Laboratory Study of Phosphogypsum, Stabilizers, and Tropical Soil Mixtures

de Rezende, Lilian Ribeiro; Curado, Tallyta da Silva; Silva, Millena Vasconcelos; Mascarenha, Márcia Maria dos Anjos; Metogo, Daniel Arthur Nnang; Neto, Manoel Porfirio Cordão; Bernucci, Liedi Legi Bariani

doi:10.1061/(ASCE)MT.1943-5533.0001711

Open access

Technical Papers

Aug 10, 2016

Laboratory Study of Phosphogypsum, Stabilizers, and Tropical Soil Mixtures

Authors: Lilian Ribeiro de Rezende [email protected], Tallyta da Silva Curado [email protected], Millena Vasconcelos Silva [email protected], Márcia Maria dos Anjos Mascarenha [email protected], Daniel Arthur Nnang Metogo [email protected], Manoel Porfirio Cordão Neto [email protected], and Liedi Legi Bariani Bernucci [email protected]Author Affiliations

Publication: Journal of Materials in Civil Engineering

Volume 29, Issue 1

https://doi.org/10.1061/(ASCE)MT.1943-5533.0001711

PDF

Abstract

Phosphogypsum (PG) is a by-product of the phosphoric acid fertilizer industries that has possible applications in paving when stabilized with cement or lime. This study’s aim is to compare the mechanical performance of the hemi-hydrate (HH) and dihydrate (DH) mixes with tropical soil, lime, and cement. The results of physical, mechanical, and mineralogical tests are presented, which demonstrate that the use of HH in cement and lime mixes reduces the swelling of this material compared to the use of DH; in addition, HH provides better mechanical behavior. These findings indicate that DH has limited application in pavement construction, but when converted into HH and mixed with tropical soil or chemical stabilizers, its alternative uses are possible, and more content (approximately 90%) of this material can be used.

Introduction

Phosphogypsum (PG) is a by-product of the phosphate fertilizer industry that is obtained from the production of phosphate rock and phosphoric acid. The composition varies as a function of the rock type and the treatment process. The greatest concern regarding this material is related to its environmental characteristics, primarily the existence of radioactive radon, derived from the parent rock.

However, since the 1980s, many studies have focused on the use of PG since it has a low radioactive emanation rate in civil engineering applications, such as in soil stabilization (Degirmenci et al. 2007; James and Pandian 2014; James et al. 2014), pavement layers (Lloyd Jr. 1985; Conklin 1992), the cement industry (Altun and Sert 2004), brickmaking (Yang et al. 2009), and concrete (Smadi et al. 1999). These research results are promising and demonstrate that this by-product can be widely used, rather than restricted to agriculture or disposed as waste.

To technically and scientifically contribute to this discussion, this paper addresses the application of PG generated in the midwest region of Brazil to asphalt pavement layers. This study was motivated by the presence of a fertilizer plant located in Catalao, Goias, which generates 720,000 t of PG a year and has targeted this material solely for agricultural purposes (soil improvement). Moreover, the state of Goias requires a large amount of input to build and maintain its infrastructure because more than 90% of the region’s production is transported via highways. It should also be noted that the natural granular material for building a pavement subbase and base, lateritic gravel, is no longer easily found in the region, and for economic and environmental reasons, its use in construction is becoming unfeasible. Thus, there is a strong trend towards using the regionally abundant lateritic clayey soils and wastes from different types of industries, including PG, if they are technically, economically, and environmentally viable.

It is thus important to consider some aspects about tropical soils in pavement construction. Camapum de Carvalho et al. (2015) discussed many points related to lateritic soils, including the peculiarities and considerations for using them in highway layers. This soil has

{Fe}_{2} O_{3}

distributed in two forms—nodules and dispersed—which interferes in soil aggregation and behavior. The issue here is that while granular tropical soils comply with the standard specifications, the fine tropical soils can present some mechanical limitations. However, these are easily found where the pavement is constructed and can be favored when they have high iron content, since techniques such as chemical stabilization can leverage the use of lateritic clays as subbase and base layers. As the Brazilian fine lateritic soils are not expansive, the stabilization process aims to strengthen the soil. However, microstructural aspects have to be observed in this process. Sometimes, chemical additives such as lime can influence the microaggregation and worsen mechanical behavior (Camapum de Carvalho et al. 2015). In other cases, when the soil has free aluminum atoms to react with, there is a significant strength gain (Rezende 2003). Thus, this issue is complex and needs to be analyzed for each situation.

Ahmed et al. (2011) noted that when calcium sulfate mixes with soil it dissociates into

{Ca}^{++}

and

{SO}_{4}^{- -}

. The attraction between

{Ca}^{++}

and clay induces the cementation between soil particles and improves the soil strength. In addition, the reactions between the silica and alumina that make up the clay minerals contribute to flocculation, and the pozzolanic reactions developed in the process add to the strength developed. This phenomenon will be influenced by the quantity of cementitious gel produced and consequently on the amount of lime consumed (Bell 1996).

Because of the Brazilian regional condition, the mixtures with lateritic clay, PG, and other stabilizers need to be understood. Thus, this paper presents the results of physical and mechanical characterization laboratory tests conducted with dihydrate (DH) PG produced in Goias, Brazil. The material was subjected to a thermal treatment in the laboratory to transform it into the hemi-hydrate (HH) condition. This transformation was done in laboratory scale by calcining samples of DH at 130°C, similar to the procedure presented by Ortiz Oliva (1997). These materials were evaluated in mixes produced with lateritic clay, which prevents swelling, and lime and cement to improve their properties, because PG alone does not usually exhibit satisfactory mechanical strength or stability in the presence of water (Chang et al. 1989). The instability due to water also can be verified in Fig. 1, which depicts PG after soaking in water for 3 h (Matos 2011).

