Breakdown in Liquid Dielectrics

Gases are clean and self-restoring, but they are sparse, and sparse media insulate only so well per millimetre. When a designer needs to pack a great deal of insulation into a small, heat-soaked volume — the interior of a power transformer, a cable joint, a capacitor — a liquid earns its place. Mineral (transformer) oil, the workhorse, brings three advantages at once. It has a high dielectric strength, several times that of air at the same spacing. It is a coolant, carrying the windings' heat to the tank walls by convection — a role no solid or gas can match in the same package. And it largely self-heals: after a small discharge the liquid flows back into the disturbed region and recovers most of its strength, unlike a solid, which is permanently scarred.

Beyond mineral oil, the family includes synthetic and natural ester fluids (favoured now for their biodegradability and high moisture tolerance), silicone oils for fire-safety, and specialised liquids for pulsed-power and capacitor use. All share the same essential physics, and the same essential weakness: their performance is governed far less by the chemistry of the pure liquid than by what is dissolved or suspended in it.

The central fact of liquid insulation is a gap of nearly two orders of magnitude. A pure, degassed, carefully filtered liquid measured in a tiny laboratory gap can withstand fields of the order of \(1~\mathrm{MV/cm}\) (\(100~\mathrm{kV/mm}\)) — comparable to a good solid. But the commercial oil in a transformer, in a practical centimetre-scale gap, breaks down at only a few tens of \(\mathrm{kV/mm}\) at best, and often far less once it has aged. The difference is entirely the impurities: water, fibres, metal particles, oxidation products and dissolved gases that a real liquid inevitably carries.

Because the pure-liquid value is almost never reached in practice, the engineering question is not "how strong is this oil?" but "what is the weakest defect inside it, and how does that defect grow into a breakdown?" Three theories answer that question for three kinds of defect, and a real breakdown is usually some combination of them.

For a genuinely pure liquid, breakdown is electronic, and the picture is a direct descendant of the gas avalanche of Chapter 4. An electron injected from the cathode (by field emission at very high stress) is accelerated by the field and gains energy \(eE\lambda\) between collisions with the closely packed molecules. If the gap is large enough and the field high enough that the energy gathered reaches the liquid's ionization threshold, the electron ionizes a molecule, and an avalanche grows much as it would in a gas:

Because a liquid is so dense, the mean free path \(\lambda\) is extremely short, so the field needed to deliver the ionization energy in one path is enormous — which is exactly why the intrinsic strength is so high, around \(1~\mathrm{MV/cm}\). This electronic mechanism sets the ceiling a liquid could in principle reach. In any practical liquid, however, one of the impurity mechanisms intervenes at a far lower field, and the intrinsic limit is never seen. The remaining two theories describe how.

Real oil carries tiny solid particles — cellulose fibres shed by paper insulation, metal flakes from windings, dust. A particle of permittivity \(\varepsilon_2\) suspended in a liquid of permittivity \(\varepsilon_1\), sitting in a non-uniform field, feels a net force. For a small spherical particle of volume \(v\), the dielectrophoretic force is:

The sign of the bracketed factor decides everything. When \(\varepsilon_2 > \varepsilon_1\) — true for a water droplet (\(\varepsilon_2 \approx 80\)) or a conducting fibre — the force points up the field gradient, dragging the particle into the highest-field region, typically the inter-electrode axis. There the particles, polarised end-to-end, attract one another and link into a chain. As more arrive, the chain lengthens into a continuous, relatively conducting bridge spanning the gap, and breakdown follows along it at a voltage far below the intrinsic value.

This is why filtering oil to remove particulates, and keeping cellulose fibres out of the oil stream, raises the breakdown voltage so dramatically — often the single most effective maintenance action. Conversely, a gas bubble (\(\varepsilon_2 \approx 1 < \varepsilon_1\)) feels the opposite force and is pushed toward the low-field region — yet bubbles cause breakdown anyway, by the quite different route of the next section.

The most important practical mechanism in service is the gas bubble. Bubbles form readily: from dissolved gas coming out of solution, from local boiling at a hot spot or a small discharge, from vapour at electrode micro-points, or from gas pockets trapped on surfaces. Once present, a bubble is dangerous for two compounding reasons.

First, the field inside the bubble is higher than the field in the surrounding liquid. For a spherical cavity of permittivity \(\varepsilon_2\) (gas, \(\approx 1\)) in a liquid of permittivity \(\varepsilon_1\), the internal field is:

Second, the gas inside has a far lower dielectric strength than the liquid. The combination is fatal: the bubble's interior, stressed above the liquid's average and far weaker, ionizes first. The little internal discharge heats and vaporises the surrounding liquid, the bubble elongates into a filament along the field, the discharge extends, and a channel propagates across the gap. The whole gap fails, but the bubble was the seed.

This explains why degassing the oil and pressurising the tank (which suppresses bubble formation) both raise the breakdown strength, and why a transformer must never be energised while gas is still evolving after a fault. It also connects forward to Chapter 9: the field-in-a-cavity formula above is exactly the one that governs partial discharge in the voids of a solid dielectric.

Pulling the theories together, a short list of contaminants accounts for almost all the loss of strength in service oil — and for the maintenance regime that fights them.

Moisture is the worst offender. Even a few tens of parts per million of dissolved water sharply lowers the breakdown voltage, and free water (above saturation) is catastrophic, because water droplets are high-permittivity particles that bridge readily. Keeping oil dry — typically below \(10\text{–}20~\mathrm{ppm}\) — is paramount.

Particles and fibres provide the bridge of §8-4; cellulose fibres are especially harmful because they absorb water and become conducting. Dissolved gases seed the bubbles of §8-5. Oxidation and ageing products (acids, sludge) raise the loss angle and conductivity, accelerating further decay. And the geometry matters too: a larger stressed volume or electrode area is statistically more likely to contain a weak defect, so big gaps test weaker than small ones — a weakest-link effect that makes laboratory values optimistic for full-size apparatus.

Because the strength of oil is a cleanliness property, it is monitored by routine testing rather than assumed from a datasheet. The standard breakdown test, IEC 60156, fills a cell with the oil under test, immerses a pair of polished electrodes \(2.5~\mathrm{mm}\) apart, and raises the voltage at \(2~\mathrm{kV/s}\) until breakdown; the reported value is the mean of several successive breakdowns. New, dry, filtered oil typically withstands well above \(30~\mathrm{kV}\) (often \(60\text{–}70~\mathrm{kV}\)) across that gap, while a service minimum of around \(30~\mathrm{kV}\) is a common red line below which the oil must be reconditioned.

When oil fails the test it is reconditioned rather than discarded: filtering removes particles, vacuum dehydration and heating drive off water and dissolved gas, and adsorbent treatment (Fuller's earth) strips out acids and oxidation products, restoring much of the original strength. Alongside the breakdown test, oil is monitored by dissolved-gas analysis — the gases evolved by incipient faults reveal arcing, overheating or partial discharge long before failure — and by loss-angle and acidity measurements, tools we return to in Part 6 on diagnostics.

For each item, first name the idea it tests — the role of liquids, pure-vs-commercial strength, the electronic, particle or bubble theory, the degrading factors, or oil testing — then work it through. Difficulty rises down the list.