Radio PW Paper
Radio PW Paper
Radio PW Paper
H I G H L I G H T S G R A P H I C A L A B S T R A C T
art ic l e i nf o a b s t r a c t
Article history: Natural gas has become an essential energy resource in the U.S. due to the increasing demand of energy,
Received 29 December 2014 the high oil prices, and the need of foreign oil independency. The improvement in the drilling technology
Received in revised form has allowed the rapid expansion in gas production, especially for unconventional gas such as shale gas.
10 June 2015
Shale gas is natural gas trapped within fine-grained sedimentary rocks called shale formations.
Accepted 10 July 2015
Hydraulic fracturing is used to extract natural gas from these formations. Although natural gas is
cleaner-energy source than coal or oil, there is a lot of controversy due to the environmental impact
Keywords: related to the water consumption and treatment. Hydraulic fracturing generates significant volumes of
Natural gas wastewater that contain dissolved chemicals, high content of salts, and significant levels of natural
Hydraulic fracturing
occurring radioactive material (NORM). Hence, one of the biggest challenges of this industry is to
Produced water
develop techniques for the prevention, remediation, and appropriate disposal of NORM. The overall
Natural occurring radioactive material
(NORM) objective of this work is to develop and implement a novel computational tool for high-throughput
Group contribution method (GCM) screening and selection of new adsorbents for NORM removal. In the first part of this paper series, we
Adsorbate solid solution theory (ASST) study the adsorption theory for NORM removal and applied group contribution methods (GCMs) to
predict specific properties of adsorbents based on their thermodynamics. Then, in the second part of this
paper we develop a computer-aided molecular design (CAMD) framework that generates potential
adsorbent candidates according to the properties developed by the GCMs.
& 2015 Elsevier Ltd. All rights reserved.
1. Introduction
The increase of the world's energy demand and the high oil
n
Corresponding author. prices has led to the exploration of new resources such as the
E-mail address: urmila@vri-custom.org (U. Diwekar). extraction of natural gas from unconventional shale plays like
http://dx.doi.org/10.1016/j.ces.2015.07.012
0009-2509/& 2015 Elsevier Ltd. All rights reserved.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
2 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
shale gas. Natural gas production from shale gas formation is one concentrated amount of hardness species (e.g. Ca þ þ , Mg þ þ , Mn þ þ ,
of the fastest-growing divisions in the U.S. oil and gas industry, Br þ þ , and Cl ) and metals such as barium and strontium. As stated
and it is a promising resource to achieve the foreign oil indepen- earlier, added to this high salinity, many flowback and produced
dency. The U.S. Energy Information Administration estimates U.S waters contain significant levels of NORM in the form of radium
natural gas reserves at 2203 trillion cubic feet (tcf), enough to last isotopes with activities of 185–592Bq/L (Warner et al., 2013). Although
about 92 years (U.S. Department of Energy Website, 2014a, 2014b). 95% of the flowback water is reused, the supply of flowback and
Advances in drilling technology have rapidly expanded natural gas produced water is expected to exceed the reuse capacity due to the
production. According to a reference case reported on the Annual rapid increase in shale gas extraction. This situation has become a big
Energy Outlook of 2014 (U.S. Department of Energy Website, issue in the safe disposal of liquid waste in the U.S. given their large
2014a, 2014b), the production of natural gas will grow steadily, volume and typically high levels of contaminants. Most oil–gas field
with a 56% increase between 2012 and 2040, when production will NORM waste is stored at production sites awaiting for disposal in
reach 37.6 tcf. This substantial growth in shale gas production specially designated and permitted landfills, disposal wells, injections
comes with new environmental concerns. wells, or treat it in a publicly owned treatment plant or wastewater
Shale gas is a natural gas trapped within fine-grained sedi- treatment plants. If the production site cannot handle the final
mentary rocks called shale formations. Two important shale plays disposal of these waters, then these waters must be sent to other
are the Barnett Shale in Texas and the Marcellus Shale in the water treatment facilities which increase the costs of operation.
