You are on page 1of 8

Structural Assessment of Ancient Building

Components, the Temple of Artemis at Corfu

Georg Herdt, Aykut Erkal, Dina D’Ayala and Mark Wilson Jones
University of Bath, UK

Abstract:
The pediment of the temple of Artemis at Corfu is one of the very earliest remains of monumental Greek
construction. This temple, which dates towards the beginning of the 6th century BC (to judge from the
style of the sculptures of the pediment), has suffered extensively since antiquity, leaving us with only a
highly fragmented image. Excavated by Gerhardt Rodenwaldt at the beginning of the 20th century, the
Artemision was reconstructed on paper by Hans Schleif, in their 2 volume publication. The reconstruction
has been criticized in minor respects, but the plausibility of the pedimental sculptures in their proposed
position has been generally accepted. In this paper however structural analysis suggests that the current
position of these sculptures is inconsistent with stability. This triggers a revised reading of the architectural
reconstruction, calling into question the whole plan, with its supposed grand pseudo-dipteral layout.

Keywords:
Ancient Greece, Structural Analysis, Finite Element Method, Corfu, Conservation, Architectural Sculpture

1. Introduction stone-architecture” (Gruben 2001, 112).


Erected in the early 6th century, it holds a key
Due to the ruined state of so many place in the development of the Doric order that
ancient structures, the efforts of architects and went on to govern the appearance of so much
archaeologists are essential to restitute their Greek architecture. The pediment of the temple
original appearance by means of traditional is important as the first surviving substantially
drawings, and, increasingly, using virtual complete assembly of Greek architectural
techniques. However, the consideration sculpture. In addition the monument is striking
of whether proposed reconstructions are in terms of its structure and dimensions. As
structurally stable is often not confronted reconstructed by the excavators, it measures
adequately. Unless physical reconstruction 47.90 metres long, 22.41 m wide and about
(anastylosis) is involved, issues of structural 14 m high, incorporating a complete circuit of
feasibility may not be addressed, or only columns (peristasis) with 8 on the ends and 17
indirectly by virtue of common sense and on the flanks (Fig. 1).
reference to comparanda. On account of
the structural conservatism of many Greek The significance of the Artemision is
trabeated buildings this situation is generally highlighted by the fact that only a generation
unproblematic; yet there are some cases where or two earlier the great majority of temples
structural analysis can be usefully employed to in Greek territories were much smaller and
validate – or not – a particular reconstruction, narrower; they were built of perishable
even one that has been widely accepted and materials and involved relatively modest
reproduced in the literature. spans; their appearance was characterized by
ornament and sculpture that by comparison
must have been much less elaborate (even if in
2. Temple of Artemis truth we lack the information to be absolutely
certain on this point). In the decades of
The Temple of Artemis on the island of transition between humble structures of this
Corfu, according to Gottfried Gruben, is the sort and Artemis’s impressive temple there
“first and mightiest example of developing did exist a few other temples that likewise
Corresponding author: gherdt@gmx.de
1
CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012

Figure 2. Elevation of the temple of Artemis. Existing


Figure 1. Plan of the temple of Artemis at Corfu.
fragments are indicated by tone.