Fig. 1. Phosphogypsum DH immersed in water (reprinted from Matos 2011, with permission)

The first studies conducted with PG in Brazil to analyze the by-product existing in the state of Minas Gerais began in the 1990s and reported the viability of using it in pavement applications (Parreira et al. 2003; Oliveira 2005). In these papers, the PG was produced in Minas Gerais and was mixed only with cement or lime. Thus, the use of tropical soils in the mixtures was not evaluated. The studies with the Goias PG started after 2000, and the results have stimulated the applications of PG (Mesquita 2007; Rufo 2009; Metogo 2011). Both PGs generated in the two regions (Minas Gerais and Goias) are of the DH type, with grain sizes that fit into the silt range and do not exhibit plasticity. This study differs from the others because of the DH transformation in HH, the use of tropical soil, and the impact on the material properties necessary for pavement design.

Table 1 lists the physical properties and the elements found in the chemical analysis performed on pPG from the two regions after leaching and solubilization (Mesquita 2007; Rufo 2009; Pavanin et al. 2007). These results may vary as a function of the parent rock and the treatment process. However, they allow environmental classification of the two types of PG as noninert and nondangerous (Class II A) according to the specific standard of the Brazilian Association of Technical Standards (ABNT 2004) and may thus undergo coprocessing or reuse.

Table 1. Physicochemical Characteristics Observed for Samples of Brazilian Phosphogypsum

Parameter	Units	Minas Gerais phosphogypsum (Oliveira 2005; Pavanin et al. 2007)	Goias phosphogypsum (Mesquita 2007)	Brazilian specification limit (ABNT 2004)
pH	—	5.0	6.8	—
Specific gravity of grains	$g / {cm}^{3}$	2.526	2.676	—
Particles pass on the sieve of 0.075 mm	%	90	80	—
Aluminum	$mg / L$	2.50	4.61	0.2
Arsenic	$mg / L$	—	$< 0.01$	0.01
Barium	$mg / L$	—	0.81	0.7
Cadmium	$mg / L$	—	$< 0.005$	0.005
Lead	$mg / L$	—	$< 0.01$	0.01
Copper	$mg / L$	0.04	0.15	2.0
Total chromium	$mg / L$	—	$< 0.05$	0.05
Iron	$mg / L$	1.67	0.1	0.3
Fluoride	$mg / L$	1.86	10.4	1.5
Manganese	$mg / L$	1.50	0.30	0.1
Mercury	$mg / L$	—	$< 0.001$	0.001
Silver	$mg / L$	—	$< 0.05$	0.05
Selenium	$mg / L$	—	$< 0.01$	0.01
Sodium	$mg / L$	1.24	10.5	200.0
Sulfate	$mg / L$	2.14	2,750	250.0
Zinc	$mg / L$	0,18	1.47	5.0

Table 1 also shows that the fluoride leachability is greater than the specification value for the DH PG made in Goias. However, Matos (2011) performed chemical analysis of the water that percolated in the permeability tests of the mixes of compacted samples containing DH PG and observed that the water contained no fluorite. Thus, for roadway engineering applications, this problem can be avoided by using the DH PG mixture with other materials and compaction. Ahmed et al. (2011) also noted that the presence of cement on soil-gypsum mixtures can reduce the release of fluorine. However, they did not study samples without cement. Considering that asphalt surface is a waterproofing material and that the drainage design must be implemented with the pavement construction, when used in pavement subbase and base layers, leaching and solubilization of the compacted layers and subsequent soil contamination are not expected to occur.

Saueia et al. (2013) presented a recent study on the PG from different industries in Brazil and concluded that the metal concentration obtained from the samples studied in within the safe use parameters. Mazzilli et al. (2000) verified that the radioactive element concentration ranges found in Brazilian PGare wide (

{}^{226}{Ra}

from 22 to

695 Bq / kg

,

{}^{210}{Pb}

from 47 to

894 Bq / kg

,

{}^{210}{Po}

from 53 to

677 Bq / kg

, and

{}^{232}{Th}

from 7 to

175 Bq / kg

); however, the results match the values found in the literature for PG from other locations.

There are results from radiometry evaluations conducted both in the industrial plant and in specimens produced in the laboratory with the Goias PG (Mesquita 2007). The average exposition rate obtained at the plant was

0.09 μR / h

or

2.32 \times 10^{- 8} C / kg / h

, whereas in the specimens compacted in the laboratory, the rate was

0.00 μR / h

, which can be considered low (Chang et al. 1989). Thus, there is no radioactive problem when this material is used in on-site pavement construction.

Objective

The PG produced in Goias, Brazil, does not present environmental risks, and its use in pavement layers will occur in the region where it is generated, which makes it economically viable. The objective of this paper is to present the physical characteristics and the mechanical behavior of samples composed of PG, tropical soil, and lime or cement mixes through laboratory tests for use in asphalt pavement subbase and base layers. With these results, engineers can choose the best way to reuse the PG without environmental and mechanical risks.

Materials and Methods

The materials studied in the research and in the laboratory tests conducted are presented as follows.

Materials

Ten samples were studied, eight of which comprised mixes containing PG, one comprised only tropical soil, and one contained soil-lime. A fixed percentage of 9% lime was adopted based on the results presented by Faria et al. (2007). In the study for cement, the same percentage was used as an analogy. Table 2 shows a detailed identification of the mixtures for which the type of PG used (dihydrate, DH, or hemi-hydrate, HH) varied along with the type of chemical stabilizer incorporated (lime or cement) and the presence of tropical caly in the mixture. In this study, the PG and the soil used were the same as in Matos (2011).

Table 2. Definition of Samples for Laboratory Studies

Sample	Identification	Symbol
1	$91 % phosphogypsum dihydrate + 9 % lime$	91DH-9L
2	$11 % phosphogypsum dihydrate + 80 % soil + 9 % lime$	11DH-80S-9L
3	$91 % phosphogypsum dihydrate + 9 % cement$	91DH-9C
4	$11 % phosphogypsum dihydrate + 80 % soil + 9 % cement$	11DH-80S-9C
5	$91 % phosphogypsum hemi - hydrate + 9 % lime$	91HH-9L
6	$11 % phosphogypsum hemi - hydrate + 80 % soil + 9 % lime$	11HH-80S-9L
7	$91 % phosphogypsum hemi - hydrate + 9 % cement$	91HH-9C
8	$11 % phosphogypsum hemi - hydrate + 80 % soil + 9 % cement$	11HH-80S-9C
9	100% soil	100S
10	$91 % soil + 9 % lime$	91S-9L

Two types of PG were used: DH from the Catalao industry (Goias, Brazil) and HH obtained from a thermal treatment of DH performed in the laboratory. For this, the PG collected at the company underwent a calcination process, during which it was kept in an oven at 130°C while its moisture was monitored every 24 h. Only after verifying the stabilization of the material mass, which would ensure that all of the water had been extracted, was the PG removed from the oven. The transformation of 5 kg of DH into HH under these conditions was verified to take approximately 4 days, which reduced the DH mass 50% until it transformed into HH.

