Skip to main content Skip to footer

Unventilated Roof Tile Underlayments In Denmark

May 15, 2007

For some time, it has been common
practice in Denmark to use underlayments
in conjunction with sloping roofs, especially
in constructions where the ceiling and
the roof covering are parallel. The purpose
of the underlayment is to act as
an extra barrier against penetration of
water and drifting snow from the exterior.
Traditionally, roofs with underlayments
have been ventilated with a gap
between the insulation and the roof
underlayment with openings to the
surroundings (see Figure 1, left). The
ventilation removes moisture which, by
diffusion or convection, has penetrated
from the interior of the building.
Some 10 or more years ago, a new
type of material was introduced to the
market. These products were used in a
different way than previously, as they
were placed as a roof tile underlayment
directly on the insulation. The missing
ventilation gap meant that moisture
from the interior had to be removed in
a way other than ventilation, namely
diffusion. In order to achieve this, the
new products must not only be watertight,
but at the same time be very
open to water vapor diffusion, similar
to the way Gore-tex acts in clothing.
The materials are mainly thin membranes
that can be produced in a variety
of ways and are often laminated
and produced with a “carrier” material
to reinforce and protect the membrane
itself. Other types of materials act in the
same way – e.g., gypsum board, bitumenimpregnated
plywood (only in wood-based
roof elements), and oil-impregnated or coated
fiberboard. Oil-impregnated fiberboard
has been used in Denmark as a roof tile
underlayment for more than 30 years.
Figure 1 – Cross section of ventilated roof tile underlayment with a gap between insulation and
underlayment (left) and unventilated roof tile underlayment with the underlayment directly on
the substrate/insulation (right).
OC T O B E R 2007 I N T E R FA C E • 2 1
Shortly after introduction of the new
materials, a project was launched to investigate
the reasons for the problems, including
identification and assessment of properties
necessary to ensure a satisfactory performance.
The project was primarily
focused on properties related to moisture
transport in the roofs, as this is normally
considered to be crucial to performance and
Analysis of performance properties
An analysis was conducted in order to
identify the properties necessary for a satisfactory
performance of a roof underlayment.
This was based on the functions the underlayment
should fulfill and all the agents
anticipated to act on it, as described in CIB
Publication 64.1 It should be noted that
some of the functions and acting agents are
only relevant for some materials and/or
constructions; and, similarly, some properties
are only relevant for certain materials
and/or in certain periods of the service life,
e.g., the construction period.
For this particular purpose, only properties
related to moisture transportation
and buildup have been investigated. The
most important properties are found to be:
• Tightness against precipitation –
This property is especially important
during the construction period until
the primary roof has been laid. (In
Denmark, the underlayment often
serves as a temporary roof during
the construction period.)
• Tightness against water – This
property covers standing water as
well as water running on the surface.
When well constructed, no
ponding should occur on the underlayment,
i.e., it should have a welldefined
• No tent effect – It has been found
that in several cases, water penetration
occurs for underlayments laid
directly on wood or insulation. In
Denmark, this is called “tent effect,”
with reference to the well-known
fact that touching the inside of a
tent during rain may cause penetration
of water. For roof underlayments,
no tent effect should occur,
as it would impair the watertightness.
• Water-vapor permeability – For
unventilated constructions, it is evident
that moisture from the interior
of the building can only escape
through the underlayment by diffusion.
Consequently, the material
must be very permeable to diffusion
of water vapor.
• Moisture-accumulating properties
– These can be supplementary
assets if all other requirements of
the roof are fulfilled. These allow the
takeup and accumulation of moisture
during periods of high exposure.
The moisture is allowed to be
removed during other periods.
• Dimensional stability against
changes in RH – This is an important
property, but it is not considered
a problem with the current materials.
Based on the findings of the analysis, it
was decided to perform exposure tests in a
full-scale, test-house, computer simulation
of the moisture conditions in unventilated
roofs with a number of laboratory tests. The
investigations dealt with the new types of
materials, as well as the changed require-
RCI, Inc.
1500 Sunday Drive, Suite 204, Raleigh, NC 27607-5151
Phone: 800-828-1902 • Fax: 919-859-1328 •
A monthly* publication from
the industry’s experts, featuring
the latest technical information:
• avoiding trapped moisture
• energy efficient roof designs
• sustainable building products
• roof warranties
• protected membrane systems
• steep roofs vs. low-slope roofs
• bituminous roof systems
• cold-applied roof systems
• insulating cellular concrete
• wind resistance
• and more
the technical journal of RCI, Inc.