eastern of the U.S which is shown in Fig. 1 (Replica Engineering However, as the natural gas production increases, these alternatives
INC, 2014). It was deposited around 390 million years ago in the will not be enough to support its final –safe-disposal. In some states,
shallow sea that once covered the region. Shale is composed of these alternatives are not available or allowed due to the lack of
tiny mud particles and organic matter, which, because of the regulations or appropriate geology for deep-well injection sites.
properties of radioactive elements present in the sea water, often Moreover, there is still potential for drinking water contamination
contains concentrations of naturally occurring radioactive material due to spill in the surface and leaks from improperly cemented well
(NORM). Black shale, such as that contained in the Marcellus, often casting. Even though multiple layers of steel casting are inserted into
contains trace levels of uranium (U), thorium (Th), and potassium the wellbore to avoid contamination, there has been found some
(K), and in higher concentrations than found in less organic-rich documented cases of localized releases of fluids at the surface causes
gray shale. Natural gas fracking in the Marcellus shale brings out by spill and casting ruptures (Lustgarten, 2009). Therefore, one of the
NORM to the surface in a concentrated form, which could pose a biggest challenges of this industry is to develop techniques for the
radiation safety hazard and it is the focus of these series of papers. prediction, prevention, remediation, and appropriate disposal of oil-
The first stage of shale gas production requires large volumes of field NORM. Currently, it is not possible to selectively remove NORM
fresh water (e.g. 4–5 million of gallons according to Smith (2012)) that from produced water so that brines could be reduced in volume to
may affect the availability of water for other uses as well as aquatic concentrated salt solutions or dry salts producing a waste material
habitats. About 15–25% of the water injected during hydraulic that does not have radionuclide safety issues.
fracturing returns to surface as “flowback water” within the first 30 In this work, we concentrate on the adsorption process for NORM
days while produced water, which is the water produced at the end of removal specifically the radium and barium isotopes. Computer-aided
the flowback, continues to flow for the life of the well at a reduced molecular design (CAMD) is used to develop new adsorbents with
rate (Silva et al., 2012). Moreover, hydraulic fracturing generates desired properties that can remove these materials. CAMD is gen-
significant volumes of contaminated wastewater which may contain erally the reverse use of the group contribution method (GCM) that is
dissolved chemicals and other contaminants that require a special used to generate molecules having desirable properties. A GCM is a
treatment before final disposal or reuse. In Marcellus gas shale, the technique to estimate and predict thermodynamic and other proper-
overall estimated volume of wastewater (3.1 106–3.8 106 cubic ties from molecular structures of compounds. In the first part of this
meter per year) has increased during the last years and there is an paper, we focus on developing group GCMs to predict the desired
average of 0.15–0.3 MM gallons per day of produced water that must properties based on their thermodynamics. We use experimental data
be properly disposed (Silva et al., 2012). It has been found that the of known adsorbents and their adsorption equilibrium to predict the
salinity of shale gas wastewater varies from 5000 mg/L to properties such as activity coefficient. A modified UNIFAC method is
4200,000 mg/L (Warner et al., 2013). Much of the flowback and used for calculation of activity coefficient. In the second part of this
produced waters such as those generated in the Marcellus contain paper (Benavides et al., 2015), we use CAMD framework to generate
new adsorbent having the desirable properties. As a result, the GCM
along with CAMD based optimization methods identify existing and
new potential groups of adsorbents for the NORM removal.
