Skip to main content Skip to footer

Lessons Learned from Curtain Wall Failure Investigations

July 15, 2010

S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 0 9 MC C OWA N A N D K I V E L A • 9
LESSONS LEARNED FROM CURTAIN WALL
FAILURE INVESTIGATIONS
DEREK B. MCCOWAN, PE
AND
JOSHUA B. KIVELA, PE
SIMPSON GUMPERTZ & HEGER, INC.
41 Seyon St., Building 1, Suite 500, Waltham, MA 02453
Phone: 781-907-9000 • Fax: 781-907-9009 • E-mail: dbmccowan@sgh.com and jkivela@sgh.com
1 0 • MC C OWA N A N D K I V E L A S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0
ABSTRACT
This presentation will discuss failures and other problems encountered during forensic
investigations of ailing curtain wall systems, with primary focus on glass and metal (aluminum)
curtain-wall systems. Failures include air and water leakage, glass breakage, loss
of (falling) metal components, fogging glass, and other serious deterioration. The authors
will explain the reasons for the failures and explain how to avoid such problems.
We will also summarize a few recent successful curtain wall installation projects and
some of the key design, preconstruction, and construction procedures that led to the successes
and some critical tasks that are often overlooked.
SPEAKERS
DEREK B. MCCOWAN, PE — SIMPSON GUMPERTZ & HEGER, INC.
Derek B. McCowan has eight years of experience in investigation, design and consultation,
and construction monitoring and administration of historic and contemporary buildings.
He specializes in curtain wall and window systems and also has experience with foundations,
opaque wall systems, skylights, and steep- and low-slope roofing. He is currently
pursuing a master’s of science in civil engineering/construction management at
Northeastern University, and he received a bachelor’s in civil engineering from Union
College. McCowan has authored various articles about window and curtain wall systems
and has also lectured on these topics. He is a registered professional engineer in
Massachusetts, Connecticut, Maine, Vermont, and New Hampshire.
JOSHUA B. KIVELA, PE — SIMPSON GUMPERTZ & HEGER, INC.
Joshua B. Kivela has more than nine years of experience in design, investigation, evaluation,
restoration, and construction administration and monitoring of historic and contemporary
buildings. He specializes in waterproofing design of building envelope systems,
including foundations; wall systems; curtain walls; windows/glazing; skylights; and flat-,
steep-, and low-slope roofing and has a background in construction and laboratory testing.
Kivela is an active member on ASTM International Committee D08 — Roofing and
Waterproofing. He received his bachelor’s degree in civil engineering from the University of
Rhode Island and his master’s in mechanical engineering/materials science from
Northeastern University. He is a registered professional engineer in Connecticut as well as
a registered professional engineer and civil engineer in Massachusetts.
INTRODUCTION
Curtain walls are a form of exterior
cladding that do not support floor or roof
loads—they “hang” off of the building structure
like a curtain. Most contemporary exterior
wall systems are technically curtain
walls, though most people in the architecture/
engineering/design community envision
multistory glazed systems when they
think of curtain walls. These glazed systems
form an integral part of the building enclosure,
and they must be designed and constructed
to achieve various structural and
nonstructural performance requirements,
such as the following:
• Water penetration resistance,
• Air infiltration resistance,
• Structural adequacy (transfer all
loads back to building structure),
• Energy efficiency,
• Aesthetics, and
• Durability and maintainability.
Other design criteria include thermal
movement, condensation, sound attenuation,
fire resistance, and blast resistance.
These performance requirements apply
whether the curtain walls are field-constructed
(i.e., stick-built), partially prefabricated
(i.e., ladder systems), or fully prefabricated
in a factory (i.e., modular
or unitized systems). The
authors have observed various
problems with all types of curtain
wall systems during the
fabrication, installation, and
building usage stages.
This paper discusses failures
and other problems
encountered during recent
forensic investigations of curtain
walls, with primary focus
on glass and aluminum systems.