The aforementioned procedure was adopted for this study to ensure that the HH was obtained in a small scale, but other authors (Smadi et al. 1999; Yang et al. 2009; Rentería-Villalobos et al. 2010) used different temperatures (90 to 170°C), periods (3 to 24 h), or autoclave. For large-scale transformation, other processes should be done to test different temperatures and pressures together. If HH is manufactured reliably in large scale, it can be obtained with less time in the industry and the processing cost should not be so expensive. The design of this conversion process is already known by the industries. On the other hand, as some fertilizer industries obtained the HH first in their plants, if HH can be applied for engineering constructions it could be marketed and shipped in bags as done for lime and cement.

The soil used was from Aparecida de Goiania (Goias, Brazil) and was collected at a depth of 0.50 m. The soil was classified as SM (United Soil Classification System), A-4 (Transportation Research Board), lateritic clay LG’ (Brazilian specification for tropical soils—MCT Methodology), and is not expansive nor soft (Metogo 2011). It was chosen because it had been used in mixtures with PG to construct a segment of experimental pavement in 2009 that is undergoing the monitoring phase (Metogo 2011).

The lime was of the CH-III hydrated type, and the portland cement was CPII Z-32 RS (pozzolanic compound resistant to sulfate), both of which were classified according to the Brazilian standards (ABNT 1991, 2003) and were easily found in the region.

Methods

Most laboratory tests for determining the sample properties were performed for all of the mixes following the Brazilian ABNT or the National Department of Transportation Infrastructure (DNIT) standards. Some tests also used the ASTM standards. These tests were performed to understand the behavior of the two PG types (DH and HH) when it is mixed with other materials (X-ray diffraction, characterization, compaction, and scanning electron microscopy) and to determine important mechanical properties for pavement construction (swelling, unconfined compression, and resilient moduli). Table 3 presents the tests that were performed in each material or sample type. The X-ray tests were also conducted to identify the soil characteristics, the PG type generated in the thermal treatment process and for samples composed of PG, and lime or cement that exhibited special mechanical behavior.

Table 3. Summary of Laboratory Tests

Type of test	Sample
Type of test	91DH-9L (1)	11DH-80S-9L (2)	91DH-9C (3)	11DH-80S-9C (4)	91HH-9L (5)	11HH-80S-9L (6)	91HH-9C (7)	11HH-80S-9C (8)	100S (9)	91S-9L (10)
X-ray diffraction	X	—	X	—	—	—	—	—	—	—
Characterization	X	X	X	X	X	X	X	X	X	X
Compaction	X	X	X	X	X	X	X	X	X	X
Swelling	X	X	X	X	X	X	X	X	—	—
Unconfined compression	X	X	X	X	X	X	X	X	—	—
Dynamic triaxial	—	—	—	—	X	X	X	X	—	—

Characterization and Compaction

The first characterization test was performed to determine the specific masses of the grains. Conventional tests could not be performed for the samples containing PG mixed with either lime or cement, i.e., Sample 1 (91DH-9L) and Sample 3 (91DH-9C) presented mixture stiffening (ABNT 1984b). For this reason, a gas picnometer equipment type, the Pentapyc 5200e (ASTM 2006), was used. Tests were also conducted for determining the grain-size curve, with and without using dispersant (ABNT 1984a), and the Atterberg limits (ABNT 1984c, d).

For determining the compaction curves, the specimens were compacted, without reusing the material, using the energy defined in Brazil as intermediary Proctor: a small cylinder (

1,000 {cm}^{3}

), large tamper (4,536 g), with three layers and 21 blows per layer (ABNT 1984e). This energy was chosen as it is the most frequently used in paving works in the state of Goias, Brazil.

Swelling

The mixtures’ swelling was determined by volume variation measurement of the material fraction put through a 0.42-mm sieve, when in defined compaction conditions water was absorbed by capillarity through a porous plate. This index is expressed in percentage. A DNIT standard was used for this test, in which samples are compacted in two nearly equal layers; each layer undergoes 50 evenly distributed compressions with a frequency of one compression per second (DNIT 2012). In this method, periodical readings of the extensometer are made until two readings with a 2-h interval yield the same value or decreasing values. Since in conventional tests for analyzing swelling, the soil and soil-lime, Samples 9 (100S) and 10 (91S-9L), presented values below 0.1% (Rufo 2009), the swelling tests were conducted solely for the other samples.

Unconfined Compression

Specimens of the samples were compacted at the optimum water content and at the maximum dry density defined from the compaction curves obtained for the intermediary Proctor energy. After this process, some specimens were taken to the press, to determine the compressive strength without curing. For the mixtures, other specimens were wrapped in plastic film, a process known as sealed curing, and placed into a sealed polystyrene foam box where they were kept for 7 and 28 days to be tested. These tests were conducted based on the ABNT (2012) standard.

Dynamic Triaxial

The dynamic triaxial test allows determining the resilient modulus for different stresses and is important in the study of pavements for simulating the loads applied to the structure due to the traffic. The dynamic triaxial tests were conducted according to the DNIT (2010) standard. The specimens measuring

10 \times 20 cm

were molded at optimal water content and at the maximum dry density of the intermediary Proctor energy. The compacted specimens were covered by PVC tubes, placed in bags, and kept in a humidity chamber during the curing time. Before the tests, specimens underwent gypsum capping to make the bases even. The tests were conducted at 0, 7, and 28 days of curing. To optimize the use of the equipment, the samples that had presented the best results, based on the other mechanical tests results, were selected for this test.

