Slab curling is a function of the moisture gradient and temperature differential across the slab thickness. Moist subgrades increase slab curling, but even a so- called dry subgrade keeps the bottom of the slab damper than the slab sur- Figure 4. Relat- face. This article continues ing slab curling to a series started in the No- concrete shrink- age shows that vember issue (page 825), the greater the and figures and tables are shrinkage, the numbered consecutively greater the curl- ing (from Ref. 7). with that beginning article.
lab edges that curl upward and cen-
S ters of slabs slightly depressed from their original elevation are common in slabs on grade. Curling makes a mockery of the commonly specified 1⁄8-inch flatness tolerance because slab edges can curl up- ward 1⁄8 inch or more above a spot only 4 to 5 feet from the edge. At slab corners the uplift can be greater—3⁄16 to 1⁄4 inch above a location 4 to 5 feet from the slab corner. F i g u re 5. O n e Heavy loads on the slab can cause the un- year of shrink- age data for a 15- supported, curled-edge cantilever to crack. inch-thick slab as Curling is caused by drying shrinkage a function of rela- and by moisture or temperature differ- tive humidity with ences across the thickness of the slab. only the top sur- Negative moisture gradients are usual in face exposed to slabs on grade and they cause upward drying (from Kee- edge curling. Such moisture gradients and ton as cited in Ref. 6). curling increase when: • The slab is made from high-shrinkage concrete Figure 6. Enclosed slabs (bottom) curl only upward from the negative moisture gradient. Outdoor slabs exposed to the sun (top) may curl downward at the edges.
ments, insulation, restrictions on the rate
and time of placement, or admixtures. Long-term relative humidity at the sur- face of the slab has a tremendous effect on the moisture gradient. Figure 5 shows shrinkage at various slab depths at 1 year, making it clear that drying shrinkage and upward curling of slabs on grade increase as the ambient relative humidity decreases. • The slab surface is exposed to low- Either temperature or moisture gradi- Drying shrinkage can continue for humidity air ents or both can cause curling in slabs on years. Even when dried from both faces, • The subgrade or subbase under the grade, and researchers frequently express it takes 28 months for a relative humidity hardened slab has a high moisture moisture gradients in terms of equivalent of 50% to be reached at mid-depth in a 6- content temperature gradients. Temperature gradi- inch-thick concrete member with 35% ents are easier to measure than moisture relative humidity on both faces (Ref. 16). • The concrete has a high modulus of gradients. The latter are caused by the In a typical enclosed slab on grade (Fig- elasticity ambient relative humidity at the upper ure 6, bottom), the absence of heat from The most common positive tempera- slab surface, the free water in the concrete the sun on the upper surface leaves only a ture gradient with its downward edge (drying shrinkage potential), and the moisture gradient (partially offset by slab curling is that caused by heat from the sun moisture content of the subbase or sub- weight) which produces upward curling. on the upper slab surface. Upward edge grade below the slab. The moisture content is much higher in curling already caused by negative mois- Heat of hydration of cement can cause the lower half than in the top half. The ture gradients can be increased by cold early upward curling (negative tempera- slab can only dry from its upper surface slab surface temperatures or by hotter slab ture gradient) in slabs on grade (Ref. 15). and therefore shrinkage is greatest at the bottom temperatures. Vertical deflection In high-cement-content slabs the hydra- upper surface and the slab curls upward. is greater at slab corners than at slab tion process creates an exotherm, which heats the slab bottom more than the top Shape of slab curling edges simply because that is what hap- pens when two adjacent, perpendicular and can cause significant internal stresses ACI’s Concrete Craftsman booklet, edges in a flat plane both roll upward. and cracks. The exotherm can be con- Slabs on Grade (Ref. 17), incorrectly trolled with ice, post-cooling, special ce- states on page 61 that the amount of curl- Definition of terms The terms curling and warping are Table 2. Drying Shrinkage Gradients Expressed as Equivalent used interchangeably in this article, in Temperature Gradients, and Critical Slab Lengths for conformance with ACI terminology (Ref. Various Temperature Gradients (From Ref. 7) 14), which defines them as follows: Curling – The distortion of an original- Equivalent Equivalent temperature Shrinkage strain gradient for various slab ly essentially linear or planar member into temperature gradient through thicknesses (° F per in.) a curved shape such as the warping of a gradient through slab (millionths) slab due to creep or to differences in tem- slab (° F) 6 in. 8 in. 10 in. perature or moisture content in the zones adjacent to its opposite faces. 275 50 – – 5.0 Warping – A deviation of a slab or wall 220 40 – 5.0 4.0 surface from its original shape, usually 165 30 5.0 3.8 3.0 caused by temperature or moisture differ- 110 20 3.3 – – entials or both within the slab or wall. Critical slab length (ft.) for various temperature gradients Slab Curling and differential shrinkage (T, ° F) thickness 6 Although not all slab-on-grade curl- (in.) E 5 10 psi E 3 106 psi ing is caused by drying shrinkage, Fig- T 20 T 30 T 40 T 20 T 30 T 40 ure 4 shows that the potential for curl- 6 25 28 – 26 27 – ing will be reduced if drying shrinkage 8 – 34 37 – 34 35 of the concrete is reduced. 10 – 39 41 – 38 40 Figure 7. Comparison of curling de- flections for a 30° F gradient. Note that deflections are less for the slab represented on the right, which has the lower modulus of elasticity (data from Ref. 18).