Failures include air and
water leakage, glass breakage,
loss of (falling) metal components,
and fogging glass. The
reasons for failure and lessons
learned are provided to help
the reader avoid such problems
in their practice.
We also discuss some of the
more interesting forensic work completed
recently by members of national engineering
firm Simpson Gumpertz & Heger Inc.
(SGH) on problematic curtain wall systems.
THE CASE OF THE MISSING
SEALANT
An owner asked SGH to investigate
widespread air and water leakage at his new
14-story office building. The construction of
the building exterior had recently been
completed, and the office space was approximately
40% occupied. The building
includes multistory-height curtain wall
“bays” set in large punched openings in
exterior walls. Surrounding wall areas consist
mainly of brick veneer cavity wall systems.
The stick-built curtain wall system
was produced by a large, reputable manufacturer.
The system includes a combination
of pressure glazing (exterior pressure
bars at horizontal mullions) and drop-in
glazing (fixed exterior bars/stops at vertical
mullions) that allow reglazing of vision lites
from either the interior or exterior.
SGH performed a series of water tests,
typically spray rack tests followed by hand
nozzle tests for tracing specific leakage
paths, followed by disassembly of various
wall areas. We discovered a variety of problems,
both with the perimeter flashing of the
wall and with the curtain wall itself. With
regard to the leakage through the field of
the wall, the frame seals were systemically
deficient (i.e., missing or otherwise discontinuous),
which led to widespread leakage
throughout the building. The manufacturer’s
installation instructions, which were
readily available and very clear about the
frame-seal requirements, had not been followed.
The primary deficiencies of the curtain
wall itself existed at the horizontal-to-vertical
mullion intersections (i.e., frame corners)
and at the splice joints in the vertical
mullions (i.e., splices). Frame corners
required application of silicone sealant and
foam rubber joint plugs (end dams) to fill
and seal the joinery. The intent of these
materials is to create a watertight pocket so
that any water that enters the glazing pocket
area exits harmlessly through the weep
holes in the exterior pressure bar. SGH
found missing and deficient end dams and
missing and deficient sealant at the metalto-
metal joints at these conditions (Photo 1).
Unsealed joinery located at the low point in
the glazing pocket allowed water to travel
inward to the building interior as easily as it
traveled outward through the weep holes.
LESSONS LEARNED FROM CURTAIN WALL
FAILURE INVESTIGATIONS
Photo 1 – Unsealed metal-to-metal joint.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0 MC C OWA N A N D K I V E L A • 1 1
Unsealed fastener penetrations at shear
blocks at these locations also served as a
leakage path. Splices were in similar condition
as frame corners: negligible weatherproofing
provisions had been provided, with
the exception of an occasional piece of
department-store-grade silver duct tape.
Splice joints required silicone sealant
applied to the glazing pocket area that was
occasionally exposed to water, though this
had not been installed. Therefore, joints
acted as open funnels for any water traveling
down the vertical glazing pocket. Splice
joints were so poorly constructed that daylight
was clearly visible from the interior at
these areas when interior finishes and
spandrel insulation were removed.
These various leakage paths also served
as avenues for increased air infiltration,
though air leakage was a secondary concern
for the owner, as the water leakage was
causing damage to interior finishes.
Corrective action included repairing
100% of the frame seals at the curtain wall.
Unfortunately for all parties involved, this
required removal of all of the building’s
glass units (over 1,500) to expose the frame
corners that required the repairs. The entire
curtain wall was reglazed. Various other
repairs were also performed, including
replacement of perimeter membrane flashing,
removal of portions of the surrounding
cladding systems to allow the perimeter
flashing repairs (similar to the work performed
below, in the Case of the Projecting
Curtain Wall Bays), and roof repairs.
Lesson Learned
Deficient frame-corner seals can be catastrophic
with respect to leakage and are
extremely difficult and costly to access for
repairs. These recommendations will assist
in ensuring that all frame joints are sealed
properly during the curtain wall construction/
installation phase.
