Construction and Building Materials 181 (2018) 510–526

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Measuring the effective Young’s modulus of structural silicone sealant

in moment-resisting glazing joints
Adam D. Lee a,b,⇑, Paul Shepherd a, Mark C. Evernden a, David Metcalfe c
a
Department of Architecture and Civil Engineering, University of Bath, Claverton Down, Bath BA2 7AY, UK
b
PTCC Facade Design, Telecom Plaza, 316 Senator Gil Puyat Ave., Makati City, Metro Manila 1200, Philippines
c
Centre for Window and Cladding Technology (CWCT), The Studio, Entry Hill, Bath BA2 5LY, UK

h i g h l i g h t s

Structural silicone sealants bond glass panes to glazing frames and curtain walls.

These sealant joints can provide torsional bracing for structural members.

An new method of analysis is proposed, and verified experimentally.

The new model may result in more efficient use of metal in glazing systems.

a r t i c l e i n f o a b s t r a c t

Article history: Structural silicone sealants are synthetic rubber adhesives used in the construction industry to bond glass
Received 29 January 2018 and other sheet infill materials to the frames of windows and curtain walls. In this paper, two different
Received in revised form 25 May 2018 algebraic expressions are proposed to describe the way in which the rotational stiffness of the adhesive
Accepted 6 June 2018
connection – resistance to moments acting about the axis of the joint – varies with the sealant’s cross-
sectional dimensions and elastic modulus. Laboratory testing of DC-983, a two-component structural sil-
icone sealant used widely in factory prefabricated glazing applications, has, with some caveats, validated
Keywords:
the mathematical models.
Facade design
Structural silicone sealant
Ó 2018 Elsevier Ltd. All rights reserved.
Structural glazing
Elastic modulus
Curtain wall
Mullion

1. Introduction tions are illustrated in Fig. 1, in four cross-sections through

extruded aluminium ‘‘mullion” profiles, which are those that span
In the middle of last century there began to emerge, hand in vertically from one floor to another.
hand with the glass box architectural style, a new method of con- The first of the mullions, Fig. 1-A, is a simple box section. Alu-
structing tall buildings. First, a frame made up of columns and minium curtain wall profiles of this sort, which must be cut and
beams was erected, to support the floors; then, to keep out the assembled at the construction site, became popular in the 1960s
weather, the freestanding structure was enclosed with a light- [e.g. [2]] and, for some applications, are still in use today. Glass is
weight, metal framed, skin [1–3]. Since that time, designers of retained at the face of the wall system using a mechanical clamp.
these exterior ‘‘curtain walls” have been using the same set of Rubber gaskets permit relative movement, in the plane of the wall,
assumptions when modelling the forces that are transferred from between the glass and the metal frame. So, the glass does not pre-
the sheet material covering the facade, which is often glass, to vent the aluminium members from moving laterally. Also, the
the members in the supporting frame. These structural idealiza- mechanical clamp at the edge of the glass permits rotation, as
shown in Fig. 1-A, so glass deflections do not cause the framing
member to twist about its longitudinal axis.
⇑ Corresponding author at: PTCC Facade Design, Telecom Plaza, 316 Senator Gil Another means of securing glass to its frame is to use an adhe-
Puyat Ave., Makati City, Metro Manila 1200, Philippines.
sive. This approach, known as ‘‘structural glazing” and shown in
E-mail addresses: [email protected] (A.D. Lee), [email protected]
Fig. 1-B, is relatively new. The first high-rise tower with a struc-
(P. Shepherd), [email protected] (M.C. Evernden), [email protected]
(D. Metcalfe). turally glazed curtain wall was completed as recently as 1986

https://doi.org/10.1016/j.conbuildmat.2018.06.038
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 511

A B C D
Fig. 1. Cross-sectional shapes of vertical mullions. In detail ‘‘A”, glass is held mechanically to a simple box section. Structural sealant, in detail ‘‘B”, bonds glass to the male and
female profiles of a unitized wall’s split mullion: this design’s structural idealization is shown in ‘‘C”, although ‘‘D” may be a better represent the wall’s actual behaviour when
subjected to wind load.

