by J E Gordon
Chapter 11 The advantage of being a beam
-with observations on roofs, trusses and masts
Solomon... built the House of the Forest of Lebanon, a hundred cubits long, fifty broad, and thirty high, constructed of four rows of cedar columns, over which were laid lengths of cedar. It had a cedar roof, extending over the beams, which rested on the columns, fifteen in each row; and the number of the beams was forty-five.
1 Kings 7.1-3 (New English Bible)
A solid roof over one’s head is one of the prime requirements of a civilized existence, but permanent roofs are heavy and the problem of supporting them is really as old as civilization itself. When one looks at a famous and beautiful building – or indeed at any building – it is illuminating to bear in mind that the way in which the architect has chosen to solve his roofing problem has affected not only the appearance of the roof itself, but also the design of the walls and the windows and indeed the whole character of the building.
In fact, the problem of supporting a roof is essentially similar in its nature to the problem of making a bridge, with the difference that, since the walls of buildings are likely to be thinner and weaker than the piers of bridges, any sideways thrust which the roof may impose must be considered even more carefully. As we saw in Chapter 9, if the roof pushes outwards too hard upon the tops of the walls on which it rests, the line of the thrust in the masonry will be displaced to a dangerous extent and the walls will collapse.
Many Roman buildings and practically all Byzantine formal architecture made use of vaulted or domed roofs. These arch-like structures thrust vigorously outwards upon their supports, and in most cases this was catered for by resting the roofs upon very thick walls within which the thrust line had generally plenty of room to wander about in safety. As we have seen, these thick walls were often made from mass concrete, sometimes lightened and thickened by the incorporation of empty wine jars. Such walls were structurally stable, and they had the additional advantage of providing excellent heat insulation in hot climates: a Byzantine church is often the only cool place in a Greek village. However, it is not easy to make windows in very thick walls, and such windows as existed in Roman and Byzantine buildings were usually small and placed high up.
The medieval castles were built pretty much in the Roman tradition, frequently, as at Corfe Castle, from mass concrete many yards thick. Such walls were well able to resist the thrusts set up by the vaulted roofs; and, for military reasons, the defenders did not really want windows anyway. The earlier Norman or ‘Romanesque’ churches were not very different, and their thick walls, little, rounded arches and small windows derive directly from late Roman prototypes. Most of the early Romanesque churches were satisfactory enough, and many of them survive today.* The difficulties and complications arose later on and are to a great extent associated with the growing fashion for bigger and better windows.
Understandably, people living in sun-drenched countries do not feel quite the same about windows as northerners, and even today many of them seem to dwell, apparently from choice, in a perpetual shuttered twilight. No doubt this is all part of a long Mediterranean tradition, for in Greek and Roman and Byzantine times such windows as existed were generally small and rather ineffectual.† As far as one can see, this was by no means entirely due to a shortage of glass.
In northern Europe, even warlike knights and barons did not want to spend all their time in gloomy and nearly windowless castles. What they wanted was light and sunshine, and so they tired of architectural forms based upon dark Roman models. The cult of windows became an obsession, and, as time went on, builders competed in constructing both halls and churches with larger and larger and ever more splendid windows. The medieval craftsmen may have been hopelessly unscientific but they were sometimes much more creative than we generally recognize. In particular we owe them a great debt for showing us what beautiful and exciting things can be done with windows.
However, much of the effect of an impressive and expensive window is lost if it has to be inserted into a tunnel-like opening in a thick wall. Inevitably, attempts to provide bigger windows set in thinner walls ran into trouble with thrust lines. Norman architecture was basically Roman architecture and cannot be made to do this sort of thing, because it depends for its stability and safety on the use of thick walls. But this did not stop builders from trying, and it has been said of late Romanesque architecture that the question to ask of any particular building is ‘not whether, but when, the Great Tower fell’.
Figure 1. King’s College Chapel, Cambridge.
Just how clearly the medieval masons appreciated what was happening is not certain. Most probably their understanding of the situation was muddled and subjective; otherwise they would not have gone on making the same mistakes for several generations. Sooner or later, however, somebody realized that the way to deal with a demand for large windows and thin walls was to make use of buttresses, which could prop the wall against the outward thrust of the roof by pushing against it from outside.* Effectively, buttresses make the wall thicker, and so they do the same job as the Roman wine bottles, only in a different way.
Figure 2. The introduction of side aisles and a clerestory required the invention of the flying buttress.
The ordinary solid buttress is really no more than a local thickening of the wall between the windows. Where there is only a single aisle, as in King’s College Chapel (Figure 1 and Plate 13), it is very effective. Difficulties arose, however, with side aisles. In order to prop the roof of the nave without unduly shading the clerestory windows, the Gothic masons had to invent the flying buttress (Figure 2). In this case the vertical part of the buttress is separated from the wall by a series of arches, which transmit the thrust without intercepting much of the light.
