Cast-in-Place Cellulose Fiber-Reinforced Cement Paste, Mortar, and Concrete

INTRODUCTION

Cellulose fibers are inexpensive, easily produced forms of reinforcement for cement-based materials. Much research has inspected the properties of cellulose fiber-reinforced thin-sheet elements. Because of their low cost and demonstrated reinforcing ability, cellulose fibers are currently being investigated as reinforcement for cast-in-place concrete.

In general, many types of fibers can enhance mechanical properties such as tensile strength, flexural strength, and flexural toughness in concrete and cementitious materials (Kim, Saeki, and Horiguchi 1999; Mobasher and Li 1996; Cheng et al. 2000; Shah 1991; Bentur 1989; Banthia et al. 1994; Morgan 1989; Li and Stang 1997). The work performed in this study investigates how cellulose fibers alter mechanical properties of cementitious materials.

Fiber dispersion controls the effectiveness of the fiber reinforcement. If fibers clump, little reinforcement is provided in the material. If fibers are uniformly dispersed, maximum reinforcement is provided (Akkaya 2000; Lawler 2001).

Fiber dispersion also affects rheological properties of fresh cement, mortar, or concrete. The addition of fibers to a cementitious material stiffens the mixture. The yield stress-stress necessary to start material flow-and viscosity-resistance to continued flow-are both increased. The change in workability due to the addition of fibers is of concern because even if a material shows excellent mechanical performance, it will not be used if it cannot be placed in molds. It is hypothesized that well-dispersed fibers will affect rheological parameters differently than poorly dispersed fibers. In addition, it is considered that changes in the rheology of the matrix might alter fiber dispersion. That is, a more (or less) viscous material might prevent fibers from dispersing well-likewise for changes in yield stress.

Fiber dispersion has been analyzed primarily qualitatively. Researchers have inspected the fresh mixture for fiber clumps (Yang 2002; Ogawa et al. 2001; Flink and Stenberg 1989) or otherwise performed burnout tests and measured changes in weight (Stroeven, Shui, and Cheng 2000). While qualitative analysis is an acceptable screening test, it is not sufficiently rigorous to correlate numerical fiber dispersion and material properties such as strength, toughness, and workability. To analyze fiber dispersion quantitatively, certain statistical processes have been adapted to the cementitious systems under study.

Previous researchers have shown that cellulose fibers at volumes ranging from 0.06% to 0.5% significantly reduce restrained drying-shrinkage cracking in concrete (Buch, Rheman, and Hiller 1999; Sarigaphuti, Shah, and Vinson 1993; Soroushian 1997). In this study, restrained ring shrinkage tests are performed on fiber-reinforced mortar to evaluate the potential of the fibers to reduce shrinkage cracking at a range of fiber volumes. Fiber dispersion and shrinkage cracking reduction are compared. In addition, the effect of different matrix yield stresses and viscosities on fiber dispersion and shrinkage crack reduction is inspected.

RESEARCH SIGNIFICANCE

The purpose of this research was: to determine appropriate applications for cellulose fiber-reinforced cementitious materials; to correlate rheological parameters (fresh properties) with fiber dispersion, as well as correlate fiber dispersion with shrinkage crack reduction (hardened properties); and to correlate rheology and shrinkage performance directly.

MATERIALS AND MIXING

Shrinkage and rheology tests

Fibers used in this study were bleached, kraft-processed cellulose fibers, with an average length of 2.5 mm (0.1 in.) and an average diameter of 15 m (0.006 in.). Type I ordinary portland cement, tap water, and 1 mm (0.039 in.) monosized silica sand were used in mortar. High-range water-reducing admixture and methylcellulose (MC) were used at dosages of 0.5% and 0.1%, by weight of cement, respectively. These admixtures and additives were chosen to alter paste rheology. The high-range water-reducing admixture is designed to reduce yield stress, though it also reduces viscosity somewhat. MC is used to increase viscosity. It has little effect on yield stress at low dosages. The volume of fibers was varied in the yield stress study.