measuring 24 feet between free edges.
Note that the slab is depressed into the subgrade over a distance of about 16 feet with only 4 feet on each end curled above the original finished floor elevation. The vertical uplift of the slab ends ing depends on joint spacing: “ . . . slabs stress nor will it increase the vertical curling above the original elevation is not as great with 30-foot joint spacing can be expect- deflection as is incorrectly stated in ACI’s as the elevation difference between the ed to curl about four times as much as Slabs on Grade (Ref. 17). Instead, specify- top of the slab surface at the free edge and similar slabs with joints 15 feet apart.” ing shorter spacing between shrinkage con- at the center, because the slab has partially This statement implies that slabs curl in trol joints will simply increase the number buried itself into the subgrade. the shape of a circular arc with some of locations where curling can occur. Slab-on-grade thickness is calculated monumental radius and therefore vertical Joint spacing should be either less than by three methods (Ref. 9), assuming that uplift at the slab edge increases with the the critical length to reduce vertical curl- the slab is fully supported by the subgrade square of the distance between control ing deflection and curling stresses, or it in accordance with Westergaard’s as- joints. This is incorrect. Enclosed slabs on could be much greater than the critical sumption (Ref. 19). The three commonly grade do not curl upward in a circular arc, length without further increasing curling used methods are the Portland Cement because gravity prevents it. The vertical deflection or stresses. Association (PCA) method, the Wire Re- displacement at the edges of a 6-inch slab inforcement Institute (WRI) method, and does not increase as slab length increases Warping stress and the Corps of Engineers (COE) method. beyond about 20 to 27 feet. modulus of elasticity Instead of using these commonly accept- As the free ends of an enclosed slab lift Since warping stress is directly related ed methods, the thickness of the outer 5 off the subgrade, the unit gravity load in- to the modulus of elasticity E of the con- feet of slab-on-grade panels should proba- creases on the remaining supporting sub- crete, it is desirable for slabs on grade to bly be based on cantilever design. grade and the center of the slab buries it- be cast from low-modulus concrete (Fig- The fact that the free edges of most self into the subgrade an amount which ure 7). The modulus of elasticity of con- slabs on grade are only partially support- varies inversely with the modulus of sub- crete seems to depend mostly on the ed is not well understood. Increasing the grade reaction, k. The weight of the can- modulus of the aggregates used in the subgrade modulus k also increases the un- tilevered ends prevents them from assum- concrete. However, the specifier must supported length of slab edges because ing the shape of a circular arc. Only a few make a trade-off when calling for lower the slab cannot sink into the subgrade as feet of the slab ends actually are can- E values for aggregate, since these ag- easily. This puts into doubt the standard tilevered above the subgrade. gregates may tend to increase concrete PCA, WRI, and COE slab-on-grade shrinkage (as pointed out earlier). thickness design methods, which general- Critical slab length Figure 8 shows some computed and ly require less slab thickness as the sub- Large stresses develop in concrete slabs observed deflections of a 6-inch-thick slab grade modulus increases. on grade when they warp or curl, because the curling is restrained by the weight of the slab. A study of curling stress data (Ref. 6) shows that vertical displacement and warp- Figure 8. Compar- ing stresses increase as slab length increases ing observed and only up to a certain distance, called the crit- computed deflec- ical slab length. Critical slab lengths (Table tions in a 6-inch slab showed only 4 feet 2) vary from about 15 to 40 feet, depending curled above origi- on temperature/moisture gradients, slab nal floor elevation. thickness, moduli of subgrade reaction Note that the slab (Ref. 18), and concrete elasticity. The dis- has buried itself in placement and stresses do not increase as the subgrade at the the slab length increases beyond the critical center (from Ref. 