A curtain wall does not sit on the building: it hangs from it. Everything it does rests on two invisible ideas — one load-bearing bracket per mullion, and a joint that drains instead of sealing.
Vertical mullions hung from the slab edges, horizontal transoms forming the grid, glass units and panels inserted from outside and clamped by pressure plates: this is the stick system, assembled piece by piece at height. It costs less than a unitised facade and bends to irregular geometry, but it shifts all the quality onto workmanship, because every joint is made on site. The least intuitive part is watertightness: it is not achieved by sealing the outside, but by letting a little water into a chamber held at the same pressure as the wind, where it has no reason to travel further, and draining it out by gravity. The glass, for its part, never touches metal: it floats on setting blocks inside its rebate, and the perimeter clearance is its only defence against thermal movement and structural settlement.
click a beat · arrow keys ← → walk the site · dashed pauses are the chemistry at work
The detail of the craftthe load-bearing bracket and the restraint bracket
Fixing the adjustable brackets to the slab edge and erecting the vertical mullions, with an expansion joint at every floor.
The mullion has to breathe: one bracket holds it, the others just guide it. Lock them all and July will tell you.site notebook — editorial synthesis
The counter-intuitive principle
Instinct says: seal everything. Physics says the opposite. Water enters a joint only if something drives it — capillarity, gravity and, above all, a pressure difference between outside and inside. Remove the pressure difference and you remove the drive: the drained, pressure-equalised joint does exactly that, and hands real sealing to a gasket kept out of the rain.
Schematic. It applies to curtain walling as much as to windows and rainscreen cladding: wherever there is an exposed joint, the right strategy is almost always to drain, not to close.
The millimetre rule
Between a January night and a July afternoon a dark aluminium mullion can go from −5 to +70 °C. Over twelve metres that is nearly two centimetres of growth. The slab, meanwhile, deflects over time through concrete creep. The glass perimeter clearance and the mullion expansion joint exist to absorb these two independent movements: they are empty spaces, and they must be defended from anyone wanting to fill them.
Order-of-magnitude figures, based on the expansion coefficient of aluminium (about 0.024 mm/m°C). The detailed design sets the actual clearances according to height, colour and seismic zone.
It depends on how much the facade repeats. Unitised panels are built in the factory, arrive finished and go up in a day: consistent quality, a very short site programme, higher cost and no tolerance for improvisation. The stick system costs less, accepts non-repeating geometry and lets you correct at height, but its quality is the quality of the team fitting it. Below three or four storeys and with irregular facades, stick almost always wins; above ten repetitive storeys, almost never.
Because the system is designed that way. The glass rebate is not a dry space: it is the outer part of the double barrier, and a film of water running down to the drains is normal operation. The defect is water that stops, not water that passes. It is worth explaining before handover, with the detail in hand: it is one of the most common and most avoidable complaints.
Largely yes, and it is the least discussed advantage of curtain walling: the package runs in front of the slab instead of being interrupted by it. What remains is the bracket, a piece of steel crossing the insulation: it is mitigated with thermal isolators under the base plate and by keeping the number of anchors to what is needed, not by removing brackets.
Anodised or powder-coated aluminium comfortably passes forty years. Two things expire sooner: the gaskets, which harden and lose elasticity in twenty to twenty-five years, and the edge seal of the insulating units, whose life depends mostly on how dry they have been kept. Real maintenance is trivial and almost always skipped: check once a year that the drains are clear.