[[4] p. 53]. Since then, architects have embraced the new aesthetic, sional resistance is small. In a typical, unitized curtain wall system,
using the technology to create large, flat facades, without any the panes of glass or other sheet infill materials that are connected
metal components to the exterior of the glass. Structural glazing to the mullion’s outer flanges have ample structural capacity to
has become a common and conspicuous feature of large buildings serve a torsional braces. However, current design guides advise
around the world. Current structural design methods and usage that, even in structurally glazed systems, glass and infill materials
guidelines for the adhesives – structural silicone sealants – are should not be considered to be restraints [e.g. [10]]. Consequently,
detailed in ASTM C1401 [5]. The reasons for inclusion of the ‘‘glaz- in structural analysis, mullions are modelled [as in [11] Part VIII,
ing tape” shown in Figs. 1-B and -D, and the effect that this tape has pp. 56–61] with the assumption that no moment is imparted to
upon structural performance, are explained in Section 7.3. them by the glass. The structural idealization of the glass support
Fig. 1-B shows the E-shaped male and female extrusions that is a hinge, as shown in Fig. 1-C.
together form the split mullion of a modern unitized curtain wall. In reality, because the structural sealant joint has stiffness,
In such designs, the facade is made up of discrete panels that can Fig. 1-D might better describe a unitized mullion’s mid-span con-
be prefabricated. Because of this, and other, practical advantages dition under wind load. In this diagram, positive wind pressure
[[6] p. 4–5; [7] p. 86], the great majority of the world’s new curtain causes the glass to deflect toward the interior of the building
wall is of this type [[8] p. 82]. From a structural standpoint how- and, as a consequence, moment is transferred through the struc-
ever, the split mullion’s narrow profiles are, in torsion, less rigid tural silicone sealant to the mullion profiles, whose inner flanges
than the box sections they replace (Fig. 1-A). Consequently, it is move toward each other. The onset of LTB in the mullion profiles
frequently the case that prevention of buckling is the dominant will, therefore, be affected by the moment resistance of the sealant
concern for today’s facade engineers. joint.
Lateral torsional buckling (LTB) is the mode of structural failure Facade engineers are interested in improving current methods
caused and characterized by extreme axial rotation of a flexural of predicting LTB [12,13] because, with continuing advances in
member’s cross section. At the onset of failure by LTB, a glazing the sizes of the panes that can be processed by glass fabricators,
system’s profiles deflect in the manner shown in part D of Fig. 1. the structural members in exterior wall systems are becoming
The analysis of LTB is complex [e.g. [9] Chapter 5], and is affected increasingly slender [10]. Research by others [e.g. [14]] suggests
by parameters aside from bending moment distribution, material that a structural silicone joint may provide sufficient support to
properties, cross-sectional shape and distance between supports. prevent LTB in some cases, but a recent survey [12] showed that
Other significant particulars are the member’s initial straightness, facade design professionals have insufficient information to assess
and also the load eccentricities, which may themselves be func- whether an attachment to a glazing system’s frame will be effec-
tions of the profile’s rotation. tive as an LTB restraint. The analytical steps proposed in this pre-
If one of a member’s flanges is restrained to prevent it from sent paper might therefore be incorporated in a more
rotating about its long axis, then LTB can be prevented. The comprehensive model, to predict the angle through which a fram-
moments that are transferred to such braces can be estimated ana- ing extrusion will rotate when full design load is applied, and thus
lytically [e.g. [9] Eq. 12.10], and the magnitude of the required tor- demonstrate that the frame’s resistance to LTB is adequate.
512 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Intuitively, it might seem dangerous to create structures using tural sealant joint, and to collect experimental data with which to
metal flexural members that remain stable only because they are validate the mathematical model.
supported by restraints made of glass. Glass is, after all, a brittle The terms ‘‘elastic modulus”, ‘‘Young’s modulus” and ‘‘modulus
material, and building facades must be designed with the expecta- of elasticity” are equivalent, and may be used interchangeably. For
tion that occasional breakages will occur. It is therefore worth a sealant, the published value of this property is usually that mea-
explaining that the governing design loads acting upon a building’s sured 14 days [e.g. [17]] or 21 days [18] after creation of the sam-
facade are usually wind pressures. So, if breakage destroys a pane, ple. Research [[17] p. 967] has, however, revealed that the Young’s
it is true that its frame will no longer be restrained by the glass but, modulus of a particular structural silicone sealant, DC-995, can rise
at the same time, the frame will no longer receive wind load. dramatically during the 100 days following the assessment on day
Therefore, failure of the glass will not cause a failure of the metal 14. In that single-component sealant – one that does not need to be
structure that it restrains. mixed with a separate catalyst prior to application – the increase in
With greater understanding of the joint’s behaviour and with a elastic modulus ‘‘was in the range of 850–950%”. If the results of
more sophisticated structural model, it may be possible to reduce this study are to be applied in practice then there is a need to
the mass of metal required to construct a curtain wall. The oppor- understand the changes in Young’s modulus that occur as a sealant
tunity for metal savings exists because, in the design of modern ages. Therefore, the rotational stiffness of a two-component sea-
mullion extrusion shapes, it is common that stability – in particu- lant, DC-983 – a type that is commonly used when structurally
lar, resistance to lateral torsional buckling – is the governing struc- glazed curtain wall panels are prefabricated in a factory – has been
tural design consideration. If structural sealant joints can be shown assessed at 14, 114 and 214 days. Also, each sample’s indentation
to provide effective lateral or torsional restraint for their frames, hardness was measured at 214 days.
then facade engineers will be able to make use of lighter profiles Apart from the rubber pads used in vibration-absorbing mount-
[e.g. [14]]. The pursuit of efficient curtain wall design solutions is ings, elastomeric materials are rarely encountered in building
worthwhile, not only because of the cost savings that can be structures, and it may therefore be helpful to summarize their spe-
attained by reducing material usage, but also because, amongst cial properties. Structural silicone sealants are viscoelastic, mean-
common building materials, the embodied energy in aluminium ing that an applied load causes damped elastic deflection, and so
is unusually high [[15] p. 74] and so significant environmental ben- the magnitude of stress depends not just upon the magnitude of
efit can be achieved by controlling metal content [16]. strain, but also upon the rate of change in strain. Fig. 2, shows
The manner in which a mullion profile might be caused to the relationship between tensile stress and strain [[19] Table 1]
rotate about its lengthwise axis is illustrated in Fig. 1-D. When for the particular sealant that has been tested in this study, DC-
the magnitude of this rotation is sufficiently large then, aside from 983. It can be seen that, even when strain is increased at a constant
any structural consequences, other functions of the wall can be rate, in accordance with ASTM C1135 [18], the material’s beha-
affected. For example, it is possible that the interior flanges of viour is not linear-elastic. If, however, the product is used within
the male and female profiles will disengage, breaching the weather the allowable range defined by the manufacturer, so that strain
seal. Curtain wall system failures of this sort, caused by excessive does not exceed 25%, then a linear-elastic idealization results in a
rotation of a member about its extrusion axis, could be predicted maximum error of less than 9%.
during the design process if facade engineers were provided with While considering the precision with which a sealant’s elastic-
a model of the relationship between moment and rotation in a ity can be modelled, it should be noted that the expected level of
structural silicone joint. For these two reasons – firstly to help cur- variability in laboratory measurements is high. If a sealant’s stress
tain wall designers determine whether a particular structural sea- is measured twice, at 10% tensile strain, using the ASTM C1135
lant connection can provide torsional bracing for a mullion, and method, there is 95% probability that the two measurements will
secondly to quantify the axial moment transmitted – the objective differ by less than 0.041 MPa (6 lbf/in2) if the tests are carried
of the research described in this paper has been to develop a sim- out in the same laboratory, or by 0.090 MPa (13 lbf/in2) if carried
ple algebraic method to describe the rotational stiffness of a struc- out in different laboratories [[18] Table 1]. For the DC-983 sealant

Fig. 2. Stress with respect to strain in DC-983 structural sealant, measured by the ASTM C1135 laboratory method [18]. Also plotted is a linear idealization, for the useable
range defined by the sealant manufacturer, in which strain is limited to 25%.
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 513

