Title No.

57-19

Shell at Denver-
Hyperbolic Paraboloidal
Structure of Wide Span

By ANTON TEDESKO

· A free spanning reinforced concrete shell gives an attractive shape to an

exhibit pavilion for a Denver department store. The roof is made up of four
hyperbolic paraboloidal surfaces and is supported through steel hinges on
buttresses in the four corners of a rectangle, I 12 x 132 ft; the 3-in. shell
rises to a height of 28 ft. This paper describes design and construction
features of the structure .

• A HYPERBOLIC PARABOLOIDAL SHELL ROOF has recently been completed

in Denver which increases the span of such shells beyond that of earlier
applications. The structure was chosen by the architects, I. M. Pei &
Associates, as a dramatic entrance feature for the new department store
which is part of the Denver Court House Square Development. A modern
hotel, an underground garage, and a flag-lined skating rink are other
parts of this development by Webb & Knapp.
In current architectural usage "h.p." has come to mean "hyperbolic
parabola," a phrase hard to pronounce. Thus, superior phonetical force
403
404 JOURNAL OF THE AMERICAN CONCRETE INSTITUTE October J 960

ACI member Anton Tedesko is a vice-president of Roberts and Schaefer Co.,

New York and Chicago, who are consulting engineers on major industrial, institutional,
and government projects. Among the first to advocate the use of industrial production
methods in shell roof construction, he has been directly responsible for the design
and development of over 60 major shell projects.
Dr. Tedesko is chairman of ACI Committee 334, Concrete Shell Structures, and a
member of ACI Committees 314 and 315. He is a member of the Reinforced Concrete
Research Council and is chairman of its Task Committee on Composite Construction.

has made an "h.p. shell" out of a "hyperbolic paraboloidal shell." The

h.p. shape and construction are less complicated than the name. Assume
two parallel guide lines in space. If another straight line travels along
these guide lines, the surface thus generated becomes a plane. Now
assume that our straight lines are not parallel and do not intersect
each other and that a straight line again travels along the two guide
lines. A simple warped surface will be generated- the Germans call it a
"ruled surface," as if produced by the simple movement of a ruler. If
the travel of our straight generating line is uniform so that this line
always remains parallel to a given plane, we end up with a saddle
surface which is part of a hyperbolic paraboloid. Such a hyperbolic
paraboloid could also be generated by a second system of straight lines,
because there are always two sets of straight lines in such a doubly-
curved h.p. surface.
Doubly-curved shells are much stiffer than those singly-curved. A
singly-curved shell (a cylinder, a cone) requires stiffening members
to keep it from "unrolling," so as to hold the shell to its original shape
and thereby maintain its great carrying capacity. A doubly-curved
shell (an egg, a dome) is a nondevelopable surface and cannot "unroll"
because it is stiffened by its double curvature. Concrete can be cast in a
doubly-curved mold as easily as in a cylindrical shape, but most of the
great variety of doubly-curved
shells are difficult to calculate and
costly to form.
' There is one notable exception:
the h.p. shell. It is doubly-curved;
it is inversely-curved; the curva-
ture in one direction (convex) is
of stiffening value for the other
direction (in which the curvature
is concave) and vice versa. The
h.p. shell has good carrying ca-
pacity without extra stiffeners. It
is comparatively easy to form,
and the membrane stresses are not
Fig. I - Diagrammatic view of hyper- difficult to calculate. This means
bolic paraboloidal shell that for moderate spans, where
 DENVER SHELL 405

engineers are not much concerned about secondary stresses, it is a

useful shape, as proven by the ingenious and inspiring structures
designed by Baroni, Candela, and others.
DESIGN FEATURES
In this Denver shell, where the span is greater than in previous
designs, secondary stresses had to be considered but the design and
construction were still relatively simple. '
The public enters the department store through the one-story, column-
free exhibit building, which is covered by the shell structure made up
of four h.p. surfaces. The roof is supported at four points in the corners
of a rectangle, 112 x 132 ft, and rises to a height of 28 ft. The shape of the
structure is shown in Fig. 1. The four h.p. surfaces are joined along
ridge bands which form a big horizontal cross. Straight-line marginal
ribs provide the vertical gable framing along each of the sides of the
rectangle.
The shells are generated by sliding a straight line along the horizontal
ridge as well as along the inclined marginal rib. Vertical planes
parallel to the edges would therefore cut the shell along straight lines.
Horizontal planes would cut the shells along hyperbolas (Line c). Vertical
planes in diagonal direction would lead to convex parabolic arcs (Line a).
A vertical plane through a wire stretched between peaks of adjoining
gables would give a concave parabolic cut (Line b). This counter
curvature and the direct compression and tension forces, for which
these surfaces are known, help give the shell an extraordinary balance.
(There is compression along Line a, tension along Line b).

