging the insulator, or an insulating sleeve shaved during make-up all leave you with a fitting that looks correct and conducts. None of that is visible after assembly.

Worked case: two failures, one fitting, no change

A shop inherited a recurring leak at the connection between a water heater and the building's galvanized piping. The prior contractor had replaced the failed nipple and, on the second visit, installed a dielectric union, which is the standard reflex and was not an unreasonable move.

The record gave two intervals around the union. The original connection failed at about 26 months. After the dielectric union went in, the next failure at the same point came at about 24 months. Two intervals is a thin basis for a trend and the shop said so, but 24 against 26 months is the wrong direction to argue that the union helped, and there was no other change in that window to explain it away.

They looked for parallel paths first, because that is the cheapest check. Two were live. The piping was bonded, correctly, with a jumper across the union, so the two metals were electrically connected by design and were going to stay that way. Separately, both pipes were clamped to a common metal support within arm's reach of the joint, a second path nobody had noticed because it was not plumbing.

Then they questioned the mechanism itself, which is the step that had never been taken. The failed nipple showed general thinning with heavy tuberculation rather than concentrated loss on one side of the joint, and the same rust product was present at fittings elsewhere in the run, well away from any dissimilar-metal connection. That pattern points at oxygen and water chemistry, not at a galvanic couple at this fitting. The union had been installed against a mechanism that was not running.

The corrective work followed the actual finding: the water chemistry and the make-up water rate were addressed, the failed section was replaced in a material appropriate to the water rather than matching what came out, the flow-direction relationship between the metals was corrected so the more noble metal sits downstream, and the shared support was isolated where it spanned the two metals, which costs almost nothing and removes one path. The bonding jumper stayed exactly where it was.

The shop was careful about what it claimed afterward. The system has run past the point where the prior two failures occurred, which is evidence and not proof, and they wrote in the record that the galvanic hypothesis had been tested and did not fit rather than that it had been fixed. The general lesson is the one that had cost two visits: a dielectric union installed without naming the mechanism is a guess wearing hardware.

When it is the right call

None of the above makes the fitting useless. It earns its place in three situations, and it is worth knowing them precisely so you neither over-install it nor refuse it when it is correct.

• The equipment manufacturer requires it. Some appliance and heater installation instructions specify dielectric connections, and following the manufacturer's installation instructions is what keeps the equipment warranty intact. Read the instructions rather than the tradition.
• The transition is genuinely isolated otherwise. Where the two metals are not bonded together elsewhere, are not sharing a support, and the water is not aggressive, the fitting removes the one condition it is built to remove and it works.
• You want a defined, inspectable interface. A purpose-made transition puts the joint at a known point that can be checked and replaced, rather than leaving one metal cut directly into another somewhere behind a wall. Even where the isolation is imperfect, a designed interface is easier to service than an accidental one.

Where you install one, a longer non-metallic transition section gives you more than a short one, because it lengthens the parallel water path rather than only breaking the metal path.

How to check that yours is doing anything

• Confirm the isolation at install, before the jumper. With the system out of service, meter across the two halves and expect an open reading. Once a required bonding jumper is installed, this measurement no longer tells you about the fitting, so it is done at the right moment or not at all. Do not remove an installed bonding conductor to run a test.
• Walk the two pipes and hunt for parallel metal. Supports, hangers, strut, cabinets, pump bodies, braided connectors. This takes a few minutes and finds the reason more often than anything else on this list.
• Read the corrosion before you specify the fix. One side of the joint eaten and the other clean supports galvanic attack. General loss through the system, or a directional wash pattern, sends you to chemistry or velocity instead.
• Put the fitting on the inspection list. It is a wear item with an aging liner, not permanent hardware, and it should be looked at on the same cadence as any other serviceable component in that line.

References

• NFPA 70 (National Electrical Code), Article 250, bonding of metal water piping and equipment
• 29 CFR 1910.333(b)(2), selection and use of work practices: de-energizing and lockout or tagging before electrical work
• NFPA 70E-2021, 120.5, verification of an electrically safe work condition, including the instrument check before and after the test
• Equipment manufacturer installation instructions, which may specifically require dielectric connections
• See related: Galvanic Corrosion and the Metals That Fight; Dissimilar Metals and Corrosion; How to Join Two Materials That Do Not Like Each Other; The Compatibility Question on a Mixed-Material System