tested in this present study, these 95% probability ranges are, length of sealant joint, M, can be obtained by integrating torque
respectively, equivalent to 20.7% and 44.8% of the published value over the width of the triangular compression and tension zones that
of stress at 10% strain, which is 0.2 MPa [20]. have been shaded in Fig. 3:
Z B
2Ess 2 2x2 Dg
2. Algebraic model of moment resistance M¼ dx; ð2Þ
g x¼0 B
In the model that is commonly used by practising facade engi- if the x-axis is horizontal in the right hand diagram of Fig. 3. Hence:
neers to assess the load capacity of a structural sealant joint [[5]
Section 30], it is assumed that the glass remains flat when wind B2 DgEss
M¼ : ð3Þ
pressure is applied, and that the sealant experiences only tensile 6g
or only compressive stress. Several researchers [e.g. [21]] have car- If the eccentricity of F from the joint’s centreline is Q then M ¼ 2FQ .
ried out sealant analyses that take glass deflection and the sealant’s Substituting for F and M from Eqs. (1) and (3) gives the value of Q,
shear stresses into account. In these more sophisticated studies, which is B=3. This is, plainly, the standard result for a triangular
solutions for particular design conditions have been obtained load zone, in which the amplitude of force varies linearly with
numerically, using finite difference or finite element techniques. distance.
While the results obtained in this way are comprehensive – the If the angle through which the glass edge rotates is a then, from
entire stress field is revealed – the process of preparing numerical Fig. 3, tan a ¼ 2Dg=B. These rotations are small – usually only a few
models is time consuming, making it impractical to use this degrees – and so a is a close approximation to tan a, when a is
approach to investigate the many and varied design cases that measured in radians. This small angle assumption is reasonable
might be encountered in a real building’s facade. It has therefore while a is less than 0.176 radians (10°). In this range, a ¼ 2Dg=B,
been this present study’s aim to find, and to validate experimen- and Eq. (3) can be rewritten:
tally, a set of simple, closed-form algebraic expressions that may,
in the future, be incorporated in a design code. B3 Ess
M¼a : ð4Þ
For the sake of simplicity, here the assumption will be made 12g
that structural sealants obey Hooke’s law. The validity of this
approach is discussed later, in Section 7.1. Adopting a linear-
elastic model makes it possible to consider separately the different 3. A more sophisticated model of moment resistance
components of load, and to sum their effects using superposition.
The application of a pure moment causes a unit length of sealant Wolf and Cleland-Host [22] used a polynomial expressions to
joint to deform elastically from its original rectangular cross sec- describe relationship between stress and strain in two-part struc-
tion into a trapezoid, as shown in Fig. 3, then the force in the equiv- tural silicone sealants. Their coefficients, below, were chosen to
alent couple, F, can be expressed in terms of the sealant’s Young’s fit the experimentally determined responses of two commercial
modulus, Ess . Considering either side of the sealant joint – the area products, A and B, that had been tested at 22  C after one year of
in compression or the area in tension – the mean stress is half of aging. For a given strain, e, the corresponding stress, f, was found
the extreme fiber stress. Also, recalling that the force required to to be:
extend a linear-elastic material is the product of cross sectional f A ðeÞ ¼ 0:87244e5  1:74222e4 þ 1:59336e3  1:17958e2 þ 1:01308e ð5Þ
area, strain and elastic modulus:
f B ðeÞ ¼ 3:80874e5  7:37014e4 þ 4:44015e3  1:02037e2 þ 0:93614e ð6Þ
BDgEss
F¼ ; ð1Þ These two stress-strain curves are plotted in Fig. 4.
4g
The market names of the sealant products tested by Wolf and
where B is the ‘‘bite” or width of the adhesive plane, g is the ‘‘glue- Cleland-Host were not revealed, and so it cannot be assumed that
line” or sealant thickness, and Dg is the maximum distance through their elastic properties will match those of the sealant tested in this
which the sealant extends. Similarly, the sum total moment per unit study, or those of the sealant used in a given glazing system. How-

Fig. 3. The edges of a glass pane rotate (Left) under the action of wind pressure. A section through the loaded structural sealant joint (Right) shows its trapezoidal shape.
514 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Fig. 4. Curves chosen by Wolf and Cleland-Host [[22] Table 2] to fit experimental measurements of stress and strain in two different two-part structural silicone sealants.

ever, in Fig. 4 it can be seen that the sealants’ elastic moduli are tension zone, and also in the compression zone. Because of its sim-
greater in compression than they are in tension, and in this discus- plicity, that previous model still may be of interest to engineers,
sion the presumption is that such asymmetry is typical amongst all but, in reality, Young’s modulus is a function of strain, and is
structural silicone sealants. A further supposition is that the stress- greater in compression than in tension. Therefore, when a pure
strain curve for any particular structural glazing sealant, or a rea- moment is applied about the axis of a sealant joint, it will deflect
sonable approximation to it, can be obtained by applying a con- in the manner sketched in Fig. 5.
stant scaling factor to the stress function in Eq. (5) or (6). In order to determine the joint’s rotational stiffness, a first step
Expressed another way, for a sealant Z: is to develop an expression for the width of the tension zone, l.
Because the joint is in static equilibrium, the total tensile force is
f Z ðeÞ ’ kZ f A ðeÞ; ð7Þ
equal to the total compressive force, so:
where kZ is a constant. Z  
l
x Dg
While formulating the previous rotational stiffness model, 0¼ f dx: ð8Þ
described in Section 2, the stress-strain curve from a tensile test x¼ðBlÞ l g
was examined, and it was argued that, within the range between The result may be used to determine l, and then the total moment
zero and 25% strain, a linear-elastic approximation is sufficiently for a given Dg can be found by integration:
accurate for engineering design purposes. A constant value of Z  
Young’s modulus was then assumed to apply in the sealant joint’s
l
x Dg
M¼ xf dx: ð9Þ
x¼ðBlÞ l g

When the sealant’s stress-strain curve is given in polynomial form,

as in Eq. (5), mathematical integration is easy. However, the results
cannot readily be reduced to a simple closed-form algebraic for-
mula relating rotation to moment. What is needed is an approxima-
tion that can be expressed simply, and which is, at the same time,
less crude than the model proposed in Section 2.
In the analysis that follows, it is assumed that a constant mod-
ulus of elasticity, Ess , applies throughout the tension zone, and that
its value is independent of strain. It is also assumed that the mod-
ulus of elasticity within the region that is in compression, EC , is
constant for any given angle of joint rotation, but it’s value
increases with rotation. The applicability of this set of assumptions
is discussed in Section 7.1.
Applying these new premises to the sealant joint shown in
Fig. 5, the equivalent concentrated force in the triangular tension
zone, F T , and in the triangular compression zone, F C , can be found
in the same way that they were found for Eq. (1). Static equilibrium
is achieved when F T þ F C ¼ 0, hence:
l Dg ðB  lÞ ðB  lÞ Dg
Ess þ EC ¼ 0; ð10Þ
2 g 2 l g

Fig. 5. Cross section through structural sealant deflecting solely because of a which reduces to:
moment about the joint’s axis, with asymmetric tension and compression zones (cf. 2
Fig. 3). The positions of equivalent concentrated forces are based on the premise Ess ðB  lÞ
¼ 2
: ð11Þ
that Young’s modulus is constant in the tension zone and constant in the EC l
compression zone.
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 515

By definition, if e is a tensile strain, and if f is a stress function sim- to the axis of the extrusions, as shown in Fig. 7: the width of the
ilar to Eq. (5) or (6): contact surface between the sealant and the painted metal
substrate or ‘‘bite”, B, was 24.2 mm, and the joint’s length was
f ðeÞ
Ess ¼ : ð12Þ 110 mm. The glazing tape, seen in Fig. 7, was not removed: it
e remained in place throughout the test process.
The strains on the compression side are smaller in magnitude than The thickness of the metal in the extruded aluminium box
those on the tension side, and e is a tensile strain, so: sections was sufficiently great – approximately 7 mm – that the
 