X-Ray Diffraction and Scanning Electron Microscopy

To perform the X-ray tests, the samples were oven-dried at 40°C (DH) and 150°C (HH), disaggregated, and run through a 0.075-mm sieve. These tests were conducted on a Bruker D8 Discover diffractometer (Bruker, Belgium) using

{Kα}_{1}

radiation from a tube with a copper anode at 1,600 W, a Bragg-Bentano

θ / 2 θ

configuration, a Lynxeie linear detector, a measuring span at

2 θ

from 5° to 100° with a 0.005° step, and 1-s counting time for each step. The sample underwent a 15-rpm rotation during the measurement.

For obtaining additional information about the mineralogy of the soil, DH and HH PG, and of the mixes of these materials with tropical soils and chemical stabilizers, scanning electron microscopy (SEM) tests were performed. The tests were performed with Jeol JSM-6610 equipment, which allows capturing high-resolution images from a sample surface.

To conduct the SEM, the samples were molded into small cubic pieces of about 1 cm in edge and vacuum-dried in oven at 60°C, for about 48 h. After drying, the samples were stored in containers with blue silica, which acts as a desiccator, to ensure that the material did not absorb moisture. Later, the sample was removed with a spatula, fit into a sample holder, and placed in the equipment for performing the test. The X-ray and SEM results will help to understand the changes in behavior of these materials.

Results and Discussion

The results from this research, as well as the analyses and discussions, are presented in this section.

Characterization and Compaction

The results of the characterization tests are presented in Table 4. The sample granulometric analyses showed that for the mixes prepared with HH PG, the percentage of fines found in the test without dispersant was greater than for the tests with dispersant. This did not occur with the mixes prepared with DH PG. These results suggest that the DH transformation into HH changes the shapes of the material’s particles, therefore impacting its mechanical behavior.

Table 4. Summary of Characterization Tests Results

Property	Sample
Property	91DH-9L (1)	11DH-80S-9L (2)	91DH-9C (3)	11DH-80S-9C (4)	91HH-9L (5)	11HH-80S-9L (6)	91HH-9C (7)	11HH-80S-9C (8)	100S (9)	91S-9L (10)
Particle size with dispersant
Gravel (%)	0.00	0.14	0.00	0.07	0.00	0.00	0.00	0.00	0.80	0.79
Sand (%)	11.21	52.69	13.41	53.66	47.19	54.34	67.22	78.98	54.35	58.16
Silt (%)	88.79	47.17	86.6	46.28	50.79	32.19	32.59	17.59	13.63	37.05
Clay (%)	0.00	0.00	0.00	0.00	2.02	13.56	0.19	3.42	34.23	3.99
Particle size without dispersant
Gravel (%)	0.00	0.14	0.00	0.07	0.00	0.00	0.00	0.00	0.80	0.79
Sand (%)	30.28	65.21	23.97	65.92	22.45	83.97	45.08	78.46	71.44	79.69
Silt (%)	69.72	34.64	76.03	34.01	74.89	16.03	54.92	15.42	27.56	19.21
Clay (%)	0.00	0.00	0.00	0.00	2.66	0.00	0.00	6.13	0.21	0.31
Specific gravity of grains and Atterberg limits
$ρ (g / {cm}^{3})$	2.542	2.902	2.461	2.733	$2.771 \pm 0.02 %$ ^a	2.515	$2.831 \pm 0.06 %$ ^a	2.695	2.697	2.775
$w_{L}$ (%)	—	31	—	32	—	34	—	34	34	31
$w_{P}$ (%)	—	25	—	23	—	27	—	27	29	28
PI (%)	NP	6	NP	9	NP	7	NP	79	5	3
Classification systems
TRB	A-4	A-4	A-4	A-4	A-4	A-4	A-4	A-4	A-4	A-4
USCS	ML	SM	ML	SM	ML	SM	SM	SM	SM	SM

Note: PI = plasticity index; TRB = Transportation Research Board; USCS = Unified Soil Classification System; $w_{L}$ = liquid limit; $w_{P}$ = plasticity limit; $ρ$ = specific gravity of grains.

a

Determined with gas piconometer.

Table 5 presents the results of the compaction tests. These tests were carried out just to obtain the materials’ optimum compaction conditions. The values of the maximum dry density (

γ_{d \max}

) of the samples were observed to remain practically the same, independently of the type of PG incorporated into the mixes. The HH PG mixes showed optimum water content values (

w_{opt}

), much better than the DH PG mixes. This increase can be due to the dihydrate state of HH PG, which demands a higher amount of water.

Table 5. Summary of Compaction Tests Results

Property	Sample
Property	91DH-9L (1)	11DH-80S-9L (2)	91DH-9C (3)	11DH-80S-9C (4)	91HH-9L (5)	11HH-80S-9L (6)	91HH-9C (7)	11HH-80S-9C (8)	100S (9)	91S-9L (10)
$w_{opt}$ (%)	24.4	21.9	25.6	20.9	33.1	24.9	27.9	22.6	21.0	25.8
$γ_{d \max}$ ( $kN / m^{2}$ )	13.30	15.60	13.30	15.70	13.42	15.05	13.40	15.73	16.00	14.30

Note: $w_{opt}$ = optimum water content; $γ_{d \max}$ = dry maximum density.

Swelling

Fig. 2 depicts the data of the samples’ volume variation over time obtained in the swelling tests. The samples containing DH PG had their volumes stabilized from 2 to 5 days. In turn, those containing HH PG stabilized on the very first test day, except for Sample 6 (11HH-80S-9L), which took 3 days. Only the latter sample presented greater swelling than the sample containing DH PG (Sample 2 11DH-80S-9L). These results show that HH PG is more stable in the presence of water than DH and that the reasons for these differences in behavior should be assessed through complementary tests, such as X-ray and SEM.

Comparing the results from the 10% limit established in the Brazilian standard (DNIT 2007) for the lateritic gravel commonly used in the region, only Sample 3 (91DH-9C) is observed not to abide by this criterion. Using another methodology, Parreira et al. (2003) also observed a continuous swelling on the PG DH sample stabilized with 10% of cement even after 84 days of curing. However, the authors concluded that this swelling value is in the same range of traditional base pavement materials.