deserve to be called monumental. Two of the 3. Documentation


most notable are the temple of Poseidon at
Isthmia (ca. 650 BC) and the temple of Hera at Following excavations in 1911 to 1914,
Olympia (ca. 600 BC). The former had a cella the two volume report on the sanctuary of
measuring approximately 32 m by 7.9 m, but it Artemis at Corfu was published by Gerhard
is unclear whether there really was a peristasis Rodenwaldt and Hans Schleif in 1939 and
of wooden columns/posts (Gruben 2001, 106). 1940. Their reconstruction of the temple and
The Heraion measures 50 m by 18.76 m and is its dimensions is based on the study of two
the first building known with certainty to have main aspects; the foundations and the sculpted
had a monumental and regular peristasis, one pediment. Presumably because they were
consisting of 6 by 16 columns, though initially of little use to later builders, the sculptures
these were made of wood (Gruben 2001, 52). were not destroyed and remain surprisingly
Soon after our Artemision there followed in complete. Rodenwaldt and Schleif’s restitution
Sicily a succession of large religious structures, of their original disposition is carefully thought
some of them truly gigantic (Gruben 2001, through and represents an important advance in
288). The Corfu temple was not necessarily the the knowledge of Greek architectural sculpture.
project that launched the new monumentality The stonework of the foundations on the other
(this distinction could belong to a different hand, was almost completely removed by later
building that happens to be lost to us), but builders in the locality (Gruben 1996, 91). On
it clearly stands at a critical moment in the this basis Schleif restores the eight columned
development of Greek temple architecture. front as 22.41 m wide, which is about 3½ m
wider than the slightly older temple of Hera
The chief focus of this paper is the detailed at Olympia. The proposed column-spacing is
reconstruction of the pedimented front of the quite dense, with 17 columns reconstructed for
temple and its sculpture as proposed by the the 47.90 m long flanks. Since so little is known
excavators. The association of the sculptures about the substructures (the layout of the
with this building is not in doubt; however foundations being determined by the change of
their capacity to bear load requires evaluation. fill in the presumed robbed trenches – hardly
A revised reconstruction has the advantage of the most certain method), Schleif sought to
being more plausible from the point of view of give order to the plan by applying a foot grid.
statics and constructional soundness, and this The result was questioned in 1943 by Hans
in turn has implications for the layout. Riemann, who suggested a different value for
the foot. However, for seventy years no doubts
have been raised over the more fundamental

2
Structural Assessment of Ancient Building Components, the Temple of Artemis at Corfu
Georg Herdt et al.

section steps in higher up, as can be observed


at the left panther-lion, and then continues to
narrow towards the top, presumably to lighten
the blocks for the sake of lifting them (Fig. 3).
Around the heads of the two panther-lions the
stone has been trimmed down to as little as 5 cm
thick behind the feline’s mane. And even worse,
the block displaying the Pegasus is reported to
narrow down to 4 cm (Schleif 1940, 25).

Four blocks of the raking stone geison


remain, each measuring about 2 m in length.
The surfaces of these stones show a series of
cuttings, allowing a definite allocation of their
Figure 3. The block with of the left lion-panther according position and that of the adjoining blocks. At
to Rodenwaldt. their lower surface an erosion mark (67.5
cm from the front) indicates the junction
points of Rodenwaldt and Schleif’s work, and in with the next components. Connections were
particular the monumental scale of the edifice, effected by means of dowels, the cavities for
the extent of which matches their reconstruction which still remain visible on each geison-
of the pediment (Kähler 1949, 42). block. The upper surfaces of the geison blocks
also display a smoothed area as well as holes
The pediment and its sculptures are for connecting to the sima, with dove-tail
indeed impressive. The Gorgon itself in the connectors for joining each block and a large
centre of the tympanon measures more than 3 socket at the rear for the timber beams that
m in height. The overall length of the pediment once spanned across the porch. The smoothed
is about 20 m, though there is an element of surface matches the dimensions of the second
estimation due to some missing parts (Fig. sima, the marble sima, of which some elements
2). Generally, the sculptural blocks are in survive. This must have substituted an earlier
pieces but in a good condition; many surfaces sima, fragments of which also survive made of
are intact, and for some blocks the outline of terracotta and about 80 cm in height, as well as
missing parts can be confidently reconstructed. some tiles.
Each block incorporates a decorated framing
strip, one at the top and one at the bottom; the To define the cross-section of the pediment,
upper follows the pitch of the roof under the Schleif placed the sculptures centrally below the
raking geison, the other sits on the horizontal raking-geison; their position is set at 67.5 cm
geison. The decoration of these two strips is from the front by the limit of the erosion due to
different; a chevron pattern for the upper and a weathering. The sculptures were thus sheltered
meander pattern about 30 cm tall for the lower. from rain, but two problems arise with this
position. The first concerns the sequence of
The sculpted taller blocks at the centre construction, since the geison is placed on
of the pediment present an increased width at top of the sculptures. For this to be the case
the bottom, forming a stand which is preserved these must have been inserted into position
up to a maximum depth of about 55.5 cm. before the roof could have been placed onto the
The depth of the sculptures is 17 cm at the building, the inverse of usual practice. But more
sides and roughly 29 cm near the centre. This importantly: Schleif’s reconstruction presumes
leaves a surplus of the stone material of 26.5 that load from the roof can be applied to the
cm to form the background at the bottom. The sculpted blocks, and he supplies a section that