• Obtain the manufacturer’s installation
instructions regarding frame
seals, and enforce them. Require the
installers to follow the instructions
to the letter. Focus on horizontal-tovertical
mullion intersections, splice
joints in vertical mullions, and wallperimeter
conditions, as well as
other areas noted in the instructions.
• Failing to properly install ¼ oz. of
silicone sealant can lead to leakage
that costs thousands of dollars to
access and repair. Take whatever
quality-control measures are necessary
to ensure proper installation of
these seals, such as inspections and
performance tests by the manufacturer,
design team, consultants,
and/or third parties.
THE CASE OF
PROJECTING
CURTAIN
WALL BAYS
A nine-story,
mixed-use building
enclosed with
projecting curtain wall bays, brick veneer,
and Exterior Insulation Finishing System
(EIFS) was built in the greater Boston area.
Shortly after the building was constructed
and before all parties were paid for their
work, the owners complained of water leakage
at the curtain wall bays, and SGH was
asked to investigate the problems. Water
testing with and without applied differential
air pressure (conducted in accordance with
ASTM E1105) and subsequent partial disassembly
of the curtain wall showed that
the system leaked and did not perform to
the specified requirements.
Initially, leaks occurred through the
curtain wall during water testing at air
pressures of 2.1 psf and above. (The curtain
wall is rated for 10 psf due to blocked or
misplaced weep holes in the pressure bar,
missing seals around mullions and joint
plugs, and poor drainage (Photo 2) from the
over application of sealant.) After the water
ran for about 45 minutes to an hour with no
applied air-pressure difference, we
observed higher volume leaks at the perimeter
of the curtain wall. Discussions with the
owners and the building maintenance personnel
revealed that similar higher-volume
leaks had occurred in the past but only during
long (about two or more days) rainstorms
with and without high winds.
We removed the rowlock brick from the
base of the curtain wall and observed the
membrane sill flashing of the curtain wall
turned up against the brick veneer to form
an end dam (Photo 3). The through-wall
flashing under the curtain wall and the
brick wall was made from a combination of
sheet metal drip-edge and self-adhering
membrane flashing. The upturned rear leg
of the membrane flashing was supported by
Photo 2 – Water ponding in curtain wall.
Photo 3 – Curtain-wall sill flashing
not sealed to through-wall flashing.
1 2 • MC C OWA N A N D K I V E L A S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0
the backup wall at the brick veneer, but it
lacked support under the curtain wall; the
transverse seams in the membrane
through-wall flashing were open, and the
flashing sagged under its own weight. The
greater problem, however, was that the
brick wall and curtain wall flashing did not
connect or seal to each other (Photo 3), and
the curtain wall lacked jamb flashing altogether.
This discontinuity at the throughwall
flashing level allowed water to leak into
the building from the wall cavity.
Our review of the design drawings
showed vague details, and the specifications
were not explicit on flashing integration.
Review of construction photographs
and discussions with construction personnel
showed that the brick veneer was
installed before many of the curtain walls,
the self-adhering air-barrier membrane ran
long in many areas, and an 8- to 12-in
“flap” was visible in the curtain wall rough
openings. The construction manager claims
he told the curtain wall installer to seal the
membrane to the jamb of the curtain wall
during installation, but the installer either
cut off the membrane or simply folded it
into the rough opening. Without a clear
design, proper coordination of subcontractors,
and a mock-up installation, the
installer assumed he did not own the flashing
installation.
Repairs involved removing
the rowlock brick below the curtain
walls, “toothing-out” the
running-bond brick at the sides
of the curtain wall, and
installing new continuous
through-wall and jamb flashing
(Photo 4). Labor costs could have
been significantly reduced by
saw-cutting a straight vertical
joint through the head joints in
the brick veneer and reinstalling
stacked-bond masonry. The
repairs were complicated by the
lack of working room created by
the inside corners between the
brick walls and the curtain wall
bays.