The decorative possibilities of flying buttresses in conjunction with large windows are very great, and, as we have said, they are still further enhanced by the judicious introduction of statues and pinnacles, whose weight, as the masons must somehow have realized, helps the buttresses in the tricky task of guiding the thrust lines safely down through the lace-like forest of masonry. In the end the windows became so large that not very much actual solid wall was left to support the building. Like a modern mast, these narrow strips of stonework depended entirely upon lateral support. As a tall thin mast relies upon a network of sophisticated rigging, so these slender walls depend entirely for their stability upon the bracing afforded by arches and buttresses.
By whatever mental process all this was accomplished, the structural and artistic achievement was immense. By the time the master masons had created the Gothic buildings of the high Middle Ages, architecture had lost any visible connection with its classical origins. Few things could look much more different than, say, Canterbury Cathedral and a Roman basilica. Yet the line of descent is clear and simple.
Although buildings like these are often very beautiful, they are always horribly expensive, and in any case arched or domed roofs are usually unsuitable for private houses. Rather than using arches it is much cheaper and simpler to support the roof of a building by using beams of one sort or another. If the spaces to be covered are spanned with long poles or joists, then such beams can transmit the weight of the roof from their ends, vertically downwards into the masonry of the walls, without any need to push sideways and outwards. Thus no unwelcome disturbance is caused to the thrust line and so the walls can be made quite thin and will not need buttressing (Figure 3).
Figure 3. Simply supported roof-truss. This one is shown mounted on rollers to emphasize that there need be no outward thrust upon the supporting walls.
For this reason alone, the beam is one of the most important devices in the whole of structural engineering. In fact, however, the applications of the beam – and of its equivalent, the truss -extend far beyond the roofing of buildings; and beams and beam theory have played a very important part indeed in making technological civilization possible. Similar ideas are also continually cropping up in b
iology.
The word ‘beam’ means a tree in Old English, and this usage still survives in tree names like ‘whitebeam’ and ‘hornbeam’. Although nowadays beams are very commonly made from steel or reinforced concrete, for a great many years a ‘beam’, in the structural sense, implied a baulk of timber, very often a whole tree-trunk. Although it is cheaper and much less trouble to cut down a tree than to build a masonry arch or vault, the supply of suitable large trees is not unlimited and a time arrives when long pieces of timber become scarce. When this happens one may be forced to try to construct roofs from short lengths of material.
Roof trusses
To the modern mind it might seem fairly clear that the most promising way to try to bridge a roof-span using short pieces of timber would be to join the short members together, Meccano-fashion, so as to make a triangulated structure, something like Figure 4. This is really the beginning of a lattice girder. We are all familiar with lattice girders in steel railway bridges. Any triangulated lattice structure of this kind is called a ‘truss’. Like a long solid beam, when a roof-truss is properly designed it allows considerable spans to be roofed economically and without putting any dangerous outward thrusts upon the supporting walls. As with beams and beam theory, the applications of trussing in modern technology are far wider than this and extend to ships and bridges and aeroplanes and to all manner of other structural devices. As we have seen in the last chapter, the tied arch is really another example of the same idea.
Figure 4. If long pieces of timber cannot be got, then a roof-truss could be built up, Meccano-fashion, from short pieces.
However, in the history of architecture the concept of the truss or lattice beam was surprisingly slow in developing. The most primitive form of this idea, the ordinary wooden roof-truss, may seem obvious to us but it took our ancestors a long, long time to think of it. But then they had never seen a railway bridge or played with Meccano. As it turned out, architectural trussing was a late Roman invention, although it never really caught on properly until the Middle Ages. During most of antiquity architects just managed without trusses.
Greek builders never thought of trusses at all. Vastly eminent Athenian architects, such as Mnesicles, who built the Propylaea, and Ictinus, who designed the Parthenon and the Temple of Apollo at Bassae, consciously rejected arches and vaults as a method of roofing their buildings, and yet they failed conspicuously to invent the roof-truss or to devise any really adequate substitute for it. The brilliance of Hellenic architecture seems to come to a stop, rather suddenly, when one gets to the architrave. Greek roofs can only be described as intellectually squalid.
Figure 5. Roof of an archaic Greek temple.
Simple stone beams or lintels cannot safely be used to span distances of more than about eight feet (2-5 metres); otherwise they are liable to crack. Therefore, in order to provide practicable roofs for temples and other buildings, it was necessary to use wooden beams, in spite of the fact that in classical Greece timber had become nearly as scarce as it is in modern Greece.
In Greek temples for which the necessary number of full-span wooden roof-beams could be found, the beams were simply laid horizontally right across the tops of the walls and of the stone lintels of the peristyle. These beams or joists were then boarded over so as to provide a continuous flat ceiling over the whole area of the building (Figure 5). Naturally this flat ceiling, which was only made from ordinary planks, was anything but weatherproof. So a great mound of clay soil, mixed with water and straw, was heaped up on top of it. For an average-sized temple this pile of clay must have weighed something like 3,000 tons. When they had got all this agricultural material up there and tamped it down properly, the mound was trimmed off as accurately as might be to the triangular shape of a pitched or sloping roof. After this the roofing tiles were brought up and simply laid directly on top of the clay, very much like laying paving stones for a garden path. Presumably one hoped that this great mass of wet clay would dry out before the wooden ceiling which supported it began to rot. When dry, it must have made a wonderful sanctuary for vermin; but the excellent thermal insulation would, no doubt, have been welcome in hot weather.
