Unsaturated hydraulic properties of cemented tailings backfill that contains sodium silicate

Engineering Geology 123 (2011) 288–301

Contents lists available at SciVerse ScienceDirect

Engineering Geology journal homepage: www.elsevier.com/locate/enggeo

Unsaturated hydraulic properties of cemented tailings backfill that contains sodium silicate N. Abdul-Hussain, M. Fall ⁎ Department of Civil Engineering, University of Ottawa, Canada

a r t i c l e

i n f o

Article history: Received 7 December 2010 Received in revised form 14 April 2011 Accepted 29 July 2011 Available online 23 August 2011 Keywords: Unsaturated hydraulic conductivity Water retention curve Air-entry value Residual water content Gelfill Cemented paste backfill tailings

a b s t r a c t Recycling the mine waste (tailings) into cemented tailings backfill has economical and environmental advantages for the mining industry. One of the most recent types of cemented tailings backfill is gelfill (GF), a backfill that contains sodium silicate as chemical additive. GF is typically made of tailings, water, binder and chemical additives (sodium silicate gel). It is a promising mine tailings backfill technology. From a design point of view, the environmental performance or durability of GF structures is considered as a key factor. Due to the fact that GF structures are cementitious tailings, their durability and environmental performance depend on their ability to resist the flow of aggressive elements (water and oxygen). Thus, understanding the unsaturated hydraulic properties of GF is essential for a cost-effective, environmentally friendly and durable design of GF structures. However, there is a lack of information with regards to unsaturated hydraulic properties of GF, the factors that affect them and their evolution with time. Hence, the unsaturated hydraulic properties (water retention curve (WRC) or water characteristic curve, air entry value (AEV), residual water content, unsaturated hydraulic conductivity) of GF are investigated in this paper. GF samples of various compositions and cured in room temperature for different times (3, 7, 28, and 90 days) are considered. Saturated hydraulic conductivity and microstructural tests have been conducted; WRCs are measured by using a WP4-T dewpoint potentiameter and the saline solution method. Unsaturated hydraulic conductivity is predicted using the van Genuchten (1980) equation. The water retention curve (WRC) is determined as the relationship between volumetric water content and suction for each GF mix and curing time. The van Genuchten (1980) equation is used to simulate the WRC to best-fit the experimental data. AEV and residual water content are also computed for each mix and curing time. Furthermore, functions are developed to predict the evolution of AEVs, residual water content and fitting parameters of the van Genuchten model with degree of hydration. Important outcomes have been achieved with regards to unsaturated hydraulic properties. The unsaturated hydraulic conductivity of GF was calculated to decrease when the suction, binder content, and degree of hydration increase. The effects of binder content and degree of hydration are more obvious at low suction ranges. The obtained results would contribute to a better design and assessment of the durability and environmental performance of GF structures. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The generation of acid mine drainage (AMD) which results from the oxidation of sulphide-bearing tailings is considered as one of the major environmental problems that is challenging the mining industry. A massive amount of sulphide-bearing tailings are generated by mine activities around the world each day. These tailings should be perfectly controlled to avoid destructive impacts on the environment. Therefore, ways to effectively and economically limit the environmental impacts caused by AMD have always been a main issue for all mining operations (Yilmaz, 2007; Orejarena and Fall, 2008; Fisseha et al., 2010). It is worth mentioning that Georgius Agricola wrote in his 16th century classic, De Re Metallica, that ‘when the ores are washed, the water ⁎ Corresponding author at: Department of Civil Engineering, University of Ottawa, 161 Colonel by, Ottawa (Ontario), Canada K1N 6N5. Tel.: + 1 613 562 5800#6558. E-mail address: [email protected] (M. Fall). 0013-7952/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.enggeo.2011.07.011

that has been used poisons the brooks and streams and either destroys the fish or drives them away’(Davis and Ritchie, 1986). Several techniques have been developed in recent years to mitigate the problem associated with AMD. One of the techniques that have become increasingly popular in underground mining operations around the world is recycling the tailings in the form of cemented paste backfill (CPBs). CPBs are heterogeneous material produced by mixing tailings (70–85 wt.%) with fresh or mine processed water, and hydraulic binder (3–7 wt.%). The cost of the binder consumption is high. It can represent up to 75% of the cost of cemented backfill (Fall et al., 2008, 2009; Nasir and Fall, 2009). This factor has compelled mining companies to seek for alternatives that can increase the strength of the fill and reduce the binder content. Sodium silicate is the most recent chemical additive that is used to reduce the binder content in CPBs. This new product is named gelfill (GF). GF is a new cemented tailings backfill material. It is typically made of tailings, water, binder and chemical additives (sodium silicate gel).

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

According to previous studies on GF (Doucet et al., 2007; Hassani et al., 2007), the chemical additive, sodium silicate, (like a gel, therefore the name GF) has the ability to absorb a large amount of water (cemented tailings are always prepared with excessive water to allow easy transport to underground voids), to accelerate binder hydration and promote the formation of additional binder hydration products. This contributes to a substantial increase in GF strength. Mechanical, environmental performance and durability are the most important design criteria of cemented tailings backfill structures (Grabinsky et al., 2008; Nasir and Fall, 2008; Fall et al., 2009). Permeability is the key parameter that most affects the environmental performance and durability of cemented backfill (e.g., GF). Vulnerability of GF to AMD is dependent on the reactivity of tailings contained in the GF. The reactivity in turn is reliant on the types and quantity of sulphide minerals present in the GF as well as hydraulic properties of the GF (seepage of water and diffusion of oxygen through the GF matrix). These properties can be evaluated by investigating the hydraulic conductivity of GF, in particular, the unsaturated hydraulic conductivity. In addition, the hydraulic conductivity controls the leaching and transport of pollutants through the GF to the ground water (Fall and Samb, 2008; 2009). The hydraulic conductivity can also give information about the GF pore structure, such as coarseness, connectivity, and cracking. Mechanical properties (shear strength and uniaxial compressive strength (UCS)) are affected by the unsaturated conditions in GF structures. Increasing the suction in GF structures leads to an increase in the strength. Self-desiccation, which is mainly responsible for the development of suction at early ages of the GF, and occurs in all types of cementitious materials due to chemical shrinkage, has an influence on the properties of fresh GF as well as the long-term behaviour of the GF. During self-desiccation, the largest pores empty first. The menisci formed in these partially empty pores create capillary tension within the pore solution and reduce the internal relative humidity (RH) of GF structures (Bentz, 2008). As a result, the compression strength of the GF structure increases, which influences the stability and durability of the GF. Decreasing the moisture content of GF structures due to higher suction (low RH) can lead to more economical GF designs. A good understanding of the unsaturated properties of GF is needed to evaluate the mechanical properties of GF in unsaturated conditions. As GF is a new cemented material, only a few studies have discussed its mechanical properties. However, no studies have addressed the relevance of unsaturated hydraulic properties. There is a lack of knowledge on the unsaturated hydraulic properties of GF for different curing times and binder contents. The determination of the unsaturated properties of GF is being a challenge for geotechnical engineers and engineering geologists. This has inspired the author to conduct the current study. The objectives of this study are: – to determine the water retention curve (WRC) for the GF and its evolution with time as well as to simulate the WRC to best-fit the experimental data; and – to predict the unsaturated hydraulic conductivity of GF sample by using saturated hydraulic conductivity and WRC; – to better understand the unsaturated hydraulic properties of GF. The results presented in this paper will contribute to the design of safer, more economically and environmentally friendly GF structures. 2. Experimental programs