2. Background
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3
respectively (Marcellus Shale, 2011). On the other hand, it has in Eq. (1):
been found that mixtures of carbonate and sulfate minerals are 1 1 1
present in the coating and sediments in the production equip- ¼ þ ð1Þ
q b ðbKÞC
ment. One of these sulfate minerals is Barite (i.e. Barium Sulfate),
which is known as the primary host NORM (Zielinski and Otton, where q the amount adsorbate adsorbed in a unit mass of
1999). Therefore, the focus of this work is on the adsorption of adsorbent and C is the equilibrium concentration of adsorbate in
NORM specifically radium and barium isotopes. the solution. b and K are constant values. Another typical equation
Radium (Ra) is a radioactive element that may represent a used to describe the isotherm is the Freundlich equation as shown
potential health hazard (i.e. as carcinogen and other disorders) if it in Eq. (2). In this case, kf and nf are the Freundlich parameters
is released into the environment because radium and its decays determined from correlation of adsorption data.
product radon emit alpha particles that can kill and mutate cells. q ¼ k f C nf ð2Þ
In produced water, NORM species typically comprise two forms:
226
Ra and 228Ra (Zielinski and Otton, 1999). Due to the long half- Adsorption can perform many separations that are impossible
life of 1600 years, 226Ra is of principal interest compared to 228Ra or impractical by conventional techniques such as distillation,
which has a half-life of 5.8 years. Although it is expected that both absorption, and even membrane-based systems. However, adsorp-
forms will behave similarly in the removal process, we focus on tion is rarely employed as a separation process due to the lack of
the adsorption of 228Ra. Radioactive isotopes are commonly both experimental data and suitable models used to predict the
quantified in terms of activity concentration or simply “activity”. equilibrium behavior of multicomponent mixtures (Berti et al.,
These concentrations are generally reported as picocuries/gram 2000). In addition, the calculation of the measurements involved
(pCi/g) of solid material or picocuries/liter (pCi/L) of water or air. A in the equilibrium is very time-consuming and cost-intensive.
picocurie equals 0.037 Bq which is another form to express the Therefore, distillation processes have assumed a dominant role in
activity. Marcellus shale produced water samples contain more separation technology. GCM method has shown a great success for
radium than non-Marcellus samples. For instance, the range of the design of the most common separation processes and similar
radium activities for samples from Marcellus shale (median value approaches are suggested for the calculation of adsorption
is 2460 pCi/L) overlaps the range for non-Marcellus reservoirs processes.
(median value 734 pCi/L) (Rowan et al., 2011).
Barium is a major fission product resulting from the fission of 2.3. Adsorbate solid solution theory (ASST)
plutonium and uranium. High concentration of barium has been
associated with multiple sclerosis and neurodegenerative diseases In contrast to the classical formulation of adsorption equili-
(Purdey, 2004). The toxicity of barium is related to its solubility brium, the adsorbate solid solution theory (ASST) describes the
and for insoluble salts the solubility increases with decreasing pH. adsorption behavior of liquid mixtures on solid surfaces using GE-
The EPA allows a maximum of 0.002 mg/g of barium in drinking models (i.e. Gibbs excess free energy models). The approach,
water (U.S.EPA, 2014). discussed by Berti et al. (1999, 2000) considers the adsorbed
phase to be a mixture containing the adsorbed species and the
adsorbent as additional component (see Fig. 2a). The relation of
the phase equilibrium for component i is shown in Eq. (3):
!