On the other hand, some authors did not study swelling on soils stabilized with low PG values (less than 25%) and solidification agents (Degirmenci et al. 2007; James and Pandia 2014; Ahmed 2015). For PG mixtures prepared with expansive soil and without solidification agents, James et al. (2014) observed less swelling than in the natural soil. Thus, the swelling questions for these mixtures are strongly dependent of the soil type, the PG type, the PG content, and the presence of other stabilizers.

Unconfined Compression

When preparing the specimens for the simple compression tests, it was difficult to mold compacted specimens for Samples 5 (91HH-9L) and 7 (91HH-9C) in the lathe, because soon after being compacted, they already presented great stiffness. This behavior is due to the rapid reaction of the HH PG with binder and water, potentialized by the application of the compaction energy. Samples 6 (11HH-80S-9L) and 8 (11HH-80S-9C) were relatively easy to mold. However, during the curing time, these mixtures stiffened considerably. Note that molding a specimen of the test size without using a lathe may provide higher strength values.

Fig. 3 presents the maximum strength values obtained in the tests as a function of the curing time for each sample. For the mixtures containing HH PG, mainly Samples 5 (91HH-9L) and 7 (91HH-9C), high strength values were observed. For Sample 5, the value obtained was 20.9 times higher than the one in Sample 1 (91DH-9L) after 28 days of curing. The presence of HH PG influences the increase in strength, even in mixtures containing soil because at 28 days, Sample 8 (11HH-80S-9C) reached twice the value obtained by Sample 4 (11DH-80S-9C). This shows the extent to which HH PG has better mechanical behavior than DH. The greatest variation in strength over time was verified in Sample 6 (11HH-80S-9L).

Fig. 3. Results of unconfined compression tests

Ahmed et al. (2011) explain that the PG HH mechanical behavior is better than that of DH because the former absorbs water from the soil, reducing the voids and strengthening the soil. This mechanism can also explain the lower expansion observed in HH PG.

When PG is added to the soil stabilization process, an ettringite precipitation may appear and strengthen the soil by filling up pores. However, according James and Pandian (2014), for higher PG levels, more large clusters of ettringite can be formed and this can cause a decrease in strength. On the other hand, Ahmed (2015) studied the microstructure and mineralogical compositions of soft clay soil stabilized with recycled bassanite, produced from gypsum waste plasterboards by heating. The results showed various cementation compounds in the soil matrix when recycled bassanite was added, which improved the compressive strength. The author also noted that, as the admixture increased, the ettringite formation would also increase. It is also notable that the admixture ratio would influence ettringite formation differently in each bassanite mixture: in bassanite-cement, it would reduce ettringite formation, while in bassinte-lime, there would be no influence. Thus, the impact of ettringite in the mechanical behavior of mixtures with PG depends on the soil type and solidification agents. The discussion about this issue should to be based on other test results like X-ray and SEM.

The fine tropical soil (Sample 9) exhibited 0.37 MPa compressive strength. When stabilized with 9% lime (Sample 10), the initial strength increased to 0.53 MPa. At 21 days, the strength reached 1.30 MPa, and its behavior was similar to that of Sample 2 (11DH-80S-9L) and was better than that of Sample 1 (91DH-9L), as shown in Fig. 3. This finding may induce smaller gains in strength in mixes of DH PG and lime. This smaller gain should be evaluated to determine whether it is sufficient for the good performance of this material as a pavement layer.

Unconfined compression tests were carried out with mixtures composed of gypsum or PG, lime or cement, in different ratios and curing time by several authors (Parreira et al. 2003; Oliveira 2005; Ahmed et al. 2011; Ahmed and Issa 2014; James and Pandian 2014; Ahmed 2015). Table 6 presents the tests results after 28 days. The stabilizers percentages were presented as soil or PG dry mass. Thus, the material contents for this study and Oliveira (2005) samples were recalculated using this definition for better understanding.

Table 6. Comparison of Unconfined Compression Test Data for 28 Days

Author	Materials	Compacted type	Unconfined compression (MPa)
Ahmed and Issa (2014)	$Soil + 22.5 % (GHH + cement ratio 1 ∶ 1)$	Static (n.d.)	1.8
Ahmed and Issa (2014)	$Soil + 22.5 % (GHH + lime ratio 1 ∶ 1)$	Static (n.d.)	1.9
Ahmed et al. (2011)	$Soil + 5 % GHH + 3 % cement$	Static (pressure 33 kPa)	0.6
James and Padian (2014)	$Soil + 2 % PG + 7 % lime$	Static (n.d.)	2.2
Parreira et al. (2003)	$PG DH + 10 % cement$	Static (standard energy)	0.9
Parreira et al. (2003)	$PG DH + 10 % cement$	Static (modified energy)	1.8
Oliveira (2005)	$PG DH + 11 % lime$	Static (modified energy)	1.0
This study	$PG DH + 10 % lime$ (Sample 1)	Dynamic (intermediary energy)	0.3
	$PG DH + 10 % cement$ (Sample 3)		2.1
	$Soil + 14 % PG DH + 11 %$ lime (Sample 2)		1.6
	$Soil + 14 % PG DH + 11 %$ cement (Sample 4)		3.2
	$PG HH + 10 %$ lime (Sample 5)		8.3
	$PG HH + 10 %$ cement (Sample 7)		10.9
	$Soil + 14 % PG HH + 11 %$ lime (Sample 6)		1.7
	$Soil + 14 % PG HH + 11 %$ cement (Sample 8)		6.6

Note: DH = dihydrate; G = gypsum; HH = hemi-hydrate; n.d. = not determined; PG = phosphogypsum.

The results for PG

DH + 10 %

cement presented by Parreira et al. (2003) are coherent with the values herein (PG

DH + 10 %

cement). However, for mixtures with lime (Oliveira 2005), the strength was lower, but the compacted energy was different (

PG DH + 10 %

lime).

