Misted sealed units and why they go

What is actually inside the edge of a sealed unit

Cut a sealed unit in half and the edge is a small piece of engineering with five components in about 15mm of depth.

Two or three panes of glass, held apart by a spacer bar that runs the full perimeter. Inside the spacer, a desiccant, which on a modern unit is a molecular sieve with a pore size chosen to take up water without also taking up argon. Bonded between the spacer and each pane, a thin continuous bead of primary seal. Filling the channel outboard of the spacer, a much thicker secondary seal. And in the cavity between the panes, either dry air or an inert gas fill, usually argon at 90 per cent purity or better.

The spacer itself is a thermal component as much as a structural one, and its effect on where condensation appears on the room side is set out on warm edge spacers and where condensation starts. What matters here is that the spacer is the surface both seals bond to, and that its corners are where a metal bar is joined with a key or a weld. Everything that follows happens at those bond lines.

Polyisobutylene, and the job the primary seal does every hour

The primary seal is polyisobutylene, a butyl rubber, applied as a bead perhaps 3 to 5mm wide and a fraction of a millimetre thick between the spacer and the glass. It is chosen for one property above all others: it has extremely low permeability to water vapour and to gas, lower by orders of magnitude than any of the materials used for the secondary seal.

What it does not have is strength. Butyl creeps. Put a sustained load on it and it flows slowly rather than resisting, which is why it can never be the component holding the unit together. It stays put because the secondary seal outboard of it carries the load and keeps the panes at a fixed distance, and the butyl simply fills the gap and blocks the path.

So the failure sequence is nearly always the same. The secondary seal loses stiffness, or the glass flexes more than it should, or the bond at a corner gives up a few millimetres. The butyl, which was never structural, is pulled thin or peeled off the glass along a short length. It only takes a path a few millimetres long. From that point the cavity is connected to outside air.

Vapour pressure, and the direction water always travels

It helps to understand what the seal is holding back, because it is not liquid water and it is not really air.

Water vapour exerts its own partial pressure, independent of the air it is mixed with, and it moves from high pressure to low. A newly manufactured sealed unit has a cavity that has been dried by its desiccant to a dew point well below minus 40 degrees. Outside, on any ordinary day in the Severn Vale, the air holds a great deal more moisture than that. So there is a permanent, continuous pressure difference driving vapour inward across the seal, twenty four hours a day, for the whole life of the unit. It never reverses and it never pauses.

That is why a sealed unit does not fail slowly and then recover. Once a path exists the traffic is one-way. The desiccant absorbs what comes in until it cannot, and then the vapour has nowhere to go except onto the coldest glass surface available, which is the cavity face of the outer pane.

Pumping, and the breathing a sealed cavity is not supposed to do

The cavity is a fixed volume of gas with two flexible walls. Heat it and the gas wants to expand; the panes bow outward until the pressure balances. Cool it and the panes bow inward. On a sunny March afternoon you can see it: a large pane in direct sun will show a slightly distorted reflection because it is genuinely convex.

That flexing is what fatigues the edge. Every cycle puts the panes in bending, and the load lands at the perimeter where the panes are restrained, which is exactly where the two seals live. In a wall unit the daily swing is modest. Overhead it is not, and the reason is geometry: a rooflight at five to fifteen degrees is close to perpendicular to a summer sun at the middle of the day, so it collects far more energy per square metre than a vertical pane ever does. At night it faces the sky, and radiative loss to a clear sky takes the outer pane several degrees below the air temperature around it.

Where a small breach already exists, the same flexing pumps air. Contracting cavity draws air in through the breach, expanding cavity pushes it back out, and each inward stroke brings moisture that only partly leaves again. That is why a unit with a hairline path fails much faster than the size of the path suggests. It is not diffusion; it is a pump running twice a day.

Thermal cycling on a roof, in numbers

The numbers make the case better than the argument does.

A south-facing rooflight on a Cheltenham loft room can put its outer pane surface at 55 to 70 degrees on a hot July afternoon, higher again behind a dark internal blind that traps heat between the blind and the glass. That same pane, on a clear frosty January night, sits at minus 5 or lower once radiative loss to the sky is counted. That is a swing of around 70 degrees across a year and 25 to 35 degrees across a single clear spring day, when the pane goes from frosted at dawn to hot by two in the afternoon.

Glass expands about 9 micrometres per metre per degree. Over a 1.4 metre pane a 30 degree swing is roughly 0.38mm of movement in the length of the glass, and the inner and outer panes are not moving by the same amount at the same time because they are at different temperatures. The seal accommodates that difference every day. Multiply by 365 and then by twenty years and you have somewhere in the order of seven to ten thousand full cycles on a bond line a few tenths of a millimetre thick.

Ultraviolet at an upward-facing edge

Ultraviolet is the second attacker, and it works on the secondary seal rather than the primary one.

