Low-emissivity coatings and argon fills

Emissivity, and the one number a coating changes

Every surface above absolute zero radiates energy. Emissivity is the fraction of the theoretical maximum a given surface actually radiates, on a scale from zero to one, and for a given material it is also the fraction of incoming radiation it absorbs rather than reflects.

Ordinary float glass has an emissivity of about 0.837 in the long-wave infrared band that matters at room temperature. It is very nearly a perfect radiator, which is exactly the wrong property for the inner pane of a window. Polished metals sit at the other end: aluminium foil is around 0.05, silver lower still. A low-emissivity coating is a way of giving a sheet of glass the infrared behaviour of a metal while leaving it transparent to visible light, and it does that by including an extremely thin silver layer, on the order of ten nanometres, sandwiched between metal oxide layers that protect it and tune its optical behaviour. A sputtered soft coat reaches an emissivity of 0.03 to 0.05. A pyrolytic hard coat, fused into the glass on the float line, reaches around 0.15 to 0.20.

How much heat radiation carries across a cavity

Radiant exchange between two parallel surfaces depends on both of their emissivities combined into an effective figure, calculated as one divided by the sum of the reciprocals of each emissivity minus one.

Two uncoated glass surfaces at 0.837 each give an effective emissivity of about 0.72. Coat one of them at 0.04 and the effective figure collapses to about 0.040, because the reflective surface refuses to absorb what the other one is radiating and sends it straight back. The radiant transfer coefficient across the gap is roughly four times the Stefan-Boltzmann constant, multiplied by that effective emissivity, multiplied by the cube of the absolute temperature. At around 10 degrees Celsius that works out at about 5.1 W/m2K multiplied by the effective emissivity.

Putting numbers on an uncoated cavity against a coated one

Run those two figures and the whole subject becomes concrete.

Uncoated cavity: 5.1 multiplied by 0.72 gives roughly 3.7 W/m2K of radiant transfer. Meanwhile conduction through a 16mm argon layer at 0.017 W/mK is 0.017 divided by 0.016, or about 1.06 W/m2K. Radiation is carrying more than three times what conduction carries.

Coated cavity: 5.1 multiplied by 0.040 gives roughly 0.20 W/m2K. The radiant path has essentially been switched off, and conduction through the gas is now the dominant term by a factor of five. That reversal explains the whole order of specification: coat first, then improve the gas, then narrow or split the cavity to deal with convection. Improving the gas without a coating is close to pointless, which is why nobody sells an uncoated argon unit.

Surface numbering, and the only unambiguous way to specify a coating

Glass surfaces are numbered from the outside inward, starting at the weather face. In a double unit, surface one faces the sky, surface two is the cavity face of the outer pane, surface three is the cavity face of the inner pane, and surface four faces the room. A triple unit runs one to six on the same rule.

That numbering is how a coating gets specified without ambiguity, and it belongs on the order in that form. “Low-e on the inner pane” is not precise enough, because the inner pane has two faces and the difference between them is substantial. “Soft coat low-emissivity on surface three, spectrally selective solar control on surface two” leaves nothing open to interpretation, and a fabricator will read it immediately.

Why the low-emissivity coating belongs on surface three here

In a heating-dominated climate like Gloucestershire’s, the low-emissivity coating goes on surface three, the cavity face of the inner pane. That position is not arbitrary and it has a consequence people notice.

On surface three the coating faces the cavity and reflects long-wave heat radiating off the room-side pane back towards the room. The inner pane consequently stays warmer, which is what raises its internal surface temperature and pushes it further above dew point. Move the same coating to surface two and it would still cut the radiant exchange by a similar amount, but it would keep the outer pane warmer and the inner pane colder, which is the arrangement a cooling-dominated climate wants and precisely what a British winter does not. Solar control coatings, by contrast, belong on surface two, because their job is to reflect incoming solar energy before it is absorbed into the assembly. Two coatings, two jobs, two positions.

Surface four coatings, and why they are wrong overhead

A hard coat low-emissivity layer can be applied to surface four, facing directly into the room, because pyrolytic coatings are durable enough to be exposed. It reduces the whole-unit U-value by roughly 0.2 by cutting the radiant loss from the room-side glass surface to the room itself.