289

backfill. Table 1 shows the physical and chemical properties of the binder. Tap water was used to mix binders and tailings. 2.1.2. Tailings Silica tailings (ground silica, TS) are used in this study. TS can be classified as medium tailings with 41–45 wt.% fine particles (b20 μm). The use of silica tailings allows accurate control of the mineralogical and chemical compositions of the tailings. This maintains the level of uncertainties at a minimum level. Indeed, natural tailings can contain several reactive chemical elements, and often, sulphide minerals (which oxide and produce sulphate during contact with oxygen) that can interact with sodium silicate and/or the cement and thus affect the interpretation of the results as well as study outcomes. TS contain 99.8% SiO2 and show a grain size distribution close to the average of nine Canadian mine tailings (Fig. 1). Tables 2 and 3 show the physical and chemical properties of TS. TS are non-plastic and classified as sandy silts of low plasticity (Celestin, 2008; Fall et al., 2010) and the ML category in the UCCS. ML is typical for tailings from hard rock mines as also determined by Vick (1990). 2.1.3. Sodium silicate Soluble sodium silicates are water soluble glasses generally manufactured from varied proportions of an alkali metal and SiO2. Soluble sodium silicates are silicate polymers. It is manufactured by smelting sand with sodium carbonate at 1100 °C to 1200 °C. It is an inorganic chemical, a clear, colourless, and viscous liquid. Aside from being used as an admixture for cement, soluble sodium silicate is also used for a number of applications in various industries or fields, such as the paper industry (e.g., for binding packaging), geotechnical engineering (e.g., soil grouting, mine tailings backfill), soap and detergent manufacturing, waterproofing, textile processing, and foundries. In the present study, soluble sodium silicate is used as an admixture for the cement or binder for the preparation of cemented tailings backfill. A commercial solution of sodium silicate (type N) was used, in which the ratio of SiO2 to Na2O is 3:2. Table 4 shows the sodium silicate properties. Sodium silicate was added to the mix in a liquid form. 2.2. Preparation of materials and mix proportions Around fifty GF samples were prepared to study the hydraulic properties. Tailings, binder, water, and sodium silicate were mixed using a B20F food mixer. The mixing time was 10 min for all mixes. Tailings, binder, and water were mixed first, followed by the addition of sodium silicate. Various binder proportions were used to prepare the mixes; 2%, 4.5%, and 6% PCI by weight with a solid mass concentration of 73% which correspond to water–cement (w/c) ratios of 18.5, 8.2, and 6.2, respectively. Then, 0.4% sodium silicate by weight of solids was added to the mixes. All mixes have the same slump. The produced GF was poured into curing cylinders which are 5 cm in diameter and 10 cm in height. The prepared samples were then sealed and cured at 20 °C ± 2 °C for periods of 3, 7, 28, and 90 days. 2.3. Testing and analysis methods 2.3.1. Water retention curve (WRC) A WP4-T Dewpoint PotentiaMeter and the saline solution method were used to measure the WRC of the GF samples. The WP4-T was

2.1. Materials The materials used to prepare GF include binder, tailings, sodium silicate, and water. 2.1.1. Binder and mixing water Portland cement type I (PCI) was used as the binder. It is the most common material used by the mining industry to produce cemented

Table 1 Characteristics of Portland cement type I. Types of binder

MgO

CaO

SiO2

Al2O3

Fe2O3

SO3

Relative density

Specific surface (m2/g)

PCI

2.65

62.82

18.03

4.53

2.70

3.82

3.10

1.30

290

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

Volume percent (cumulative)

100 90 80

Table 3 Main chemical elements of TS.

TS 9 mines

Element Al wt.% Ca wt.% Si wt.% Fe wt.% Na wt.% Pb wt.% S wt.% K wt.%

70

TS

0.10

b0.01

99.80

b 0.01

b0.01

0.00

0.00

0.00

60 50

trimmed until they were 1 cm in thickness. The volume of the desiccators was high enough compared to the volume and the number of samples. This meant that the specimen load did not disturb the ambience controlled by the saline solution and significantly increase the time needed to reach equilibrium. Two samples were used from each mix; and the curing time and average value were taken. The weight of the samples was monitored up to the equilibrium state. At the end of the testing, the samples were placed in an oven to determine the dry weight. The water content was also calculated for each sample.

40 30 20 10 0 0.01

0.1

1

10

100

Grain size (µm) Fig. 1. Grain size distribution of the tailings used and the average grain size distribution of tailings from nine Canadian mines.

used to measure the WRC of samples for periods of 3, 7, and 28 days whilst the saline solution method was used to determine the WRC of samples cured for 90 days. Both methods were used to measure the total suction of the samples. The reason that the WP4-T test was adopted for early age samples is to obtain all results on the same testing day because the microstructure of GF changes as hydration proceeds. 2.3.1.1. WRC determination by WP4-T Dewpoint PotentiaMeter. The WP4T is considered to be the fastest instrument for measuring water potential within five minutes. It measures water potential from 0 to 60 MPa with an accuracy of ±0.1 MPa from 0 to 10 MPa and ±1% from 10 to 60 MPa (WP4 and WP4-T manual, 2003). The method is used to only measure the total suction. Calibration was done on the WP4-T according to the procedures provided by the manufacturer. The WP4-T was placed in a temperature and humidity controlled room. It was ensured that the bottom of the PVC cup was fully covered with GF and the cup was half empty as suggested by the WP4-T manual (2003). Two samples were cut, trimmed, and levelled by using a knife and then sealed in the PVC cups for at least 16 h before measuring the suction. After each suction measurement, the specimen was taken out of the WP4 chamber, and then the specimen was weighed and left to air dry for about 10– 15 min before starting on the next suction measurement. The process was repeated several times until the total suction was close to 20 MPa. When the test ended, the specimens were placed in an oven to determine the dry weight. With the dry and saturated weights of the specimens, the gravimetric water content (w) was calculated for each suction measurement. 2.3.1.2. Saline solution method. After 90 days of curing, the saline solution method was employed. The hydration reactions were almost over and no further significant changes would occur in the microstructure of the GF. As well, higher suction was needed. The saline solution method measures high suction ranges and requires 1–2 months for samples to reach equilibrium. Seven salt solutions are selected for this study (Table 5). Saturated solutions were prepared and poured into desiccators. The covers of the desiccators were sealed to prevent any changes in the equilibrium condition. RH and temperature were monitored using a Traceable® digital hygrometer thermometer manufactured by the Control Company. When the RH hit equilibrium, the samples were placed inside the desiccators. The samples were cut and Table 2 Physical properties of silica tailings (TS). Element

Gs

D10 (μm)

D30 (μm)

D50 (μm)

D60 (μm)

D90 (μm)

Cu

Cc

TS

2.7

1.9

9.0

22.5

31.5

88.9

16.2

1.3

2.3.2. Unsaturated and saturated hydraulic conductivity Several experimental methods are available to measure the unsaturated hydraulic conductivity (also termed as unsaturated water coefficient of permeability in ASTM standard D5084) of porous media in the laboratory or in situ; namely, the steady-state methods, the instantaneous profile methods, and the parameter estimation methods. Li et al. (2009) provided a comprehensive review of these methods. Steady state methods, whether using short columns, long columns or steady state infiltration, are based on a time-invariant, one dimensional flow determined for various known water contents or suctions (Plagge et al., 1989). Apart from problems related to discontinuities between the sample and the ceramic disk as well as the suction and hydraulic conductivity limitation induced by the ceramic disk, a major disadvantage of these methods is the large amount of time needed to perform measurements, as the time necessary to achieve steady state conditions can be considerable (Plagge et al., 1989; Li et al., 2009). Consequently, these methods are not appropriate to measure the unsaturated hydraulic conductivity of hydrating GF, since the pore structure of the GF quickly changes with time due to the cement hydration. The instantaneous profile method is an unsteady state method and can be used either in the laboratory or in situ. The method uses a cylindrical specimen of porous media that is subjected to a continuous water flow from one end of the specimen (Li et al., 2009). As there is no need to attain steady state conditions, the instantaneous profile method is faster than steady state methods. However, this method can be very time-consuming for low-permeability porous media (e.g., clay, hardened cemented materials) and high suction ranges (N100 kPa). Cui et al. (2008) reported a testing time of 2200 h (approx. 92 days) to complete a test using this method for measuring unsaturated hydraulic conductivity of sand/bentonite mixtures over a wide suction range, especially at high suctions (N100 kPa). Thus, since the GF is a lowpermeability porous medium and can experience high suction during its lifetime, a significant amount of time will be necessary to measure the unsaturated hydraulic conductivity of GF. The rapid evolution of the cement hydration within the GF porous medium makes the instantaneous profile method unsuitable for measuring the unsaturated conductivity of hydrating GF. The parameter estimation method is an indirect approach to estimate the unsaturated hydraulic conductivity of porous medium by using the results from transient or steady-state flow data (Li et al., Table 4 Sodium silicate properties (National Silicates Ltd.). Properties

