Environ. Sci. Technol. 2002, 36, 1584-1591
Fractional Factorial Design To Investigate the Influence of Heavy Metals and Anions on Acid Neutralization Behavior of Cement-Based Products A . P O L E T T I N I , * ,† R . P O M I , † A N D P. SIRINI‡ Department of Hydraulics, Transportation and Roads, University of Rome “La Sapienza”, Via Eudossiana 18, 00184 Rome, Italy, and Department Civil and Environmental Engineering, University of Florence, Via di S. Marta 3, 50139 Florence, Italy
A major concern of cement-based solidification/stabilization of hazardous wastes is the interaction of waste contaminants on cement properties. Literature contains many examples of studies on the interference of individual contaminants on cement properties. Conversely, little information is available on how the interactions between contaminants affect the properties of cement/waste systems. This paper provides a discussion on the interference mechanisms exerted by seven contaminants, five heavy metals and two anions, on cement hydration. The seven contaminants were selected on the basis of the typical composition of municipal solid waste incineration (MSWI) fly ash. Spiking experiments using pure compounds were performed according to a 27-3 IV fractional factorial design to simulate addition of MSWI fly ash to ordinary Portland cement. The acid neutralization behavior of the laboratory cement-contaminant mixtures was studied to detect the presence of solid phases responsible for the buffering capacity of the solid matrix. The results from the experimental work showed that Zn, Cl-, and SO42- were the major factors influencing, occasionally in combination with other contaminants, strength and acid neutralization capacity of the cementitious products. The release of Cd, Cr, Cu, and Pb in the eluates as a function of pH also suggested possible chemical immobilization mechanisms of such metals within the hardened matrix.
Introduction The use of industrial byproducts and hazardous wastes in cementitious systems for both waste reuse and safe disposal purposes is well documented. Various contaminants present in the waste materials affect the performance of such systems because of either synergistic or antagonistic interactions. Consequently, observed system behavior can potentially differ from that which is expected if one accounts only for the independent effects of the individual contaminants (1). The uncertainty of the final properties of cementitious * Corresponding author: phone/fax: +39.06.44.585.037; e-mail:
[email protected]. † University of Rome “La Sapienza”. ‡ University of Florence. 1584
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002
systems in the presence of contaminants has hindered both hazardous waste solidification/stabilization and recycling of industrial byproducts in cement-based applications because the risk of failure is far greater than the benefits of the application. Of all waste contaminants, inorganic species such as heavy metals and anions are known to interfere with cement hydration. Specifically, such species can affect the setting and hardening of cement, when measured in terms of the kinetics or degree of completion of hydration reactions. Also, it is well documented (2-6) that waste contaminant ions can be incorporated into the cement matrix during the hydration process, through either addition, substitution, precipitation, or sorption mechanisms. For example, soluble Zn, one of the studied contaminants, is widely known (5-10) to greatly delay cement hydration through formation of a layer of amorphous Zn(OH)2 on cement grains (coating), thereby preventing the unhydrated cement grains from reacting further with water. According to Tashiro (10), metals, such as Zn and even Pb and Cu, that have ionic potentials close to that of Ca2+ remarkably hinder hydration due to cement grains coating. Moreover, set retardation can also be related to the formation of double salts with Ca2+ ions released during cement hydration, such as CaZn2(OH)6‚2H2O (11). Cadmium is reported to be efficiently fixed in cementbased waste forms, in that precipitated Cd(OH)2 provides nucleation sites for C-S-H and portlandite (7). Cd is also recognized to affect set, although the modes of interference are related both to its chemical form and to its concentration. CdO is known to act as a set retarder, but other Cd compounds are recognized as retarders, accelerators, or without effect (12). Cr(III) is known to be able to increase the amount of heat liberated during early hydration, promoting the formation of ettringite crystals, probably by reducing the liquid phase supersaturation degree (6). Addition of 1 and 5% mol of Cr2O3 to pure C3A-gypsum pastes resulted in as much C3AH6 formed due to C3A hydration as was measured for the control paste (9). Higher Cr additions (20% mol) produced a retardation in early C3A hydration. However, the presence of hydrated C3AH6 was still well recognizable after 28 days of curing. Various researchers (4, 7, 11, 13) also provided evidence for the ability of C-S-H to