f ðBlÞ
l
e magnitude of deflections occurring within the metal parts during
EC ¼ : ð13Þ testing was negligible.
ðBlÞ
l
e DC-983 is available in two colors, gray and black, and the two
varieties have different physical properties [20]. The samples used
Substituting the above two expressions, for Ess and EC in Eq. (11),
in this study were black.
and simplifying:
  After the sealant samples had been created, and during the peri-
ðB  lÞ ðB  lÞ ods between tests, they were stored in a covered outdoor location
f ðeÞ ¼ f e : ð14Þ
l l where they were not exposed to rain or direct sunlight. There, the
ambient temperature varied between 20  C and 36  C, and the rel-
The way in which the width of the tensile zone varies with strain
ative humidity ranged from 65% to 85%. Commentary in Section 6
has been determined numerically, using Eq. (14), and the results
considers the effect that environmental conditions have upon the
are presented in Fig. 6. The graph shows that, at the limit of allow-
sealant’s physical properties.
able strain, when Dg=g ¼ 0:25, which is the condition that will be of
For each test, one of the sample’s two extrusions was clamped
greatest interest to designers, ðB  lÞ=l ’ 0:78, which is to say that
to a rigid support as indicated in Fig. 7. A counterbalance was posi-
l ’ 0:56B.
tioned so that the axis of the suspended extrusion, which was sup-
The modified versions of Eqs. (1) and (3), for force and moment,
ported only by the structural silicone joint, was also horizontal. The
are:
load, which varied as a function of time, was applied to the sus-
l Dg pended extrusion, at a distance of 100 mm from the sealant joint’s
F T ¼ F C ¼ Ess ð15Þ
2 g centreline. Force was increased by 3.2 N every 30 s until the free
beam’s deflection was near to the maximum that the instruments
and, could measure, or until the strain in the sealant was near to the
Bl Dg 25% limit established by the sealant’s manufacturer [20]. There-
M¼ Ess : ð16Þ
3 g after, force was removed incrementally, at the same rate as it
had been applied. Immediately prior to each change in load, the
When the glass edge rotation is a then, from Fig. 5, a ¼ Dg=l,
position of the moving side of the sample was measured using dial
where a is less than 0.176 radians. Eq. (16) now can be rewritten:
gauges, graduated in hundredths of a millimeter, located as shown
2
Bl Ess in Fig. 7.
M¼a : ð17Þ Each of the three sealant specimens was held and loaded, in
3g
sequence, in four different orientations, as shown in Table 1. There-
4. Experimental method & results fore, on each of the occasions that the samples’ properties were
measured – after 14, 114 and 214 days – the test procedure
Three identical samples of structural silicone sealant were pre- described above was carried out twelve times in total.
pared specifically for this study. In each sample, DC-983 sealant Fig. 8 shows the magnitude of rotation about the axis of the sea-
was applied between two painted aluminium extrusions, to create lant joint, plotted against applied moment, for two example sets of
a joint in which the thickness of the joint or ‘‘glueline”, g, was data – test numbers 9 and 10. The readings presented are those
6 mm. The sealant joint was orientated with its axis perpendicular recorded while the load was being increased and while the load

Fig. 6. Width of the compression zone, as a proportion of the width of the tension zone, in a structural sealant joint subject to axial moment. Using the variable names shown
in Fig. 5, this is ðB  lÞ=l.
516 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Fig. 7. Experimental apparatus used to measure rotation of structural sealant joint with respect to applied moment. The box section on the left is rigidly fixed, while that on
the right is supported only by the structural sealant sample. Load F is applied to the free side, while deflection is measured using dial gauges.

was being reduced. The hysteresis patterns are typical of those The sequence of tests was repeated one hundred days later, on
seen in the other test results. day 114, and then repeated again after a further one hundred days,
The measurements plotted in Fig. 9-A are similar – rotation on day 214. These two sets of rotation measurements, plotted
with respect to moment – but records of all 12 tests carried out against applied moment, are shown in Figs. 10-A and 11-A respec-
on the fourteenth day are presented. For clarity, the part of each tively. For completeness, the corresponding direct stresses are
sample’s response that has been plotted is that recorded during given in Figs. 10-B and 11-B.
the initial phase of the test, while the applied load was increasing. In summary, the torsional resistance of sealant joints has been
Fig. 9-B is the mean tensile stress, averaged over the sealant joint’s measured in the laboratory. Each of three specimens was tested
contact area, caused by the applied load together with the dead in four different orientations, as shown in Table 1, after 14, 114
load of the suspended part of the test sample and its counterbal- and 214 days. All 36 measured rotational stiffness are presented,
ance. At any given test load, the magnitude of this direct stress is in the form of a histogram, in Fig. 12.
much less than the stress caused by the applied moment. The
graph showing direct stress, Fig. 9-B, has been provided beneath
the rotation-moment plot so that the description of stresses is 5. Young’s modulus of structural silicone sealant
complete. This information may be of interest because, even
though this paper’s models presume that the relationship between After rotational resistance had been measured for the third
stress and strain is linear, it is known [22] that the elasticity of a time, on day 214, each specimen was partially disassembled by
structural sealant does, in reality, vary with stress. separating the sealant from one of its metal substrates using a
When the sample’s fixed beam was positioned below its free sharp blade. At the newly-exposed surfaces of sealant and glazing
beam, which was the case during tests 1, 4, 5, 7, 9 and 11 (see tape, hardness was found using a Shore A indentation tester
Table 1), the counterbalance was not of the same length and mass [23,24]. For each of the three specimens, sealant hardness mea-
as that used during the other tests. Hence, the absolute magnitudes surements were taken at six locations. For specimens 1 and 2, glaz-
of mean tensile or compressive stresses in the sealant during tests ing tape hardness measurements were taken at six locations. As
1, 4, 5, 7, 9 and 11 were lower than those during other tests. the glazing tape in specimen 3 had been damaged when it was
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 517

Table 1
In a series of twelve tests, each of the three sealant specimens was tested in four different orientations. To make it easier to differentiate the two box
sections, one has been drawn with shading. The enlarged views show the position of glazing tape in the joint.

separated from its substrate, it was possible to measure its hard- In various studies documented in the existing literature, for a
ness in only one location. Minimum, mean and maximum values wide range of different materials, values of Young’s modulus
are shown in Table 2. Because the thickness of material tested determined by tensile testing have been compared with values
was equal to the sealant joint’s glueline, 6 mm, it was not neces- based on indentation hardness measurements. Regression analy-
sary to apply any correction to the gauge readings [[25] Fig. 3]. sis of the laboratory data [e.g. [27]] has shown that, away from
As the glazing tape is narrow – only 6 mm in width – the centre the extreme ends of the hardness scale, the relationship between
of the tip of the indentation tester could not be positioned more Young’s modulus and Shore A hardness is in close agreement
than 3 mm from the tape’s edge. The close proximity of the inden- with that predicted by Gent (Eq. (18)). There is however signifi-
ter to the edge is likely to have influenced the measurements but, cant scatter in the experimental data, and for this reason values
as discussed in Section 7.3, the tape’s properties are not considered of Young’s modulus derived from individual indentation hardness
in this paper’s algebraic models of sealant joint behaviour. measurements should be considered to be indicative rather than
The behaviour of an elastic material under the tip of a hardness precise.
instrument’s indenter was considered by Gent [26], who proposed According to the manufacturer’s technical data sheet [20], the
the following theoretical relationship between Young’s modulus, E, hardness of the sealant used in this study, DC-983, should be in
and Shore A hardness, S; the range between 35 and 45 on the Shore A scale. Gent’s conver-
sion method, Eq. (18), implies that the Young’s modulus will be in
0:0981ð56 þ 7:62336SÞ
E¼ : ð18Þ the range between 1.39 and 2.03 MPa.
0:137505ð254  2:54SÞ
518 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Fig. 8. Rotation, as a function of the moment applied during tests 9 and 10, fourteen days after creation of the sealant samples.

