Roof lantern thermal performance and Part L

Picture the two side by side. A flat rooflight is a sealed glazed unit dropped into a slim frame: one sheet of glass, four edges, one perimeter seal. A roof lantern is a small glazed roof in its own right. It has a ridge bar along the top, hip bars running down to each corner, eaves bars along the base, and between them anything from four to a dozen individual glass panes, each with its own sealed edge. The classic shape over a south Essex extension is a hipped rectangle, wider than it is tall, with sloped facets meeting a central ridge.

That extra structure buys you the look people want, and it costs you thermally in three ways. There is more frame per square metre of roof opening, and frame usually conducts heat faster than glass. There is far more sealed edge, and the edge of a glazed unit, where the spacer bar sits, is always its weakest thermal line. And the raised profile puts more of the glazed surface up in the moving air above the roof, where wind strips heat away faster than it would from a flush unit. So before you have chosen a single pane, a lantern starts from behind. The job of a good specification is to claw that back.

None of this is a reason to avoid lanterns. Fitted properly they are one of the best ways to light a deep single-storey room, which is exactly why they sit over so many knocked-through kitchens. It is a reason to read the whole-unit figures rather than the glass figures, and to know which parts of the frame the manufacturer has actually engineered.

Part L of the Building Regulations sets the thermal rules, and a new or enlarged roof lantern is notifiable work. The figure it polices is the whole-unit U-value: watts of heat lost per square metre of the unit, per degree of temperature difference between the room and the outside air, written W/m²K. Lower is warmer. The regulations do not set a G-value or solar limit for a lantern in an existing home, which is the gap that lets a fully compliant lantern still cook the room beneath it. More on that further down.

The current England edition of Approved Document L gives limiting U-values, meaning the worst a unit is allowed to be, and separate standards for replacement work in a home you already own. The headline numbers, quoted on the standard vertical-plane basis explained in the next chapter, are these.

Two things stand out. First, the bar for a lantern, 2.2, is generous compared with a wall window, and far softer than the opaque roof around it. A lantern is a hole in your insulation by design, so the regulations accept a much higher loss through it than through the tiles beside it. Second, 2.2 is a backstop, not a target. A basic double-glazed lantern will scrape past it. A well made one will beat it comfortably, and the difference between 2.2 and 1.4 is the difference between a lantern that feels cold to sit under in January and one that does not. Meeting Part L and specifying well are not the same thing, and we make the Building Control notification either way, to Southend-on-Sea City Council or the relevant local authority for your address.

Here is a quirk that catches people out. Part L compares rooflights on their vertical-plane U-value. A manufacturer tests the unit standing upright, like a window, and the figure on the data sheet is that vertical value. But your lantern is not standing upright. Its facets are sloped, and the whole assembly faces broadly upwards. A glazed unit tilted towards the horizontal loses slightly more heat than the same unit stood vertical, because the warm cavity gas convects more freely when the panes are laid over. The horizontal, in-use loss is therefore a touch worse than the printed number.

This is not a trick, it is a shared convention so that every product is measured the same way. But it means the vertical figure is the comparison figure, and the number you check against the 2.2 limit. When someone quotes you a U-value for a lantern, the useful question is whether it is the whole-unit vertical value, the centre-of-glass value, or the glass-only value from the sealed unit supplier. They can be a long way apart.

Centre-of-glass ignores the frame and the edge seal entirely, so it always flatters the unit. A lantern might carry a centre-glass figure near 1.0 and a whole-unit figure closer to 1.6 once its many bars and edges are counted in. Because a lantern has so much more frame and edge than a flat rooflight, the gap between the glass number and the honest number is wider for a lantern than for almost anything else you can put in a roof. Ask for the whole-unit value, on the vertical plane, and you are comparing like with like.

Trace the heat as it leaves and the weak points sort themselves into an order.

The bars. Ridge, hips and eaves bars are the skeleton, and their material decides a lot. Older timber and standard uPVC lanterns use chunky bars that are visually heavy and thermally middling. Modern aluminium lanterns went the other way, using very slim sightlines for the view, but bare aluminium conducts heat quickly, so a slim aluminium bar with no thermal break is a cold bridge running the length of the ridge. The good ones use a thermal break, a low-conductivity insert splitting the outer metal from the inner, so the bar can be slim and still not run cold. That single detail separates a lantern that beats Part L with room to spare from one that only just clears it.

The sealed edges. Every glass facet has a spacer bar around its perimeter holding the two panes apart. Traditional aluminium spacer conducts heat straight across the gap and chills the glass edge, which is where condensation forms first on a cold morning. A warm-edge spacer, made from a composite or stainless strip, cuts that edge loss and lifts the whole-unit U-value measurably. On a lantern, with its long total run of edge, warm-edge spacer earns its place more than on almost any other rooflight.

The ridge and hip joints. Where bars meet, there is a junction to seal and insulate. Cheap lanterns leave gaps that whistle in an estuary wind. Well engineered ones use gasket systems and insulated ridge sections so the joints are not open thermal holes.

The cavity gas. Argon between the panes slows convection compared with plain air, for a small cost. On a multi-pane lantern that saving is repeated across every facet, so it adds up.