Company __________________________________________
Address ____________________________________________
City ____________________State ________Zip __________
Phone __________________E-mail ______________________
Amount Enclosed $ __________________________________
❑ Check (payable to RCI) ❑ Visa ❑ Mastercard ❑ American Express
Card No. ______________________Exp. Date __________
Name on Card ______________________________________
Signature __________________________________________
MAIL OR FAX TO: *RCI publishes 11 issues of Interface a year
22 • I N T E R FA C E OC T O B E R 2007
ments to the entire roof
construction to which
this might lead.
To investigate the
behavior of unventilated
roofs, a new roof construction
was made on
the test house at the
Danish Building Research
Institute (SBi).
The roof has a slope of
40º (1:1.2) and consists
of 11 pairs of elements,
each pair with an element
oriented toward both the
north and south.
Each element is 1 m
wide and has a height of
240 mm. Ten pairs are
unventilated, and the last
one – acting as a reference – is ventilated.
The elements are made with two timber
members as sides and a gypsum board as
interior surface. There is no vapor barrier,
but the gypsum board is painted to achieve
a desired water vapor permeability. The elements
are totally filled with mineral wool as
thermal insulation. On the outside, the
underlayment is placed directly on top of
the insulation material. The underlayment
is fastened to the rafters, and a distance
batten of 22 mm is attached to the rafters
over the underlayment. The roof is finished
with battens and roof tiles.
The climate in the test house is controlled
in winter to 23˚C (73˚F) and 60 percent
RH, a very high humidity level compared
to the expected 30 percent RH or less
under winter conditions. (See Figures 2 and
Measurements in the test house
The test house was equipped with sensors
and datatakers to monitor temperature,
humidity, etc., according to the following.
The roof elements are supplied with
humidity sensors in the timber members so
humidity can be recorded at 1) the bottom
(inside), 2) just below the underlayment,
and 3) in the counter (or distance) batten.
Some of the humidity sensors are supplemented
with temperature sensors. A number
of the elements are supplied with
humidity sensors glued to the underside of
the underlayment. These sensors, made by
Wetcor, were developed to measure time of
wetness and are herein used to measure
whether condensation occurs on the inside
OC T O B E R 2007 I N T E R FA C E • 2 3
Figure 2 – Mounting of
roof elements on the test
Figure 3 – Moisture
sensors are embedded
in the wood near the
bottom and in the top
sides of the element.
of the underlayment.
In two pairs of elements,
sensors are
mounted to monitor
the amount of water
running on the underlayment
in use. Similarly,
for two pairs of
elements, the temperature
on the inside of
the tiles is measured
with sensors mounted
on the surface of the
Supplementary to
the measurements of
the roof elements, the
temperature inside
and outside the house,
the relative humidity
inside and outside, the
number of hours of
sunshine, the wind
velocity, and the wind
direction are monitored.
Results of the
exposure in the
test house
As expected, the
humidity level in the
roof construction
showed a variation
over the year with a
rather high level in
winter, drying to a
lower level during the
summer. Even though
humidity level in winter
is high, it is not
assessed to be alarming,
considering the
use conditions, with
23˚C and 60 percent
RH in the test house.
Besides, it should be remembered that
the construction has no vapor barrier, but
only a surface treatment, providing a water
vapor resistance of approximately 17 GPa s
m2/kg (1.0 perm) in contrast to at least 25-
30 (0.58-0.7 perm) provided by a normal
vapor retarder. Finally, it should be noted
that the last winter of exposure was very
Only the reference roof was found to be
very wet, which is explained by unsatisfactory
ventilation, as there is no connection
from eave to eave; i.e., the ventilation gaps
in the two elements on each side of the
house have no connection, and consequently
there is very restricted possibility for wind
to pass through the construction and
remove moist air. (See Figure 4.)
Computer simulation of moisture transfer
Planning of the full-scale tests was supported
by computer simulation of the moisture
transfer in a roof construction by
means of the simulation program MATCH.