Values

Na2O% by weight SiO2% by weight Weight ratio, %SiO2/%Na2O Specific gravity @ 20 °C Solids %

8.90 28.66 3.22 1.39 37.56

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

291

Table 5 Salt solution properties. Saturated salt solution

T (°C)

RH% at 25 °C

Total suction (kPa)

Solubility in water (g/ml)

Supersaturated solution (g/ml)

BaCl2·2H2O barium chloride dihydrate K2SO4 potassium sulphate NaCl sodium chloride Mg(NO3)2 6H2O magnesium nitrate hexahydrate MgCl2·6H2O magnesium chloride hexahydrate KCl potassium chloride LiCl lithium chloride

5–60 15–60 5–60 0–50 10–50 5–40 20–70

85.7 95.0 70.8 51.0 32.7 82.9 11.7

22,603 6987 46,940 91,937 152,570 25,545 291,564

37.5 g/100 ml 11.1 g/100 ml 35.9 g/100 ml 125 g/100 ml 54.2 g/100 ml 34.0 g/100 ml 83 g/100 ml

47.5 g/100 ml 21.1 g/100 ml 45.9 g/100 ml 135 g/100 ml 64.2 g/100 ml 44.0 g/100 ml 93.0 g/100 ml

2.3.3. Determination of the binder hydration degree The degree of hydration (α) was approximated from the results of UCS tests on the GF samples with different binder contents and curing times up to 120 days. Fig. 2 presents the results of the calculations. The degree of hydration is a function of time, which ranges between 0% at the beginning of hydration, and 100% when the hydration was entirely accomplished. It is estimated using the following equation: UCSn αn ¼ UCSfinal

ð1Þ

UCSn uniaxial compression strength at n curing time UCSfinal uniaxial compression strength of mature GF (in this study, it is considered that the strength of GF does not increase after 120 days) αn α at n curing time.

It should be emphasised that this method of estimation of the degree of hydration is affected by the curing temperature and selfdesiccation mechanism, i.e. the relationship between the compressive strength and α is influenced by temperatures and the development of negative pore water pressure due to self-desiccation. Nonetheless, this limitation is not applicable here because same curing temperatures are used for all mixes, the GFs are prepared with high w/c ratios and the size of the GF samples is small (10 × 5 cm). 2.3.4. Microstructural analysis The microstructure of the studied GF samples was investigated by thermal analyses (Differential thermogravimetry (DTG), Differential thermal analysis (DTA), and Differential Scanning Calorimetry (DSC)). Tests were carried out on various prepared hardened cement pastes of GF with a high w/c ratio (w/c = 2; to simulate the cement matrix of GF). The thermal analyses were undertaken using an SDT apparatus from TA Instruments which allows simultaneous registration of weight loss and heat flow along the thermal treatment of the sample. The various (dried) samples (about 30 mg each) were heated in an inert nitrogen atmosphere at the rate of 10 °C per minute up to a temperature of 1000 °C. Moreover, the total (overall) porosity (n) and void ratio (e) of the GF samples were also determined. 3. Results and discussion 3.1. Water retention curves 3.1.1. Time-dependent evolution of WRC A fundamental property of an unsaturated porous medium related to its ability to attract water at various water contents and suction is characterised by the water retention curve (WRC). This curve is also called soil water characteristic curve (if the porous medium is a soil), capillary pressure saturation curve, moisture retention curve, moisture characteristic curve, desorption and/or sorption isotherm curve, and the suction volumetric water content curve. The evolution of WRC with

120

Degree of hydration (%)

2009). However, it should be emphasised that this method cannot directly measure the unsaturated hydraulic conductivity of porous medium and there is limited understanding of the reliability of the computed unsaturated hydraulic conductivity curve to a wider suction range (Li et al., 2009). Furthermore, due to the time associated with the determination of the unsaturated hydraulic conductivity using this method, the unsaturated water coefficient of permeability of hydrating cementitious materials cannot be correctly estimated using the parameter estimation method. It can be concluded from the analysis presented above that none of the established direct methods of measuring unsaturated hydraulic conductivity of soils, such as the steady-state methods and the instantaneous profile methods are applicable to hydrating GF. Therefore, indirect techniques were used in this study to estimate or predict the unsaturated hydraulic conductivity of the GF from early to advanced ages. It is common practice to empirically predict unsaturated hydraulic conductivity by using the saturated hydraulic conductivity coefficient and WRC (Fredlund and Xing, 1994). The van Genuchten (1980) equation was used to calculate the unsaturated hydraulic conductivity of GF. Due to the presence of cement, the microstructure of GF changes with time and variation of the binder content. In order to make this model applicable for GF at any curing time, each mix was considered as a different kind of porous media and the fitting parameters were individually calculated. The flexible wall technique was used to determine the saturated hydraulic conductivity of the GF samples with different binder contents (2%, 4.5%, 6%) and cured for 7, 28, and 90 days. The procedure for this method is described in ASTM D5084-00 and was conducted in the constant head mode. Each hydraulic conductivity test was repeated at least twice and the average value was considered as the saturated hydraulic conductivity of the sample tested. It should be emphasised that the results of the saturated hydraulic tests are only needed to represent the curve of unsaturated hydraulic conductivity vs. suction for the samples studied. This paper focuses on the unsaturated hydraulic properties of GF.

(26 °C) (20 °C) (25 °C) at 25 °C at 20 °C (20 °C) at 20 °C

100 80 60 40 6% PCI

20

4.5% PCI 2% PCI

0

0

20

40

60

80

100

120

Curing time (days) Fig. 2. Evolution of the degree of binder hydration with time.

140

292

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

degree of hydration for different binder contents (6%, 4.5%, 2% PCI) is illustrated in Fig. 3. For the same binder content, it can be seen that despite the differences between the obtained curves, the overall trend of the WRCs is similar. In general, samples cured for 90 days have higher suction than samples cured for 3, 7, and 28 days. For example, for 15% volumetric water content (θw), the suction for the 4.5% PCI-GF cured at 3 days (α3 = 37%), 7 days (α7 = 44%), 28 days (α28 = 67%), and 90 days (α90 = 88%) is 1000, 1800, 3300, and 7000 kPa, respectively (Fig. 3b). It can be seen that the effects of curing times or degree of hydration on WRC are obvious at both low and high suction ranges (Fig. 3). These results indicate that GF will display less moisture loss associated with external drying as the curing time increases. This will influence the durability of these materials. The observed time-dependent evolution of the WRC of GF cannot be explained by the mechanisms that control the water characteristic