Water and high-range water-reducing admixture were premixed by hand. Cement and MC were hand-blended. Unless otherwise specified, wet ingredients were added to dry ingredients and mixed in a small, planetary mixer for 1 min, after which the paddle and sides of the mixing bowl were scraped down. Mixing resumed for 4 min, with another pause for cleaning after 2 min.

Yield stress tests were performed on fiber-reinforced paste with w/c of 0.4 and 0.5. Viscosity tests were performed on unreinforced pasted with a w/c of 0.4. At higher w/c in the viscosity test, some settlement occurred in which the cement grains drifted to the bottom of the testing apparatus. This caused concern that the viscosity of the material was being measured inaccurately, thus in subsequent tests, a lower w/c was used.

Flexural tests

Flexural studies were performed on concrete beams both with and without fibers. Concrete was formed with tap water, the cement described previously, graded river sand, and 9 mm (3/8 in.) pea gravel. Cement, coarse aggregate, fine aggregate, and water were mixed with weight ratios of 1:2:2:0.5. One percent by cement weight of high-range water-reducing admixture was used. Fibers were added at 0.25% volume.

The water and high-range water-reducing admixture were hand-mixed initially and placed in the mixer. The fibers were added and mixed with the water for 5 to 30 s. Coarse aggregates were added and mixed for 1 min. Fine aggregates were added and mixed for 3 min. Cement was added and mixed for 2 min. The concrete was placed in the molds, rodded 25 times, and vibrated for 1 min.

TESTING PROCEDURES

Restrained ring shrinkage tests

To determine whether cellulose fibers reinforce against drying-shrinkage cracking, mortar shrinkage rings were cast at a range of fiber volumes. The restrained ring shrinkage apparatus was used. Figure 1 shows a schematic of this setup. Only paste and mortar were tested for drying-shrinkage cracking. As a result, smaller steel rings than those used for concrete could be used. This facilitated casting and ring movement and also shortened the time of the tests. Smaller rings cracked faster.

A 400 mm (4 in.) tall, 12.5 mm (0.5 in.) thick, 175 mm (7 in.) outer diameter ring of steel was placed around pegs on a plywood board. The pegs were placed to ensure that the ring was centered on the board. A 3 mm (1/8 in.) thick groove was cut into the board, exactly 12.5 mm (0.5 in.) outside of the steel ring. Cardboard ring molds-the type used for casting concrete columns-400 mm (4 in.) tall, 3 mm (1/8 in.) thick were placed into the grooves. Paste or mortar was cast into the gap between the cardboard and steel to form a ring 400 mm (4 in.) tall, 12.5 mm (0.5 in.) thick, 200 mm (8 in.) in the outer diameter. Three rings were cast for each mixture. After casting, rings were covered with plastic wrap and cured in ambient lab conditions. After 2 days, the plastic wrap and cardboard molds were removed and the tops of the paste or mortar rings were sealed with silicone caulking to ensure drying only occurred laterally. Rings cured at 22 C (72 F), 50% relative humidity for the duration of the test. Rings were checked daily until cracking occurred. Crack widths were measured several times at three locations along the crack height during testing. Crack widths reported are the average of the three measurement locations and the three rings for each mixture. In cases where multiple cracking was observed, the largest crack width is reported because that crack controls permeability.

To observe whether alterations in rheological properties would change material mechanical performance, mortar shrinkage rings with different viscosities and yield stresses were cast. Rings with matrixes with low and high viscosities (NM and MC, respectively) were cast with fiber volumes of 0.75%, 0.375%, and 0.14%. The 0.75% fiber volume mixtures were also cast with and without 0.5% high-range water-reducing admixture to observe the effect of yield stress on mechanical performance.

Flexural tests

To inspect the influence of cellulose fibers on concrete flexural properties, three 50 100 450 mm (2 4 18 in.) concrete beams were cast at 0.25% fiber volume and 0% fiber volume (control). Beams were left in molds overnight and then stripped and cured at 100% relative humidity and room temperature for 7 days before testing.

Beams were tested in three-point bending to determine flexural strength and toughness. A notch was saw-cut in the middle of the specimen to approximately 1/3 the height of the specimen. A strain gauge extensometer, with a 12.5 mm (0.5 in.) range, was glued onto the specimen, with the knife edges of the extensometer on either side of the notch. The crack-mouth-opening displacement (CMOD) of the extensometer was used as the feedback signal. Center-point displacement was measured with yoke-mounted linear variable differential transducers (LVDTs).