The different estimated values for the Young’s modulus of the sion – is an increase in the sealant’s apparent elastic modulus in
DC-983 sealant samples tested in this study, obtained by the meth- tension.
ods outlined in this section, are summarized in Table 3. (b) Taking laboratory measurements of rotation with respect to
moment, and finding the elastic moduli that are implied by the
6. Discussion of experimental findings asymmetric tension–compression model that is the basis for
Eq. (16), leads to the values shown in the fifth row. The mean
It is known that the apparent elasticity of a viscoelastic mate- of these implied elastic moduli is 2.68 N/mm2, which is, in com-
rial varies with the magnitude of strain, strain rate and the direc- parison to the elastic moduli obtained using Eqs. (3) and (4),
tion of loading [for structural sealant, see [22]], and it is therefore closer to the value published by sealant manufacturer
to be expected that any quantification of Young’s modulus will be (2.00 N/mm2, in the second row of Table 3).
influenced by the manner in which measurement is made. In
Table 3, the values of Young’s modulus in the second and third As noted previously, the sealant samples were cured, stored,
rows have been obtained from tensile tests, those in the first and tested in conditions ranging between 20  C and 36  C, and
and last rows are based on indentation hardness measurements, 65% to 85% relative humidity. This environment was therefore war-
while the estimates in the fourth and fifth rows are those implied mer and more humid than the reference conditions – a constant
by the rotational stiffness equations that were developed in Sec- 25  C and 50% relative humidity – in which the sealant manufac-
tions 2 and 3. Therefore it is true that, here, apples have not been turer’s own test specimens cured [20]. The manufacturer indicates
compared with apples. However, the following observations can [also [20]] that elevated temperatures do increase the rate of cur-
be made: ing, but published test results [[28] Table 1] suggest that the
impact of the non-standard storage temperature and humidity
(a) The behavioural model that underlies the estimates in the upon the modulus of elasticity of this particular sealant, DC-983,
fourth row of Table 3, that is the basis for Eqs. (3) and (4), is small. Also, it should be remembered that any deviation from
assumes that the relationship between stress and strain is the the ideal laboratory conditions has been modest in comparison
same whether the sealant is in tension or compression. The with the extremes that are experienced by structural sealants in
result of this simplification – that is to say, ignoring the increase service, where facade surface temperatures of 20  C through
in elastic modulus that occurs when the sealant is in compres- +80  C can be encountered.
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 519

Fig. 9. (Graph A) Rotation with respect to applied moment, measured 14 days after creation of the sealant joint, for each of the twelve test cases. (Graph B) Mean direct stress,
averaged over the sealant-to-substrate surface area.

7. Validity of the mathematical model Further deviation from the theoretical model is to be expected in
service, where, in wind storms, the rate of change in pressure or
The rotation of an idealized structural sealant joint can be pre- the duration of load application might differ greatly from this
dicted using Eq. (3). In this model, the relationship between stress study’s conditions. In addition, the material’s elastic modulus
and strain in the sealant is linear, and the glazing tape, shown in may be influenced by its cyclic loading history [29]. In short, a
Figs. 1-B, -D and 7, does not affect the moment resistance. It is structural designer should be aware that the effective value of a
worth revisiting these two premises in the light of the experimen- structural sealant’s Young’s modulus may vary with factors such
tal findings. as age, temperature and loading history, and the range may be a
large proportion of the mean. This is not to say, however, that a
7.1. Variability of the sealant’s apparent modulus of elasticity structural sealant cannot function effectively as a restraint for
the metal member to which it is bonded.
The algebraic model presented in Section 2 is based upon the The mathematical representations of rotational resistance are
assumption that structural sealant obeys Hooke’s law. However, based upon premises that are not perfectly consistent with each
it has been noted already that actual stresses, measured in condi- other. For example, in Section 3, the modulus of elasticity on the
tions of steadily increasing strain, differ from the linear-elastic compressed side of joint is assumed to vary with glass rotation
ideal by around 9% in the range of strain between zero and 25%. but, at the same time, it is assumed that elasticity at any given
520 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Fig. 10. (Graph A) Rotation with respect to applied moment, measured 114 days after creation of the sealant joint, for each of the twelve test cases. (Graph B) Mean stress,
averaged over the sealant-to-substrate surface area.

angle of rotation is constant within the triangular compression Poisson’s stresses are resisted by the adhesive connection. Because
zone, even though strain varies in that region. There is therefore changes in cross-sectional shape are inhibited in this way, in vicin-
an element of arbitrariness in the formulation of the models. ity of the substrate, the sealant’s effective elasticity is reduced.
Rather than try to correct the inconsistencies or to create a more Changing the shape of a sealant joint therefore effects the mate-
realistic and more complex model, a more practical approach rial’s apparent rigidity, Er . As the ratio of bite to glueline or, using
might be to use an empirical constant to adjust the crude relation- the variable names shown in Figs. 3 and 5, B=g, increases, so the
ships, such as Eq. (3), offered in Section 2. sealant’s apparent rigidity will increase.
The value of a sealant’s elastic modulus that is published by the
7.2. Influence of sealant joint’s cross sectional aspect ratio product’s manufacturer is, commonly, determined by the ASTM
C1135 method [18], using sealant specimens with a bite to glueline
Normally, when a piece of incompressible or partially com- ratio of 1:1. Laboratory tests have shown [[30] p. 42] that changing
pressible solid is stretched in one direction, there is a reduction the joint’s aspect ratio from 1:1 to 2:1 ‘‘more than doubles” the
in the area of its cross section in the perpendicular plane. This, of apparent modulus of elasticity in a tensile test. Recently Descamps,
course, is Poisson’s effect. However, when sealant is bonded to a Hayez and Chabih [31] studied the influence that aspect ratio has
rigid substrate then, in the material adjacent to the contact surface, upon the rigidity of a particular, two-part, structural silicone
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 521

Fig. 11. (Graph A) Rotation with respect to applied moment, measured 214 days after creation of the sealant joint, for each of the twelve test cases. (Graph B) Mean stress,
averaged over the sealant-to-substrate surface area.

sealant, DC-993, and they concluded that the relationship is is 40, while that of the DC-983 [20] tested in this investigation is in
described by the second order polynomial below; the range between 35 and 45. It might appear, therefore, that Eq.
"  2 # (19) could be applied to this current analysis. However, if the sea-
B B lant’s modulus of elasticity is in the range between 1.36 and 2.0
Er ¼ 0:1506 þ 0:3409 þ 1:0852 E: ð19Þ
g g MPa (see Table 3), then the modulus of rigidity of the joints tested
in the laboratory (Section 4) will, from Eq. (19), be in the range
The expressions for a sealant joint’s moment resistance, which were from 6.68 to 9.8 MPa. This range is well above the mean apparent
developed in Sections 2 and 3, can be modified to model the signif- value implied by the rotational stiffness test measurements, which
icance of joint shape by replacing Young’s modulus, E, with the is less than 3.5 MPa (Table 3, row 4).
effective rigidity of the sealant, Er , from Eq. (19). It remains reasonable to argue that the apparent modulus of
The modulus of elasticity of the structural sealant used in this elasticity based on torsional rigidity measurement is greater than
present study’s laboratory tests, DC-983, is similar to that of the the modulus of elasticity inferred from indentation hardness mea-
product considered by Descamps et al. According to the technical surements because the sealant joint samples have a high bite to
data published by the manufacturer [32], the Shore A hardness of glueline ratio. However, Eq. (19) was developed by Descamps
DC-993, measured after 7 days at 25  C and 50% relative humidity, and coauthors for the analysis of structural sealant joints in
522 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