curve of conventional soil, but rather by those that control the desorption isotherm of hardening cement paste (hcp). There are three main states of water in hcp that need to be recalled in order to better understand the evolution of WRC with degree of hydration: (i) first, capillary water which contains free water when the capillary pores are greater than 50 nm and water held by capillary tension when the capillary pores range between 5 and 50 nm. The capillary pores are emptied when the ambient RH falls below about 45% (Neville, 1995); (ii) secondly, gel water which is divided into adsorbed water held by the surface force of the gel particles (removing this water causes shrinkage when the RH is less than 30%) and interlayer water held between the surfaces of C–S–H planes which can be lost only upon strong drying (11% to 0% RH) (Beaudoin, 1982) and leads to a considerable shrinkage of the C–S–H structure; and (iii) third, chemical combined water, which is non-evaporable water, and can be only lost when hydration products decompose. This observed increase in suction of GF as the curing time increases is explained by the fact that as hydration proceeds, the capillary pores decrease due to the increase in hydration products as demonstrated by Fig. 4 (increasing the solids content of GF matrix leads to blockage of the capillary pores). Decreasing the porosity of the capillary pores has, as a consequence, the reduction of connectivity of these pores and thus moisture loss processes will be more controlled by the gel pores. As a result, higher suction is needed in order to remove the water from the fine pores (gel pores) due to the fact that water in these pores is held by surface force (adsorptive force). Fig. 4a and b illustrates the results of a thermal analysis (TGA/DTA) of cemented paste of GF cured at 7 days and 28 days, respectively. On the whole, typical endothermic peaks can be observed in the range of 50–150 °C, 450 °C, and 750 °C. These peaks at different curing times indicate the presence of CSH gel, ettringite, CH, and CaCO3. A comparison of the TG/DTA diagrams of the cement pastes cured at 7 days and 28 days clearly shows that the amounts of CSH, ettringite, CH, and CaCO3 that are formed in the 28 day samples are higher those formed in the 7 day sample. Indeed, it can be seen from Fig. 4 that the mass loss which occurs in the temperature ranges of 50–150 °C, 450 °C, and 750 °C is higher with increasing curing time. 3.1.2. Effect of binder contents on the WRC Fig. 5 shows the effect of various binder contents (6%, 4.5%, 2% PCI) on the WRC for GF samples of different ages. It can be observed that the binder proportion affects the WRC. The results show that for the same volumetric water content, a significant change in suction occurs for samples prepared with different binder contents and cured for 3, 7, 28, 90 days. However, the effect of binder on the WRCs becomes less noticeable or negligible for high suction ranges or low RH ranges. For example, when the suction is 25,000 kPa (RH = 83%, calculated by using the Kelvin equation, see Eq. (2)), the volumetric water content (θw) of GF samples cured for 90 days with 6%, 4.5%, and 2% PCI are 6.8%, 6.6%, 3.3%, respectively (Fig. 4d); the θw is 3.9%, 5.4%, and 2.7% when the suction is 46,940 kPa (RH = 70%); and when the suction value is 150,000 kPa (RH = 32%), the θw values are 1.6%, 1.5%, and 1.1%. ψ¼−

RT lnðRHÞ vwo wv

where

Fig. 3. The evolution of WRC with degree of hydration for GF samples with (a) 6% PCI, (b) 4.5% PCI, and (c) 2% PCI.

ψ R T t° vwo wv RH

total suction (kPa) universal (molar) gas constant [8.31432 J/mol K] absolute temperature [273.16 + t° (K)] temperature (°C) specific volume of water [(1/ρw) (m 3/kg)] molecular mass of water vapour (18.016 kg/ml) relative humidity.

ð2Þ

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

293

Fig. 4. Thermal behaviour (TGA/DTA) of the hardened Portland cement pastes of GF with w/c = 2 cured at: a) 7 days and b) 28 days.

This observed reduced or negligible impact of binder content on the WRC at high suction ranges (i.e. low RH) is consistent with the experimental observations made by Baroghel-Bouny et al. (1999) on hcp, who found that with a given cement below a given RH (i.e. above a given suction), the WRCs of hcp are identical or close to each other regardless of the mix (i.e. the mix parameters such as cement content, w/c has no significant influence on the WRC anymore). The results from Baroghel-Bouny showed that for RH b 44%, the isotherms of all the hcp remain close, which illustrate that this range of RH (or suction) moisture equilibrium takes place in a pore structure (internal to C–S–H gel) that is not influenced by the mix parameters (Baroghel-Bouny and Chaussadent, 1995; Baroghel-Bouny, 1998). The author demonstrated that the mix parameters have important effects on the WRC only when RHN 76% (low range suction) whilst they have only a mitigated influence when 50% ≤ RH≥ 76%. The observed decrease in the volumetric water content of the GF samples at low suction ranges as the cement content decreases can be attributed to the fact that increases in the binder content will increase the rate of binder hydration (Fig. 2) and the quantity of binder hydration

products formed, such as C–S–H and CH. Formation of more hydration products is associated with the refinement of the pore structures of the GF (Figs. 6 and 7). Consequently, GF samples with lower cement content will have larger and high proportions of capillary pores than the samples with higher cement content. Lower suction is needed to remove the water from the capillary pores. This refinement of the GF pore structure as the binder content increases is graphically demonstrated by Figs. 6 and 7. The variations of void ratio and porosity with respect to curing time for different binder proportions of GF samples are shown in Fig. 6a and b, respectively. It can be noted that increases in the binder content significantly decreases the void ratio and the total porosity of the GF samples (up to 28 days). Furthermore, Fig. 7, which illustrates the effect of binder content on the MIP pore size distribution of GF, shows that increases in the binder content are associated with higher proportions of macroporosity (N1 μm). 3.1.3. Air-entry value and residual water content The key features of WRC are the AEV and the residual water content (RWC). AEV and residual state conditions are essential parameters to

294

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

a

a

1.04 6% PCI

45

1.02

40 35

Void ratio

Volumetric water content %

50

30 25 20

4.5% PCI 2% PCI

1 0.98 0.96

15 0.94

10 5 0

6% PCI

1

4.5% PCI

10

0.92

2% PCI

100

1000

10000

100000

0

20

Suction (kPa)

60

80

100

b

b 55 50 45 40 35 30 25 20 15 10 5 0

0.51 6% PCI 4.5% PCI 2% PCI

0.5

Porosity

Volumetric water content %

40

Curing time (days)

0.49

0.48 6% PCI

1

4.5% PCI

10

2% PCI

100

0 1000

10000

20

40

60

80

100

Curing time (days)

100000

Suction (kPa)

Volumetric water content %

c

Fig. 6. Evolution of the a) void ratio and b) porosity of GF with curing time for various binder contents.

50 45 40 35 30 25 20 15 10 5 0

6% PCI

1

4.5% PCI

10

2% PCI

100

1000

10000

100000

Suction (kPa)

Volumetric water content %

d

50 45 40 35 30 25 20 15

geotechnical and geo-environmental engineers. The AEV represents the suction that is required to cause desaturation of the largest pores. It is important to define the AEV because it indicates the initiation of the desaturation state. It is also necessary to define the residual state condition in order to obtain the fitting parameters that are used in numerical models to predict unsaturated hydraulic conductivity. The AEV of GF is obtained by extending the maximum slope portion of WRC to intersect with the line extending from saturated volumetric content (θs). The corresponding value of suction is taken as the AEV. The residual water content is considered as the degree of saturation at which the liquid phase becomes discontinuous. The residual water content can be defined as the point where a line that extends along the curve from 1,000,000 kPa intersects with the tangent to the maximum slope portion (Barbour, 1998; Lu and William, 2004). The evolution of AEVs with curing time and degree of hydration for different binder proportions is plotted in Fig. 8. It can be noted that the AEVs increase with curing time and degree of hydration. This can be explained by the refinement of the pore structure during hydration process (Fig. 6). As hydration proceeds, the solid content of the paste increases and blocks the capillary pores. Therefore, these capillaries turn into capillary pores interconnected by gel pores (Neville, 1995). The relationship between suction and pore radius is inversely proportional according to the Laplace law (Eq. (3)) (Mokarem, 2002).

10 5 0

6% PCI

1

10

4.5% PCI

100

pv −pc ¼

2% PCI

1000

10000

100000

2σ cosθ r

ð3Þ

1000000

Suction (kPa) Fig. 5. Effect of various binder contents (6%, 4.5%, 2% PCI) on WRC for curing time of: (a) 3 days, (b) 7 days, (c) 28 days, and (d) 90 days.

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

Hg intrusion porosity (%)

25

(Eq. (4)). Non-linear regression (Engel, 1982; Snedecor and Cochran, 1986) techniques (with Excel built-in functions) were used to correlate the AEV (dependent variable) with the binder hydration degree (independent variable). The quality of the fit to the experimental data was assessed based on the the coefficient of determination (R2). The exponential function (Eq. (4)) was found to best fit the experimental data.