It does that by making the inner glass surface colder, which is a poor trade in any window and a bad one in a roof. The coldest surface in the room becomes colder still, condensation forms sooner and more often, and on overhead glazing water on the inside face is more visible and more of a nuisance than it is on a wall. The coating also carries a slight haze that is noticeable when you are looking directly up through it, and it is exposed to cleaning. Surface four coatings have their place in specific commercial specifications. Above a kitchen table in the Severn Vale they are the wrong product.

Coating positions in a triple unit

A triple has four cavity faces available: two, three, four and five. Two coatings is the normal arrangement, and there are two conventions.

For a heat-retention specification, coatings go on surfaces two and five, or three and five, so each cavity has one reflective face. Surface five is the direct equivalent of surface three in a double, keeping the inner pane warm. The second coating handles the outer cavity. For a unit that has to control summer gain as well, the spectrally selective solar control coating takes surface two and the low-emissivity coating takes surface five, which reaches a g-value near 0.25 to 0.30 alongside a whole-unit U-value of 0.8 to 1.0. The cost is daylight, and the arithmetic is worked through on triple glazing in a rooflight and when it pays.

Edge deletion, and the millimetres removed before sealing

A soft coat cannot run to the edge of the pane, and understanding why explains a lot about how sealed units fail.

The silver layer oxidises when it meets moisture or air. The primary butyl seal also needs to bond to glass rather than to a coating, because the bond to a metal oxide stack is weaker and creeps. So before a coated pane is made into a unit, the coating is ground or laser-removed from a border around the perimeter, typically 8 to 15mm wide, so the seal lands on clean glass. That strip is called the edge deletion, and on a very close inspection at the right angle you can sometimes see its faint boundary just inboard of the sightline.

Where edge deletion is skipped or done badly, the coating corrodes inward from the perimeter over several years, appearing as a creeping cloudy or iridescent border. That is a fabrication defect rather than a weathering one, and it is one reason the edge seal and the coating are specified together rather than as separate items.

Handling, toughening and the shelf life of coated stock

Soft coated glass is delicate before it becomes a unit. It has to be stored in controlled conditions, interleaved, and used within a limited period, because the coating degrades on exposure to humid air. A fingerprint on a soft coat is permanent.

Toughening adds a complication. Standard soft coats will not survive a toughening furnace, so a toughenable grade has to be specified when the coated pane also has to be heat treated, which in overhead glazing it very often does. Toughenable coatings sometimes need a heat treatment to reach their final optical state, so an untoughened offcut can look different from the finished pane. Hard coats survive toughening without difficulty, which is one of the few remaining reasons they are used at all. None of this is a homeowner’s problem to manage, and all of it is a reason that overhead glazing specifications get written down at order rather than agreed verbally.

What the coating does to daylight, colour and ultraviolet

A single soft coat costs a few percentage points of visible light transmittance, taking a clear double unit from around 81 per cent uncoated to 78 or 80. The visible tint is very slight, usually reading as a faint neutral or slightly green cast, and it is most obvious when a coated unit sits alongside an uncoated one on the same elevation.

Colour rendering is worth a thought where daylight is the reason the opening exists. Low iron glass, with the iron oxide content reduced, removes the green that ordinary float carries and returns two or three percentage points of transmittance, which matters most on a north-facing Stroud valley cottage or a Gloucester back addition kitchen where daylight is scarce. On the ultraviolet side, a low-emissivity stack blocks a substantial share of UV and a laminated inner pane blocks essentially all of it, which slows the fading of floors and furnishings under a rooflight considerably. It also means houseplants directly under overhead glazing get less ultraviolet than they would under clear single glass, which is occasionally noticed and rarely a problem.

Optimum cavity width by gas, and why wider is not better

Each gas has a cavity width at which it performs best, and going beyond it makes the unit worse rather than better.

Narrow the gap and conduction through the gas dominates, because the heat has less distance to cross. Widen it and at some point the gas starts circulating in a convective loop, and convection carries heat faster than conduction does. The optimum sits where the two curves cross. Argon, being denser and more viscous than air, resists convection a little longer, and its optimum lands around 15 to 16mm. Krypton is denser again and starts convecting sooner, so its optimum is nearer 10 to 12mm, which is exactly why it suits a slim unit: it gives its best performance in a narrow gap rather than being handicapped by one. A 20mm argon cavity performs worse than a 16mm one, which reliably surprises people who assume more space means more insulation. Lay the unit down towards horizontal and the whole convection curve shifts, which is dealt with on the hub page.