Lessons Learned
Pay special attention when
designing and installing projecting
curtain wall bays because
their shape creates more corners
and intersections between adjacent
materials and systems than
one would see on flush-built curtain
walls. Continuity of perimeter
flashing is a critical design consideration
that is often forgotten. The following are
other tips to keep in mind:
• Continuous perimeter flashings that
connect to adjacent building components
should be fully designed and
described in the construction documents.
Do not rely on the subcontractors
to develop such critical
details on their own.
• Mock-ups of these intersections
should be built to vet out potential
c o o r dina t i on
issues between
trades and to
confirm trade
responsibilities.
• Consider the
risks of leaving
“flaps” of airbarrier
membrane
in rough
openings that
will later seal to
curtain wall
mullions. Flaps
of air-barrier
membranes are
easily damaged
if left to hang
out of the wall
for any length
of time. Further, most air-barrier
membranes are not designed to
span unsupported across gaps and
are not designed to accommodate
differential movement between
backup walls and curtain walls; for
these conditions, specialty transition
flashing membranes should be
considered.
• Provide continuous support for
membrane through-wall flashings.
THE CASE OF THE FALLING TRIM
COVERS
After maintenance workers began noticing
unusual metal components on the
ground near a 19-story building in New
Jersey, SGH was asked to investigate the
cause. Metal trim covers were falling off the
building, presenting a danger to people and
property below. The curtain wall included
various snap-cover sizes and profiles, ranging
from 3/8-in, low profile covers to projecting
covers more than 2 in deep.
SGH performed a 100% survey of the
façade, which consists almost exclusively of
unitized curtain wall panels. Investigation
found a handful of areas where covers were
missing, though there were dozens of areas
where covers were slightly disengaged. The
disengagement was often visible via a small
(1/16-in to 1/8-in) joint between the inside
edge of the snap cover and the exterior glazing
gasket at the pressure bar (Photo 5).
This open joint was not present at properly
engaged covers.
These partially engaged covers were
reviewed up close, and the ad-hoc “yank”
test was performed. Many covers detached
from the wall with little effort (Photo 6).
Photo 4 – Continuous copper and membrane sill,
through-wall, and jamb flashing.
Photo 5 – Space between glazing gasket and trim cover.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0 MC C OWA N A N D K I V E L A • 1 3
Some of the disengaging covers did not
immediately release due to the presence of a
few daubs of silicone sealant that temporarily
held the cover in place; though after
applying light pressure, the covers readily
came loose.
During the survey, SGH noted that the
corners of the pressure bars below the disengaged
covers were usually bent upward,
preventing proper snap engagement of the
covers (Photo 7). Physical damage, such as
dents, scratches, and other evidence of
abuse, was also apparent on dozens of covers.
We did not conclusively identify the
cause of the pressure-bar cover damage,
though we suspect that it was due to abuse
during erection onto the building. Other
theories with merit include poor cutting/
fabrication operations in the factory
and bending of covers when they were
removed for other reasons, such as to allow
reglazing of a cracked glass unit.
One additional factor in the falling cover
problem included the use of a suspended
scaffold (swing-stage) window-washing and
maintenance rig that bumped the deeper
horizontal covers on its way up and down
the building. The house rig included clips
designed to engage a vertical rail mounted
to occasional vertical curtain wall mullions
to help secure
the rig to the
wall. Unfortunately,
the projecting
wheels on the rig were not considered
in the staging and curtain wall design.
As such, the rail system was ultimately
abandoned. On a related note, window
washers not using a scaffold had been seen
standing on the horizontal covers.
Perhaps the most disheartening factor
for the owner was that the bent pressurebar
corners could be fixed quickly and easily
by simply bending the bent covers back
into place with a pair of common pliers
(Photo 8). Unfortunately, this remedy was
not implemented during the original construction
process (which included an
aggressive schedule for curtain wall erection),
so a 100% survey and widespread
repair campaign were implemented within
five years of construction of the building.
We added fasteners to all snap covers as
part of the repair process, just to provide an
additional safety factor and more comfort
for the owner.