3
CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012

sima directly onto the massive geison block,


while Solution B engages an additional layer
in between these two components. Solution B
therefore raises the sima about 25 cm, which
is consistent with a decorative bead at the edge
of the geison. According to Schleif it is the
presence of the bead that makes Solution B more
probable than Solution A (Schleif 1940, 57 Abb.
43.). However, this fill layer brings additional
load onto the sculptures and the slender wall
to which they are attached. In order to check
the level of potential stresses, structural
analysis using Finite Element Modelling has
been performed. The presumed properties of
the materials and loads, necessary to perform
the analysis, are given in Table 1. These have
been chosen from the literature (Schleif 1940,
21) on the basis of the archaeological evidence
using limit conditions assumptions. In addition
to the self-load, tiles have been applied as
Figure 4. a) Temple of Artemis, section through the uniformly distributed loads over the roof and
pediment at the ‘dying-giant’ (Schleif) and b) at the
terracotta as a line load on the outer edge of the
‘panther-lion’ (Herdt).
roof. More accurate results could be obtained
makes it look plausible (Fig. 4a). However, he by a detailed survey of the remaining fragment
does not supply a section through other blocks and characterisation of the original materials.
which have been tapered by the sculptors to
the point that their structural capability must
be called into doubt (Fig. 4b). Indeed it seems 4. Structural Analysis
doubtful that Greek sculptors ever intended
such a slim profile to bear load. Hence the need The Finite Element Method (FEM) has
to test its feasibility by means of structural been used to structurally analyse the proposed
analysis. reconstruction of the pediment. The model
replicates the left lion-panther sculpture and
For this purpose, the weight of the roof the objective is to verify whether the thinner
has to be assessed. The surviving blocks were portion of the stone section would be able to
combined into two alternative reconstructions withstand the roof loads as proposed in Schleif’s
by Schleif, of which one results in a lighter hypothesis B. After generating its geometry
construction for the roof than the other. in AutoCAD, the model has been transferred
Solution A places the monumental terracotta into the finite element software ALGOR (Fig.

Structural Materials Non-structural Materials Table 1. Material


Properties and
Properties Lime Stone Fill Layer Mass Roof Tile Cover Sima
Loads.
(raking (Timber, Density and (Compacted Clay) (Compacted Clay)
geison) Clay, Earth, Resultant Uniformly Uniformly
Pebble) Dead Loads Distributed Distributed Line
Surface Load Load
Mass Density 3500 kg/m3 2200 kg/m3 Mass Density 2500 kg/m3 2500 kg/m3
Elasticity Modulus 30 Gpa 0.2 Gpa Dead Loads 2250 N/m2 25000 N/m
Poisson Ratio 0.25 0.30

4
Structural Assessment of Ancient Building Components, the Temple of Artemis at Corfu
Georg Herdt et al.

parts have been assigned


coarser meshes (Fig.
6). The model was
constructed using 8-node
3-dimensional brick
finite elements having 3
translational degrees of
freedom at each node. The
advantage of using these
elements is to provide
a complete 3-D state
of stress in the stones.
Boundary conditions
at the base of the stone
have been assumed as
fixed, i.e. the base of the
stone is not able to move
downward under the
applied load. Although
Figure 5. Model generated in AutoCAD. a) Schematic Model; b) CAD Model this is not necessarily the
case, as some deflection in
5). Meshing (discretization) has been applied, the bottom of the pediment between columns
paying attention to each part in the model. As might have occurred under the weight of the
such, the sculpture and the background stone stones above, the error committed is of the
have been finely meshed to better analyse same order of magnitude as the other numerical
the critical thin section where higher stress assumptions in the model.
concentrations are expected, while the other

Figure 6. Model geometry including the structural parts and meshing. a) Perspective View; b) Front View; c) Right-
side View.
5
CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012

tensile strength is in the range of 1 to 2 MPa.