incorporate Cr3+ ions through Si substitution in poorly crystalline structures. Calcium aluminate hydrate phases can also be formed where octahedrally coordinated Al is partly replaced by Cr(III). The effect of lead is generally recognized as a set retardation due to precipitation of sparingly soluble compounds onto the surface of silicate phases, thus preventing their access to water (7). However, the retardation effect is concentration-dependent, so that Pb was also shown to be a set accelerator in some cases (14). Little information concerning the interference of Cu with cement hydration is available in the literature. Tashiro et al. (6, 9, 10) showed that when dosing 5% mol Cu(OH)2 to pure C3S mortars, the hindering effect of Cu on strength development was very pronounced up to 28 days, and Ca(OH)2 was not detectable in the hardened mortars. Also, the total pore volume increased. As to the other clinker constituents, addition of 1, 5, and 20% mol of Cu(OH)2 to pure C3A-gypsum pastes strongly retarded early C3A hydration, the extent of retardation increasing with dosage. Concerning the effects of anions on cement setting and hardening, chlorides are variously reported as set retarders or accelerators depending on their concentration (7, 11). At 10.1021/es010002z CCC: $22.00
2002 American Chemical Society Published on Web 02/22/2002
TABLE 1. Alias Relationships for the Fractional Factorial Design I ) ABCE ) BCDF ) ADEF ) ACDG ) BDEG ) ABFG ) CEFG A ) BCE ) DEF ) CDG ) BFG E ) ABC ) ADF ) BDG ) CFG AB ) CE ) FG B ) ACE ) CDF ) DEG ) AFG F ) BCD ) ADE ) ABG ) CEG AC ) BE ) DG C ) ABE ) BDF ) ADG ) EFG G ) ACD ) BDE ) ABF ) CEF AD ) EF ) CG D ) BCF ) AEF ) ACG ) BEG AE ) BC ) DF ABD ) CDE ) ACF ) BEF ) BCG ) AEG ) DFG
high concentrations (3-5 wt %) the presence of chlorides can even result in a flash set (7). However, the influence of Cl- on cement strongly depends on the associated cation, which affects the chemical mobility of the chloride ion. In particular, the diffusion coefficient of Cl- increases (as well as the degree of set acceleration) according to the order NaCl < KCl < LiCl < CaCl2 < MgCl2 (7). Likewise, cations that are capable of forming highly soluble sulfate salts increase the solubility and mobility of the sulfate ion, which inhibits C3A hydration (15). Sulfates are known to strongly affect set, and calcium sulfate is commonly added to clinker to control the hydration rate of C3A by forcing ettringite to be formed instead of hydrated calcium aluminates (16, 17). If ettringite production occurs in the hardened matrix, the expansion caused by the large ettringite crystals can lead to detrimental effects such as cracking and, eventually, matrix disruption (16, 17).
Materials and Methods The experimental program was arranged according to a fractional factorial design (18, 19), the details of which are provided in the Supporting Information. Seven contaminants (the factors of the design), consisting of five heavy metals [Cd, Cr(III), Cu, Pb, and Zn] and two anions (Cl- and SO42-), were selected on the basis of the composition of an electrostatic precipitator ash from an Italian municipal solid waste (MSW) incineration plant. The study was designed to investigate the acid neutralization behavior of various cementitious mixtures, which were prepared by using different combinations of the above contaminants, supported by some results from mechanical characterization. The mixtures were obtained through spiking experiments in which pure compounds containing the contaminants of interest [NaCl, Cd(NO3)2‚4H2O, Cr(NO3)3‚ 9H2O, CuCl2‚2H2O, Pb(NO3)2, K2SO4, and Zn(NO3)2‚6H2O] were added to class 42.5R ordinary Portland cement. The concentration of each contaminant was selected to simulate the addition of 20% fly ash to cement. Distilled water was used as mixing water at a water/solids ratio of 0.4. For the experiments, two concentration levels for each factor were used, one [indicated with a minus (-) sign] corresponding to an absence of the contaminant in the mixture, and the second [indicated with a plus (+) sign] corresponding to the selected concentration. To include all of the combinations of the seven factors at each of two levels, 27 ) 128 different runs for each replicate would be required. Accordingly, to reduce the number of runs and at the same time allow for the interactions between the factors to be estimated, a one-eighth fraction of the full factorial design was selected, resulting in 27-3 ) 16 runs. For the sake of preliminary screening between factors, it was assumed that interactions between three or more factors would be negligible compared to the effects demonstrated by the individual factors (main effects) and by combinations of two (second-order interactions). The fractional factorial design had a resolution of 4, meaning that the main effects and second-order interactions can be separated out. However, such a design resolution did not allow second (and higher)-order interactions to be individually estimated; in other words, the effects were confounded.