Fig. 12. The torsional resistances of three structural silicone sealant specimens, were measured in four different orientations (Table 1), after 14, 114 and 214 days. The
experimental results are presented in this histogram. Rotational stiffnesses are the gradients of straight lines fitted to the test results shown in Figs. 9–11, and the units are
degrees of rotation per N.mm of applied moment, per mm of axial length of sealant joint.

Table 2
Hardness of the DC-983 structural sealant samples, as well as the hardness of the glazing tape, measured with a Shore A indentation gauge 214 days after creating the specimens.

Material tested Specimen number Shore A hardness

Min. Mean Max.
Structural sealant. 1 34 35.3 38
Structural sealant. 2 32 33.2 34
Structural sealant. 3 34 34.5 36
Glazing tape. 1 30 31.2 35
Glazing tape. 2 21 25.3 28
Glazing tape. 3 32 32.0 32

Table 3
Different estimated values of Young’s modulus for the DC-983 structural silicone samples tested in this study.

Basis for estimate Age Shore A hardness Implied Young’s Modulus

Min. Mean Max. Min. Mean Max.
Days N/mm2 N/mm2 N/mm2
Sealant manufacturer’s published Shore A range [20], 7 35 – 45 1.39 1.71 2.03
converted to Young’s modulus using Eq. (18)
Young’s modulus implied by sealant manufacturer’s 21 – – – – 2.00 –
published stress at 10% strain [20]
Linear elastic model based on laboratory tension 21 – – – – 1.60 –
measurements [19] at up to 25% strain (Fig. 2)
Mean measured joint stiffness (Figs. 9, 10, 11) converted 14 – – – 2.31 3.49 5.77
to Young’s modulus using Eq. (4) 114 – – – 1.94 3.21 5.79
214 – – – 2.22 3.48 5.49
Mean measured joint stiffness (Figs. 9, 10, 11) converted 14 – – – 1.83 2.76 4.56
to Young’s modulus using Eq. (17) 114 – – – 1.54 2.54 4.57
214 – – – 1.75 2.75 4.34
Measured Shore A hardness (mean of values in Table 2) 214 32 34.3 38 1.24 1.36 1.57
converted to Young’s modulus using Eq. (18)

tension, and it appears to overestimate the rigidity of joint in section, with acrylic adhesive on the two sides in contact with
which, as in this instance, the predominant load is torsion. the substrates. Its purpose is to maintain the required distance
between the bonding surfaces while the sealant is being applied
7.3. The influence of glazing tape and while it is curing but, as is the usual practice in glazing sys-
tems, the tape was not subsequently removed from the joint.
The ‘‘glazing tape” adjacent to the sealant in the test specimens In the proposed mathematical models for the joint’s rotational
(see Figs. 1-D and 7) is an open-cell foam spacer, square in cross- stiffness, established in Sections 2 and 3, the glazing tape has been
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 523

ignored. When the load applied to the test specimens caused ten- 8. Quantifying axial moment in framing members
sion in the sealant beside the glazing tape – during tests 3, 8 and 12
(see diagram in Table 3, – it was obvious from inspection that the When a rectangular pane of glass is simply supported at its
tape’s adhesive held it to only one side of the joint, and the tape perimeter and subjected to a uniform pressure acting normal to
therefore played no part in the transfer of moment. This condition its plane, the deflected shape of its surface, measured along any
is outlined in the left hand diagram of Fig. 13. Conversely, during line parallel to an edge, is similar to one half period of a sinusoid
tests 4, 7 and 11, the glazing tape was placed in compression, in [[33] Eq. 36] if the sides of the pane are not greatly different in
the manner shown in the right side of Fig. 13. However, the test length. It is to be noted that this approach will under-estimate
results do not show a consistent increase in the joint’s rotational edge rotations away from the centre of the sides of plates with
stiffness. The experimental data therefore support the assumption higher aspect ratios [[34] Figs. 56–71]. This point is best explained
that the contribution of glazing tape can be ignored in analysis of with reference to Fig. 14, which shows the actual deflected shape
moment resistance. of a rectangular plate with aspect ratio of three, and that shape’s
In some glazing systems, the material used to separate the deviation from the sinusoidal idealization. At any point along the
metal and glass may be much harder than the glazing tape that centreline parallel to the long sides, other than at the plate’s geo-
has been used in this study. When such a separator is placed in metric centre, the actual deflection exceeds the sinusoidal model’s
compression, the deflected shape of the sealant joint will be similar prediction.
to that shown in right hand diagram of Fig. 13, except that the ful- Even though the actual rotations at both the long and the short
crum for glass rotation might be at, or near to, the upper right hand edges of a rectangular plate will, if the plate’s edges are disparate
corner of the sealant’s initial, unloaded cross section. Here, all of in size, be larger than those predicted by the sinusoidal surface
the sealant is in placed in tension, and the joint’s theoretical model, the model is still useful because its mathematical descrip-
moment resistance can be determined using the expression below, tion is simple. If the lengths of a panes’ short and long sides are,
which has been derived in the same way as Eq. (4): respectively, a in the x direction and b in the y direction, if wc is
3gM the central deflection, and if wx;y is the deflection at point ðx; yÞ,
Ess ¼ : ð20Þ then an expression for the sinusoidal surface is;
aB3

Fig. 13. Glazing tape separates from one of its substrates when the adjacent sealant is in tension (Left). However, when the adjacent sealant is compressed (Right), the glazing
tape also is placed in compression.