A horizontal or near-horizontal surface receives substantially more ultraviolet over a year than a vertical one, because it sees the whole sky dome rather than half of it and because it takes the summer sun at a favourable angle. Diffuse sky ultraviolet matters as much as direct: an overcast Gloucestershire day still delivers a meaningful dose, which is why the effect accumulates through a climate that is not notably sunny.

What ultraviolet does to polysulphide is embrittle the surface layer, which then crazes into a fine network of cracks. Those cracks let water and oxygen deeper into the material, the process continues inward, and the seal loses the elasticity it needs to keep holding the panes against daily movement. Once the secondary seal has stopped behaving elastically, the primary bead behind it starts taking load it was never designed for.

Where a unit is edge-capped by an aluminium retainer, the retainer shades the seal and the problem largely goes away. On a slim-sightline flush glazed unit, where the appeal is precisely that there is very little frame, the edge is out in the weather. That is a trade-off worth knowing about at the point of specifying, not afterwards.

The low edge that never dries

Third attacker, and on a shallow rooflight it is often the decisive one.

Rain runs down the glass and collects at the bottom, in the rebate, against the seal. Dust, pollen and the fine organic film that lands on any horizontal surface build up in the same place and hold water between showers. On a north-facing unit at eight degrees, in a damp vale winter, that edge can stay wet for weeks at a time.

Sustained water contact attacks all three secondary seal chemistries, silicone least and polyurethane most, and it also works on the adhesion between the seal and the glass rather than on the bulk material. Two things protect it. The unit needs a genuine fall, which is why about five degrees is the practical minimum on made-to-size glass. And the frame needs a drained, ventilated rebate with clear weep holes at the low corners, so the channel empties onto the glass below rather than becoming a shallow tank. A blocked weep hole is a small maintenance item with a large effect on service life, and it is on the list at cleaning and maintaining a rooflight.

Desiccant capacity, and the day the sieve is full

The desiccant is the reason a new unit stays clear despite the moisture that was inside it when the panes were pressed together. It has a finite capacity, sized by the fabricator against the volume of the cavity with a margin for the vapour that will diffuse in slowly across a normal service life.

The important thing about it is that it does not fail gradually in a way you can see. It absorbs, and absorbs, and stays fully effective right up to the point where it is saturated, and then it does nothing at all. That is why misting appears abruptly. A homeowner will say the unit was completely clear last winter and cloudy this one, and that is usually literally true: the desiccant crossed its capacity somewhere in between.

Two design details are worth knowing. A molecular sieve with a correctly chosen pore size takes up water while leaving the argon molecules alone, whereas a cheaper blend will adsorb argon as well and quietly drop the fill purity over the years. And a thermoplastic spacer, extruded as a continuous bead with the desiccant dispersed through the polymer, distributes it round the whole perimeter with no corner joints, which is a different geometry from a hollow metal bar with the desiccant poured into it.

The first winter of misting, and the fifth

Misting is not one appearance. It changes over years, and where a unit sits in that progression tells you how long ago the seal went.

In the first months, the cavity fogs only in the worst conditions: a cold clear morning with the sun on the glass, clearing by mid-morning as the pane warms and the moisture evaporates back into the cavity. The middle of the pane clears first. A homeowner often notices it, then does not see it again for weeks, and concludes it was nothing.

Within a year or two it is fogging on most cold days and taking longer to clear. Streaks and runs appear, because the condensed water is now enough to form drops that run down the inner face of the outer pane and dry in place. By the third or fourth year there are visible tide marks, and in a shallow unit a permanent damp zone along the low edge where water collects in the cavity and never fully leaves.

After that comes the stage nothing reverses.

Mineral bloom, and why the glass itself is the casualty

Water sitting on a glass surface for years does not simply sit there. It leaches sodium and calcium ions out of the glass, which is a soda-lime silicate, and leaves behind a silica-rich, slightly porous surface layer.

Two things follow. The leached ions form alkaline solution in the cavity, and alkaline water attacks glass considerably faster than neutral water does, so the process speeds itself up. And the dissolved minerals redeposit as the water evaporates and recondenses each day, building a white or iridescent film. That film is the permanent white bloom people see on an old failed unit, and it is not dirt on the surface: it is deposit bonded to, and etched into, the glass.

The practical consequence is that a badly failed unit cannot be brought back even in principle. Once the surface is etched, the pane is scattering light whether or not there is any moisture present. A unit at that stage looks milky on a dry summer day with no condensation anywhere in it. That is the point at which the glass has genuinely gone rather than merely misted.

Argon loss runs alongside, invisibly

The same edge seal that stops vapour coming in is what stops the argon getting out, so a unit that is misting has usually lost most of its fill as well.

The product standard for sealed units allows a loss of less than one per cent of the fill per year. At that rate a unit filled to 90 per cent is still above 80 per cent after a decade and the U-value has drifted by a few hundredths, which is immaterial. A unit with a compromised seal can lose several per cent a year, and after fifteen years the cavity is essentially air.