Frequently, of course, it was necessary to use beams or rafters of shorter length. King Solomon had made special political arrangements* with King Hiram for the supply of cedar from the Lebanon, but even so his roof-beams were only about 25 feet long (7 metres or 17 cubits). Many Greek temple beams were shorter than this. In the Greek temples, as in Solomon’s building, these short rafters were supported directly from underneath, by rows of pillars, regardless of architectural convenience. In one of the great Doric temples (c. 550 B.C.) at Paestum in southern Italy, there is a line of columns right down the middle of the nave, dividing it into two equal aisles. This must have made any kind of religious ceremony very awkward. In most of the later temples more seemly and more symmetrical arrangements were generally achieved (Figure 6), but even the interior of the Parthenon was cluttered up with pillars which we should think unnecessary.
The simplest form of roof-truss, which was an ‘A’-shaped affair, was developed during the Middle Ages. The horizontal tension member or tie-bar across the bottom of the truss is called by builders the ‘collar’. For short spans it was generally easy enough to find timbers for the collar which were sufficiently long to make a simple triangular truss like Figure 7, but for a small two-storied house this arrangement often results in rather clumsy architectural proportions; moreover, a good deal of spacemay be wasted in the roof. For these reasons builders often attached the collar higher up – in effect putting the upstairs rooms partially within the roof and using dormer windows where necessary. This is all very well, but, if the collar is put high up on the truss, there is a tendency for the rafters to bend or spring outwards under the weight of the roof. This pushes the wall outwards at the same time (Figure 8), very possibly with expensive results. Naturally, the higher the collar is placed, the worse the effect is likely to be.
Figure 6. The more sophisticated temples of the fifth century managed to support their roofs without the use of trusses.
Figure 7. Simple two-storey house with the collar of the roof-truss level with the tops of the walls.
Figure 8. The effect of raising the collar too high in order to save space and cost. (Exaggerated – but not much.)
In large medieval halls and churches, which were often of considerable span, roofing was a serious problem. A trussed roof might be cheaper than an arched or vaulted masonry one, but, even if timbers long enough to make full-length tie-bars or collars could be found, the presence of these collars comparatively low down in the building spoilt the architectural effect of the nave or hall, and, in particular, they blocked the view of the great east and west windows. Since people in those days were often so backward as to pay more attention to appearances than to ‘efficiency’, Continental builders stuck to masonry vaulting, supporting their arched roofs by means of elaborate and expensive buttressing.
Figure 9. Simple hammer-beam roof. The effect is to move the point of application of the outward thrust (which results from the distortion of the truss) further down the walls so that it has less effect on the thrust line. At the same time the view of the end window is kept clear.
Characteristically, the English builders produced a compromise or palliative type of timber roof, which has been described as ‘more ingenious than scientific’. This was the ‘hammer-beam’ roof (Figure 9). Hammer-beam roofs became comparatively popular for large buildings in England, and they can be seen in Westminster Hall, in many Oxbridge colleges and in some large private houses today. They are much admired by the artistic, perhaps partly because of the opportunities which the ‘knuckles’ of the trusses afforded to imaginative wood-carvers. Dorothy Sayers addicts will remember the adventures of Lord Peter Wimsey among the angels and cherubim carved upon the hammer-beams in the church of Fenchurch St Paul.*
In structural terms the main effect of a hammer-beam truss, as compared with any simi
lar large truss with a high collar, is to shift the point of application of the outward thrust further down the supporting walls, so that its effect upon the all-important thrust line is less disastrous. Although this has worked well in practice, the hammer-beam truss has never appealed to the logical Continental mind and there are few examples of it outside this country.
In traditional wooden roof-trusses the joints were made by means of wooden pegs, or sometimes with iron straps. Although these joints were not particularly efficient, the main requirement in such structures was for stiffness rather than strength, and so weak joints did not matter very much. In large modern buildings, such as factories and sheds and barns, roof-trusses are often made up from steel sections such as angle-bars, in which case no particular problems may arise. In small modern houses, however, the roof-truss is nearly always of wood, and the thickness of the timbers has often been cut to the minimum – or even beyond it. The ceiling-joists, in particular, may be barely stiff enough to support the ceiling without causing the plaster to crack. If we are tempted to indulge in the fashionable activity of turning a modern attic into an extra bedroom, the most serious problem is likely to be the stiffness of the floor. Although the roof-truss is unlikely to break, the deflections caused by the extra weight of people and furniture may well cause serious and expensive damage to the house. Amateur handymen, please take note.