Lesson Learned
While the primary function of snap covers
is visual (i.e., trim), falling metal is a
serious problem that warrants consideration
and respect. Do not completely disregard
the design of exterior snap covers as
one might with interior trim. When using
snap covers, keep in mind the following:
• Snap engagement alone of unusually
deep or otherwise precariously
projecting metal components cannot
be relied upon for permanent
attachment.
• Be mindful of haphazard erection
techniques and the risk of damage
Photo 6 – Trim cover removed from curtain wall by hand with minimal effort.
Photo 8 – Fixing pressure bar with pliers.
Photo 7 – Bent corner of pressure bar.
1 4 • MC C OWA N A N D K I V E L A S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0
to weakly secured
components. Be particularly
careful with
unitized wall assemblies
due to hoisting/
craning erection
techniques.
• Consider maintenance
and related
access needs of the
wall systems when
designing exterior
covers, sunshades,
and other projecting
elements.
THE CASE OF THE
WINDOW FILM
SGH investigated the
cause of insulated glass (IG)
units that were breaking at a
recently renovated military
office building (Photo 9). The
window installer removed
several IG units and
observed that glass-to-metal
contact was the cause of
some of the breaks, but the
installer could not provide
an explanation for all the
cracks. SGH visited the site
and found two crack patterns
in the glass. About
three quarters of the IG
units had cracks that ran
perpendicular to the edge and surface of
glass (indicative of thermally driven breaks)
and about one quarter had cracks that did
not run perpendicular to the edge. The IG
units had a post-applied tint film on the
interior surface (for increased occupant
comfort) and vertical blinds in the offices.
Cracks were occurring on the south and
east elevations only and in IG units with
and without post-applied film.
Review of the original shop drawings
and specifications showed the IG units were
to consist of a ¼-in-thick exterior lite, a ½-
in-wide air space, and a ¼-in-thick laminated
glass inner lite with a 0.030-in-thick
interlayer (Figure 1). The glass surfaces on
Figure 1 are labeled No. 1 through No. 6. All
glass was to be annealed (i.e., not heatstrengthened
or tempered). Setting blocks
were shown in the sill of the glazing pocket;
antiwalk blocks were not shown. Antiwalk
blocks are important because they prevent
the glass from moving or walking in the
frame and thus bearing against hard metal
surfaces and/or reducing glazing gasket
contact area. The specifications required
that the probability of failure of the IG
units, upon first application of the design
wind load, would not exceed 8 lites per
1,000. (Glass is specified
according to probability of
breakage because of its susceptibility
to the stress-concentrating
effects of flaws
and the statistical nature of
flaw severity and distribution.)
The submittals showed
that two different tinted
films were used.
A thermal analysis was
performed on the IG units
(film included) based on
ASTM E2431, Standard
Practice for Determining the
Resistance of Single-Glazed,
Annealed, Architectural Flat
Glass to Thermal Loadings.
This analysis assesses the
probable edge stress in the
glass as a result of temperature
differential between the
central regions of the glass
and the concealed edge
(Figure 1). Current industry
standards for determining
thermal-edge stresses in
annealed glass apply only to
monolithic glass, so other
analyses were required to
assess the more complex IG
unit with applied tint.1
Before using the ASTM
method, the solar load for
the IG units had to be calculated
using a glass/optics computer program.
This step took into account the solar
transmittance and absorption of the different
components (glass, air space, interlayer,
Photo 9 – Crack in IG unit.
Figure 1 – Thermally driven breaks in IG units.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0 MC C OWA N A N D K I V E L A • 1 5
film). Investigators essentially
built monolithic glass models
with similar optical properties to
those of the specified IG units.
Investigators calculated the
allowable thermal stress for various-
sized IG units on the building,
based on the permissible
breakage rates dictated by the
specifications. The predicted inservice
thermal stresses were
acceptable on the north elevation,
but they were excessive on the
west, east, and south. Analysis
showed that a south-facing unit
with an angular shadow pattern
would reach the highest stress.