Going from argon to air in a 16mm cavity costs something in the order of 0.2 to 0.3 W/m2K on the glass, and it happens with no visible sign at all until the misting starts. So the room was getting colder for years before anybody noticed a haze. It also means that a unit which mists in a cold snap is not simply an appearance problem: it is a component that stopped performing to its specification some time earlier. The glazing numbers are on U-values and glazing specification and low emissivity coatings and argon fills.

Why they go in groups across one elevation

Homeowners often report that one unit misted and then two more followed within a couple of years, and treat it as bad luck. It is not.

Every unit on a roof was almost certainly glazed within days of every other one, from the same production batch, with the same seal chemistry and the same desiccant charge. They have then been exposed to the same weather for the same number of years. The only variables left are elevation and shading, and on a single slope those are close to identical. So the units are not independent items that happen to be the same age. They are one batch reaching the end of one design life together, and the one that failed first was simply the one with the smallest manufacturing margin.

The arithmetic that follows from that, including whether it makes sense to bring the others forward onto the same scaffold, is on replacing several rooflights at once.

Drilled vents and the products that claim to clear a cavity

There is a market in treatments that promise to clear a misted unit, and it is worth being straightforward about what they do.

The method is always the same underneath the branding: one or two small holes are drilled through the outer pane at the corners, the cavity is flushed or a cleaning solution is worked through it, and a small vented plug is fitted. Warm dry air can then move through the cavity, and on a unit at an early stage the visible fogging genuinely reduces for a while.

What you have afterwards is a deliberately vented cavity. The unit is no longer sealed, so the argon has gone permanently, the cavity is now connected to outside air and to the dust and humidity in it, and the desiccant is saturated and inert. The thermal performance is that of an air-filled, ventilated double skin, which is a long way below the unit’s original figure and below the current limiting standard for a replacement. It does nothing whatever about mineral bloom already etched into the glass, because that is a change to the surface rather than something suspended in the cavity. And drilling a toughened pane is not possible at all, so on a rooflight, where a toughened outer is standard, the technique frequently cannot be applied.

The honest position is that a sealed unit is a consumable component with a design life, and when it reaches the end of it the glass is renewed. That is a straightforward, planned piece of work rather than a failure of the roof.

New glass into the old frame, or a new complete unit

Once the diagnosis is confirmed, and separating a failed unit from a failed perimeter is set out on telling a failed unit from a failed installation, there are two routes.

New glass into the existing frame makes sense where the frame is sound, the mechanism works properly, the gaskets still have spring in them, the flashing kit is in good order, and the unit is broadly under fifteen years old. On a manufactured roof window the sash is designed to be re-glazed, and on a made-to-size flat roof unit with a mechanically retained bead it is often achievable too. On a structurally bonded flush unit, where the glass is the structure, it is generally not, and the answer is a new unit.

A new complete unit is the better outcome where the frame is the same age as the glass, because everything else in it is at the same point in its life. The distinction, and what each involves, is covered at can you replace just the glass in a rooflight. Either way, a replacement is also the one moment when a glass upgrade costs almost nothing in extra labour, which is argued at upgrading the glass while the roof is open.

Which roofs in this county mist first

Elevation and exposure decide the order, and the county gives a fairly consistent pattern.

South and west-facing slopes go first, everywhere. On the estates at Quedgeley, Abbeydale, Bishops Cleeve, Churchdown and Ashchurch, where developer-fitted units went in across whole streets in the same few months, the rear south-facing slopes are typically five to eight years ahead of the north-facing fronts on identical houses. In the Forest of Dean around Coleford and Cinderford, the combination of a steeper pitch, more rainfall and more wind puts more water on the low edge for more of the year, and shallow-pitched units there suffer disproportionately.

Across the Severn Vale terraces of Gloucester, Longford and Tewkesbury, the classic case is a flat rooflight over a rear addition kitchen, sitting at five to eight degrees with a kitchen underneath. High indoor humidity below, standing water on the low edge above, and a unit that was probably specified without much thought about the secondary seal. In central Cheltenham, behind the parapets at Lansdown, Montpellier and Pittville, the shallow hidden flats do the same thing, and the units on them are often older than anybody realises because nobody has looked at them from the street in forty years.

On Cotswold stone slate through Painswick, Bibury and Northleach, conservation rooflights are the case to watch. A slim steel frame with a central glazing bar produces two small panes rather than one large one, which doubles the length of edge seal per square metre of glass, and the frame conducts. Those units earn a silicone secondary seal and a warm edge spacer more than anything else in the range, and it is worth specifying both at order. Send photographs of the unit and its data plate to info@severnskylightsgloucester.co.uk, and the costs page sets out how a range by email becomes a fixed price after a survey.