These are typical published whole-unit figures for common lantern build-ups, on the vertical-plane basis Part L uses. Exact values move with size, bar count and manufacturer, so read the pattern rather than the decimal. A bigger lantern with fewer, larger panes generally posts a slightly better whole-unit figure than a small one chopped into many small facets, because there is proportionally less edge and bar.

Look at the first two rows against the third. The regulations accept the top line at 2.2, but it is a lantern you can feel across the room on a cold night, and its cold bars invite condensation to bead along the ridge. Moving to a thermally broken, warm-edge, argon unit takes the loss down by a third for a modest premium, and it is the change that matters most. Beyond that, triple glazing buys you a warmer winter at the cost of weight and price, and whether it is worth it depends on the room and the aspect. A lantern over a rarely heated garden room does not need triple glass. A south-facing lantern over a room you live in every day is a different conversation, and that conversation is mostly about the third column, not the second.

A lantern does not sit on the roof. It sits on a kerb, an upstand, usually a timber box built up off the flat roof deck so the glass clears the water and the covering laps up the sides. That upstand is part of the thermal envelope, and it is the part most likely to be built as an afterthought.

Think about what it is: a ring of timber, often 150mm or more tall, wrapping the whole opening, sitting directly in the airflow. If it is framed from solid timber and left uninsulated, it is a continuous cold bridge around the base of your beautiful new lantern, exactly where the ridge detailing worked so hard to save heat. The inside face of a cold upstand is also the first surface in the room to reach dew point, so it is where you get the damp line and the black spotting, not on the glass at all, on the plaster reveal below it.

A proper upstand is insulated on its outer face and continuously connected to the roof insulation below and the frame above, so the warm line runs unbroken from the ceiling, up the reveal, into the frame, without a cold gap. This is a build detail, not a product you can buy, and it is why the U-value on the data sheet only tells you half the story. The finished performance is the unit plus the upstand plus the workmanship joining them. When we set a lantern, the upstand and its insulation are surveyed and detailed as part of the roof lantern installation, because a 1.5 unit on a cold kerb performs like a worse one.

Everything above is about keeping heat in during winter. On a south-facing roof in Southend the bigger problem runs the other way. This corner of Essex is unusual: it faces south across the Thames Estuary, with an open horizon and a wide reflective sheet of water throwing extra light up onto the roofs along the front. A lantern over a south-facing rear extension, which describes a great many of the 1930s semis across the patch, is a large area of upward-facing glass pointed at the strongest sun of the day.

Because a lantern faces broadly upward, it gathers midsummer sun far more directly than a wall window on the same house, and it gathers it at ceiling height where the heat then sits. This is where the G-value comes in, the share of solar energy the glass lets through. Ordinary clear double glazing runs around 0.6, so more than half the sun’s energy arrives inside. A solar-control coating drops that towards 0.3 while keeping the room bright, because it blocks the invisible heat and passes the visible light. On a big south lantern that swing is worth several hundred watts across a July afternoon, which is the difference between a room you can use at four o’clock and one you cannot.

Part L will not push you towards this, because it sets no G-value limit for a lantern in an existing home. Part O, the newer rule that does address summer overheating, applies to new residential buildings rather than to an extension on the house you already own. So the regulations that make you insulate the lantern say nothing about the sun coming through it, and a lantern can pass every thermal check and still overheat. Picking the G-value to suit the aspect is the point of solar-control glazing, and on a south or west facet it matters more than the last decimal of the U-value. On a north-facing facet, where there is no direct sun to manage, we do the opposite and keep the glass clear so the room gets all the daylight going.

You can settle the thermal side of a lantern in one honest conversation with whoever is quoting. The questions that separate a considered specification from a catalogue price are short.

1. “Is that U-value whole-unit or centre-of-glass?” On a lantern the two can differ by half a point or more, because of all the bars and edges. You want the whole-unit figure, on the vertical plane, which is the one Part L checks.
2. “Are the bars thermally broken?” Slim aluminium bars with no thermal break run cold and invite condensation along the ridge. A thermal break lets the lantern be both slim and warm.
3. “What spacer sits around the glass?” Warm-edge spacer lifts the whole-unit value and keeps the glass edges above dew point, which matters more on a lantern than almost anywhere, given how much edge it has.
4. “Which way does the extension face, and what is the glass G-value?” If the answer to the first is south or west and the answer to the second is around 0.6, the glass is wrong for the aspect whatever its U-value says.
5. “How is the upstand insulated?” A cold kerb undoes a warm unit. You want insulation on the outer face, continuous with the roof insulation below and the frame above.
6. “Who makes the Building Control notification?” A new or enlarged lantern is notifiable. We make that notification as part of the job.

The pattern across all six is the same. The data sheet U-value is a starting point, not the finished answer, and on a lantern more than on a flat unit the difference between the printed figure and the lived performance is decided by the frame, the edge, the upstand and the aspect. If you want the numbers worked out for your own roof, facet by facet, before you commit to anything, you can compare a lantern against a run of flat-roof skylights for the same opening, or ask us for a specification with the whole-unit U-value and the glass G-value written down for the aspect you actually have. We have been installing skylights and rooflights across south Essex for over fifteen years, and every one is backed by a ten-year workmanship guarantee, with the survey of aspect and upstand where the job begins.