This allows only a very modest safety factor,
which would be in contrast with the usual
overdesign of Doric temples. Figure 9 shows
the distribution of minimum principal stresses,
i.e. the value of maximum compression at any
point in the stone. In general the values are
very low and far from the typical compressive
strength of a limestone, usually in the range
between 20 and 200 MPa, for un-weathered
un-fractured limestone. However it should be
noted that where the cross section is reduced,
just above the head of the panther, the stress
is at its worst, 4.45 MPa. Given the way in
which the load is transferred from the geison,
performance is best simulated by considering a
concentrated load over the thin slab of stone.
The strength to point load of limestone, is much
lower than its uniaxial compressive strength at
Figure 7. Displacement Magnitude (Contours of resultant about 2 MPa to 10 MPa, hence again creating
displacements). a condition which would lead to either a very
small safety factor or even failure. The results,
5. Analysis Results and Discussion therefore, suggest that the structural stability of
the structure is not fully ensured for Schleif’s
The results show that the maximum hypothesis B.
displacement magnitude is as expected at the
edge of the pediment overhang, a very modest Thus numerical analysis, thanks to the 3D
0.156 mm (Fig. 7). The critical stresses are modelling, is able to identify the concentration
detected at a position just below the strip of stresses resulting from the specific proposed
decorated with the chevron, where the depth geometry and construction, and it shows that the
of the background stone is at its
minimum.

The distribution of the


maximum principal stresses
shown in figure 8 demonstrates
that tensile stresses are
generally very low except for a
concentration of stress at the
necking of the block between
the body and the head of the
panther where the cross-section
reduces. The value in this area
is as high as 0.71 MPa. Values
of strength for Corfu Limestone
have not been found in the
literature, though a typical
Figure 8. Maximum Principal Stresses. a) Front and right-side views;
average value for limestone b) Back and left-side views.

6
Structural Assessment of Ancient Building Components, the Temple of Artemis at Corfu
Georg Herdt et al.

Figure 9. Figure 10. Temple


Minimum of Artemis, proposed
Principal new position for the
Stresses. sculptures of the
pediment, divorcing
them from the structural
system.

sculptures would not able to bear the assumed


loads. Inclusion of likely additional components
not included by Schleif - for example a roof
acroterion - would make the situation even minor change has significant implications. The
more severe. Indeed such a feature might well supposed horizontal geison with an inclined
be expected, and in this period acroteria were leading edge has to be altered into a more
large and prominent and therefore heavy. The typical flat-topped geison, the common shape
weight of any such embellishment would arrive for horizontal geisa, which would now host the
at the edge of the verge, increasing stresses on sculptures up to near its edge. This in fact was
the sculpted blocks underneath. Furthermore, the normal condition judging from later well-
if the designer of the temple intended to preserved temples (Schwandner 1985, 108).
bring load upon these elements, he surely
would not want to weaken them by tapering. Due to the geometry of the corner of the
Schleif’s solution opposes normal principles of pediment, the modification of a shallower
craftsmanship by transmitting structural forces geison results in a more compact elevation,
through fragile parts of great value, due to an which would be at least 1 m shorter and could
artistic effort invested in the sculpture. perhaps be as much as 3 m shorter. Such a
reduction would require rethinking not just
the façade but also the plan as a whole. A more
6. Discussion compact elevation favours a façade with six
instead of eight columns, like so many other
As the structural analysis indicates, temples of that period. A front with seven
Schleif’s proposal loses credibility due to the columns is also hypothetically possible. As
risk of failure involving the most precious part of his eight-column-wide proposal Schleif
components of the building, inviting alternative imagined what the Roman architect-writer
solutions. Indeed, it seems more likely that Vitruvius called a pseudo-dipteral layout, and
the sculptures were disconnected from the both the resulting elevation and the plan have
structural system. This can be achieved by been widely reproduced since their publication
reconstructing the sculptures closer to the front seventy years ago. A pseudo-dipteros fits within
of the pediment and placing a proper load- the overall envelope of a dipteral or double
bearing wall at the back (Fig. 10). This relatively colonnade, except that there are no inner