MATCH is a dynamic, one-dimensional
model for combined heat and moisture
transfer in composite constructions. The
model makes it possible to simulate the
temperature and
moisture conditions
on an hourly basis in
the individual layers of
the construction
based on knowledge of
the materials, properties
of the individual
layers, and the boundary
conditions to
which the construction
is subjected. The
model takes vapor as
well as liquid transfer
(but not precipitation)
into account. The climatic
data used are
from “The Danish Test
Reference Year.”
For practical purposes,
a vapor retarder
installed in a
building might be
expected to have a Zvalue
of 30 GPa s
m2/kg (0.58 perm) [Z
denotes water resistance,
not water vapor
resistance as is used
in North America.
Resistance and permeability
are reciprocal.] On the safe side,
a value of eight was
chosen for the calculations.
With this
assumption, the simulation
showed that,
dependent on the roof
underlayment material
used, condensation
might occasionally occur.
(See Figure 5.)
Laboratory tests
Laboratory tests
were made for the most important of the
performance properties related to moisture
transfer, i.e.:
• Water vapor permeability
• Tightness against precipitation
• Tent effect
The water vapor permeability was tested
according to ASTM E-96, wet-cup – i.e.,
with 100 percent RH in the cup and 50 percent
outside. Some 25 materials were tested,
including the new type of underlayment
materials as well as the previous types.
Those new types, available on a commercial
Figure 4 – Moisture in the top of two roof elements as measured with the sensors.
Note that the properties for the underlayments are not the same.
24 • I N T E R FA C E OC T O B E R 2007
basis, all had Z values below 3 GPa s m2/kg
(5.8 perm), whereas the old type had Z values
ranging from 30 to 500 GPa s m2/kg
(0.003 – 0.58 perm).
Tightness against precipitation was
tested at Velux’s (a major European skylight
producer) wind tunnel. Two different investigations
were carried out.
The first was primarily intended to
assess the differences in the amounts of
water penetrating through various roof coverings
and underlayments. The tests were
made on a test roof of 2.6 x 3.6 m. The tests
on underlayments were made with slopes of
25˚ and 45˚ and wind directions of 0˚ and
22˚ (0˚ is wind perpendicular to the building/
eave). For roof coverings, slopes of 25˚,
35˚, and 45˚ and wind directions 0˚, 22˚,
and 45˚ were used.
For this first test round, a test simulating
driving rain was used. The water is
equivalent to 120 mm per hour and the
wind is a dynamic wind profile with gusts
up to 20 m/s and 30 m/s for underlayments
and roof coverings, respectively.
The results were that considerably more
water penetrated when the underlayment
was mounted perpendicular instead of parallel
to the rafters. For board materials,
water penetrated the joints, irrespective of
whether these were made with overlapping
or with joint profiles. Some of the roof coverings
had virtually no water penetrating,
with a few drops at the most, whereas considerable
amounts penetrated others (i.e.,
water was running in at some joints). The
amount of water penetrating the roof covering
depends on slope and wind direction,
but the dependence varies from one roof
covering to another.
In the second test round, the underlayment
mounted perpendicular to the rafters
was tested, together with the roof covering,
where the highest amount of water penetrated
in the first test. The result was that
no water penetrated the inside of the underlayment
when the roof covering was present.
Water penetration of the underlayment
is therefore considered to be a
problem mainly during the building period.
For the testing of tent effect, no suitable
existing methods were found. Instead, a
proposal for a test method was made by the
SBi. The test method is comprised of accelerated
aging followed by a test of watertightness
of the product when placed in
contact with the substrate. The aging is
assessed to simulate roughly six months’
exposure to sunlight in Denmark. A flat aluminum
“tray” is used to collect any water
Figure 5 – Relative humidity and moisture content just beneath the vapor-permeable
underlayment as calculated with MATCH for a five-year period. Two underlayments with Z
values of 3 and 25 GPa s m2/kg, respectively, are compared. The upper figure shows the
relative humidity in the insulation (Iso) just beneath the underlayment, as well as in the
gypsum board (Gips) on the underside of the element. The lower figure shows that
equilibrium is reached for both products, though on different moisture levels.
26 • I N T E R FA C E OC T O B E R 2007
penetrating the underlayment during the
water exposure. A batt of mineral wool is
placed in the lower end of the tray. In the
upper end, a piece of plywood is placed.