~ SYMM.
t SYMM.

I ~,......__I·~'·o·
3·

II "
.,.
.

"'
z
ii I
...z SECTION "A·A" 'Y

-i
;: I tSYMM.
"'
0 3'·6" 3'-0"

~
0.
0.
::> I I i .
"'
"
1--·
t;_ ~B I
0
1-
SECTION "B·B"
I

..,..,
u.J
til I
J
A

131' • 7 12" C. TO C. SUPPORTING PINS J

SECTION "c-c"

PLAN

Fig. 2 - Dimensions of Denver shell

406 jOURNAL OF THE AMERICAN CONCRETE INSTITUTE October 1960

.. Shear forces are transferred

from shell to ribs and ridges. The
ribs and ridge bands, therefore, are
subject to compression forces from
the accumulating shear transferred
from the shell. The compression in
the gable members increases in
downward direction. The ribs in-
crease in section towards the cor-
ner, thereby keeping the unit stress
fairly constant. The ribs are of
rectangular section, 9 x 20 in. at
their peaks, slightly twisted for
Fig. 3 - Corner detail showing con- architectural effect, remaining al-
crete resting on steel shoe, hinge, and ways at right angles to the tan-
steel buttress
gential plane of the shell (Fig. 2).
The ridge bands take their greatest compression in the crown; here
on account of lack of curvature a critical point exists from the standpoint
of buckling. The architects did not want to provide projecting ribs
along the ridges but wished to have a nonprojecting rib band. The rib
band requires a good deal of reinforcing in the center where it is 10 in.
deep, against 8 in. at the gables.
The shell has a minimum thickness of 3 in. and a basic reinforcing
of two layers of #4 gage mesh with #3 diagonal bars in between. Its
thickness is gradually increased to 5 in. toward the ribs over a distance
of 3 ft. Reinforcing is added in this area of the shell, top and bottom, to
provide for stresses due to flexure.
The shell is supported at the four corners by buttresses of structural
steel (Fig. 3) which are connected to the steel framing of the two-story
underground garage beneath, designed by engineers Weiskopf and Pick-
worth. The four corners of the shell are equipped with shoes of structural
steel. Cylindrical pins of stainless steel provide hinges between shell and
buttresses. The buttresses are oriented in the direction of the thrust,
about 300 tons (including 40 psf snow load) acting approximately in the
diagonal direction, at 35 deg with the horizontal. Fig. 4 shows details of
the hinge and buttress assembly.
CONSTRUCTION FEATURES
The contractor built the falsework after getting his men familiar with
the h.p. problems by first constructing a scale model of the centering.
Fig. 5 shows that 10-in. joints on 2-ft centers were placed along one of
the sets of straight lines in the h.p. surface. Diagonal sheathing, 1 x 8 in.,
tongue-and-groove was nailed to the joists and took the curvature.
The joists rested on 4 x 10-in. caps which were supported by 4 x 4 in.
posts spaced at 10-ft centers, and screw jacks were recessed in the
 DENVER SHELL 407

bottom of these posts. Amount of camber was proportional to the antici-

pated deflection; a camber of llh in. was provided at the crown.
The reinforcing pattern for the shell can be seen in Fig. 6 to 8. The
lines in the diagonal direction are the forming, the lines at right angles
represent the steel which takes the principal tension.
The concreting operation was planned for July 3, to be completed
prior to the July 4 holiday weekend. The temperature rose to 100 F on
the preceding day, and the humidity was low. The engineers recom-
mended that special precautions be taken for hot weather concreting,
such as cooling of aggregates by inundation, addition of ice or a
retarding agent, and of an admixture to reduce drying shrinkage.
However, Webb & Knapp Construction Corp., the contractor, found it
too late to change the mix. Instead, and as a concession to the heat, it
was decided to start work at 4 a.m. when the temperature was 80 F.
Fig. 6 to 8 illustrate screeds, runways, and the concrete placing opera-
tion. A 6-bag mix with 3/4 in. maximum aggregate size produced the
specified 3750 psi concrete strength which would permit decentering
after 7 days. The concrete was placed simultaneously, starting from the
four earners, by one crew in each quadrant. The concrete was fur-
nished by 6-cu yd trans-
it mixers. The two north
quadrants could be
reached by cranes (with
100-ft booms and 20-ft
jibs) and concrete could
be placed by bucket.
Concrete for the south
quadrants was hoisted in
TOP VIEW
the adjoining multistory
structure and carted over
a runway to elephant BUTTRESS