The owner replaced the
cracked IG units with heatstrengthened
glass of the same
thickness and size because the
analysis showed that this
stronger glass could handle the
high thermal stresses. They
included antiwalk blocks during
reglazing to prevent the glass
from moving laterally and contacting
metal.
Lessons Learned
For projects requiring use of IG units
and an applied film, remember the following:
• Consider the requirements for glass
film (for occupant comfort, energy
efficiency, or blast resistance) during
the design-development process
and for retrofit projects, and follow
industry guidelines for IG unit construction
(e.g., the Glass Association
of North America) and installation
guidelines for applied films (e.g.,
technical documents by glass manufacturers).
• If tinted films, reflective blinds,
insulating drapes, or other components
that could increase the center-
to-edge temperature difference
and thermal stress in the glass are
used, consider using heat-strengthened
glass. The stress analysis to
determine the appropriate type of
glass should be based on ASTM
E2431 and modified as described
above or, if a more precise stress
assessment is required or complex
glass configurations and loadings
are encountered, the analysis
should be based on finite element
modeling.
• Review curtain wall shop
drawings and require antiwalk
blocks at all vertical
glass edges.
THE CASE OF THE
EXPLODING GLASS
The authors witnessed this
glass breakage firsthand during
performance testing of a large
mock-up assembly at a testing
laboratory (Photo 10). The failure
occurred during a 150% ASTM
E330 overload test (150% of
design wind load for façade)
under negative/suction pressure.
(It actually occurred during a second,
unspecified overload test that the contractor
elected to complete; we noted no visible
damage during the first overload test.)
The curtain wall system that “failed”
includes an intermittent aluminum clip
that engages a channel at the perimeter of
the IG unit. Butt seals are installed at glass
joints, similar to those present at common
structural silicone-glazing (SSG) curtain
wall system joints. Structural silicone adhesive
and pressure bars are not present
along edges where the clips secure the
glass. The custom-designed outside corner
condition is “mullionless”; vertical corner
framing consists only of a 1.5-in by 1.5-in
aluminum tube that is adhered to the glass
edges with structural silicone sealant.
SGH reviewed the remnants of the glass
and glazing materials at the opening of the
failed unit (Photo 11). The spacer bar along
the noncorner jamb was fully disengaged
from the clips, and the spacer bar was bent
at various clip locations. We noted damage
to the interior pane of glass at more than
one clip location, with the most severe damage
located approximately 12 inches up
from the sill corner. The structural silicone
and low-profile aluminum tube at the corner
were undamaged.
We also reviewed a video of the glass
breakage frame by frame after returning to
1 6 • MC C OWA N A N D K I V E L A S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0
Photo 10 – Glass breakage during
mock-up test.
Photo 11 – Curtain wall after
150% design load test.
the office. Analyses are being performed by
the curtain wall manufacturer, insulating
glass manufacturer, and the structural
engineer who designed the curtain wall
framing. These analyses are currently in
progress, though it is becoming clear, from
the calculations and the video images, that
the glass disengaged from the mechanical
clips along the jamb prior to fracturing. The
analyses show that under the 150%
extreme suction load and assuming (1) no
deflection of the glass supports (i.e., the
mullion and the low-profile tube) and (2) the
glass edge is perfectly perpendicular to the
clips, approximately 1/8 in of engagement
is lost at the clip condition.
We are confident that the deflection of
the corner tube (visible in the photos) combined
with the rotation of the glass edge
(and the bending/damage at the spacer bar)
result in full loss of engagement/“bite” at
the clip conditions (i.e., loss of more than ¼
in of bite). This is supported by the images
that appear to show the jamb of the glass
unit free of the mullion prior to glass breakage.
It appears that the edge of the interior
pane of glass contacted (“caught”) the edge
of one of the aluminum clips as the glass
unit exited the opening, we suspect causing
the breakage.