Filling, purity, and what belongs on the order

Units are filled either by displacing the air through a lance while the unit is being assembled, or by evacuating and refilling in a chamber. Chamber filling reaches higher and more consistent purity.

Ninety per cent initial fill purity is the accepted minimum and good fabricators achieve 92 to 95. The number matters less than most people assume, because the conductivity of an argon and air mixture varies almost linearly with the proportion: a 90 per cent fill delivers around 90 per cent of the available benefit, and the difference between 90 and 95 per cent is a few hundredths on the U-value. What matters far more is that the initial fill is genuinely there and that the edge seal holds it. So the order should state the gas, the stated initial fill purity, and the cavity width in millimetres, and the U-value quoted should be the one calculated with that fill rather than an aspirational figure.

Losing the fill, and the units that are effectively air filled

Gas escapes through the edge seal slowly, and the product standard for sealed units limits the loss to less than one per cent of the fill per year. At that rate a unit is still performing close to its original figure after fifteen years.

A poorly constructed edge seal can lose several times that. The unit looks perfect, the glass is clear, nothing has misted, and the cavity is essentially air. Whole-unit performance drifts from 1.2 towards 1.5 or worse without a single visible symptom, and the first sign is usually that the glass edge starts running with condensation on mornings when it never used to. Argon content can be measured non-destructively with a spark emission analyser, which puts a small electrical discharge across the cavity and reads the emission spectrum, and glass fabricators and some surveyors have them. Across the county’s 1990s and early 2000s rooflights, the working assumption at survey is that anything of that age is at or near air-filled, which is part of why the step to a current unit feels larger than the two datasheets suggest.

Two checks anyone can make from inside the room

The first tells you whether there is a coating and which surface it is on. In a dark room, hold a small flame or a bright pen light close to the glass at an angle and count the reflections. A double unit gives four images, one from each surface. If all four are the same colour, the unit is uncoated. If one image is noticeably different, usually reading slightly violet, reddish or copper against the others, that is the coated surface, and its position in the sequence, counted from the room side, tells you which surface number it sits on.

The second is simpler and needs no equipment. On a still autumn morning, look for dew on the top face of a rooflight, outdoors, clearing by mid-morning. That is a sign the unit is working: so little heat now escapes through it that the outer pane is no longer being warmed from inside and falls below the dew point of the outside air, exactly as a car roof does. It gets mistaken for a failed unit regularly and it is the opposite. Water sitting between the panes, where no cloth can reach it, is the failure, and that is covered on replacing a failed or misted rooflight.

What to specify across the Gloucestershire stock

The county’s roofs sort into a few recognisable cases and the coating decision follows the slope more than the building.

On north-facing slopes, common on the front elevations along Gloucester’s Victorian terraces and throughout the Stroud valleys where houses sit against the hillside, specify a single soft coat on surface three, argon fill, and take the highest light transmittance available. There is no solar gain to control and solar control glass there simply removes daylight that was the point of the job. Consider low iron glass.

On south and west facing slopes across the interwar estates at Longlevens, Hucclecote and Charlton Kings and on the vale extensions, add spectrally selective solar control on surface two. On the wolds villages where a conservation rooflight is the only acceptable product and the rebate is shallow, a krypton-filled slim unit is the fill that earns its extra cost, because it gives its best performance in exactly the narrow cavity the frame will take. And on the near-flat units behind Cheltenham’s parapets, coatings on both cavity faces of a triple, with the fill chosen to suit the cavity depth the frame allows.

All of it goes on the order in writing: coating type, surface number, fill gas, stated purity and cavity width, alongside the U-value and g-value the fabricator calculated from them. Email info@severnskylightsgloucester.co.uk with the direction the slope faces and photographs of the roof and the room, and an indicative range comes back before anyone visits. The costs page sets out how that becomes a fixed price, and the written guides go further into individual questions.