Fig. 14. The graph’s heavy line shows the deflected shape of an initially-flat rectangular plate with aspect ratio b=a ¼ 3, measured at the centreline parallel to the long sides,
when a uniform pressure is applied to the plate’s surface [traced from 34 Fig. 56].
524 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

px py
wx;y ¼ wc sin sin : ð21Þ maximum allowable stress, F t . So, considering a unit length of sea-
a b lant joint and substituting for M in Eq. (3);
The gradient in the x-direction is obtained by differentiating B2 DgEss
partially; F t Be ¼ ; ð25Þ
6g
@wx;y pwc px py then substituting for Dg from Eq. 24, and simplifying;
¼ cos sin : ð22Þ
@x a a b
pB2 Ess wc
e¼ : ð26Þ
so, at the middle of the pane’s longer side, where x ¼ 0 and y ¼ b=2; 12F t ag

@wx;y pwc In practice, the maximum allowable glass deflection is often speci-
¼ : ð23Þ fied as a proportion of the length of the pane’s shorter side. A limit
@x a
commonly found in design codes and technical specifications for
As is evident in the diagram at the right hand side of Fig. 3, the gra- glazing – for example, the Australian design code for glass in build-
dient of the glass surface at the sealant joint is 2Dg=B. Substituting ings [[35] Section 3.3.3] – is wc ¼ a=60. At this deflection condition,
for gradient in Eq. (23) gives; Eq. (26) becomes;

pBwc pB2 Ess

Dg ¼ : ð24Þ e¼ : ð27Þ
2a 720F t g
In reality, when wind pressure acts upon a structurally glazed Typical numerical values can be assigned to the variables in
pane, the sealant at its perimeter experiences both a direct force Eq. (27) in order to assess whether the magnitude of the turning
and a moment. However, it is easiest to convey the relative impor- moment – that imparted by the glass, through the structural sea-
tance of these two components of load if the force is considered to lant, to the glazing system’s metal frame – is of practical signifi-
act, as shown in Fig. 15, through an imaginary lever attached to the cance. Assuming that the structural sealant industry’s standard
frame, at a distance e from the centreline of the sealant joint. The tensile stress limit is observed, F t ¼ 139 kPa [[5] Section 27.5],
significance of a given eccentricity will be immediately apparent and that the geometry and elastic modulus of the joint are similar
to curtain wall designers, who are accustomed to seeing similar to those in the samples that were tested in this study, with glueline
lever arms in the shapes of brackets used to connect mullions to g ¼ 6 mm, bite B ¼ 22:5 mm, and Ess ¼ 4:0 MPa (see Table 3), then
a building’s primary structure. the eccentricity of the load, e, measured from the joint’s centreline,
In a structurally optimized design, the width of the sealant bite is 10.6 mm. The effect of the moment upon the glazing frame will
will be minimized so that the actual tensile stress is equal to the of course depend upon the cross-sectional shape and span of the

F F

Fig. 15. The current analytical convention is that loads transferred from glass to mullion are considered to act through the centre of the structural sealant (Left), but in reality,
because rotation occurs at the glass edge (Right), there is an effective eccentricity, e, between the force and the sealant joint.
 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526 525

mullion profile, but it is probable that, in most cases, an eccentric- sible values for the structural sealant’s elastic modulus, some
ity of this magnitude could safely be neglected by the framing degree of inaccuracy in the predictive model will not be of
designer. However, if the sealant’s elastic modulus were to be at consequence.
the upper end of the range that has been observed in one-part Others [17] have recorded a manyfold increase in the Young’s
structural silicones, say Ess ¼ 20:0 MPa [e.g. [17] Fig. 6], then the modulus of single-part structural silicone sealant, occurring during
value of e would be 53.0 mm. Such a large eccentricity certainly the months after the initial fortnight of curing. The DC-983 two-
would concern the designer of a typical unitized curtain wall sys- part sealant specimens prepared for this investigation were tested
tem with open, E-shaped mullion extrusions. at 14, 114 and 214 days, but no significant change in modulus was
observed. This observation shows that, at 14 days, the cross-linking
9. Conclusions of polymer chains within the sealant – the curing process – was
complete or substantially complete.
When wind load causes the glass or other sheet infill material at Using the formulae that have been presented in this paper,
the face of a structurally glazed curtain wall to deflect, moments designers of curtain wall systems will be able to estimate quickly,
are induced about the longitudinal axes of the framing members. without a requirement for numerical modelling, load eccentricities
Using the analytical methods proposed in this paper, facade that are ignored in standard structural analyses. The information
engineers can incorporate mathematical expressions – for a sealant will be of greatest interest when the metal framing members to
joint’s rotational stiffness and, hence, for axial moment – in their which structural glazing is to be applied are susceptible to rotation
structural models of curtain wall mullions. There are at least two about their longitudinal axes – for example, if unsupported spans
practical incentives to do so. Firstly, it will be possible to identify, are long, and if the profiles have low torsional rigidity.
during the design process, cases in which excessive axial rotation
of a proposed mullion extrusion would impair its non-structural Declarations of interest
functions, such as the effectiveness of its air seals. Secondly, if anal-
ysis shows that the torsional restraint provided by a sealant-to- None.
glass connection is sufficient to prevent lateral torsional buckling,
then there will be an opportunity to create framing members con- Acknowledgements
taining less aluminium than would be needed if the established
structural design conventions were to be observed. For this second The authors are grateful to Mr. James Chant, Chairman of Sea-
application, however, further research into the stability of struc- pac Philippines Inc., who provided the material samples used in
turally glazed members will be needed to demonstrate that they these tests. Mr. Lawrence Carbary of Dow Corning reviewed a draft
can be reliably restrained by structural sealant in the all of the con- of this paper and made helpful comments. Staff of PTCC Facade
ditions that might be experienced by the facades of real buildings. Design, led by Engineer Warren Tan and Mr. Rene Ramiscal helped
Data sets have been obtained by physical testing and the results to commission the test rig and collect data. PTCC’s Mr. Jhun Fabrero
show that, for any one sample sealant joint, the relationship transformed the authors’ hand sketches into CAD drawings.
between rotation and applied moment is practically linear. The
variation between one sample and another is, however, consider- References
able: in a population of sealant joints, one standard deviation is
approximately one third of the mean value. So, when estimating [1] D. Yeomans, The pre-history of the curtain wall 14 (1998) 59–82. URL:http://
the magnitude of the moment transferred to the frame of a struc- www.arct.cam.ac.uk/Downloads/chs/final-chs-vol.14/chs-vol.14-pp.59-to-82.
pdf.
turally glazed pane, a pragmatic approach will be to consider a [2] I. Ábalos, J. Herreros, Tower and Office: From Modernist Theory to
range of values for the sealant joint’s stiffness. When establishing Contemporary Practice, MIT Press, 2003. ISBN 0-262-01191-3.
that range, it should be kept in mind that the variability in rota- [3] D. Yeomans, The origins of the modern curtain wall, APT Bull. 32 (2001) 13–18,
https://doi.org/10.2307/1504688. URL:http://www.jstor.org/stable/1504688.
tional stiffness of structural sealant joints in a real building’s cur- [4] S. Murray, Contemporary Curtain Wall Architecture, Princeton Architectural
tain wall panels will, because of normal batch-to-batch Press, 2009. ISBN 978-1-56898-797-2.
inconsistencies in the sealant properties, dimensional tolerances, [5] ASTM C1401. Standard Guide for Structural Silicone Sealant Glazing, 2014.
DOI: 10.1520/C1401-14.
differences in loading history and other factors, be greater than
[6] American Architectural Manufacturers Association, Curtain Wall Design Guide
that observed in these tests, which have been carried out in rela- Manual, 2005.
tively tightly controlled conditions. [7] A. Watts, Modern Construction Envelopes, second ed., AMBRA—V, 2014. ISBN
The hardness of this study’s sealant samples, measured by the 978-3-99043-559-5.
[8] Yuanda China Holdings Limited, Global Offering, 2011.www.hkexnews.hk/
indentation method (Table 3, last row), was found to lie within listedco/listconews/sehk/2011/0420/027891057382/EWF114.pdf. (Document
the range published by the sealant manufacturer (Table 3, first available from Hong Kong Stock Exchange).
row). However, the mean value of this study’s laboratory measure- [9] R.D. Ziemaian (Ed.), Guide to Stability Design Criteria for Metal Structures,
sixth ed., John Wiley & Sons, 2010, ISBN 978-0-470-08525-7.
ments of rotational stiffness, when used with the simple mathe- [10] Center For Window and Cladding Technology, Technical Update No. 7:
matical model proposed in Section 2, suggests that the elastic Buckling of Curtain Wall Mullions, 2002.
modulus of the sealant samples (Table 3, fourth row) is approxi- [11] Aluminum Association, Aluminum Design Manual, 2005.
[12] L.T. Goco, Rationalizing assumptions for evaluating the lateral torsional
mately 70% greater than the value published by the sealant manu- buckling strength of aluminium mullion extrusions by EN1999 and AA ADM
facturer (Table 3, second row). The more refined behavioural (Master’s thesis), Bath University, UK, 2017. URL:https://www.bath.ac.uk/
model, laid out in Section 3, which acknowledges a difference library/dissertations/index.php?programme=MSc+Architectural+Engineering.
[13] D. Skejić, M. Lukić, N. Buljan, H. Vido, Lateral torsional buckling of split
between the sealant’s properties in tension and compression, also aluminium mullion, Key Eng. Mater. 710 (2016) 445–450, https://doi.org/
leads to a value of Young’s modulus that is an apparent overesti- 10.4028/www.scientific.net/KEM.710.445.
mate: the computed value (Table 3, fifth row) is about 34% greater [14] C.D. Clift, W.J. Austin, Lateral buckling in curtain wall systems, J. Struct. Eng.
115 (10) (1989) 2481–2495, https://doi.org/10.1061/(ASCE)0733-9445(1989)
than that indicated by the sealant manufacturer (Table 3, second
115:10(2481).
row). Nonetheless, even if predictions of rotational stiffness based [15] G. Hammond, C. Jones, Embodied Carbon: The Inventory of Carbon and Energy
upon this paper’s algebraic models are not in precise agreement (ICE), Bath University and BSRIA, 2011. ISBN 978 0 86022 703 8.
with the laboratory results, the models are still useful to facade [16] A.D. Lee, P. Shepherd, M.C. Evernden, D. Metcalfe, Optimizing the architectural
layouts and technical specifications of curtain walls to minimize use of
designers. Providing the analysis of an adhesive joint’s rotational aluminium, Structures 13 (2017) 8–25, https://doi.org/10.1016/j.istruc.2017.
stiffness will take into consideration a suitably wide range of pos- 10.004.
526 A.D. Lee et al. / Construction and Building Materials 181 (2018) 510–526

