Part I: A look at the curling mechanism and the effect of moisture and

shrinkage gradients on the amount of curling

BY BRUCE A. SUPRENANT
 mo~stumand shrinkage aradient I
M . ost slabs on grade curl. Sometimes t h e curl is
considered objectionable, sometimes it's not. When
t h e curl is considered objectionable, we consider the
constant tlRIIn@snvlronmont
In 1934. Carison obtained experimental data showing
causes. Sometimes we understand what caused the curl, moisture contents and shrinkage gradients for slabs
sometimes we don't. When we believe we know what drying only from the top.' He tested 6-in. (300 mm)
caused the curl. we are ready to attribute blame and seek concrete cubes cast in copper forms and exposed the
remedy. Sometimes we're right, sometimes we're wrong. tops of the cubes to a drying environment of 80 "F
Some research has been done over the years to (27 'C) and 50%relative humidity. He used concrete
determine why slabs curl and how curling can be mixtures made with varying combinations of three
reduced or controlled. Unfortunately. because much of differentcements and two aggregate types.
that work is so scattered, it hasn't received the attention Figure 1 shows the moisture loss and shrinkage in the
it deserves nor has it contributed to a better under- . specimens at 600 days, measured at various distances
standing of why slabs curl. The following is a summary from t h e exposed concrete surface. Greater moisture
of the research and what it may mean to designers, loss and shrinkage occurred near the top, and exposed
specifiers. and contractors. concrete surface and less moisture loss and shrinkage
occurred near the bottom. This shrinkage gradient due
Mechanlcs of curling to moisture-content differences applies a curling
Slab curling is caused primarily by differencesin moment to the slab, causing it to deflect upward. The
moisture content or temperature between the top and larger the shrinkage gradient, the greater the differ-
bottom of the slab. The slab edges curl upward when the ence between top and bottom Shrinkage, and the larger
surface is drier and shrinks more, or is cooler and the applied curling moment. The larger the applied
contracts more than the bottom. Curling is most notice- curling moment, the greater the upward deflections.
able at construction joints, but it can also occur at It's interesting to note that the shrinkage gradients
sawcut joints or random cracks. The curl can result in a provided by Carlson are different for different cements.
loss of contact between the slab and subbase. Generally, We typically use the drying-shrinkage potential of
the length of lost subbase contact is about ¡O!% of the concrete when evaluating the amount of curl, but we
slab length (measured between joints) at joints that have don't consider the effects of the concrete ingredients on
load transfer (doweled or sawcut joints), and about 20% the shrinkage gradient because not enough test data are
at joints with no load transfer. However. these values are available. Carlson's work showed that differences in
also a function of joint spacing, concrete properties, slab measured shrinkage due to changing cements or aggre-
thickness, and subbase stiffness. Upward curl at slab gates were small near the concrete cube's surface but
corners can be as high a s 1 in. (25 mni). but is typically could be much larger near the bottom.
about 114 in. (6 mm).
The term curling is used to refer to upward vertical
deflections in slabs caused by moisture-content differ-
ences and the associated shrinkage. The term warping is
sometimes used to refer to upward vertical deflections
caused by temperature differences between the top and
bottom slab surfaces. In practice, a measured vertical
slab deflection is a combination of both temperature
and moisture differences, so the resulting upward
deflections from both sources are typically referred to
as curling. Both the temperature and shrinkage differ-
ences apply an upward curling moment to the slab that
causes the upward deflections.
The curling moment that lifts the d a b is greater
near the ends of a slab and then decreases to almost
zero at the slab center. Due to gravity forces, t h e
internal stresses caused by curling are smallest near
the slab end and highest over a large center area.'

Molsture and shrlnkaae gradients

To assess the relationship between moisture and
shrinkage gradients in concrete and the aniount of nisanre I b m a p d I=< iiiihor