17). length. Increasing the spacing of shrinkage control joints beyond the critical slab length will not increase the amount of warping Subbase, vapor barrier, slipsheet Table 3. PCA’s Suggested Spacing of Control Joints in A permeable (porous) dry subgrade re- Plain (Unreinforced) Slabs on Grade (From Ref. 1) duces curling. It allows some of the water Slab Spacing, ft., Spacing, ft., Spacing, ft., to leave the bottom of the slab before the thickness, when aggregate when aggregate when slump is concrete takes its final set. This results in in. is less than 3⁄4 in. is larger than 3⁄4 in. less than 4 in. denser, lower-shrinkage concrete at the slab bottom. Serious cracking may occur 5 10 13 15 when concrete is cast on an impervious 6 12 15 18 base (such as polyethylene, clay, or tight- 7 14 18 21 ly compacted soil) because excess water 8 16 20 24 cannot leave the bottom of the slab be- 9 18 23 27 fore final set. Floor slabs on grade should 10 20 25 30 be placed over a 1⁄2-inch layer of sand on Spacing of control joints though there may be hairline-width drying top of a dry subgrade. If the designer shrinkage cracks between the joints. feels he should specify a granular fill, it Upward curling of indoor slabs on should be “choked-off” with 1⁄2 inch of grade occurs at free slab edges. Construc- Subgrade moisture control sand or a smaller crushed stone material tion joints that surround the perimeter of Another reason that shrinkage cracking so as to provide a smooth surface that al- each daily slab-on-grade placement are and curling problems have become more lows the slab to slide as it shrinks or ex- considered to be free edges. By definition, common in recent years is that slabs on pands. The subbase or subgrade should control joints (also known as contraction grade are being constructed on less desir- be designed so it does not need to be, nor joints, usually sawcut) must be free to able, higher-moisture-content subgrades should it be, soaked with water before the open as the slab shrinks. PCA (Ref. 1, as the cost of good industrial land has slab is placed. page 19) states that distributed steel “must risen. A higher moisture content directly If the subgrade can become moist be- be discontinued at all control joints.” under the slab increases slab curling. De- cause of groundwater, ACI 302 (Ref. 2, Therefore, control joints also can become signers seldom engineer vapor barriers to Section 2.4.1) recommends a vapor bar- free edges, just like perimeter construc- offset high subgrade moisture contents. rier covered with 3 inches of sand. How- tion joints, if the slab shrinks so much that Nor is slab thickness often increased to ever, a 3-inch layer of sand can easily be the control joints open. offset these problems because slab thick- displaced during slab placement, thus Notes to Table 3 (from the same PCA ness design methods account for subgrade causing the slab thickness to vary. I rec- reference) state that there will be a marked moisture only indirectly, through the sub- ommend a minimum 50-mil imperme- loss of effectiveness of aggregate interlock grade modulus k. Figure 9 shows that able vapor barrier covered with about 6 at shrinkage control joints in the slab if the edge curling increased about 50% when inches of crushed stone and topped with joints are spaced more than 15 feet apart. the subbase under the slab was saturated a 1⁄2-inch layer of sand. The vapor barrier Table 3 shows that for crack control, joint with water. and crushed stone fill should be placed spacing should be from 12 to 18 feet for a The subgrade should be dry so that it after the roof is tight (for enclosed slabs) 6-inch slab. There is no mention of curl- does not add to the already large moisture or it should be pitched (for outside ing, or the possibility that slab edges at gradient that exists in all slabs on ground. paving) to prevent the fill from becom- joints may rise off the ground. Reducing the slab length between control joints to 15 feet reduces, but does not eliminate, vertical uplift from curling in the typical 6- inch slab on grade. However, it will in- Figure 9. Upward crease the number of curled joints. curling deflection These PCA recommendations that slab of a slab during lengths between control joints should be its initial drying cycle on a dry so short that shrinkage cracking cannot subbase, com- occur, and that distributed steel be cut