Detailing a Building Expansion Joint

Grid 14, Where the Drawing Says One Building and the Frame Says Two

The tender drawings show a single 190 metre block. The structural general arrangement shows two, with a double column line at grid 14 and a note reading EXPANSION JOINT — REFER TO STRUCTURAL ENGINEER. Nobody has drawn what happens where that line reaches the cladding, and the facade specification mentions it in one clause that says to allow for proprietary movement joint covers as required. That clause is where a job loses money.

The line at grid 14 is not a detail. It is a surface that runs the full height and the full depth of the building: down through the foundations, across every floor plate, up both faces of the external wall, over the roof, through every rated compartment wall it happens to cross, and out through the paving in the yard. Every element the building is made of gets cut by it, and every one of those cuts is a separate manufactured item with its own listing, its own movement rating and its own price. The facade package usually owns the vertical run on both elevations, the transitions at ground and parapet, and — depending on how the roofing package was written — the run across the roof as well.

Then there is a second, entirely different joint on the same job, and confusing the two is the classic error. The joint at grid 14 separates structure from structure. The joint where the frame meets the slab edge separates cladding from structure, all the way round the building, on every floor. The first is a structural decision made by an engineer and executed with a proprietary cover system. The second is a sealant joint whose width you are expected to calculate, and it is the one the estimator is more likely to get wrong because it looks like ordinary perimeter caulking.

The Joint Is a Plane, Not a Line

Take a highlighter to the sections and follow the separation from the lowest slab upward. It passes through the ground slab, where it needs a floor joint system rated for whatever traffic runs over it. It passes through each suspended floor, where it needs a floor cover plus a rated joint system where the floor is a fire barrier. It passes through the external wall, where it needs a watertight, insulated, air-sealed cover on the outside and usually a second finish cover on the inside. It crosses the roof, where it needs a raised joint cover that is part of the waterproofing rather than a plate laid on top of it. It hits the parapet and stops, and that stop is a fabricated termination that somebody has to draw before it can be made.

The quantity that matters is not the length of the line on plan. It is the length of the developed surface plus the count of every place the surface changes direction or changes system. A joint that reads as 190 metres on a plan can carry three hundred metres of cover once both elevations, the roof run and the two internal faces are added up, and it can carry twenty or thirty fabricated transition pieces on top of that. A take-off that priced the plan length has priced roughly a third of the work.

Cut It Open at the Slab Edge

What is in the gap matters more than what covers it, and the cover is the only part most people ever see. A structural separation joint at a floor level is normally four or five separate products stacked into one blockout: the cover assembly at the traffic or weather face, a waterproofing membrane or gutter beneath it where water can reach, a compressible filler or thermal seal, a fire-rated joint system where the floor or wall it interrupts carries a rating, and the two structural edges themselves.

The accented layer below is the one that gets missed, because it is invisible once the cover is on and because it does not appear on an architectural section at 1:50. It is also the only layer in the stack whose omission is a code contravention rather than a defect, and the only one that cannot be corrected later without lifting everything above it.

Where the Movement Number Comes From

Every dimension on this page descends from one figure: how far the two sides move relative to each other, and in which directions. That figure belongs to the structural engineer, and a facade contractor who invents it has taken on a liability that was never in the price. But you do need to understand its shape, because you have to recognise when the figure you have been handed is obviously the wrong kind of number.

Thermal movement is the part you can sanity-check yourself. A frame expands from wherever it is restrained — a core, a stiff braced bay, a line of shear walls — not from its geometric centre, so the travel at the joint depends on where the stiffness sits as much as on the overall length. Take a symmetric case anyway, to get the order of magnitude: a 190 metre concrete frame split into two 95 metre halves, each restrained near its own middle, gives each free end about 47.5 metres of length to grow along. With a coefficient of around ten microstrain per degree Celsius and a forty degree operating swing, that is roughly nineteen millimetres per side, so the gap opens and closes by something in the region of thirty-eight millimetres over the year. The aggregate in the mix moves that coefficient by close to a factor of two on its own, which is exactly why the answer comes from the engineer's model and not from this paragraph.

Thermal is rarely the whole of it. Differential settlement between two foundations sized for different loads shows up as vertical offset across the joint and does not reverse. Shrinkage and creep in a concrete frame shorten it permanently over the first years, so a joint set at completion is not at mid-travel. Where the building is in a seismic region, the separation is sized so two structures cannot pound against each other under design ground motion, and that dimension is typically far larger than the thermal one — it is derived from the displacements the seismic code requires, and the cover system then has to survive movement in three axes at once rather than opening and closing on one.

Two consequences follow for the estimator. First, ask what governs before pricing anything: a joint governed by seismic displacement takes a different, more expensive class of cover than one governed by thermal movement, and the two look identical on a plan. Second, ask for the movement stated as a range about the installed width, not as a single number, because the cover has to work at both extremes and the fire-rated system inside it is listed against a percentage of its nominal width in both directions.

Pricing It by the Piece, Not by the Metre

The commercial mistake is uniform across trades and it always looks the same on a spreadsheet: one line, one length, one rate. A separation joint priced that way is wrong in three directions at once. It misses the pieces, it misses the transitions between systems, and it misses the fact that the same physical line changes product every time it changes plane or changes rating.

Count the systems first. Walk the joint from the lowest level upward and mark every place the product changes: floor to wall, wall to roof, unrated to rated, interior to exterior, trafficked to non-trafficked. Each of those is a transition where two different systems have to be made continuous, and continuity at a transition is a detail somebody has to draw and a fabricator has to make. On a mid-rise office block with a single separation joint it is routine to find eight or ten distinct systems and twenty-odd transition pieces on one line.

Then count the pieces within each system. Straight run is the cheap part and it is the only part a linear take-off sees. Corners are made units. Tees are made units. Stop ends and terminations where the joint dies into a parapet, a kerb or a ground slab are made units, and they are the ones with the longest lead times because they are frequently drawn last. Anything that has to be watertight and moving at the same time — a roof-to-wall transition, a joint passing through a gutter line — is bespoke fabrication and priced as such.

Two more things belong in the price and are usually forgotten. The blockout is one: the recess the cover sits into has to be formed correctly by the concrete or steel package, and where it has not been, the remedial work lands on whoever is fitting the cover. The other is the movement itself as a programme risk — a joint fitted in February at fifteen millimetres will be at a different width by August, so the cover has to be installed at the width the system's own installation instructions specify for the ambient conditions on the day, and the record of what it was set to is what settles the argument in year two.

Before the Order Goes In

Four documents settle almost every question above, and none of them is the architectural detail. The engineer's movement statement, giving the range and the axes rather than a single width. The fire strategy, saying which crossings are rated and to what. The manufacturer's listing for each system, which fixes the maximum gap, the movement class and the installation width. And a survey of the blockout as built, because a cover selected for a hundred millimetre gap that arrives to find a hundred and thirty is a redesign, not an adjustment.

Write the installed width and the date on the sheet as each length goes in. It costs a few seconds and it is the only thing that distinguishes, eighteen months later, a joint that has closed up from a joint that was never set open in the first place. Those two have completely different remedies and only one of them is your problem.