When designing the curtain wall system,
the structural engineer originally considered
only the aluminum framing in his or
her analysis; no consideration was given to
glass deflection. The insulating glass manufacturer
was expected to confirm the
strength of the glass units, a common practice.
With regard to the corner tube, the
structural engineer considered it only as a
dead-load support (i.e., as a hanger tube).
The structural engineer knew the corner
tube was not very stiff and simply assumed
the wind load would find its way to the horizontal
mullions designed to resist that
load. The structural engineer and the glass
manufacturer operated independently,
allowing this interaction problem to be overlooked.
The project team is currently pursuing
options for remedying this issue and moving
forward with the integrated design and
construction phases.
Lessons Learned
The causes of glass breakage can often
be traced to glass-to-metal contact, often
due to metal objects being slightly closer to
the glass edge than expected. In addition,
structural interaction among various elements
of a custom system can sometimes
lead to unforeseen movements and related
problems. Glass is a fragile material, and
glazing pockets are tiny spaces; slight deviations
in the expected dimensions and
deflections can lead to breakage. Careful
review and analysis are needed, both for
structural performance and for fieldinstalled
items that involve metal components
in close proximity to the glass. Keep
these points in mind for future projects:
• Consider all metal objects close to
glass, and evaluate the possibility
for migration and dimensional variations.
Provide cushion for the glass
as a rule to allow the glass units to
“float” within the opening.
• Perform a full structural analysis of
all typical and unusual conditions,
including consideration of all deflections.
Consider the combined effects
of deflections of multiple objects
simultaneously. Do not analyze
framing separately from glass (even if
this is convenient contractually)
because the two systems do interact.
• Consider all tolerances, including
combined effects of fabrication,
installation, structural, and all other
related tolerances.
THE CASE OF THE SPOTTED GLASS
SGH investigated large failed insulating
glass units at a waterfront curtain wall with
an unusual and complex geometry. The wall
is both sloped (tilted backward/inward) and
curved (concave). Building occupants and
maintenance staff started noticing visual
obstructions within the air space of the
glass units, including small brown spots
caused by deterioration of the metallic lowemissivity
(low-e)
coating. The expensive,
customized
wall system was in
place for less than
five years before the
occupants began
complaining about
the glass failures.
We reviewed the
curtain wall system
as well as the insulating
glass units
themselves after
deglazing and
found various
problems that were
contributing to the
failures (though the
primary cause was
defective glass units). The curtain wall
included weep holes in exterior pressure
bars and positive slope toward the weeps to
overcome the backward tilt of the wall,
though investigation revealed a significant
amount of debris in the glazing pocket that
absorbed water and slowed drainage. It
appeared that most of the debris was built
into the wall during the original construction
phase (i.e., glazing pockets were not
cleaned out prior to installation of glass).
Much of the water that entered the glazing
pocket drained harmlessly out the weep
holes, though some collected in these horizontals
due to this debris. This collected
water condition increased the relative
humidity of the glazing pocket space and
may have occasionally caused the insulating
glass unit edges to sit in water. Both
issues increased the risk of premature
glass-seal failure.
In addition to the debris issue, the insulating
glass hermetic seals were faulty at
failed units. We deglazed several units and
inspected them up close at a curtain wall
subcontractor’s shop. Upon close inspection
of the first failed unit, an open “blister”
in the silicone secondary seal at one corner
was revealed (Photo 12). The blister aligned
with an apparently unsealed keyed corner
of the spacer bar. It is clear that the blister
formed prior to curing of the sealant; the
unit may have been placed under pressure
prior to curing of the secondary sealant
and/or there may have been a mixing or
manufacturing problem with the sealant
material. Air escaped through a discontinuity
in the spacer bar (and primary seal) at
the keyed corner and exited through the
uncured silicone sealant, which resulted in
Photo 12 – Open blister in silicone secondary seal at the
corner of an IG unit.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0 MC C OWA N A N D K I V E L A • 1 7
the blister. Inspection of other failed units
revealed similar blisters.