trunks leading to the

roof. For the upper por-
tions of the shell, the
buggies dumped directly
onto the forms. Screeds
were provided for the
shell, but were not con- _ ____c-FINISHED FLOOR
sidered necessary be-
cause the layers of re- SIDE VIEW

inforcing, when prop- 0 6• I'

SCALE
2' 3'

erly placed and tied, fur-

Fig. 4-Details of hinge and buttress assembly.
nished a sufficient guide Hinge is a cylindrical pin. Buttress is welded to
to the concrete finishers. steel beam of underground garage structure
408 JOURNAL OF THE AMERICAN CONCRETE INSTITUTE October 1960

The concrete lost water rapidly with increasing temperatures and

the slump of the concrete at the truck had to be increased from 31fz in.
at 4 a.m. to 51fz in. at 7 a.m. to 61f2 in. at 9 a.m. Midday temperatures
in the shade were 98F with no shade on the roof. The humidity was about
20 percent. Water hoses kept the concrete surfaces wet and prevented
cold joints from forming. The concrete finish was good considering the
weather conditions. The yardage placed was 225 cu yd, the placing time
10 hrs.

,. 69'- o"
10'-0" 'I
i !I i
IL_--lJ----t--t- 2" X 10" (@ 2'- 0"

=
AL
0
-,._'
10

·or
-·
al
11-----JI.--H-+-+--4" X 10" CAPS

2"XIO"
QUARTER PLAN SYMM.
CENTERING

4" X 4" SHORES

~SYMM.
SECTION "A- A"
Fig. 5 - Details of centering. Contractor made model of this centering before
building full scale structure
 DENVER SHELL 409

Fig. 6 - Top view of shell showing screeds and reinforcing. At left is runway
for buggies; concrete at right was placed directly by crane bucket
The crews in each quadrant consisted of: four finishers, two vibrator
men, two steel setters, two carpenters (working on runways, chutes, and
screeds), and two men shoveling concrete, or a total of 12 men per
quadrant. In addition there were on hand: two men at two cranes each,
six men at buggies, four men at chutes, two men unloading buckets, one
hoist operator, and one to four men on water hoses. A total of 35 laborers
were used. Men handling buckets and vibrators switched to water hoses
and to placing curing mats of fiber. Except for the heat, the job would
have to be considered overstaffed; the workmanship was excellent.
On a job such as this when everyone is eager for a fast start with crews
that have not yet gained experience, the first concrete is often placed too
rapidly for sufficient vibration. This shell was no exception. The corners
of the structure were flooded with concrete when it was still rather
dark and inspection was difficult, resulting in some honeycombing in
these areas.
The structure was decentered 14 days after the concrete placing. The
14-day strength was not known, the 7-day strength of the concrete was
2600 psi, and the modulus of elasticity 2,400,000 psi. The structure was
decentered in the afternoon hours when the temperature was highest.
Nine control points were located on top of the rib bands for the meas-
urement of deflections from the adjoining store structure. Hinge move-
ments were measured by micrometer.
410 JOURNAL OF THE AMERICAN CONCRETE INSTITUTE October 1960

Fig. 9 shows schematically the

decentering sequence. The maxi-
mum measured deflection in the
crown was 1% in., close to the pre-
dicted value. Deflections at those
points of the structure which are
closer to ribs and hinges naturally
were smaller. The deflection at the
peak of gables was 1f2 in.
Shells take their worst punish-
ment (have the least safety mar-
gin) at the time of the decentering
when the concrete does not have
its full strength, even though sur-
charge loads may be insignificant.
It is important at such times to
avoid stress concentrations, stress
reversals, and resultant cracking,
byproducts of an improperly plan-
ned stripping procedure.
Fig. 7 - Concrete placing, also show- A step-by-step plan of decenter-
ing screeds and reinforcing of rib bands
ing, such as used here in Denver,
assures stresses proportional or similar to those in the decentered struc-
ture. Carpenters lowering the screw jacks on which the scaffolding
posts rested were divided into groups, and the sequence of operations
was as follows:
Jacks in Area 1 were completely loosened. Jacks in Area 2 re-
ceived one turn, while those in Area 3 received one-half turn. The
one-turn and one-half turn sequence was repeated as often as
required to free the supports from the shell in Areas 2 and 3.