[17] K. Ramesh, R.W. Tock, R.S. Narayan, C.V.G. Vallabhan, Property evaluation of 11-3716-01durometer-hardness-for-silicones.pdf, Dow Corning Form Number:
silicone elastomers used in tension-adhesion joints, Mater. Sci. Lett. 14 (1995) 11-3716-01.
964–967, https://doi.org/10.1007/bf02427478. [28] R.W. Tock, Temperature and moisture effects on the engineering properties of
[18] ASTM C1135, Standard Test Method for Determining Tensile Adhesion silicone sealants, in: T.F. O’Connor, editor, Building Sealants: Materials,
Properties of Structural Sealants, 2015. DOI: 10.1520/C1135-15. Properties and Performance. ASTM STP 1069, American Society for Testing
[19] J. Kimberlain, L. Carbary, C.D. Clift, P. Hutley, Advanced structural silicone and Materials, 1990, pp. 167–173. DOI: 10.1520/STP26803S.
glazing, Int. J. High-Rise Build. 2 (4) (2013) 345–354. [29] L. Mullins, Effect of stretching on the properties of rubber, J. Rubber Res. 16
[20] Dow Corning Corporation, Product Information: Dow Corning 983 Structural (12) (1947) 275–289.
Glazing Sealant, 2012. URL:http://www.dowcorning.com/content/ [30] C.M. Schmidt, W.J. Schoenherr, L.D. Carbary, M.S. Takish, Performance
publishedlit/63-1110.pdf, Dow Corning Form Number: 63-1110C-01. properties of silicone structural adhesives. In: C.J. Parise, editor. Science and
[21] C.V.G. Vallabhan, M.Z. Asßik, Analysis of structural glazing systems, Comput. Technology of Glazing Systems. American Society for Testing and Materials,
Struct. 65 (1997) 231–239, https://doi.org/10.1016/S0045-7949(96)00284-2. 1989, pp. 22–45. DOI: 10.1520/STP22987S.
[22] A.T. Wolf, H.L. Cleland-Host, Material properties for use in FEA modeling: [31] P. Descamps, V. Hayez, M. Chabih, Next generation calculation method for
sealant behaviour with ambient laboratory climate aging, J. ASTM Int. 1 (7) structural silicone joint dimensioning, Glass Struct. Eng. 2 (2017) 169–182,
(2004) 372–384, https://doi.org/10.1520/JAI11603. https://doi.org/10.1007/s40940-017-0044-7.
[23] ASTM C661, Standard Test Method for Indentation Hardness of Elastomeric- [32] Dow Corning Corporation, Product Information: Dow Corning 993 Structural
Type Sealants by Means of a Durometer, 2011. DOI: 10.1520/C0661-06R11. Glazing Sealant, 2011. URL:http://www.dowcorning.com/content/
[24] ASTM D2240, Standard Test Method for Rubber Property – Durometer publishedlit/62-0918.pdf, Dow Corning Form Number: 62-0918L-01.
Hardness, 2005. DOI: 10.1520/D2240-05R10. [33] D. Wang, A.I. El-Sheikh, Large deflection mathematical analysis of rectangular
[25] L. Casa, R. Mendichi, Shore A hardness and thickness, Polymer Test. 7 (1986) plates, J. Eng. Mech. 131 (2005) 809–821, https://doi.org/10.1061/(ASCE)0733-
165–175, https://doi.org/10.1016/0142-9418(87)90030-4. 9399(2005) 131:8(809).
[26] A.N. Gent, On the relation between indentation hardness and Young’s [34] R.N. Head, E.E. Sechler, Normal Pressure Tests on Unstiffened Flat Plates, 1944.
modulus, Rubber Chem. Technol. 31 (4) (1958) 896–906, https://doi.org/ URL:https://ntrs.nasa.gov/search.jsp?R=19930086088, NACA Technical Note
10.5254/1.3542351. 943.
[27] K. Larson, Can You Estimate Modulus From Durometer Hardness for [35] AS 1288, Glass in Buildings – Selection and Installation, 2006. URL:http://
Silicones? 2016. URL:http://www.dowcorning.com/content/publishedlit/ www.standards.org.au.