7%wt

20

9%wt

15 10

bα

AEV ¼ ae

5 0 <1

1 - 10

> 10

Pore size range (µm) Fig. 7. Effect of binder content on the pore size distribution of GF vs. content of fines and pore size range. Data from Hassani et al. (2007).

where σ is the surface tension of water/water vapour interface, θ is the moistening angle, pc is the pressure in water, pv is the pressure in water vapour, and r is the radius of pores where there is a meniscus. In unsaturated conditions, there is an access radius. All capillaries with a radius less than the access radius are filled with water whilst the capillaries with a radius larger than the access radius are empty (Hau et al., 1995). Hence, as the pore size becomes smaller, higher suction is needed to be applied on the GF samples. Furthermore, it can be seen that after 90 days, the binder content does not have any significant influence on the AEVs. The evolution of AEVs with degree of hydration for different binder contents was formulated based on regression analysis on data sets

a 600

AEV (kPa)

500 400 300 200 6% PCI

RWC ¼ ce

2% PCI

0

20

40

60

80

100

Curing time (days)

b

ð4Þ

where AEV is the air-entry values in kPa, α is degree of hydration (%), and a and b are the adimensional parameters to be determined for each GF mix. The values of a and b for this study are: 6% PCI, a = 61.05, b = 0.0244; 4.5% PCI, a = 42.89, b = 0.028; and 2% PCI, a = 23.64, b = 0.034. To verify the applicability of the model, the predicted function was validated by using the experimental data (not used to develop the function) obtained from the GF samples in this study. In addition, for better validation of the developed model, AEV values obtained by another researcher (Godbout, 2005) on cemented tailings backfill (CPB) samples with various mix components (binder contents and types) and cured at various times (3–28 days) were used. The outcomes of the validation are presented in Fig. 9. It can be noticed that there is a good agreement between the predicted and experimental values. Therefore, this model can be useful for estimating the AEV of cemented backfill tailings. The effect of binder proportions, curing times, and degree of hydration on residual water content is shown in Fig. 10. The effect of binder contents on residual water content is limited to early ages only. It seems that binder content and degree of hydration have only slight impacts on the residual water content. The range of variations of residual water content is 8–11% (which corresponds to high suctions of 30– 9 MPa) for all GF mixes. This observation is again consistent with results of Baroghel-Bouny et al. (1999) which indicate that below a given RH (44%), the mix parameters have no significant effect on the isotherm of hcp as explained above. Using the experimental data of the present study, the evolution of residual water content (dependent variable) with degree of hydration (independent variable) for different binder contents was formulated based on a regression analysis on data sets. This was done using the Excel built-in non-linear regression functions (Engel, 1982; Snedecor and Cochran, 1986). The regression equation that gave the best fit of the experimental data is Eq. (5), an exponential function.

4.5% PCI

100 0

295

−dα

ð5Þ

where RWC is residual water content %, α is degree of hydration %, and a and b are the adimensional parameters to be determined for each GF mix. The values of a and b for this study are: 6% PCI, c = 10.85,

600 6% PCI

600

4.5% PCI

Predicted AEV (kPa)

500

AEV (kPa)

2% PCI

400 300

Expon. (6% PCI) Expon. (4.5% PCI) Expon. (2% PCI)

200 100 0 0

20

40

60

80

100

Hydration degree %

500 400 300

6% PCI 4.5% PCI

200 2% PCI

100

Godbout(2005) Trendline

0

0

100

200

300

400

500

600

Measured AEV (kPa) Fig. 8. Evolution of air-entry value with (a) curing time, and (b) degree of hydration for various binder proportions.

Fig. 9. Predicted vs. measured AEV for various cemented backfill.

700

296

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

a Residual water content %

12 10 8 6 4 6% PCI

2

4.5% PCI 2% PCI

0

0

20

40

60

80

100

80

100

Curing time (days)

b Residual water content %

12 10 8 6% PCI

6

4.5% PCI

4

2% PCI Expon. (6% PCI)

2

Expon. (4.5% PCI) Expon. (2% PCI)

0

0

20

40

60

Hydration degree %

d = 0.003; 4.5% PCI, c = 12.16, d = 0.005; and 2% PCI, c = 11.53, d = 0.004. To confirm the applicability of the model, the predicted function was validated by using the experimental data (not used to develop the function) obtained from the GF samples in this study. The outcomes of the validation are presented in Fig. 11. It can be noticed that there is a good agreement between the predicted and experimental values. 3.1.4. Model fit to experimental data Numerous empirical equations have been proposed to represent the experimental results of the WRC into mathematical models (WRCmodels). These models have been described and/or summarised in

Predicted RWC (kPa)

12 11 10 9

6% PCI 4.5% PCI

8

2% PCI Trendline

7

8

9

10

ðθs −θr Þ ½1 þ ðαψÞn m
1 1−m 1 α¼ 2 =m −1 ψP θ ¼ θr þ

Fig. 10. Evolution of residual water content with (a) curing time, and (b) degree of hydration for various binder proportions.

7

various publications (e.g., Alonso et al., 1987; Fredlund and Rahardjo, 1993; Fredlund and Xing, 1994; Barbour, 1998; Huang et al., 1998). These models use one of three main variables to express the amount of water in the porous medium; namely, gravimetric water content, volumetric water content or degree of saturation (Fredlund, 2002, 2006). Lately, the performance of various WRC-models was thoroughly examined by Leong and Rahardjo (1997) and Sillers et al. (2001). These models included the well known expressions of Burdine (1953), Gardner (1956), Brooks and Corey (1964), Brutsaert (1966), Mualem (1976), van Genuchten (1980), Tani (1982), McKee and Bumb (1984) and Fredlund and Xing (1994). They were evaluated according to the number, the independence and the physical meaning of the parameters and the accuracy of the prediction provided (Delage, 2002). Leong and Rahardjo (1997), and Sillers et al. (2001) concluded that most of the aforementioned WRC-models behaved satisfactorily and expressed their preference to models able to reach a 1000 MPa suction. Furthermore, Gersovitch (2002) discussed the validity of the WRC-models proposed by Gardner (1956), van Genuchten (1980), and Fredlund and Xing (1994) as compared to experimental WRC from 11 Brazilian soils. They concluded that these three WRC-models were satisfactory for Brazilian soils (Delage, 2002). Comprehensive studies by Fredlund et al. (2009) have shown that the experimental data for various soils over a wide soil suction range can be well fitted using the sigmoidal and continuous WRC-equations of van Genuchten (1980) and Fredlund and Xing (1994). The van Genuchten's equation appears to be the most commonly used continuous function for a WRC (Fredlund et al., 2009). Due to the aforementioned reasons, the van Genuchten (1980) (Eq. (6)) and Fredlund and Xing (1994) models were used to fit the experimental data of this study.

11

Measured RWC (kPa) Fig. 11. Predicted vs. measured RWC of GF with various binder contents.

12

m ¼ ð1− expð−0:8SP ÞÞ 0 ≤ SP ≤ 1 1 n   1  dθ  SP ¼ θs −θr dð logψÞ    dθ  dθ   dð logψÞ ¼ lnð10Þψ dψ m ¼ 1−

ð6Þ

where: ψ θ θs θr α, n, m SP

matrix suction (kPa) volumetric water content saturated water content residual water content fitting parameters slope which is halfway between θs and θr at point P.

However, it should be emphasised that these two WRC-models as well as many of the existing WRC-equations have been developed for porous media made of soils. Therefore, the applicability of the van Genuchten (1980) and Fredlund and Xing (1994) models to cemented porous media (e.g., cemented tailings, concrete) should be questioned. Whilst the validity of the model of van Genuchten (1980) for cemented materials was experimentally demonstrated and verified since the late 1990s, the applicability of the WRC-model of Fredlund and Xing (1994) for cemented porous media has not been proven. For example, experimental evidences of the validity of the van Genuchten's WRC-model are given in Savage and Janssen (1997). These authors demonstrated that the van Genuchten (1980) is suitable for cemented materials and the soil physics principles are