curling that occurs, several researchers measured

moisture and shrinkage gradients in specimens exposed
to various drying conditions. Here are their findings:
T
Fig. I: Comparison between computed distribution of d ng and
measured distribution of shrinkage in 6411. (152 mm) cu es of
concrete drying fmm one face only (From Reference z)
 shrinkaae mradlents at dlfferent
drylnm condltlons
Research shows that the relative iiurriidity o f the
drying eiivironment can change the shrinkage gradient.
Concrete loses more water at a lower relative huniidity.
resulting i n more shrinkage. Data from Keeton. shown
in Fig. 2, provide shrinkage profiles for relative humidi-
ties from 20 to 10O'x."Asthe relative iiurnidity decreases.
the shrinkage gradient increases. thus causing a greater
applied curling moment and greater upward slab curling.
Work at the Portland Cement Association by Abrairis
and Orals4and Abrams and Monfore5also shows that
inoisturecontent gradients in 1:oncrete depend on the
relative humidity of the drying environment, with
concrete losing more moisture and at a faster rate at a
lower relative humidity (Fig. 3).
Because slabs that are exposed to a lower relative
humidity will lose moisture faster, they will curl sooner
than slabs exposed to a higher relative humidity. Thus. a
slab placed in Phoenix is likely to curl sooner than a slab
placed in Houston (assuming sirnilar concreting materi-
flg. P: One year of shrinkage data for a ipin.-thick e80mm) als, mixture proportions, subbase stiffness, and placing
slab as a function of reiative humidi wlth only the top surface
exposed to drying (From Reference 3 and curing methods are used) because t h e relative
humidity i n Phoenix i s lower.
Melstum aradlents In the fleld
Researchers at Penn State University used a stacked
disc technique to measure the moisture contents of a
pavement and two bridge decks: One of the bridge decks
was placed on stay-in-place metal fornis and the other
deck was placed on plywood forms, which were removed
after the concrete was placed and cured. Figure 4 shows
the rnoisture£ontent gradients associated with each of
the structures. These full-scale field tests verify moisture-
gradient profiles found i n other researchers' labs.
Janssen also measured nioisture-content differences
in pavements in the field.: Using lab measurements. he
then calculated shrinkage strain and stress i n the
pavement. Figure 5 shows his estimate of iiioisture
distribution in an S-in.-thick (200 nini) pavement and the
resulting stress distribution due to differential shrinkage.
I added the arrows to better illustrate the applied
upward curling moment. which Jansseii calculated as
2500 in.-lb/in. ( i 1,125 N-miri/mm) of slab width.
oemm of moisture loss
Note that in Fig. 1 through 5 , the moisttire loss is
significant only i n the top 2 in. (50 mm) of the speciiiierrs,
50x
-
TEST HUMIDITY

-- 90%
75%
regardless of the specimen size. An ofteii-used rule of
thumb for the time needed to dry concrete floors to be
covered with a moisture-sensitive covering is 1 month for
each inch (25 mm) of slab thickness. However, when
Suprenant and Maiisch measured moisturecniission rates
for concrete slabs during a Snionth-long drying period.
I they found that reduction ill the emission rate with tilile
I 2 3 was about the sanle for slabs 2-, 4-. G.and &in. (50, 100,
Distance lrom center of Slab, Inches 150, and 200 mm) thick! These measurenients were
flg. 3: E&ct of ambient humidity on humidlty gradient through repeated with four different concrete mixtures, and again
slab sections (From Reference 4) reduction in emission rate was unaffected by slab thickness.
 Slabs of all thicknesses reached
about the same emission rate at the
same drying time. This too seems to
show that moisture loss is significant
only in the top 2 in. (SO mm) of t h e
specimens or less, regardless of the
specimen size.
Effect of subbase condltions
Nagataki tested three sets of 4 x 4
x 20-in. (100 x 100 x 508 mm) con-
crete specimens cured for 7 days
and then exposed to a drying
environment of 75 'F (24 'C) and 50%
relative humidity.YH e tested one
specimen with all sides exposed to
drying, one specimen drying only AVEUAGE MOISTURE CONTENT per cent
from the top, and the final specimen
with only the top exposed to drying --