Both materials are slightly higher than the
aluminum tray. The
underlayment is mounted
on a wooden frame
that fits around the aluminum
tray. When
mounted, the frame is
placed around the tray,
causing the underlayment
to rest around the
mineral wool and the
plywood, respectively.
The underlayment is
held in place by the
weight of the frame and
the underlayment,
which is intended to
simulate the conditions
of a real roof. Water is
sprayed from a nozzle
over the entire surface of
the specimen for six
hours. The water pressure
in front of the nozzle
is very low, simulating
a fine to medium
rain. The test method
has evolved into a Nordtest
test method, NT BUILD 488, “Roof Tile
Underlays: Watertightness – Tent Effect.”
(This may be downloaded at www.nordicinnovation.
Some of the materials tested using this
method showed a considerable
penetration of water
during the test, which is
unsatisfactory, as water will
flow on the underlayment in
most roofs. (See Figure 6.)
In-situ investigations/
experience from practice
A small number of insitu
investigations were
made in the winter of 1995-
96. It was found that some
materials suffered from
water leakage – probably
due to tent effect – resulting
in a number of products
being withdrawn from the
market. The testing of tent
effect has subsequently
become common for new
products. A number of products
that failed the test
either have not been marketed
or have only been
marketed after improvements.
Some of the underlayments
have been attacked by mold growth,
and in a few cases, ice has formed on the
inside of a roof underlayment during longlasting
cold periods. This shows that the
underlayments sometimes have moisture
Figure 6 – Testing of tent effect. The underlayment is placed over a substrate, composed partly of
plywood and partly of insulation. Any water passing through the underlayment during exposure to
water spray from the nozzle is accumulated in the aluminum tray and weighed.
Figure 7 – Example of detail around ventilation duct – very poor workmanship!
OC T O B E R 2007 I N T E R FA C E • 2 7
accumulation on the inside. Visual inspections
in a number of identical houses suggest
that the problem is especially pronounced
where the vapor retarder is not
perfect; e.g., where it has been perforated or
where there are leaks around openings.
Visual inspections in the test house showed
that ice had formed on one of the two
underlayments. This stresses the importance
of an airtight construction. It is especially
important to avoid convection, which
might transport considerably more water
vapor into the roof construction than can be
removed by diffusion through the underlayment.
Even though the underlayments are
very open to diffusion, it is now considered
necessary to have at least some ventilation
of open roof spaces to prevent moist air
from accumulating in the top of the roof.
Quite a few problems are associated
with the detailing and workmanship of roof
underlayments. Some suppliers only sold
the products but did not give any advice as
to their installation. Others described just a
few of the most-used/easy details. Consequently,
the detailing was often designed by
the contractor on the building site, resulting
in mediocre quality. Over the years, a
number of leaflets have been published
showing the principles of the most-used
details. Lately, a rather large number of
details have been issued by the Danish
Timber Information Council (TOP). These
may be downloaded from TOP
is currently finishing a booklet about
underlayments with wooden materials, e.g.,
boards, plywood, and fiberboards (with or
without a supplementary, watertight membrane)
that will also have a number of
Poor workmanship often results in
problems, mainly because the work is not
done as prescribed or repairs are done
incorrectly or not at all. A common and
easy-to-detect example occurs when the
underlayment has not been mounted sufficiently
tight. It might flap, thereby creating
considerable noise, and eventually, puncture
of the underlayment where it touches
the tile clips. Otherwise, the main problem
is that details are not made watertight
because the correct procedure is time consuming.
Some of these problems might not
only be difficult to see in the finished roof
(especially if covered by a ceiling afterwards),
but also very expensive to repair.
This calls for more responsibility by the
craftsman and/or more rigorous supervision.
(See Figure 7.)
For some years, a private organization
collected information on available documentation
from all the suppliers of roof
underlayments. A couple of years ago, this
task was handed over to a new organization
called DUKO (Danish Roof Underlayment
Classification Scheme). The organization is
owned by a number of interested parties,
including contractor representatives and
building owners.
The organization’s main purpose is to
make a voluntary classification of products
with documentation for the most important
properties, including strength, watertightness,
potential for water vapor diffusion,
tent effect, and durability. Based on physical
properties, the underlayments are classified
into a number of exposure classes.
The suppliers have to provide the documentation
in the form of test reports, including
proof of constructability; i.e., they must
show how to make details such as eaves,
chimneys, skylights, etc.