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

applicable to unsaturated cemented porous media. Furthermore, the van Genuchten model has been used successfully for various cemented porous media such as concrete, cemented soils (e.g., Mainguy et al., 2001; Meschke and Grasberger, 2003; Cecconi and Russo, 2008; Khattab et al., 2008). Fig. 12a illustrates the van Genuchten (1980) and Fredlund and Xing (1994) models to fit the experimental data of this study. It can be seen that the van Genuchten (1980) model (Eq. (6)) is the best fit curve to the experimental data. This is in accordance with the findings of Savage and Janssen (1997) and the conclusions of other researchers (e.g., Mainguy et al., 2001; Cecconi and Russo, 2008; Khattab et al., 2008) with regards to the applicability of van Genuchten's WRC-equation to cemented porous media. Figs. 12 and 13 show the van Genuchten model as the best-fit curve to selected experimental data for different binder contents and curing times (3 and 90 days; only best-fit curves of the 3 and 90 day-GF are presented in this section so that to keep the number of figures is a reasonable amount). It can be seen that there is a good agreement between the predicted WRC and measured values. The presence of binders leads to changes in the pore structures of GF with curing time. Also, different binder proportions cause variations in the pore structure. As the binder hydrates, the volume of pores decreases. Furthermore, increasing the binder content leads to the production of more hydration products which reduce the interconnectivity and size of the pores (Figs. 6 and 7). Therefore, fitting parameters are individually calculated for each mix and curing time. The progress of the fitting parameters with degree of hydration is plotted and formulated (Eqs. (7), (8), and (9)) using non-linear regression analysis (Engel, 1982; Snedecor and Cochran, 1986). These functions (Eqs. (7), (8), and (9)) are meant to predict the evolution of the van Genuchten model with the degree of hydration. The least-square statistical approach was used to find the best fit to the experimental data and the quality of the fit was measured by the coefficient of determination. Eqs. (7), (8), and (9) are the best fit functions to the experimental data. Moreover, these functions were validated by using the experimental data obtained from the GF samples in this study. −bα

Alpha ¼ ae −l

m ¼ pα

−f

n ¼ eα

ð7Þ ð8Þ ð9Þ

where alpha, m, and n are the fitting parameters of the van Genuchten model, α is the binder hydration degree, a, b, d, e, f, and p are the adimensional parameters to be determined for each GF mix. For this study, a =0.006, b= 0.014, p =0.54, d =0.084, e=1.73, f =0.008 for 6% PCI; a =0.018, b= 0.035, p =0.49, d =0.046, e=1.9, f =0.02 for 4.5% PCI; and a= 0.014, b =0.024, p= 0.7, d =0.17, e= 2.5, f= 0.114 for 2% PCI. 3.1.5. Comparison of the WRCs of GF with those of uncemented tailings Experimental data on the water retention behaviour of various uncemented tailings and paste tailings have been reported. Qiu and Sego (2001) presented the WRCs for non-plastic gold and copper tailings. The gold and the copper tailings were classified as coarse tailings (they contained 6% and 2 wt.% of 20 μm particles, respectively) according to Landriault (1995), or ML and SM, respectively, according to the Unified Soil Classification System (USCS) (Table 6). Godbout (2005) studied the WRC of uncemented paste tailings. The non-plastic tailings were sampled from a polymetallic mine (Cu, Zn) and contained 53% of 20 μm particles. The tailings can be classified as medium tailings (medium tailings have between 35 wt.% and 60 wt.% of 20 μm particles Fig. 12. Model prediction for WRC for curing time of 3 days (a) 6% PCI, (b) 4.5% PCI, (c) 2% PCI, and (d) van Genuchten model for various binder contents (6%, 4.5%, 2% PCI).

297

298

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301 Table 6 Basis water retention characteristics and physical properties of the tailings and GF. Porous medium Copper tailings Gold tailings Uncemented paste tailings 2%PCI-90 days 4.5%PCI-90 days 6.0%PCI-90 days

AEV (kPa)

RWC (%)

D10 (μm)

% Fine (N 20 μm)

USCS classification

5 6 13

3.4 2.2 –

16.3 5.0 3.2

2 6 53

SM ML ML

495 500 504

7.9 8.0 8.1

– – –

– – –

– – –

according to Landriault, 1995) and ML category in the UCCS (Table 6). It should be recalled that the tailings used in the present study are nonplastic, classified as medium tailings and ML category in the UCCS. Fig. 14 illustrates the comparison of the WRCs of the gold tailings, copper tailings, paste tailings, and 90 days old GFs made of 2%, 4.5% and 6% PCI, respectively. Moreover, as a comparison, the aforementioned WRC test data were also best-fit using the van Genuchten (1980) equation. The air-entry value and residual water content of the WRCs were then determined as listed in Table 6. From this figure and table, it can be clearly observed that the cement has a significant effect on the shape and position of the WRC as well as on the AEVs and RWCs. We can notice that GFs exhibit higher AEVs and RWCs than uncemented tailings regardless of the types and the fineness of tailings. GF samples have higher suction than the uncemented tailings for given water content. This implies a different behaviour of the cemented (GF) and uncemented tailings materials during drying. The rate of desaturation of the cemented tailings is lower (particularly at lower suction range, ≤1000 kPa), in the sense that their water content decreases less rapidly for the higher suctions. This can be attributed to the fact that the precipitation of cement hydration products (e.g., C–S–H, CH) within the GF results in the refinement of the pore structure of the matrix of GF. These results suggest that the GF will show better resistance to acid mine drainage (AMD) than the uncemented tailings. Indeed, the sensitivity of cemented or uncemented tailings materials to AMD is mainly dependent on the reactivity (oxidation potential) of the sulphide minerals (e.g., pyrite, pyrrhotite) contained in the tailings (Fall et al., 2009). In turn, this reactivity is dependent not only on the types and quantity of sulphide minerals present in the tailings system, but also the ease with which fluids, such as oxygen and water, enter and move through the cemented or uncemented tailings matrix, i.e. on the saturated and unsaturated properties of the tailings systems. Finally, from Fig. 14, it can be noted that the effect of binder on the WRCs becomes less significant for high suction ranges. Again, this observation is consistent with the findings of BaroghelBouny et al. (1999) as discussed earlier. 3.2. Unsaturated hydraulic conductivity The coefficient of permeability of an unsaturated porous medium is not a constant. The coefficient is strongly influenced by the volumetric water content or suction, which, in turn depends on the porous medium suction. The suction may be either the matric suction of the porous medium or the total suction (matric plus osmotic suctions) (Fredlund et al., 1994, 1997). A large number of models, developed on physical and empirical bases, have been proposed over the years to predict the hydraulic conductivity of unsaturated soil or porous media (e.g., Brooks and Corey, 1966; Mualem, 1976; van Genuchten, 1980; Leij et al., 1997). Common to almost all predictive models is the existence of a mathematical relationship between hydraulic conductivity and the WRC. The most commonly used model, which was originally developed for soils is the one proposed by Mualem (1976) and Van Fig. 13. Model prediction for WRC for curing time of 90 days (a) 6% PCI, (b) 4.5% PCI, (c) 2% PCI, and (d) van Genuchten model for various binder contents (6%, 4.5%, 2% PCI).

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

a

6.0% PCI-90 days (this study) 4.5% PCI-90 days (this study) 2.0% PCI-90 days (this study) Copper Tailings (Qui and Sego, 2001) Gold Tailings (Qui and Sego, 2001) Paste Tailings (Godbout, 2005)

45 40 35

1.00E-01 6% PCI

30 25 20 15 10 5 0

4.50% PCI

1.00E-04

kunsat (cm/sec)

Volumetric water content %

50

299

2% PCI

1.00E-07 1.00E-10 1.00E-13 1.00E-16

1

10

100

1000

10000

100000

1000000

1.00E-19

Suction (kPa)

1

10

100

1000

10000

100000

1000000

Suction (kPa) Fig. 14. Comparision of the WRCs of 90 days old GF samples with those of uncemented tailings (copper tailings, gold tailings, paste tailings).

b

Genuchten (1980) (Eqs. (10–11)). Savage and Janssen (1997) have found that these functions can be extended to cementitious porous media. Numerous works have then been performed using this finding (e.g., Meschke and Grasberger, 2003; de Sa et al., 2008).

kr ¼

kr ¼

n   o2 1−ðαψÞn−1 1 þ ðαψÞn −m ½1 þ ðαψÞn m=3 kunsat ksat

ð10Þ

ð11Þ

Fig. 15. Unsaturated hydraulic conductivity of GF vs. suction for various binder contents and curing time of (a) 3 days and (b) 90 days.

relative coefficient of permeability matrix suction (kPa) fitting parameters (m = 1 − 1/n) unsaturated hydraulic conductivity saturated hydraulic conductivity.