To confirm that the blistered keyed corner
conditions were the root of the problem,
SGH placed certain failed units in a water
bath and applied slight pressure to the
units. Air bubbles exited the blistered corners,
confirming that there were breaches
in the continuity of the hermetic seals
(Photo 13). Small amounts of moisture
vapor reaching the sealed air space were
reacting with the metallic low-e coating in
the air space, causing the formation of the
brown spots. From a distance, the brown
spots had the appearance of condensation.
Insulating glass units cannot readily be
dried out or otherwise fixed once the hermetic
seal is breached and the interstitial
space is saturated; therefore, we recommended
replacement of all affected units.
We also suggest cleaning debris from all
glazing pockets to encourage prompt
drainage.
Lessons Learned
Hermetic seal failures may be the result
of manufacturing defects or design and construction
flaws that unnecessarily expose
the seals to water or, more often, a combination
of both of these factors. Carefully
specify, check, and enforce high-quality
hermetic-seal conditions, and design and
install draining curtain wall systems that
quickly remove water from the glazing pocket.
To do so, keep in mind the following:
• Specify durable, time-tested, insulating-
glass unit spacer and hermetic-
seal conditions, such as the following.
— Require spacer bars with bent,
soldered, or welded corners. Seal
or tape spacer-bar joints (do not
simply dry-fit joints with a splice
key). If keyed corners cannot be
avoided, flood these conditions
with sealant.
— Require continuous primary and
secondary seals, giving consideration
to joints in spacer bars as
potential breaches.
— Rigorously inspect insulating
glass units that arrive at the site,
and reject any units with seal
defects. Increase frequency of
inspections if even one bad unit
is found. Consider visiting the
insulating glass-manufacturer’s
shop to review operations. For
insulating-glass units that are
set in unitized frames, visit the
assembly plant prior to glazing
the frames.
• Install curtain wall systems that
promptly drain all water to the exterior.
— Avoid surface-sealed systems
that provide no drainage provisions.
Wet seals help minimize
water entry into the system but
do not rely on them alone to provide
waterproofing protection.
— Provide weep holes at the low
point of flat, horizontal surfaces
that may collect water. Slope sill
conditions toward weep holes
whenever possible for even faster
drainage.
— Avoid systems that drain down
the vertical mullions; instead,
drain water directly out weep
holes at the sill of every glass
lite, as a rule, as water traveling
down the verticals may contact
the edges of insulating glass
units or pond on top of insulating
glass units, increasing the
risk of premature failure.
— Be careful not to obstruct glazing
pockets with debris, excess
frame sealant, or glazing accessories,
as these can slow
drainage or even prevent it altogether.
CONCLUSIONS
Curtain walls can be effective and
attractive exterior wall assemblies if proper
consideration is given to strong design principles
and problems experienced on past
projects. Due diligence during the design,
preconstruction, testing, and installation
phases is warranted to identify and prevent
potential problems ranging from the relatively
simple (making sure that sealant is
installed correctly) to the more complex
(finite element modeling of complex glass
configurations).
REFERENCES
American Architectural Manufacturers
Association (AAMA), CW-DG-1-96,
(Rev. 2005), Curtain Wall Design
Guide Manual.
Glass Association of North America
(GANA), Glazing Manual, 2004
Edition.
D.B. McCowan, M.A. Brown, and M.J.
Louis, “curtain Wall Cautions;
Curtain-Wall Designs; Curtain-Wall
Problems,” Glass Magazine, April,
May, and June, 2007 (three-part
series).
Eric Olson, “Avoiding Water Intrusion
Problems in Field-Assembled
Glazing Systems,” presentation for
Windover Construction, January 12,
2010.
FOOTNOTE
1. A standard for evaluating thermal
stress in IG units is currently under
development by ASTM.
1 8 • MC C OWA N A N D K I V E L A S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • NO V E M B E R 2 0 1 0
Photo 13 – Air bubbling from IG unit.