AF ) DE ) BG AG ) CD ) BF BD ) CF ) EG
TABLE 2. Mixture Formulations factors mixture 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
A Cl-
B Cd
C Cr
D Cu
E Pb
F SO42-
G Zn
+ + + + + + + +
+ + + + + + + +
+ + + + + + + +
+ + + + + + + +
+ + + + + + + +
+ + + + + + + +
+ + + + + + + +
level - (mg/kg OPC) 0 0 0 0 0 0 0 level + (mg/kg OPC) 18000 1350 1000 1350 600 12000 3200
The design relationship E ) ABC, F ) BCD, and G ) ACD, that is, I ) ABCE ) BCDF ) ACDG, was chosen for the fractional factorial design. Accordingly, the effect of the factor E (Pb) was confounded with the interaction between the factors A, B, and C (Cl-, Cd, and Cr). The confounding patterns for factors F (SO42-) and G (Zn) can similarly be determined. The interactions of any order between the factors were derived by multiplying the defining relationship by each factor or factor combination, considering that any factor multiplied by the identity I gives the factor itselt and any factor multiplied by itself gives I. In Table 1 the complete defining relationship is shown along with the confounding patterns (aliases). The formulations for the 16 batches, prepared according to the above defining relation, are presented in Table 2, along with the values of the selected concentrations. Mixture preparation was made according to a randomized sequence, to prevent the effects of lurking variables from skewing the results. The results for the response variables were managed via the Yates’s algorithm (18), to allow for estimation of the main effects and second-order interactions between the seven studied factors. As fractional factorial designs are applied to reduce the number of experiments, they normally do not include replications. Thus, analysis of variance cannot be performed to apply significance tests on the experimental data. In this case, an alternative approach is generally followed, consisting of inspection of the effects plotted on a normal probability paper. If the measured data occur as the result of random (i.e., roughly normal) variation about a fixed mean, and no real effects are caused by changes in the factor levels, the effects will be expected to be normally distributed (18, 19). Any deviation from the normal distribution will therefore account for other reasons than chance occurrences. Such an approach allows for a visual estimation of the significant effects, thus providing a qualitative evaluation of the significance level. The mixtures were cast in 50 mm cubic molds and stored at ambient temperature and relative humidity >90% for 24 VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1585
h. Afterward, they were unmolded and allowed to cure for 1, 7, 28, 56, and 90 days under the same conditions. The cubes were tested at these curing times for their unconfined compressive strength (UCS). At 56 days the specimens were ground to 3 mequiv/g only. The absence of significant effects suggested that, even though the pHcontrolling phases were shown to have been altered by the presence of contaminants, variations in the amount of the hydration phases were not relevant. This was also confirmed by the fact that, despite the significant set retardation effects and low mechanical strength values at early ages caused by some contaminant combinations, an appreciable gain in the final strength was always noticed. From the plots of 7- and 56-day UCS in Figure 4 it appears that early hardening was hindered by the presence of Zn, Cu, and the combination Cl- + SO42- (confounded with the Cu + Pb and Cd + Zn combinations). Such a behavior was likely due to the occurrence of cement particle coating by metal hydroxides precipitation (5-10). A relevant reduction in the effects on
FIGURE 3. Normal probability plots for pH and amount of acid added for first and second plateaus.
FIGURE 4. Normal probability plots for UCS (curing time ) 7 and 56 days). UCS was revealed at 56 days of curing if compared to 7 days of curing, leading to the conclusion that the metal hydroxide barrier around the cement grains tends to gradually dissolve with time, allowing hydration reactions to proceed normally. Additional indicators on contaminant immobilization within the solid matrix were derived by the analysis of metal concentrations in the eluate. In Figure 5, panels a and b, the eluate concentration as a function of pH is depicted for the mixtures containing Cd and Zn, respectively. When the measured concentrations were below the detection limit, this value was used in the graphs. No statistical analysis could be carried out on the results from the eluate metal concen-
trations, because the eluate pH was different for each mixture tested, depending on its own acid neutralization behavior. For Cd, Cu, and Pb at pH values >∼10 (which is the lower limit of the theoretical stability range for C-S-H) the measured concentrations in the eluate were significantly lower than would have been expected on the basis of the solubility of the corresponding metal hydroxides. As the ANC test was carried out on finely ground samples, any physical immobilization effect due to reduction of the contact area with the leaching medium was negated. The observed leaching behavior provides an indication of chemical immobilization of the subject metals within the chemical VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1589
FIGURE 5. Eluate concentration as a function of pH for Cd and Zn. structure of C-S-H. Furthermore, the low concentrations detected for Cd at high pH values may account for its immobilization within portlandite as well, because Cd is known to provide nucleation sites both for C-S-H and for portlandite (7). Conversely, for Zn the release as a function of pH was closer to the Zn(OH)2 solubility curve. As the solubility of Zn(OH)2 is much lower than Zn(NO3)2, chemical immobilization was attained for this metal by hydroxide precipitation, although no immobilization phenomenon within the hydration products was observed. For Cr, the concentration versus pH curves showed a completely different behavior from that predicted on the basis of Cr(OH)3 solubility, suggesting an oxyanion-like behavior in solution (31). Curve maximums were observed at a pH interval of 8-9 units. Among the subject metals, chromium was by far the most strongly immobilized in the cementitious system for the whole pH range investigated (31, 32), the highest concentrations measured in the eluates being in all cases