The value of unconfined compression strength obtained by James and Pandian (2014) for

Soil + 2 % PG + 7 %

lime was higher than the value obtained for

Soil + 14 %

PG

DH + 11 %

lime. This result probably was influenced by the soil type, the level of PG, and the compaction type. Unconfined compression strength values of HH gypsum with cement statically compacted, obtained by Ahmed and Issa (2014) and Ahmed et al. (2011), were lower than the values obtained in this study for HH PG. These results can be explained by the type of cement used (furnace cement type-B versus portland cement CPII Z-32 RS) or by the cement content (3% versus more than 9%). For samples with HH and lime (

Soil + 22.5 % G HH + lime 1 ∶ 1

, Ahmed and Issa 2014), the results were similar (

Soil + 14 %

PG

HH + 11 %

lime).

Studies conducted in the United States presented unconfined compressive strength values varying between 0.60 and 5.50 MPa for PG from different sources (Chang et al. 1989). In this study, most samples had values that agreed with this range without considering the curing time. After 7 days of curing, Samples 5 (91HH-9L) and 7 (91HH-9C) had much higher values. Aiming to use the materials in a pavement base, North American specifications recommend compressive strength values greater than 1.70 MPa after 7 days of curing (Gutti et al. 1996). In this sense, Samples 4 (11DH-80S-9C), 5 (91HH-9L), 7 (91HH-9C), and 8 (11HH-80S-9C) could be used for this application.

Dynamic Triaxial

To optimize costs and time of the dynamic triaxial equipment used, not all of the samples were tested. Thus, based on the simple compressive strength results, the samples containing HH PG were selected for this assessment because they had the best mechanical results. Table 7 presents a summary of the results from the triaxial dynamic tests. Except for Sample 8 (11HH-80S-9C) without curing, the mathematical model as a function of the confining stress (

σ_{3}

) was verified to represent the data obtained from the tests more accurately.

Table 7. Summary of Dynamic Triaxial Tests Results

Sample	Days	$RM = k_{1} \times σ_{d}^{k 2}$	$R^{2}$	$RM = k_{1} \times σ_{3}^{k 2}$	$R^{2}$
5 (91HH-9L)	0	$RM = 4,260 \times σ_{d}^{0.0835}$	0.37	$RM = 5,474 \times σ_{3}^{0.1536}$	0.86
	7	$RM = 4,283 \times σ_{d}^{0.0941}$	0.41	$RM = 5,628 \times σ_{3}^{0.1696}$	0.92
	28	$RM = 4,650 \times σ_{d}^{0.0873}$	0.48	$RM = 5,816 \times σ_{3}^{0.1469}$	0.94
6 (11HH-80S-9L)	0	$RM = 612.4 \times σ_{d}^{- 0.014}$	0.01	$RM = 765.9 \times σ_{3}^{0.0672}$	0.21
	7	$RM = 1,055 \times σ_{d}^{0.106}$	0.46	$RM = 1,405 \times σ_{3}^{0.1836}$	0.94
	28	$RM = 1,507 \times σ_{d}^{0.0851}$	0.38	$RM = 1,973 \times σ_{3}^{0.1615}$	0.91
7 (91HH-9C)	0	$RM = 2,518 \times σ_{d}^{0.070}$	0.35	$RM = 3,141 \times σ_{3}^{0.133}$	0.84
	7	$RM = 3,565 \times σ_{d}^{0.023}$	0.29	$RM = 3,869 \times σ_{3}^{0.046}$	0.85
	28	$RM = 4,237 \times σ_{d}^{0.174}$	0.67	$RM = 5,907 \times σ_{3}^{0.2529}$	0.99
8 (11HH-80S-9C)	0	$RM = 1,358 \times σ_{d}^{- 0.047}$	0.64	$RM = 1,408 \times σ_{3}^{- 0.025}$	0.12
	7	$RM = 2,088 \times σ_{d}^{0.034}$	0.28	$RM = 2,374 \times σ_{3}^{0.072}$	0.82
	28	$RM = 2,727 \times σ_{d}^{0.117}$	0.58	$RM = 3,534 \times σ_{3}^{0.183}$	0.95

Note: $k_{1}$ , $k_{2}$ = constants obtained in the tests; $R^{2}$ = determination coefficient; RM = resilient modulus; $σ_{3}$ = confining stress; $σ_{d}$ = deviator stress.

The values of the resilient modulus (RM) presented in Fig. 4 were obtained for 0.371 MPa confining stress (

σ_{3}

) and 0.126 MPa deviator stress (

σ_{d}

). These stresses were determined considering the load of a tire 10.8 cm in diameter and with 0.55 MPa contact pressure on a structure commonly used in urban roads of the region composed of three homogeneous, isotropic, and elastic layers with the following characteristics:

•

3-cm-thick asphalt coating, Poisson coefficient (

μ

) equal to 0.5, and RM equal to 2,000 MPa;

•

15-cm-thick base,

μ

equal to 0.5, and RM equal to 300 MPa; and

•

Subbase with

μ

equal to 0.5 and RM equal to 50 MPa.

Fig. 4. Results of dynamic triaxial tests

The RM values obtained for the mixes containing HH PG are observed to be higher than those for the granular materials conventionally used in the region. This is the case of the lateritic gravel with RM values approximately equal to 430 MPa (Cunha 2011), which demonstrates the technical viability of using these materials and the possibility of constructing durable roadways. In the cases of DH PG and cement mixtures, Pericleos and Metcalf (1996) obtained RM values of 300 MPa after 28 days of curing. Thus, also considering this parameter, it is possible to verify the best mechanical behavior of HH PG.

The lateritic clayey soil (Sample 9) features an RM equal to 50 MPa, which is considered very low and does not show the possibility of being used as pavement base and subbase. When stabilized with 9% lime (Sample 10), the RM value rises to 360 MPa without evaluating the curing period, which shows the viability of using this stabilized soil. However, the RM values obtained from the incorporation of HH PG are much higher than the others, which shows that these mixtures have a great potential to use in regions with abundant fine lateritic soils.