Helping Make
Buildings Better™
BASF Polyurethane
Foam Enterprises LLC
28 • I N T E R FA C E OC T O B E R 2007
OC T O B E R 2007 I N T E R FA C E • 2 9
Proof of constructability is achieved by
companies mounting their roof underlayment
mock-ups, including all imaginable
details. The basic idea of the classification
system is that the underlayment is chosen
dependent on the watertightness of the roof
covering when subjected to driving rain,
slope of the roof, complexity of the roof
design, serviceability of the underlayment
after installation, and exposure to wind. For
example, some roof coverings are exposed to
UV light as well as to rain and consequently
require a very durable underlayment.
Products appearing on the classification
list from DUKO can be considered to have
sufficient documentation for their properties,
including constructability. Further
information may be downloaded from (See Figure 8.)
Roof tile underlayments have been used
for 50-plus years in Denmark. As a whole,
they have functioned properly. Roof-tile
underlayments are a prerequisite for the
use of attics over occupied spaces – i.e.,
with at least part of the ceiling and the roof
parallel (because earlier tightening of the
tiles with mortar, bitumen, etc., is not
durable and cannot be maintained).
The ventilated underlayments have
functioned properly for a long period of time
(after some early problems with durability of
the products), provided common knowledge
about the construction/ventilation is used.
In regard to unventilated underlayments,
costly errors have shown the need to
fulfill certain requirements if a well-performing
roof is to be achieved. These
include most of the following:
• The underlayment shall remain tight
when subjected to precipitation,
including driving rain. The underlayment
shall pass a Nordtest Build
118 test (or similar) without experiencing
water penetration.
• The underlayment must have no
tent effect. Only a very small
amount of water is allowed to pass
the artificially aged underlayment
(in Denmark, the requirement is less
than 15 g) when tested according to
Nordtest Build 488.
• The water-vapor permeability of the
underlayment shall be low. In Denmark,
the requirement is that the Zvalue
shall be less than 3 GPa s
• The inside of the roof construction
shall be sufficiently tight to avoid
diffusion and convection. The tightness
shall be seen in relation to the
permeability of the underlayment.
The Z-value of the inside of the construction
should be greater than 30
GPa s m2/kg. Normally, this is
achieved by using a vapor retarder
that, when well mounted with overlapping,
taped joints and no perforations,
will fulfill the requirements for
both permeability and air tightness.
• Documentation that all imaginable
details that can be made securely
shall exist.
Further, the instructions from the supplier
should be taken into account, especially
related to how long the underlayments
can be used as a temporary roof
directly exposed to UV light and precipitation.
In this connection, it should be noted
that some roofing tiles on the Danish market
will allow a penetration of as much as
five percent of the surrounding UV light.
Finally, the workmanship is crucial to the
overall performance of the roof construction.
Only when properly mounted may an
Figure 8 – Mock-up to be used for the suppliers’ proof of constructability.
underlayment be expected to fulfill its duty
for many years.
Provided that the above are fulfilled, it is
assessed that unventilated roof underlayments
are able to function in a temperate
climate such as Denmark’s.
Editor’s Note: This article is reprinted from
the Proceedings of the RCI 22nd International
Convention and Trade Show,
March 1-6, 2007, Orlando, Florida.
1. Working With the Performance
Approach in Building, CIB Report,
Publication 64 (International Council
for Research and Innovation in
Building and Construction, Rotterdam,
Building detail sheets, TOP [(Danish
Timber Information Council) in
• Roof underlayments. Concepts
and properties.
• Roof underlayments. Typical
• Unventilated roof underlayments.
Construction principles.
• Ventilated roof underlayments.
Construction principles.
Byg Erfa data sheets, Byg-Erfa
[(Building Experience Feedback
Council) in Danish].
• Flapping roof underlayments.
• Underlayments. Construction,
materials and design.
• Underlayments. Mounting and
NT Build Method 488, “Roof Tile Underlays:
Watertightness – Tent
[Undersøgelser af uventilerede undertage
(Investigations on Unventilated
Roof Underlayments), SBi Report
292, 1998 (in Danish).
Working With the Performance Approach
in Building, CIB Report, Publication
64 (International Council for Research
and Innovation in Building
and Construction, Rotterdam, 1982).