Rheology

A rheometer was used to investigate the rheological properties of cement paste, both with and without fibers. A concentric cylinder geometry was used to test the viscosity of the unreinforced matrix. A vane fixture was used to test the yield stress of the fiber-reinforced matrix.

Viscosity tests-Figure 2 shows the concentric cylinder geometry used to examine the viscosity of unreinforced paste. The gap between the inner and outer cylinders (bob and cup, respectively) was 0.8 mm, which was too small to test fiber-reinforced paste-the fibers were 2.5 mm long-thus, only the cement paste matrix was tested using this geometry.

Material was placed into the cup until it reached a line scribed inside the cup. This line is scribed so that when the bob is placed in the cup at the testing position, the material is displaced until it just reaches the top of the bob. The cup was clamped into the rheometer and the bob was lowered until the bottom of the bob was 4.2 mm (0.168 in.) from the bottom of the cup, as specified by the rheometer manufacturer.

To measure viscosity, material was placed in the rheometer cup, stirred by hand, and allowed to rest for 30 s. The shear rate was then ramped from 0 to 100 s^sup -1^ over 10 s. The bob rotated at 100 s^sup -1^ until the stress in the material was constant or for 2 min-whichever came first. This preconditioning ensured that each material started with the same shear history. The shear rate was then ramped down from 100 to 0 s^sup -1^ over 30 s. Viscosity was determined as the slope of the shear stress-shear rate curve from 20 to 80 s^sup -1^, on the downslope.

Figure 3 shows shear stress-shear rate curves for several different matrixes. The mixture with a water-binder ratio (w/b) of 0.4 and 0.1% MC has a significantly higher viscosity than the same mixture without MC. Table 1 shows the mixture proportions, viscosities, and yield stresses for the various cement paste matrixes used in the rheology study.

Yield stress tests-Figure 4 shows the vane geometry for the rheometer. The 22 mm (7/8 in.) diameter vane is inserted into a 43.4 mm (1.75 in.) diameter cylinder full of the material to be tested. A large gap exists between the vane and cylinder walls, which enables the testing of fiber-reinforced paste. Material is present between the fins of the vanes. A slow rotation rate is applied. As the vane begins to rotate, the material between the fins stretches elastically. As the vane continues to rotate, the material is stretched more and more until the elastic bonds break and the material begins to flow. The stress at which the bonds break is the yield stress of the material.

The rheometer cup was filled to within 2 to 3 mm (0.08 to 0.12 in.) of the top. The cup was clamped into the rheometer and the vane was lowered into the cylinder until the bottom of the vane was 32 mm (1.28 in.) from the bottom of the cylinder. The large distance between cylinder bottom and vane bottom ensured that no edge effects would interfere with testing (Dzuy and Boger 1985).

To acquire yield stress, the vane was rotated at 0.03 s^sup -1^ for 300 s. The shear stress-time curve was recorded. Figure 5 shows shear stress-time curves for several different matrixes. Note that cement matrixes with 0.4 w/b, with and without 0.1% MC, by weight of cement, have similar yield stresses. As has already been shown, these two matrixes have radically different viscosities. The addition of 0.5% high-range water-reducing admixture, by weight of cement, reduces yield stress almost to zero.

Figure 6 shows the yield stresses of fiber-reinforced paste, with 0.5 w/c, at several different fiber volumes. As expected, the yield stress increases with increasing fiber volume.

Dispersion

To analyze fiber dispersion, images of fiber locations were acquired as discussed in the following section. Fiber dispersion was then analyzed using statistical tools also described in the following.

Locating fibers within cement matrix-Dispersion specimens were cast into 30 mm (1.25 in.) diameter, 25 mm (1 in.) tall specimens. They were covered with plastic and cured in the lab. After 2 days, they were demolded. Specimens were cut with a band saw approximately in half height-wise (the cross section stayed circular) to expose the center section of the material. The exposed surface was ground using an polisher with 180 grit sandpaper. Specimens were washed, dried with compressed air, and then soaked for 1 h in a water solution containing 2% by weight of triethanolamine-a fluorescent whitening agent (FWA) that bonds only to the fibers and causes the fibers to fluoresce under ultraviolet light. The specimens were removed, scrubbed thoroughly, and rinsed under tap water to remove any extraneous FWA debris that might have adhered to the matrix.