In Samples 6 (11HH-80S-9L) and 8 (11HH-80S-9C), the values obtained were smaller than for the others, but they were sufficient to achieve good performance as pavement layers. In this case, depending on the type of asphalt mix used as a surface course, the base may still exhibit the structural behavior and stress distribution usually found in a flexible pavement.

In Samples 5 (91HH-9L) and 7 (91HH-9C), the RM values obtained were very high. If these mixtures are used as a base layer, depending on the type of surface course, they may even exceed the RM values of the asphalt mix, altering the structural behavior. Therefore, this should be considered when designing the pavement. These results agree with those obtained in the unconfined compression tests. In Sample 5 (91HH-9L), it is also verified that the RM value obtained in the test without curing was higher than in those with curing. This behavior was not expected, and a repetition of this test with better control of the mix homogenization process is recommended. In the international literature, there were no published data that presented RM results for mixtures with HH PG. Thus, for this parameter it was not possible to do comparisons.

X-Ray Diffraction and Scanning Electron Microscopy

Fig. 5(a) presents the results from the X-ray diffraction test conducted on the soil used in this study, where different peaks for the composition are visible, which includes quartz (

{SiO}_{2}

), kaolinite

[{Al}_{2} {Si}_{2} O_{5} (OH)_{4}]

, gibbsite

[Al (OH)_{3}]

, and hematite (

{Fe}_{2} O_{3}

). Fig. 5(b) also shows the SEM image for the natural soil (Alves 2015), where it is possible to verify the presence of microaggregations with macropores between them. In the compaction process, microporosity is preserved, but the pore distribution is altered (Camapum de Carvalho et al. 2015). These results are typical of fine lateritic soils.

Fig. 5. Mineralogical and microstructural results for fine lateritic soil: (a) X-ray diffraction; (b) SEM (reprinted from Alves 2015, with permission)

Fig. 6 presents the results from the X-ray diffraction tests of the DH PG and of the thermally treated PG. The major phase of the DH PG was observed to correspond to

Ca ({SO}_{4}) \cdot 2 H_{2} O

and that of the thermally treated PG corresponds to

Ca ({SO}_{4}) \cdot 0.5 H_{2} O

. That is, the thermal treatment process at 130°C causes the formation of HH PG. The process for transforming DH PG into HH through oven-drying was also verified by Yang et al. (2009). It is worth noting that the PG minor phases occur in low content, with relative peak intensities of 1%, which hinders identification.

Fig. 6. Results of X-ray diffraction for different types of phosphogypsum: (a) DH; (b) HH

While conducting this research, the mechanical results showed the need to perform X-ray diffractions on samples with DH containing lime or cement (Samples 1 and 3) and HH with lime or cement (Samples 5 and 7) (Figs. 7 and 8). During the DH transformation process into HH, not only bassanite but anhydrite can also be formed, as shown in Fig. 8. When HH is used on admixtures and compacted, the water addition caused its rehydration, which led to the presence of gypsum in Fig. 8.

Fig. 7. Results of X-ray diffraction for DH mixtures: (a) Sample 1 (91DH-9L); (b) Sample 3 (91DH-9C)

Fig. 8. Results of X-ray diffraction for HH mixtures: (a) Sample 5 (91HH-9L); (b) Sample 7 (91HH-9C)

For DH samples, the presence of ettringite [Figs. 7(a and b)] was observed, which can explain some mechanical characteristics that will be discussed next. For HH samples, there is no ettringite formation in lime mixtures [Fig. 8(a)], but this compound appeared in cement mixtures [Fig. 8(b)]. Ettringite is an expansive mineral that develops in the presence of sulfate, calcium, and aluminum compounds of a fine grain fraction at high pH levels. In the DH and HH samples, ettringite was a result of the reaction between aluminate, calcium in cement, lime, PG, sulfate in PG, and water in the tested sample, as discussed by Ahmed (2015). In some cases, the aluminate amount is insufficient to react and form ettringite, which in this study it happened with the HH and lime mixture. Moreover, in the HH-lime mixture calcium carbonate was formed due the reaction between calcium in PG or lime with water in the tested sample, thus producing calcium oxide that reacts with carbon dioxide in atmosphere (carbonation). Ahmed (2015) showed that the carbonation can limit the ettringite formation. Both substances, ettringite and calcium carbonate, will interfere on the mixture’s mechanical behavior.

The SEM results are presented to discuss the changes in behavior of these materials. In this test, it is possible to compare the DH PG standard observed in the images of Sample 1 (91DH-9L) with those obtained for the HH PG. As opposed to DH PG [Fig. 9(a)], which features tabular-shaped and large plates (Rutherford et al. 1994; Roy et al. 1996; Yang et al. 2009; Shen et al. 2012), HH phosphogypsum [Fig. 9(b)] has smaller lamellar plates (Yang et al. 2009; Yu and Brouwers 2011) as observed in the characterization results.

Fig. 9. SEM images of phosphogypsum: (a) dihydrate (DH); (b) hemi-hydrate (HH)

In the SEM HH PG mixtures shown in images [Figs. 10(b and d)], it was not possible to identify ettringite formation. Thus, to complement the SEM tests, energy-dispersive X-ray spectroscopy (EDS) tests were conducted. As a result, Fig. 11(a) shows that there was aluminum in the DH sample, and Fig. 11(b) shows that this element was not identified on the HH sample, which could be due to the removal of gypsum impurities, such as aluminum, when subjected to thermal treatment. These results could indicate that ettringite cannot be formed in HH mixtures. However, in the X-ray results, ettringite was identified in the HH and cement mixture. This difference can be explained by the size and shape analysis between the DH and HH plates. Figs. 9 and 10 show that the DH plates are larger than the ones found in HH, as a result of the porous microstructure and less interactions between grains. In the DH mixtures, ettringite is placed around the plates because in this region there are porous and large surfaces that facilitate its formation [Figs. 10(a and c)]. On the other hand, in the HH mixtures there are less porous and plate surfaces [Figs. 10(b and d)] that can contribute to the ettringite effect.