The unsaturated hydraulic conductivity obtained by the van Genuchten (1980) model (Eq. (10)) versus suction for various binder contents and curing times (3 days and 90 days) is plotted in Fig. 15. As expected, the unsaturated hydraulic conductivity of GF decreases as the suction increases. Furthermore, as the binder content and curing time (i.e. degree of hydration) increase, the unsaturated hydraulic conductivity of GF decreases. This is more obvious at low suction ranges. This can be explained by the fact that unsaturated hydraulic conductivity is a function of WRC and saturated hydraulic conductivity. As the hydration process proceeds, more hydration products are produced (Fig. 4). The formation of these products in the pores with curing time leads to a reduction of porosity (Fig. 6) and the refinement of the pore structure of GF materials which creates a barrier to water movement in the pores. As a result, saturated hydraulic conductivity (ksat) decreases which in turn leads to lower unsaturated hydraulic conductivity. In brief, as the degree of hydration or curing time increases, unsaturated hydraulic conductivity is reduced (Fig. 15). This argument is experimentally supported by the results of saturated hydraulic tests presented in Fig. 16. This figure shows typical results of the effects of the binder proportion (% PCI) and curing time on the saturated hydraulic conductivity (ksat) of GF. From this figure it can be clearly observed that ksat decreases when curing time increases for the GF samples made of 2% and 6% cement, respectively. The effect of binder contents on the unsaturated hydraulic conductivity of GF is easily noticed at low suction ranges. It can be seen that increasing the binder content will reduce the unsaturated hydraulic conductivity (Fig. 15). As mentioned earlier, when the binder

proportion is increased, more cement hydration products are produced, and this refines the pore structure of the GF materials. Consequently, this refinement affects the WRC (see Section 3.1) and the saturated hydraulic conductivity of GF. Indeed, from Fig. 16 it can be seen that the binder content significantly affects the ksat of the GF. The saturated hydraulic conductivity of the GF decreases as the binder content increases. It is important to state that all the factors that affect the WRC and saturated hydraulic conductivity are applicable for unsaturated hydraulic conductivity.

2.50E-05 6% PCI

2.00E-05 2% PCI

ksat (cm/sec)

kr ψ α, n, m kunsat ksat

1.50E-05

1.00E-05

5.00E-06

0.00E+00

0

20

40

60

80

100

120

140

Curing time (days) Fig. 16. Saturated hydraulic conductivity of GF vs. curing time for various binder contents (6% and 2%).

300

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301

4. Summary and conclusions The unsaturated hydraulic properties of GF have been investigated in this study. The van Genuchten (1980) model is used to estimate the unsaturated hydraulic conductivity of GF samples. In order to predict the unsaturated hydraulic conductivity, saturated hydraulic conductivity and WRC are required. The latter is measured in the laboratory by using two different techniques, which include a WP4T Dewpoint Potentiameter and the saline solution method. Based on the outcomes obtained, the following conclusions can be drawn: Whilst the effect of curing times (or degree of hydration) on WRC is obvious at both low and high suction ranges, the binder proportion influences the WRC at low suction ranges only. It is noted that the AEVs increase with an increase in curing time and decrease of void ratio values, whilst the binder content influences the AEVs at early ages only. On the other hand, binder proportions, curing times, and degree of hydration have slight effects on residual water content. Functions to predict the evolution of AEVs, residual water content, and the fitting parameters from the van Genuchten model with degree of hydration for various binder contents are proposed. These formulas can be useful for preliminary design phases of GF structures, verification of the validity of experimental results, and also estimations of the possible variations of unsaturated properties that can be expected during the service of a given GF structure. It has been calculated that the unsaturated hydraulic conductivity of GF decreases when the suction increases, as well as when binder content and degree of hydration increase. The effects of binder content and degree of hydration are more obvious at low suction ranges. Unsaturated hydraulic conductivity is a function of saturated hydraulic conductivity and WRC. For this reason, each factor that affects these parameters has an impact on unsaturated hydraulic conductivity. Valuable outcomes have been obtained with regards to unsaturated hydraulic properties. The results can be used to assess the long term durability and environmental performance of GF structures as well as the stability and behaviour of GF structures at early ages. The findings of this study can also help to predict the potential formation of AMD in GF systems. However, it is essential to mention that only the desorption (drying) part of the WRC is investigated in this study. Therefore, the hysteresis of WRC due to adsorption–desorption isotherms needs to be addressed in further studies. Acknowledgement The authors would like to acknowledge National Sciences and Engineering Research Council (NSERC), University of Ottawa and the National Research Council (NRC) of Canada. References Alonso, E.E., Gens, A., Hight, D.W., 1987. General report: special problem soils. 9th Eur. Conf. on soil Mechanics and Foundations Engineering, 3, pp. 1087–1146 (Dublin, Balkema, Rotterdam). ASTM D5084 - 00 Standard Test Methods for Measurement of Hydraulic Conductivity of Saturated Porous Materials Using a Flexible Wall Permeameter. In 2000 annual book of ASTM standard. ASTM International, West, Conshohocken, Pa. Barbour, S.L., 1998. Nineteenth Canadian geotechnical colloquium: the soil–water characteristic curve: a historical perspective. Canadian Geotechnical Journal 35, 873–894. Baroghel-Bouny, V., Mainguy, M., Lassabatere, T., Coussy, O., 1999. Characterization and identification of equilibrium and transfer moisture properties for ordinary and high-performance cementitious materials. Cement and Concrete Research 29 (8), 1225–1238. Baroghel-Bouny, V., 1998. Texture and moisture properties of ordinary and highperformance cementitious materials. Proceedings of the International RILEM Conference Concrete: From Material to structure, pp. 144–165. Baroghel-Bouny, V., Chaussadent, T., 1995. Pore structure and moisture properties of cement-based systems from water vapour sorption isotherms. Materials Research Society 245–254.