After the specimens were prepared, they were placed under a stereo microscope and images were taken at 9.75 under ultraviolet light, which was supplied by a hand-held ultraviolet wand. Light was 254 nm in wavelength at 115 V, 60 Hz, and 0.16 amps. Four pictures were required to fully image each specimen.

Acquired images were converted to grayscale. The fibers appeared as white specks. The matrix appeared as a black background. A commercially available image analysis software package was used to generate a location matrix from the positions of the white specks (fibers) in the grayscale image. In many instances, fibers were clumped, in which case the user manually input additional fibers into the location matrix by selecting positions in the software package with the mouse. In the case of large fiber clumps, a grid of evenly spaced dots was placed over the fiber blob on the screen to simulate densely packed fibers.

Figure 7(a) is an image of a treated specimen under white light. Figure 7(b) is the identical area of the treated specimen under ultraviolet light, with the image processing described previously. The fiber locations are quite clear in Fig. 7(b).

The fiber location file created by the software package was analyzed using programs written in the computer language IDL to determine fiber dispersion, as described in the following.

Statistical analyses-Fibers can be thought to disperse in three fundamental modes-uniform, random, or clumped-or combinations of them, as shown in Fig. 8. Ideally, the fibers in the materials under study would be uniformly distributed, but short of inserting each fiber into the matrix by hand, it is hard to imagine how to obtain this sort of dispersion. Fibers can also clump together. Because fibers cannot be uniformly dispersed, a random distribution is the best that can be obtained. A random distribution ensures that little clumping exists and that specimens are well-reinforced.

Statistical point processes were used to analyze the fiber dispersion in the images acquired. These processes were created by Diggle (1983) and developed for cementitious systems by Akkaya and Lawler. They are described thoroughly in Akkaya (2000) and Lawler (2001).

The K-function indicates whether fibers clump or disperse randomly. Conceptually, the K-function determines how many fibers have nearest neighbors-the fiber closest to the fiber under inspection-within a given radius. If the nearest neighbors of most fibers are quite close, then there is probably a large degree of clumping. If the distance between fibers and their nearest neighbors is greater, then the fibers are probably more randomly distributed.

The K-functions for uniform, random, and clumped point distributions are shown in Fig. 9. It can be seen that the K-function of a uniform distribution falls below that of a random distribution, whereas the K-function of a clumped distribution rises above that of a random distribution. The distance between an experimental K-function and a random K-function is used to determine fiber dispersion and clumping. Higher K-functions indicate more clumping.

RESULTS

Drying shrinkage tests

Mortar shrinkage rings with varying fiber volumes were tested to determine the effect of fiber volume on shrinkage cracking. As seen in Fig. 10, even the lowest fiber volume (0.14%) markedly decreases shrinkage crack width as compared with unreinforced mortar. For fiber volumes of 0.14% and 0.36% (equivalent to 0.1% and 0.25% in concrete), the shrinkage crack width at 14 days decreased by 37% from unreinforced mortar. At higher fiber volumes, multiple smaller cracks were induced, which has positive durability implications due to decreased crack widths. These data show that drying shrinkage crack control is a good application for cellulose fiber-reinforced concrete.

Flexural tests

Figure 11 shows the average stress-CMOD curves for unreinforced and reinforced concrete beams. CMOD values shown herein are almost identical to the center-point displacements of the beam. It can be seen that fibers do not change strength at all and increase toughness late in the post-peak region. Cellulose fibers do not enhance flexural properties.

Shrinkage cracking reinforcement and flexural reinforcement-Comparing Fig. 10 and 11, it can be seen that the influence of fibers is more pronounced in restrained ring shrinkage tests as compared with the flexural tests. There are several reasons for this apparent difference. The flexural tests were carried out only until the crack width opening displacement reached 0.4 mm (0.016 in.). However, crack widths of almost 1 mm (0.04 mm) were observed in the unreinforced shrinkage rings. Thus, it is possible that a more pronounced effect of cellulose fibers might be seen at larger CMOD values. Another possibility is that the flexural response (three-point bend test) is dominated by a single crack opening, while multiple cracks can be observed in some ring tests.