Fig. 10. SEM images of phosphogypsum mixtures ( $\times 5, 000$ ): (a) dihydrate (DH) with cement; (b) hemi-hydrate (HH) with cement; (c) dihydrate (DH) with lime; (d) hemi-hydrate (HH) with lime

Fig. 11. SEM with EDS results: (a) dihydrate (DH); (b) hemi-hydrate (HH)

According to Fig. 2, by adding cement or lime to HH PG, a reduction in swelling compared to the mixes for which DH PG will occur. This behavior can be justified by the microstructural difference between the HH and DH mixtures on ettringite formation as discussed. The ettringite formation in the DH PG and cement mixture and its relation with expansion are also discussed by Gutti et al. (1996) and Roy et al. (1996).

By analyzing Figs. 10(a and c), a larger amount of ettringite is observed in the mixture with DH cement, which can justify the greater swelling of this mixture compared to the DH lime, as verified in Fig. 2. The X-ray test results also confirmed the presence of ettringite in these samples (Fig. 7). In addition, the morphology of the ettringite observed in the mixture with cement [Fig. 10(a)] is different from that observed in the mixture with lime [Fig. 10(c)]. In the former, the elements have greater lengths; in the latter, the elements are smaller and form radial crystal spheres. Thus, the ettringite morphology in DH cement mixes appears to be more harmful in terms of swelling than in DH lime mixes.

As depicted in Fig. 3, the HH mixtures that were stabilized with cement exhibited greater strength than the mixtures stabilized with lime when considering the same content of materials and the same type of PG with samples similarly prepared. This behavior can be explained by Fig. 12 in which it can be observed that the cement and HH PG mix [Fig. 12(a)] is less porous and more homogeneous than the lime and HH PG mix [Fig. 12(b)]. These results illustrate the matrix bounding that occurs with the stabilization process. Thus, HH PG interacts better with cement than with lime, which causes an increase in strength. Hence, owing to the shape of the grains, HH PG interacts better with the components of the mixtures, resulting in materials with greater strength compared with those for which DH PG was used, as verified in Fig. 3.

Fig. 12. SEM images of HH mixtures ( $\times 1, 000$ ): (a) with cement; (b) with lime

For mixtures with the fine lateritic soil, DH or HH, cement or lime, the SEM images (Fig. 13) show that the DH soil mixtures show aggregations from the reaction of the cement or the lime with the soil, along with some pores and ettringite from the reaction of the DH and the lime or cement [Figs. 13(a and b)]. One can note that ettringite was attached only in the PG plate. For the HH soil mixtures [Figs. 13(c and d)], the images showed a more homogeneous mass, with microaggregations resulting from reactions between solidification agents and soil, but it was not possible to observe ettringite clusters. As verified from the mechanical properties presented in this paper, these mixtures had an intermediate behavior when compared with the others. However, the advantage to use the latertic soil in PG cement or lime mixtures is to reduce the swelling, which is explained by the soil’s mineralogical composition as shown in Fig. 5(a). The timely appearance of ettringite in DH mixtures is a likely result of the PG and does not generate negative consequences for the mixtures’ performance.

Fig. 13. SEM images of lateritic fine soil mixtures with: (a) DH and cement; (b) DH and lime; (c) HH and cement; (d) HH and lime

Summary and Conclusions

The behavior of mixtures with phosphogypsum (PG), stabilizers, and soils depends on the type and the content of PG (DH or HH), stabilizers (lime or cement), and soil (expansive, soft, or lateritic). Any change can impact the swelling and strength in the different forms.

Oven-drying at 130°C resulted in the transformation of dihydrate (DH) PG into hemi-hydrate (HH). This process forms smaller and more homogeneous lamellar plates that contributed to the gain in strength for the HH mixtures. For the studied PG percentages, the use of HH PG in mixtures with cement and lime reduces the swelling of this material compared to the use of DH PG. In the former case, a possible explanation is that the ettringite formation on mixtures with HH and cement is less harmful than the one formed in DH mixtures because of the HH plate type. For the HH and lime mixture, there is no ettringite formation because of the carbonation reaction and the absence of aluminate. HH PG also interacts better with cement than with lime, resulting in a less porous and more homogeneous structure, justifying the greater strengths on unconfined compression and dynamic triaxial test results. In addition, the water absorption by HH PG strengthens the soil. When the mixtures had a lateritic fine soil, the use of HH PG and cement still showed better mechanical properties than the use of DH PG and lime.

The presented results indicate that DH PG has limited applications in pavement layers. However, better parameters can be obtained if PG is used with both lateritic soil and chemical stabilizers. When transformed into HH, the alternative uses of PG multiply, and more content (approximately 90%) of PG can be used. Thus, this material may be applied in flexible or semistiff pavement layers for roadway construction with traffic varying from light to heavy, without causing environmental and mechanical problems.

This topic is still under study in Brazil, and more tests, in both the laboratory and field, have to be done to confirm and complement the results presented in this paper, such as the solubility and durability properties. The costs and the chemical and environmental risk analysis also should be further investigated. However, the results presented in this paper can contribute to defining the best way to use PG in civil engineering applications and to expand the knowledge on reusing its by-products.

Acknowledgments

The authors are thankful for the funds provided by FAPEG and Anglo American to support this research, the partnering institutions for the use of their laboratories (University of Sao Paulo, University of Brasilia, Furnas and Regional Center for Technological Development and Innovation—CRTI/UFG) and the CNPq and CAPES for the grants.

References

ABNT (Brazilian Association of Technical Standards). (1984a). “Análise granulométrica.”, Sao Paulo, Brazil (in Portuguese).

Abstract

Introduction

Objective

Materials and Methods

Materials

Methods

Characterization and Compaction

Swelling

Unconfined Compression

Dynamic Triaxial

X-Ray Diffraction and Scanning Electron Microscopy

Results and Discussion

Characterization and Compaction

Swelling

Unconfined Compression

Dynamic Triaxial

X-Ray Diffraction and Scanning Electron Microscopy

Summary and Conclusions

Acknowledgments

References

Information

Published In

Copyright

History

Authors

Affiliations

Metrics

Citations

Download citation

Cited by

Figures

Other

Share

Copy the content Link

Share with email

Share

Request Username

Create a new account

Change Password

Password Changed Successfully

Verify Phone

Congrats!