XI,
and the bottom on wet sand at a 10%
moisture c o n t e h Figure 6 shows the
shrinkage gradients h e measured.
As would be expected, Specimen I
(dried from all four sides) had only a
slight shrinkage gradient. Specimen II
(dried only from the top) had a
much larger shrinkage gradient and.
therefore, would have a larger t30 60 70 60 90 100 lío
applied curling moment. Note that
the bottom of Specimen II still dried
some as there was a resulting AVEUAGE MOISTURE CONTENT, per cent
shrinkage. Specimen 111 had the
largest shrinkage gradient because
the bottom expanded when exposed
to the moist sand subgrade.
In practice, it's often difficult to
explain differences in the amount of
slab curl for floors that are built with
similar materials, mixture propor-
tions, and construction methods.
and then exposed to similar drying I I 1 I 1
50 60 70 O; 90 100 110
environments. It's likely that the
AVERAGE MOISTURE CONTENT, per C W i
subbase moisture condition plays an
important role in determining the Fig. 4:Moisture distribution In a pavement and in bridge decks (Fmm Reference 6)
shrinkage gradient and, therefore.
the applied curling moment when all
other factors are nearly the same. ACI Committee 302, content was less than 1%. In the first hour, the moisture
Construction of Concrete Floors, suggests that for slabs content of the sand increased to 5 or 6%, but during the
to receive moisturesensitive floor coverings, it may be second hour, it decreased to 2 to 3%as the water was
best to place concrete directly on a vapor retarder reabsorbed back into the concrete. Because the
because a wet subbase can adversely affect floor- reabsorption process occurred quickly, it's difficult to
covering performance."' say whether a dry subbase would produce a moisture
Specimen II from NagataW's work represents a shrinkage gradient different than what would occur when concrete
gradient typical for slabs,placed on a vapor retarder. He is placed directly over a vapor retarder.
didn't include a dry subbase in his coniparison. In another experiment, Nagataki placed a 10-in.-thick
Suprenant and Malisch, however, measured the moisture (254 mm) concrete pavement 34 ft 4 in. long (10.5 m) by
content of a 1- and Z-in.-thick (25 and 50 nim) sand layer 2 ft 8 in. wide (0.8 m) on a heavy sheet of paper. He then
beneath freshly placed concrete." The initial moisture covered the pavement with wet burlap for 10 days.
 I, 1,;
- ESTIMATE OF
MOISTURE DISTRIBUTION
- O
TOP -

n
1-

2- APPLIED CURLING MOMENT

- 2
5 3-

z
IuI.
STRESS OlSTRlBUTlON
4- DUE TO
DRYING SRINKAGE
S-

6-
H E
7-

8-
BASE , , *

-200 o 2w 4w 600 600 1wJ

COMPRESSION I TENSION

20 10 Bo Bo 1w
(a) SATURATION (%) STRESS (PSI)

Fig. 5: (a) Estimate of moisture distribution for an 8-in.-thick (zoo mm) pavement; (b) m e stress distribution due to the molsture
gradient. A m m s show the resulting applied curling moment (Modified from Reference 7)

Carison-type striiiii qa#es were placed throughout the

UNIT SHRINKAGE DEFORMATION ( l o 6 ) concrete paveiiieiit H r t w k strain readings at different
-100 O 100 2W 300 400 500
depths. at dillereiii Iij<'ntiiinsin the pavement. and at a
tiinr t o rnirii~iiiiet t i e reiiiperature gradient in the pavr-
n i r i i t . As s h i m i i 111 Fig. 7 , Uagataki was able t o plot the
shrinkage gradirnts iir \~;iriousdistances from the frer
rnrl of ttie j>avriiiriit ' T i i t , mean shrinkage throuxh ttic.
depth of the st'i'tiim IS ;iiiiiost the same because the
a pajwr-covrrrd siihgr~iili.(dfered little resistance t o
tiorizonrnl slidiiig TIie shrinkage gradient is higher al
th? I r r e riid t h w <iiicr<.iist.s nearly linearly in proportion
I t o the (listaiiw frtiiii the. free end.

I
, / Work hy Nicti<#lsi~ii IS i.ited to prove that concrete

-100 o 100 200 300 400 500 piaced ciirt~c~tlv o 1 1 ;I v,iii(ir retarder will curl more than
wheii plarr(l OII .igraiiular subbase." For iastancr. ACI
n , 6: Dlstributlon of shrinkage defomat~on prisms 3tiIiK-'K! "1)rsiyii i l l Sl;il,s on Grade." includes the state-
di%erent drying conditions (From Reference 9) ineiit that "Wrlii,lsoii s i i ~ i w e dthat serious shrinkage
cracking - and ciirliiis call < m u rwhen concrete slabs are
cast on an impervious base."" This statement is
incorrect. Nicholson didn't measure curling or any
other flatness-related property in
his research.