Rheology tests: dispersion and rheology

Yield stress and dispersion-To correlate fiber dispersion and yield stress, two questions must be answered. First: Does fiber dispersion alter material yield stress? That is, does a material with well-dispersed fibers have a lower yield stress than a material with poorly dispersed fibers (or vice versa)? Second: Does matrix yield stress affect fiber dispersion? That is, does a matrix with a higher yield stress prevent fibers from dispersing as well as a matrix with a lower yield stress (or vice versa)? Mixtures used in this phase of testing had 0.5 w/c and were mixed in a small planetary mixer on the lowest speed, as detailed in the following.

To determine if fiber dispersion affected material yield stress, two mixtures were cast. In the first mixture P1, cement and water were mixed for 1 min, the paddle and sides of the mixing bowl were scraped down, and then the paste was mixed for an additional minute. A fiber volume of 0.5% was added and mixed for 1 min; the paddle and sides of the mixing bowl were scraped down; and then the fiber-reinforced paste was mixed for an additional minute. Sufficient material was mixed to perform the vane test and cast three dispersion specimens for the dispersion analysis. The yield stress of the material is shown in Fig. 12(a). The second mixture P3 was mixed in exactly the same way. However, prior to mixing, fibers were soaked in water and then squeezed into tiny clumps (resembling spit-balls). The fiber clumps were dried entirely to prevent the addition of extra water when they were added to the mixture. The purpose of using these clumped fibers was to ensure poor fiber dispersion. As can be seen in Fig. 12(a), the yield stress of this mixture is the same as that of P1. Figure 12(b) shows the K-functions of Mixtures P1 and P3. Recall that higher K-functions indicate more clumping. P3 has a significantly higher K-function than P1, indicating far more clumping, as expected. However, there is no change in yield stress between the mixture with well-dispersed fibers and poorly dispersed fibers. This indicates that, at least at the fiber volumes tested herein, fiber dispersion does not affect material yield stress.

To determine if the matrix yield stress influenced fiber dispersion, an additional mixture, P2, was cast. In this mixture, water and fibers were mixed for 1 min. Cement was then added and mixed for 1 min, after which the paddle and sides of the mixing bowl were scraped down. The fiber-reinforced paste was then mixed for an additional minute. As Fig. 12(a) shows, the yield stress of this material is more than three times greater than that of P1. As Fig. 12(b) shows, the K-functions for P1 and P2 are identical. Even though the yield stresses for Mixtures P1 and P2 are vastly different, their dispersions are the same. This indicates that matrix yield stress does not affect fiber dispersion. In addition, water can be considered to have zero yield stress. Therefore, the matrix into which the fibers in P2 are originally mixed has a yield stress of zero. This initial matrix yield stress for P2 is clearly less than that of P1. Yet the K-functions for P1 and P2 remain the same. When considered this way, the matrix yield stress still does not affect fiber dispersion.

Viscosity and dispersion-To correlate fiber dispersion and viscosity, the same two questions arise that occurred during yield stress testing. Namely, does fiber dispersion affect material viscosity? And, does matrix viscosity affect fiber dispersion? The former cannot be answered in this study because there is no way of testing the viscosity of fiber-reinforced paste with the available rheometer. However, the second question was explored, using a matrix with 0.4 w/b, both with and without 0.1% MC.

Figure 13(a) shows the viscosities for the matrix without any MC, neat paste (NP), and with 0.1% MC. In MC, cement and MC were hand-blended for 30 s. Water was added and the material was mixed for 4 min, with pauses after each minute to scrape down the paddle and mixing bowl. The viscosity of the matrix was tested. After material was removed for viscosity determination (and while the viscosity test was running), fibers were added to the remaining paste and mixed for 2 min, with a pause after the first minute to scrape down the paddle and sides of the bowl. Three dispersion specimens were cast for each mixture. The same mixing procedure was followed for NP except no MC was hand-blended with the cement.