Erik Brandt is a senior researcher at the Danish Building
Research Institute. He has worked for many years with building-
technology problems in general and problems dealing
with moisture, service life, building surveys, roofs, and floors.
Erik is a member of a number of national, as well as international
working groups, standardization committees, etc. He is
the author of a large number of scientific papers and publications
about building technological problems.
Erik Brandt
Tommy Bunch-Nielsen earned a master’s degree in structural
engineering in 1974 and has worked as a consulting engineer
for more than 30 years. He has specialized in building
physics and is director of the consulting company BMT A/S
and the contracting company Micro Clean A/S. The firm
works with chemical-free mold removal from buildings. He is
also the chairman of the Danish Roofing Advisory Board.
Tommy Bunch-Nielsen
RCI has received a letter of notification from the IRS confirming RCI’s
group exemption for U.S.-affiliated chapters. Each affected chapter was provided
a copy of the letter and a list of those affiliated chapters included in the
group exemption as submitted by RCI.
Chapters not shown on the submitted list that want to be included in the
group exemption should contact Catherine Moon by phone at 800-828-1902
or e-mail at to discuss the matter.
􀁕􀃊 􀀖􀂘􀁠􀁩􀂫􀁩􀂘􀁠􀁩􀂘􀃌􀂏􀃞􀃊􀃌􀁩􀃃􀃌􀁩􀁠􀃊􀀾􀂘􀁠􀃊􀂫􀃀􀂜􀃛􀁩􀂘􀃊􀃊
􀁕􀃊 􀀫􀃕􀂈􀁖􀂎􀁝􀃊􀃃􀂈􀂓􀂫􀂏􀁩􀃊􀂈􀂘􀃃􀃌􀀾􀂏􀂏􀀾􀃌􀂈􀂜􀂘􀃊􀃕􀃃􀂈􀂘􀁽􀃊􀃊
􀁕􀃊 􀀭􀃌􀀾􀂘􀁠􀀾􀃀􀁠􀃊􀃃􀁩􀂏􀁶􀂇􀁖􀂏􀂜􀃃􀂈􀂘􀁽􀃊􀁽􀀾􀃌􀁩􀃊􀁶􀁩􀀾􀃌􀃕􀃀􀁩
􀁕􀃊 􀀊􀂜􀃀􀃀􀂜􀃃􀂈􀂜􀂘􀂇􀃀􀁩􀃃􀂈􀃃􀃌􀀾􀂘􀃌􀃊􀁖􀂜􀂘􀃃􀃌􀃀􀃕􀁖􀃌􀂈􀂜􀂘􀃊􀃊
􀁕􀃊 􀀕􀂈􀁽􀂅􀂇􀃛􀂈􀃃􀂈􀁌􀂈􀂏􀂈􀃌􀃞􀃊􀃃􀀾􀁶􀁩􀃌􀃞􀃊􀃞􀁩􀂏􀂏􀂜􀃜􀃊􀁖􀂜􀂏􀂜􀃀
􀁕􀃊 􀀠􀂜􀂘􀂇􀂫􀁩􀂘􀁩􀃌􀃀􀀾􀃌􀂈􀂘􀁽􀃊􀀾􀃌􀃌􀀾􀁖􀂅􀂓􀁩􀂘􀃌

􀀪􀂈􀃛􀂜􀃌􀂈􀂘􀁽 􀀟􀂜􀃕􀂘􀃌􀂈􀂘􀁽 􀀭􀂏􀁩􀁩􀃛􀁩
􀀓􀂜􀃀 􀂈􀂘􀃃􀃌􀀾􀂏􀂏􀀾􀃌􀂈􀂜􀂘 􀁹􀁩􀃝􀂈􀁌􀂈􀂏􀂈􀃌􀃞
􀀭􀁩􀂏􀁶􀂇􀀊􀂏􀂜􀃃􀂈􀂘􀁽 􀀔􀀾􀃌􀁩 􀀕􀂈􀂘􀁽􀁩
􀀍􀂘􀃃􀃕􀃀􀁩􀃃 􀁖􀂜􀂘􀃌􀂈􀂘􀃕􀂜􀃕􀃃 􀂫􀃀􀂜􀃌􀁩􀁖􀃌􀂈􀂜􀂘
30 • I N T E R FA C E OC T O B E R 2007