Shrinkage-gradlent ettects
The cited research results indicate that:
H Curling is the result of nonuniform drying that
establishes the moisture gradient, the resulting
stress distribution, and applied curling moment
and thus the amount of curl;
H The drying takes place in the top few inches
O ! I
sin ~ - m
I 4
11-?rJ
regardless of the slab thickness or external
~g environment;
W For a given moisture gradient, differing concret-
DISTANCE FROM FREE END (FEET) DISTANCE FROM FREE END (FEET) ing materials can cause differing shrinkage gradients
within a slab, with the differences likely to be
depth of pavement sections; and greater near the hottom of the slab;
each section (Fmm Reference 9) =Placing a concrete slab on a wet subbase increases
 the shrinkage gradient and the applied curling mo- 12. Nicholson. L. P., "How to MinimizeCracking and increase
ment, and thus the amount of curl; Strength of Slabs on Grade,"Corrcrelr Conslmucrion.Sept. IYRI.
Whether concrete curls more when placed directly on pp. 73Y-741
a vapor retarder or on a granular subbase depends on 13. ACi Committee 3íüL "Design of Slabs on Grade (ACI XOR-Yi)."
t h e moisture content 01 the subbase: America81Concrete institute. Farmington Hills. Mich.. iY92. p. 35.
Slabs exposed to low relative humidities develop
greater shrinkage gradients that can increase the Received and reviewd under institute publication policies.
applied curling moment and cause more upward
deflection at joints or cracks than higher relative
humidities; and
W T h e same concrete may exhibit different amounts of
curl d u e to t h e different final environments.

Factors related to the slab's final environment-

FACl Brute A. Supmnint, P.E.. PhD., isvice
temperature a n d relative humidity at t h e surface, and
President of Engineering and Technical
moisture content in the subbase or subgrade if it's in Services for Baker Concrete Construction,
contact with t h e concrete-can affect t h e amount of curl Monroe, Ohio. He has also served as a
as much a s t h e concrete properties. However, we usually ~ structural engineer for Sverdrup E Parcel. an
attempt to control curling by modifying the concrete. analytical structural engineer forthe
In Part II o f t h i s article, 1'11 discuss other factors Portland Cement Association, and taught
alfecting the amount of curling. materials, construction, and structural
References engineering courses at several universities.
He is a member of ACI Committees 117.
1. Walker, W., and Holland. J.. "Thou Shalt Not Curl Nor Crack"
Tolerances; 301. Specifications for Concrete;
Concrere hlernutrmul. V. 21. No. 1, Jan. 1 Y 9 Y pp, 12-18.
and 302. Construction of Concrete Floors.
2. Carlcon. R. W.. "Drying Shrinkage of Concrete as Affected by
Many Factiirs." Pmceedings uf Ihe Americun Society for Tesling and
Muleritils. ASTM. West Conshohocken,Pa., V. 38. Part 11, 1938.
pp, 4194440.
:iKeeton.
. J. R.. "Shrinkage Compensating Cement forAirport
Pavement,"Reporr No. FAA-RD79-11. Phase 2. Federal Aviation
Administration. Washington D.C.. Sept. 197%
4 . Abrams. M. S..and Orals. D.L.."Concrete Drying Methods
and Their Effect of Fire Resistance." Moisture in Marerials in
Rrluti<in rv Fire Teslr. STP 385, ASTM, West Conshocken. Pa.. 1965,
pp. 52-73.
5. Abrams. M. S..and Monfore. G.E.. "Application of a Small
ProbeType Relative Humidity Gage to Research on Fire Resistance
of Concrete." Journol of the Portfond Cemenr Asswiorion Research
ond Deiwlopmenr lulmrolory. Portland Cement Association. Skokie,
ili.,Sept. iYfi5.
6. Carrier, R. E.: Pu, D. C.; and Cady, P. D.."Moisture Distribution
in Concrete Rridge Deck arid Pavements." Durnbility of Concrete.
SP-47.American Concrete Institute, Farmington Hills. Mich., 1975,
pi'. 169-19'.
i ..lanssen. D. J., *Moisture in Portland Cement Concrete,"
Tmnsp<muririnRese-orch Record i i 2 1 . Transportation Research
Board. Washington ».C., 1986. pp. 4i!-44.
8. Suprenant. B. A,. and Malisih, W. R., "Quick-Dry Concrete: A
New Market for Ready-Mix Producers.' The Concrele producer, May
IYY8. pi,. 3:30-333.
Y. Nagataki, S.,"Shrinkage Restraints in Concrete Pavements."
Journol of the Srrucfiirul Diului»n. Proceedings of the American
SricietyofCivil Engineers. July iYiü, pp. 133?-1358.
10. ACi Committee 302. "Vapor Retarder Location," Concrele
Inlernolionul. V. 23. No. 4, Apr 2üüi. pp. 72-73.
iI . Suprenant,B. A,. and Malisch, W. R.. private communication.