7
CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012

rows of columns. This involves a wide span for Barletta, B. 2001. The Origins of the Greek Architectural
the peristyle; indeed at Corfu – according to Orders. Cambridge and New York: Cambridge University
Schleif – this reaches almost 5 m, which is to Press.
say substantially greater than that the span of
the temple interior. Be it noted that Vitruvius Baykasoğlu, A., H. Güllü, H. Çanakçı, and L. Özbakır.
writes that this pseudo-dipteral scheme for 2008. “Prediction of compressive and tensile strength
temples was only invented by Hermogenes, of limestone via genetic programming.” Expert Systems
who worked in the early second century BC. with Applications 35 (1–2): 111-123.
Can we be sure that such an enterprise was
possible four centuries earlier, not long after Coulton, J. J. 1977. Ancient Greek Architects at Work.
600 BC? Greater confidence over the design of Problems of Structure and Design. Oxford: Oxbow Books.
the Artemision can probably only be achieved
by inspecting once more what remains of its Gruben, G. 1996. “Die Entstehung des griechischen
foundations. In fact, an archaeological re- Tempels.” In Klassische Bauforschung, G. Gruben (2007)
evaluation of the site now seems imperative. 66-109. München: Hirmer.

Gruben, G. 2001. Griechische Tempel und Heiligtümer.


7. Conclusions München: Hirmer.

This paper reveals that the hitherto Kähler, H. 1949. Das griechische Metopenbild. München:
accepted reconstruction of the Temple of Münchner Verlag.
Artemis should be called into doubt. Schleif’s
solution involves putting load onto the Rémondini, L., J. C. Léon, and P. Trompette. 1998.
sculptures and this does not stand up to “Towards an integrated architecture for the structural
scrutiny, nor a complete three-dimensional analysis of mechanical structures.” Computers &
structural analysis using Finite Element Structures 67 (5): 299-307.
modelling. Assessing ancient building
components structurally help us find plausible Riemann, H. 1943. “Zum Artemistempel von Korkyra.”
positions for them. It is a tool that can be used Jahrbuch des Deutschen Archäologischen Instituts 58:
to guide new reconstructions and anastylosis, 32-38.
or for scrutinizing published proposals as in
this case. By combining structural analysis Rodenwaldt, G. 1939. Die Bildwerke des Artemistempels
with attention to the archaeological details (Korkyra. Archaische Bauten und Bildwerke 2). Berlin:
of surviving blocks it should be possible to Gebr. Mann.
reconstruct ancient monuments in ways that
are structurally plausible. Schleif, H. 1940. Der Artemistempel. Architektur,
Dachterrakotten, Inschriften (Korkyra. Archaische
Bauten und Bildwerke 1). Berlin: Gebr. Mann.
References
Schwandner, E.-L. 1985. Der ältere Porostempel der
Autodesk Algor Simulation Professional. 2011. Autodesk Aphaia auf Aegina. Berlin: W. de Gruyter.
Inc, USA.
Wilson Jones, M. 2013. Origins of Classical Architecture.
ASTM, Philadelphia, Pennsylvania. 1984. “Standard Test New Heaven: Yale University Press.
Method for Unconfined Compressive Strength of Intact
Rock Core Specimens, Soil and Rock, Building Stones.”In
Annual Book of ASTM Standards 4.08. Philadelphia:
American Society of Testing and Materials.

You might also like