Because the roof was such a prominent feature, designing it so that it could be built in a cost-effective way was fundamental to the project's success.
In December 1999, with a positive outcome expected from the public inquiry, a roof design team was assembled to begin the painstaking process of value-engineering, or, as Mike Forster, BAA's design and development director for Terminal 5, puts it "designing value into the roof" to ensured that the terminal remained within budget.

BAA was helped in this task by some of the industry's elite. In addition to Richard Rogers Partnership, the framework team for the roof included structural engineer Arup, steelwork contractor Severfield-Rowen, cladding specialist Schmidlin and Hathaway Roofing. Forster says: "We got all the people around the table from day one; so the designers did not design in the absence of suppliers."

By this stage of the terminal's design, the architect's original proposal for a waveform roof supported on four rows of branched structural columns had been abandoned. In its place was a single-arched structure spanning the entire width of the terminal's concourse. The design had a roof canopy that arched in a single sweep across the width of the terminal and rippled along its length like a curved sheet of corrugated iron.

It was difficult to make the supports for this arrangement cost-effective. Initially, the architect, working with Arup, proposed a structural solution based on a diagonal grid of small steelwork sections. Its construction would have required hundreds of connecting nodes to be specially cast. It would also have had to be assembled on site, a job that Dervilla Mitchell, a director of Arup, estimates would have taken 40 welders 80 weeks.

The next proposal focused on using fewer structural elements and simplifying the connections between them. Severfield-Rowen proposed using two layers of orthogonal steelwork to form the double wave roof. The roof's construction would be based on a series of curved steel members arching the entire width of the terminal. These would be arranged at different heights so that the roof appeared to ripple along its length. They would be overlain by a second layer of steelwork arranged perpendicularly to form a rectangular grid. The steel sections in this layer would be curved so that they followed the waves created by the arches. This solution was also too expensive. "Our original design was too complex; it was beyond the capability of the contractors," says Mike Davies, the partner at RRP responsible for the terminal – which is a tactful way of saying it was too expensive for BAA. The real breakthrough in the roof design came when Severfield-Rowen proposed getting rid of the ripples and moving to a single barrel-vault form. "The fundamental change was moving from a 3D to a 2D roof form; it was an iterative process and not something the architect conceived and passed down the line," says Davies.

Simplifying the form of the roof allowed the structure to be based on a series of arched box girders spanning the width of the terminal building. "Rowen suggested box girders because this would be easier to fabricate," says Davies. Critically, it separated the roof design from the still-evolving interior.

However, the move to an arch structure introduced another challenge: managing the lateral forces in the arch.

To turn it into a stable structure, the engineers proposed connecting its feet with using structural tie-rods passing across the terminal. "We worked up versions of the long-span roof with the arch ties in different positions," says Davies. The team looked at running the ties through the floorplates to limit their visual impact, but this in turn limited the flexibility of the interior layout. In fact, the need to stiffen the arch by running tie-rods through the building contradicted the overriding imperative to free up interior space.

The next idea was to support the arch in the centre of its span – a solution that Davies calls an "assisted arch" – but the structure necessary to support it also compromised the flexibility of the interior. The idea of extending the arch beyond the perimeter of the terminal building, so that it could be bedded in the ground either side of the terminal, was considered and abandoned after BAA raised concerns about the future flexibility of aircraft stand location.

In the end, the design team came up with a solution that satisfied all criteria: a tied (or bowstring) arch supported high above the concourse on inclined structural columns – a kind of glorified version of a portal frame structure.

"The final version of a bowstring arch supported on [structural] trees works because the lateral forces are resolved within the arch itself, so only vertical forces are transmitted down the trees," says Dennis Austin, head of the roof team for RRP. This version kept the structure clear of the terminal's interior, ensuring its future flexibility – and it would also be cost-effective to construct.

The task now was to devise a way to erect the roof. At its high point, it towers 37 m above the apron. However, the airport's radar is in operation 2 m above that – an extremely inflexible height constraint. Davies refers to it as "the cursed 39 m figure" because it prohibited the use of cranes.

Instead of craning the structure into place, the roof team came up with the idea of assembling the tied-arch section on the ground. The support abutments would be constructed first, and then the roof would be jacked into place. But rather than jack the arches individually, it was decided to assemble whole sections on the airport apron before raising them into position. This would mean that most work would take place on the ground, making assembly a relatively easy and safe option. "Rowen devised a method of constructing the roof on the ground and then lifting it into place, two structural bays and one infill section at a time," says Austin.

In addition to the structure, the roof covering is being preassembled on the apron. Hathaway Roofing will deliver the components to site as 3000 preassembled cassettes. These will be slotted into position prior to the roof lift. The insulation and standing-seam cladding will not be installed until the roof is in place. However, even these components will be laid flat on the preassembled roof sections for a free lift prior to installation.

In all, five lifts of three bays each and one single-bay lift are planned. Each bay will be stitched to the next with secondary steelwork. "We could have jacked the whole lot up in one go but that would have required an enormous amount of temporary steelwork," Austin explains. As it is, each bay will be about the size of a football pitch.

The method of construction informed the way the steelwork connections had to be detailed to ensure the huge sections of roof structure could be slotted smoothly into position high above the apron. "The knuckle where everything comes together has had years of team input so that it is Tonka-Toy-simple to assemble – the whole structure is held together by bolts," says RRP's Davies.

The team were helped in the development of this detail by using a single computer model of the connection (see "The IT strategy" story).

To ensure that roof erection proceeds smoothly on site, the arch-supporting abutments are being trial-assembled by steelwork contractor Severfield-Rowen under what Davies calls "a first-run study". "We couldn't have 17 projects all running along and then find that this one was holding the others up," says BAA's Forster. The first abutment has already been constructed and, according to Davies, "it generated 70 to 100 tweaks in the design and erection methodology".

The team now plans to erect a trial section of the facade to eliminate all foreseeable snags before assembly on site.

As well as proving to the design team that the erection method was workable, the study gave Severfield-Rowen's steelwork erectors valuable practice. The contractor will be using the same team that erected the trial abutment on site at Heathrow. BAA's Forster admits that the trial run was not cheap, but reckons it was money well spent. "The first-run study tested the construction sequence – it is a sensible investment on our behalf and we'll get the money back in spades," he says. The temporary steelwork used in the study will also be reused for the construction of T5, to keep costs under control.

With roof construction programmed to start in October, fabrication of the components is now under way.

Terminal 5