Beaudoin, J.J., 1982. Effect of humidity and porosity on fracture of hardened Portland cement. Cement and Concrete Research 12, 705–716. Bentz, D.B., 2008. A review of early-age properties of cement-based materials. Cement and Concrete Research 38, 196–204. Brooks, R., Corey, A., 1964. Hydraulic properties of porous media. Hydrology Paper No. 3. Colorado State University, Fort Collins. Brooks, R.H., Corey, A.T., 1966. Hydraulic properties of porous media affecting fluid flow. Proceedings of the ASCE Journal of the Irrigation and Drainage Division 92, 61–88. Brutsaert, W., 1966. Probability laws for pore size distributions. Soil Science 101, 85–92. Burdine, N.T., 1953. Relative permeability calculations from pore size distributions data. Journal of Petroleum Technology 5, 71–78. Cecconi, M., Russo, G., 2008. Prediction of soil-water retention properties of a lime stabilized compacted silt. First European Conference on Unsaturated soils. 2–4 July 2008, Durham, United Kingdom. Taylor & Francis, pp. 271–276. Celestin, J.C. (2008). Geotechnical properties of cemented paste backfill and tailings Liners: effect of mix components and temperature. M.A.Sc. thesis, University of Ottawa, 1–221. Cui, Y.J., Tang, A.M., Loiseau, C., Delage, P., 2008. Determining the unsaturated hydraulic conductivity of a compacted sand-bentonite mixture under constant volume and free-swell conditions. Physics and Chemistry of The Earth 33, S462–S471. Davis, G.B., Ritchie, A.I.M., 1986. A model of oxidation in pyritic mine wastes: part 1 equations and approximate solution. Applied Mathematical Modelling 10, 314–322. de Sa, C., Benboudjema, F., Thiery, M., Sicard, J., 2008. Analysis of microcracking induced by differential drying shrinkage. Cement and Concrete Composite 30 (10), 947–956. Delage, P.D., 2002. Experimental unsaturated soil mechanics. 3rd international conference on Unsaturated Soils, Recife Brazil, Balkema. Doucet, C., Tarr, K., Swan, G., 2007. Evaluation of the effect of mixing method, sequence and time on the properties of gelfill. Proceedings of 9th International Symposium on Mining with Backfill, April 29-Mai 2, Montreal, CIM 2488. Engel, B., 1982. Non-Linear Regression. IWIS-TNO. (178 pp.). Fall, M., Samb, S.S., 2008. Pore structure of cemented tailings materials under natural or accidental thermal loads. Materials Characterization 59, 598–605. Fall, M., Samb, S.S., 2009. Effect of high temperature on strength and microstructural properties of cemented paste backfill. Fire Safety Journal 44 (4), 642–651. Fall, M., Benzaazoua, M., Saa, E.G., 2008. Mix proportioning of underground cemented tailings backfill. Tunnelling and Underground Space Technology 23, 80–90. Fall, M., Adrien, D., Celestin, J.C., Pokharel, M., Touré, M., 2009. Saturated hydraulic conductivity of cemented paste backfill. Minerals Engineering 22, 1307–1317. Fall, M., Celestin, J.C., Pokharel, M., Touré, M., 2010. A contribution to understanding the effect of curing temperature on the mechanical properties of cemented tailings. Engineering Geology 114 (3–4), 397–413. Fisseha, B., Bryan, R., Simms, P., 2010. Unsaturated flow and evaporation in multilayer deposits of gold paste tailings. ASCE Journal of Geotechnical and Geoenvironmental Engineering 136 (12), 1703–1712. Fredlund, D.G., 2002. Use of the soil-water characteristic curve in the implementation of unsaturated soil mechanics (Keynote address) In: Juca, J.F.T., de Campos, T.M. P., Marinho, F.A.M. (Eds.), Proceedings of the 3rd International Conference on Unsaturated Soils, UNSAT 2002, Recife, Brazil, 10–13 March 2002, Vol. 3. Taylor and Francis, London, pp. 887–902. Fredlund, D.G., 2006. The Terzaghi lecture: unsaturated soil mechanics in engineering practice. Journal of Geotechnical and Geoenvironmental Engineering 132 (3), 286–321. Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils, pp. 1–510. Fredlund, D.G., Xing, A., 1994. Equations for the soil–water characteristic curve. Canadian Geotechnical Journal 31 (3), 521–532. Fredlund, D.G., Xing, A., Huang, S., 1994. Predicting the permeability function for unsaturated soils using the soil–water characteristic curve. Canadian Geotechnical Journal 31 (3), 521–532. Fredlund, Murray D., Wilson, G.W., Fredlund, Delwyn G., 1997. Estimation of Hydraulic Properties of an Unsaturated Soil using a Knowledge-Based System. Department of Civil Engineering University of Saskatchewan, Saskatoon, Sask.,Canada, pp. 1–7. Fredlund, D.G., Sandra, L., Houston, S.L., 2009. Protocol for the assessment of unsaturated soil properties in geotechnical engineering practice. Canadian Geotechnical Journal 46, 694–707. Gardner, W., 1956. Mathematics of isothermal water conduction in unsaturated soils. Highway Research Board Special Report, 40, pp. 78–87 (Washington D.C.). Gersovitch, Sayão, 2002. Experimental unsaturated soil mechanics. 3rd international conference on Unsaturated Soils, Recife Brazil, Balkema. Godbout, J., 2005. Evolution des proprieties des remblais miniers cimentés en pate durant le curage. Mémoire de Maitrise, Ecole Polytechnique de Montréal, p. 212. Grabinsky, M., Bawden, W., Simon, D., Thompson, B., 2008. In situ properties of cemented paste backfill in an Alimak Stope. Canadian Geotechnical Conference Edmonton 1–7. Hassani, F., Razavi, S.M., Isagon, I., 2007. A study of physical and mechanical behaviour of gelfill. Proceedings of 9th International Symposium on Mining with Backfill, April 29-Mai 2, Montreal, Minefill 2007, CIM, p. 28. Hau, C., Acker, P., Ehrlacher, A., 1995. Analyses and models of the autogenous shrinkage of hardened cement paste. Cement and Concrete Research 25 (7), 1457–1468. Huang, S.Y., Barbour, S.L., Fredlund, D.G., 1998. Development and verification of a coefficient of permeability function for a deformable, unsaturated soil. Canadian Geotechnical Journal 35, 411–425. Khattab, A.A., Al-Juari, K.A.K., Al-Kiki, I.M.A., 2008. Strength, durability and hydraulic properties of clayey soil stabilized with lime and industrial waste lime. Journal Al-Rafidian Engineering 16 (10), 102–116. Landriault, D., 1995. Paste backfill mix design for Canadian underground hard rock mining. 97th Annual General Meeting of CIM. Rock Mechanics and Strata Control Session. Halifax, Nova Scotia, May 14–18, 1995, pp. 229–238.

N. Abdul-Hussain, M. Fall / Engineering Geology 123 (2011) 288–301 Leij, F.J., Russell, W.B., Lesch, S.M., 1997. Closed-form expressions for water retention and conductivity data. Ground Water 35, 848–858. Leong, E.-C., Rahardjo, H., 1997. Permeability functions for unsaturated soils. Journal of Geotechnical and Geo-environmental Engineering 1118–1126. Li, X., Zhang, L.M., Fredlund, D.G., 2009. Wetting front advancing column test for measuring unsaturated hydraulic conductivity. Canadian Geotechnical Journal 46, 1431–1445. Lu, N., William, J.L., 2004. Unsaturated Soil Mechanics. Wiley. (584 pp.). Mainguy, M., Coussy, O., Baroghel-Bouny, V., 2001. Role of air pressure in drying of weakly permeable materials. Journal of Engineering Mechanics 127 (6), 582–592. McKee, C.R., Bumb, A.C., 1984. The importance of unsaturated flow parameters in designing a monitoring system for hazardous wastes and environmental emergencies. Proceedings of the Hazardous Materials Control Research Institute, National Conference 50–58. Meschke, G., Grasberger, S., 2003. Numerical modelling of coupled hygromechanical degradation of cementitious materials. Journal of Engineering Mechanics 4, 383–392. Mokarem, D.W., 2002. Development of Concrete Shrinkage Performance Specifications. Virginia Polytechnic Institute and State University, pp. 1–225. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12, 513–522. Nasir, O., Fall, M., 2008. Shear behavior of cemented pastefill–rock interfaces. Engineering Geology 101 (3–4), 146–153. Nasir, O., Fall, M., 2009. Modeling the heat development in hydrating CPB structures. Journal of Computer and Geotechnics 36 (7), 1207–1218. Neville, A.M., 1995. Properties of Concrete, Forth Edition. , pp. 1–844. Orejarena, L., Fall, M., 2008. Mechanical response of a mine composite material to extreme heat. Bulletin of Engineering Geology and the Environment 67 (3), 387–396.

301

Plagge, R., Renger, M., Roth, C.H., 1989. A new laboratory method to quickly determine the unsaturated hydraulic conductivity of undisturbed soil cores within a wide range of textures. Zeitschrift Pflanzenernaehr: Bodenk, 153, pp. 39–45. Qiu, Y.J., Sego, D.C., 2001. Laboratory properties of mine tailings. Canadian Geotechnical Journal 38 (1), 183–190. Savage, B.M., Janssen, D.J., 1997. Soil physics principles validated for use in predicting unsaturated moisture movement in Portland cement concrete. ACI Material Journal 94 (1), 63–70. Sillers, W.S., Fredlund, D.G., Zakerzadeh, N., 2001. Mathematical attributes of some soil–water characteristic curve models. Geotechnical and Geological Engineering 19 (3–4), 243–283. Snedecor, G.W., Cochran, W.G., 1986. Statistical Methods, 8th ed. Iowa State University Press. (593 pp.). Tani, M., 1982. The properties of a water-table rise produced by a one dimensional, vertical, unsaturated flow. Journal of Japanese Society 64, 409–418 (in Japanese). van Genuchten, M.T.H., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44, 892–898. Vick, S.G., 1990. Planning, Design and Analysis of Tailings Dams. BiTech Publishers Ltd. WP4, WP4-T, 2003. Operator's manual. Version 2.2, pp. 1–83. Yilmaz, E., 2007. How efficiently the volume of wastes produced from mining operations can be reduced without causing any significant environmental impact? Environmental Synthesis Report. UQAT, Canada, pp. 1–45.