Fig. 8 - Concrete placing

by bucket and buggy nears
completion. Only rib bands
in the crown remain to be
cast. Runway in rear sup-
ports hoppers and elephant
trunks which were used for
placing concrete in lower
portion of marginal ribs
 DENVER SHELL 411

Lowering operations (on one-turn and one-half turn jacks) pro-

ceeded in Areas 4 and 5 until the centering was completely free.
The turning direction for the jacks was painted on the floor to reduce
the possibility of mistakes.
Marginal ribs of the gable, which for dead load and live load of the
shell carry compression only, deflected 3/4 in. from a straight line
between ridge and hinge, just like members in bending. They are
bending members for the dead load of the rib plus the weight of the shell
flange immediately adjacent. Also rib shortening might possibly have
contributed to the deflection. Fine hair cracks developed at the top of
the ribs at the peaks, such as might occur at the center support of a two-

Fig. 9 - Plan of shell showing posts of form centering divided into areas for
decentering sequence

Sequence of turning screw jacks at posts

Area 1 - 1 turn at a time, repeated until free
Area 2 - 1 turn at a time, repeated until free
Area 3 - ]12 turn at a time, repeated until free
Area 4 - Center jacks 1 turn at a time, repeated until free
Corner jacks Y2 turn at a time, repeated until free
Area 5 - Interior jacks 1 turn at a time, repeated until free
Exterior jacks ]12 turn at a time, repeated unti I free
412 JOURNAL OF THE AMERICAN CONCRETE INSTITUTE October 1960

span beam. As a possible explanation, it could be said that the marginal

ribs deflected downward because they shortened; however, the ridge
bands forming the intersection of two shell quadrants prevented the peak
of the gable frame from deflecting as much as it would have otherwise.
Hair cracks resulted since the ribs were not able to move freely down-
ward.
The interior of the structure received only a coat of paint. Vertical
glass walls, located 17 ft from the shell edges, have slotted connections
at roof level to provide for elastic and plastic movements. For each 10
deg F temperature difference, the calculated vertical movement of the
roof was 1/8 in. It was therefore important for the sash contractor to
consider the temperature of the day when the field measurements for
his connections were taken.
The shell received 2 in. of glass fiber insulation and a built-up roofing
with gravel rolled into the asphalt. The insulation was placed between
wood nailing strips on top of the shell and made to run up against the
rib bands with top surfaces flush so that the rib bands were hidden
after the roofing was placed. This required an increase in shell thickness
in a few places. Drainage was provided at the corners where the rain
water spills into gravel boxes at floor level below the hinges.
The cost of the concrete structure was $3 per sq ft; $4.50 per sq ft was
the cost of the building including floor, windows, insulation, roofing, etc.,
but without heating, mechanical, and electrical work. The contractor
had the advantage that when in need he could draw men from his other
construction operations. Had the roof shell been constructed by a sub-
contractor, the cost would have been higher.
SUMMARY
The structure was opened to the public August, 1958.
The shell stresses were calculated according to accepted theory. From
test information it appears that bending stresses are greater and extend
further into the shell from the edges than normally assumed. The so-
called secondary stresses can therefore by no means be considered of
secondary importance in h.p. shells of fairly large magnitude.
The structural design of the shell and supervision of the construction
were furnished by Roberts and Schaefer Co., with the writer in charge.

Received by the Institute Sept. 30, 1959. Presented at the ACI 54th annual convention, Chicago,
Ill., Feb. 26, 1958. Title No. 57-19 is a part of copyrighted Journal of the American Concrete
Institute, V. 32, No. 4, Oct. 1960 (Proceedings V. 57). Separate prints available at
SO cents each.
American Concrete Institute, P.O. Box 4754, Redford Station, Detroit 19, Mich.

Discussion of this paper should reach ACI headquarters in tripli-

cate by Jan. 1, 1961, for publication in the June, 1961 JOURNAL.