at pared with deflec- these control joints, are not beneficial. tion of the same They increase the number of curled joints slab on a saturat- in a floor, even though the amount of ver- ed subbase (from tical uplift will be less than if the joint Ref. 17). spacing were slightly greater. I believe the lowest maintenance cost is achieved by using continuous distributed reinforcement to increase joint spacing, even though these fewer joints may curl to a greater vertical deflection and even ing saturated. The advantage of this References For more information thicker stone/sand cover over the vapor barrier is that the vapor barrier will not NOTE: This reference list includes For additional information about be punctured nor will the fill material be some references from Part 1 that also slabs on grade, consult Designing easily displaced as construction equip- are cited in Part 2. Floor Slabs on Grade, by Boyd C. ment is driven over it. 1. Concrete Floors on Ground, 2nd Ringo and Robert B. Anderson. If the designer intends only to reduce ed., Portland Cement Association This comprehensive book provides (PCA), 1983. a single-source answer for selecting friction between the slab and subgrade, a polyethylene slipsheet can be installed di- 2. ACI 302.1R-89, “Guide for Concrete the most cost-effective approach for Floor and Slab Construction,” Ameri- each job in order to achieve superior rectly under the slab without a stone and can Concrete Institute (ACI), 1989. crack control, stability, flatness, and sand cover, but holes should be punched 6. Robert F. Ytterberg, “Shrinkage and overall strength. The book features in the polyethylene (or drilled before the Curling of Slabs on Grade” (published step-by-step procedures, as well as polyethylene is unrolled) to allow water in three parts), Concrete International, to escape at the bottom before the slab charts, tables, and equations. April 1987, pp. 23-31; May 1987, pp. concrete takes its final set. 54-61; and June 1987, pp.72-81. For more information and to order, call 800-323-3550, ext. 6. 7. Bailey Tremper and D. L. Spell- Curing and curling man, “Shrinkage of Concrete—Com- Since curling and drying shrinkage are parison of Laboratory and Field Per- formance,” Highway Research 19. H. M. Westergaard, “Analysis of both a function of excess “workability” Stresses in Concrete Roads Caused water in the concrete, curing methods that Record No. 3, 1963, pp. 30-61. by Variations of Temperature,” Public retain that excess water will only delay 9. ACI 360.1R-92, “Design of Slabs on Roads, May 1927, Vol. 8, No. 3, pp. the shrinkage and curling of enclosed Grade,” ACI, 1992. 54-60. slabs on grade. Chemical means, such as 14. ACI 116R-90, “Cement and Con- high-range water reducers, to reduce ex- crete Terminology,” ACI, 1990. Robert F. Ytterberg is president of cess water do not necessarily reduce 15. Ernest K. Schrader and David R. Kalman Floor Co., Evergreen, Colo., a shrinkage, as explained earlier. Lankard, “Inspection and Analysis of national subcontractor specializing in Curl in Steel Fiber Reinforced Con- construction of exposed concrete floors Remember that all cures (water, plastic crete (SFRC) Airfield Pavements,” film, paper, or curing compounds) have Bekaert Steel Wire Corp., Pittsburgh, for industrial and warehouse use. He is a limited life-spans when exposed to wear. April 13, 1983. member and former chairman of ACI Therefore, curing does not have the same 16. Ralph E. Spears, “The 80 Percent Committee 360, Design of Slabs on beneficial effect as long-term high ambi- Solution to Inadequate Curing Prob- Grade, and a former member of ACI ent relative humidity. Extended curing on- lems,” Concrete International, April Committee 302, Concrete Floor Con- ly delays shrinkage curling. Childs and 1983, Vol. 5. No. 4, pp. 15-18. struction. Kapernick, cited in Ref. 6, found that curl 17. Concrete Craftsman Series: Slabs was reduced by ponding water on curled on Grade, ACI, 1982. Editor’s note slabs, but after the water was removed, 18. G. A. Leonards and M. E. Harr, This 3-part series has been con- the slabs again curled to the same vertical “Analysis of Concrete Slabs on densed and updated from Ref. 6. Part deflection they had before the surface wa- Ground,” Proceedings, American So- 3 will appear in January. ter was applied. ciety of Civil Engineers, New York, June 1959, Vol. 85, SM3, pp. 35-58.