Figure 13(b) shows the K-functions for NP and MC. It can be seen that both mixtures have identical K-functions, even though the viscosity of MC is nearly 2.5 times greater than that of NP. This indicates that matrix viscosity does not affect fiber dispersion, at least at the viscosities and fiber volumes tested, and with the dispersion analysis approach used.

Rheology and drying-shrinkage cracking

It is assumed that mechanical performance is controlled by fiber dispersion. Even though the aforementioned results indicated that matrix rheology did not affect fiber dispersion, it was considered that perhaps the statistical analysis to determine fiber dispersion was not sensitive enough to the system under study. As such, a test series was performed to inspect directly the influence of matrix rheology on mechanical performance. Any changes in performance are still expected to be due to changes in fiber dispersion. It is simply hypothesized that such dispersion changes cannot be seen with the current dispersion analysis method.

Figure 14 shows the drying shrinkage performance of all the mortar mixtures without MC (0% NM, 0.14% NM, 0.5% NM, 0.75% NM, and 0.75% NMSP). The addition of any fibers at all drastically reduces shrinkage crack widths. Fiber volumes of 0.375% and 0.75% provide the same amount of reinforcement. This indicates that additional fibers above a certain volume do not provide extra reinforcement. Because a greater fiber volume stiffens the mixture more, this suggests that an optimal fiber volume can be found that provides maximum reinforcement with minimum stiffening.

Figure 15 shows the drying shrinkage performance of all the mortar mixtures with MC (0% MC, 0.14% MC, 0.5% MC, 0.75% MC, and 0.75% MCSP). For reference, the behavior of mortar without fibers or MC (0% NM) is also shown. Again, fibers at any volume greatly reduce shrinkage crack widths. In addition, fiber volumes of 0.375% and 0.75% provide the same amount of reinforcement, as seen in the NM test series. The mixtures with MC and fibers show the exact same behavior as the reinforced mixtures without MC at all fiber volumes (Fig. 16). This shows that changes in viscosity do not affect drying shrinkage performance and, therefore, presumably, fiber dispersion.

Figure 17 shows the shrinkage crack widths for all the mixtures with 0.75% fibers. The MC and neat mortar mixtures behave identically, both with and without high-range water-reducing admixture. This implies that changes in yield stress do not alter drying shrinkage performance and, therefore, fiber dispersion.

CONCLUSIONS

Experiments were performed to determine how cellulose fibers alter flexural and restrained drying shrinkage performance in concrete and mortar. Cellulose fibers significantly reduce drying shrinkage crack widths. Such pronounced differences were not observed in flexural performance.

In addition, relationships were inspected between rheological parameters and fiber dispersion of fiber-reinforced paste and mortar. Fiber dispersion does not affect material yield stress and matrix rheological properties do not affect fiber dispersion-at least at the levels tested herein with the dispersion and rheological methods used. This may imply that there is no correlation between fiber dispersion and material rheology.

MC was added to mortar to increase viscosity, and high-range water-reducing admixture was used to decrease yield stress. MC mixtures perform the same as mixtures without MC. In mixtures with and without high-range water-reducing admixture, the same drying shrinkage performance is observed. This shows there is no correlation between matrix yield stress or viscosity and drying shrinkage performance. This suggests that cellulose fibers disperse well under normal mixing conditions, regardless of matrix yield stress or viscosity. This is useful for field mixing, as it means that if a contractor wants to decrease yield stress (by adding high-range water-reducing admixture, for example) or change viscosity, it will not affect fiber dispersion. This means that cellulose fiber dispersion is a robust process and is not a concern when changing matrix rheology.

Mixtures with fiber volumes of 0.375% and 0.75% show the same drying shrinkage performance. This indicates that above a certain fiber volume, additional fibers do not increase reinforcement. This suggests that an optimal fiber volume can be found that provides maximum reinforcement with minimum mixture stiffening.

ACKNOWLEDGMENTS

This work was performed at Northwestern University, the headquarters of the Center for Advanced Cement Based Materials. Fibers and significant funding were provided by Weyerhaeuser Co. Additional funding was provided by ACBM.