Gaps That Must Stay Gaps: Heads, Flanks and Restraints

The class is easy to describe once the cases are set side by side. A gap at the head of a non-loadbearing wall so the structure above can move without bearing on it. A restraint that is stiff in one direction and permissive in another, so a panel is held against wind while remaining free to move along the face of the building. A single bearing per unit with everything else restraining only. A movement joint carried through every layer to the background. A tie or anchor that yields where two materials move in different directions.

What unites them is that none of them can be judged by looking at it. A gap that has been packed and a gap that has been left both disappear behind a lining or a facade, and a slotted hole with a bolt tightened through it looks exactly like one that can still slide.

The infill wall record ties the head gap and the panel's non-loadbearing status together as the same fact. The panel carries its own weight and wind load and nothing else, and the details that keep it that way are the point of the assembly. Pack the gap solid and the frame begins to load a wall that was never designed to carry it, in a way that is invisible from either face and only becomes apparent when the masonry starts to distress.

The record also answers the obvious objection, which is that the building is finished and the frame has surely stopped moving. It has not. A cast frame continues to shrink and creep for a long period, and floors continue to deflect under the loads they carry, which is why the gap exists for the life of the building rather than for the construction period. The panel built tight under a soffit early in that period is the panel most likely to end up carrying load it was not designed for.

On a light steel frame infill wall the same gap has an extra difficulty attached to it. The record describes the head as the place where structural and envelope requirements pull against each other: the track has to let the floor above move without transferring load into the studs, which means a gap whose dimension changes, while the air line, the weather layer and the insulation all have to remain continuous across that same gap.

That is why the record warns that any solution closing the joint solidly in order to keep those layers continuous has reintroduced the load path the detail existed to prevent. The two requirements are genuinely in tension, and resolving them is a design task rather than a site one. The published misunderstanding is worth quoting in substance: the head gap is not a construction tolerance to be filled.

On a facade of hung units the same principle appears as a distinction between fixings. The precast record states that support and restraint have to stay separate: if a panel is held rigidly at more than one bearing point, the structure's own shortening and deflection are no longer accommodated but transmitted into a stiff cast unit, which responds by cracking, usually at its weakest section around an opening. Slotted holes and sliding connections exist to keep the two duties apart, and they only work while their clearances remain free.

Curtain walling carries the same rule and reports its failure differently. The record describes one connection per unit taking weight and the others restraining only, so that as the structure shortens and each floor deflects the frames below are not asked to carry any of it. Where a restraint locks instead of sliding, the weight of the framing above comes to bear on the frame beneath it, and the consequence appears as distorted frames and broken glass rather than as a movement crack.

In a bonded facade the gap is a line rather than a clearance, and the failure mode is stated bluntly. The adhered tile and stone record says that a movement joint filled with grout is not a movement joint but a line of units waiting to debond. Movement in the background goes straight into the bond unless the joint grid reaches the background, and a joint stopped at the back of the units leaves the layer beneath continuous across the movement.

The render records make the same point about coats rather than units. A movement joint has to be carried through every coat to give differential movement somewhere to go, and where joints are omitted the movement still occurs but chooses its own locations, which are the corners of openings and the wall directly beneath a sill. A joint that is offset from the one in the structure behind is taking movement in the material between the two lines.

The connection across a cavity is asked to do two contradictory things at once, and one half of that is a movement duty. The veneer record describes an anchor that restrains the leaf outward and inward while yielding along the face of the wall, because the masonry and the backing move for unrelated reasons and on different timescales. An anchor stiff in every direction simply transfers that difference into the masonry as cracking.

This is why the records treat a repair specified for the wrong wall family as changing what the wall is rather than fixing it. A rigid tie where movement was expected, or a solid packing where a gap was designed, looks like a stronger connection and is a different building.

Every case on this list shares a delay. The load transfer, the locked restraint or the bridged joint happens immediately, and the evidence appears only when the structure has moved enough for the consequence to reach a surface. The infill record describes the transfer as invisible from either face; the precast record describes the evidence arriving as cracking at openings; the curtain walling record describes distorted frames and broken glass.

That delay also separates cause from appearance. A crack at the corner of an opening is a long way from the head of a panel or from a restraint on another floor, which is why the records ask for the gaps and restraints to be recorded rather than remembered. Once a facade is on, nobody can see whether a head gap is open or what a restraint at a column actually permits.

The practical rule the records support is a question rather than a prohibition. Before a gap is packed, a joint is grouted, a slot is tightened or a fixing is added, establish what that gap was provided for and who provided it. The infill record's own design question asks how wide the head gap is being left, what closes it and who is closing it, and its handover question asks how anyone will confirm later that the head gaps are still open.

For an existing building the same question goes to the structural engineer for that building rather than to a general reference. Filling, propping or cutting into an existing infill panel changes how the frame and the wall share load, and adding fixings to a hung panel changes what the panel can absorb.