2.2. Adsorption process g ad g ad
xbi γ bi ¼ xsi γ si exp oi
ð3Þ
RT Γ mi
s
Adsorption is a separation process consists adhesion of mole-
cules of liquids, gases and dissolved substances (i.e. adsorbate) into Here, xbi and xsi are the fractions of component i in the bulk
the surface of solid (i.e. adsorbent) because of physical and phase and surface phase, respectively. γ bi and γ si are the activity
chemical processes (Yang, 2003). The most fundamental property coefficients in each phase, g ad is the free energy of immersion of
in adsorption is the adsorption equilibrium which can be describe the adsorbed solution and g adoi is the free energy of immersion of
by adsorption isotherms (Suzuki, 1990). The isotherms are illu- the pure adsorbed species i. T; R and Γ smi are temperature, the
strated in many mathematical forms such as the Langmuir shown ideal gas constant, and surface phase capacity of component i,
Fig. 2. Classical description of the adsorption systems (a) the total amount of fluid molecules is divided in two independent phases that are assumed to be homogeneous: the
adsorbed or surface phase and the bulk phase. The adsorbate solid solution (b) is defined as the reference system and the bulk phase is considered to be uninfluenced by
the solid
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
4 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
respectively (Berti et al., 1999, 2000). The surface phase capacity 2.4. Group contribution method (GCM) and the modified UNIFAC
Γ smi is given by Eq. (4), when assuming pore-filling model of model
adsorption, vp is the pore volume of the adsorbent and voi is the
molar volume of the respective fluid: GCM is used to estimate properties of compounds based on
vp known experimental data of well-defined pure components and
Γ s
mi ¼ ð4Þ mixtures. Many examples can be found in the literature (Lydersen,
voi
1955; Joback and Reid, 1987; Marrero and Gani, 2001). The
Berti et al. (1999, 2000) argued that for the pure component objective of GCMs is formulating a property of a given compound
adsorption, the respective activity coefficient equals to 1, however, as a statistical correlation like multi-linear regression. The para-
the measurable enthalpy of immersion proves the existence of meters involved in this regression are determined by summing the
interactions between adsorbent and the liquid. Therefore, they frequency of each group occurring in the molecule times its
formulated a new thermodynamic framework that incorporates contribution. The ASST offers an excellent basis to incorporate
the adsorbent's properties by conceiving adsorption systems as a group contribution methodology that takes into account the
solution of the solid and the liquid adsorbates. In that sense, the influence of different structural groups of the adsorbent with
influence of different structural groups of the adsorbent on the respect to adsorption. This theory uses the UNIFAC method which
adsorption behavior is taken into account. This framework is is also used to predict information for the vapor–liquid equili-
described in Fig. 2b (Berti et al., 2000). Then, the relation of the brium (Fredenslund et al., 1977). UNIFAC stands for functional-
phase of equilibrium for this new framework is presented by groups activity coefficients and this method is resulted from the
Eq. (5) based on the chemical potential in both phases. model called UNIQUAC (Universal Quasi-Chemical). Thus, for the
n
ðφ φnoi Þ first evaluation of a group contribution method, a modified
xbi γ bi ¼ xni γ ni exp ð5Þ UNIFAC model presented by Berti et al. (2000) is used to compute
RTΓ smi
the activity coefficient in the adsorbent solid solution. In the
where xni is the fraction of component i in the adsorbate solid modified model, the pure component adsorption (γ noi ) has to be
solution, γ ni is the activity coefficient of the same component, and incorporated as shown in Eq. (12):
the expression (φn φnoi ) is the difference between the chemical
potential of the wetted solid and the chemical potential of the ln γ ni ¼ ln γ noi þ ln γ nGE;i ð12Þ
wetted solid at adsorption of the pure component i (Berti et al., where γ GE;i is the concentration dependant part of the activity
n
1999, 2000). This last term can be computed by Eq. (6) coefficient and it is divided into a combinatorial part (γ Ci n ) and
1 En residual part (γ Ri n ):
ϕn ϕnoi ¼ G GEs GE0in ð6Þ
m0 ln γ nGE;i ¼ ln γ Ci n þ ln γ Ri n ð13Þ
En
Here, m0 is the mass of adsorbent, and G ; G Es
; GE0in ;
are the Gibbs
For the combinatorial and residual part Eqs. (14) and (15) can
excess free energy of the ASS, Gibbs excess free energy of the
be used:
surface phase and the Gibbs excess free energy of the pure n n
component adsorption all in Joules, respectively. Usually, it is ln γ Ci n ln γ CGE;i
n
xi ln γ CGE;i
n
x0i ð14Þ
more convenient to express the Gibbs excess free energy in n n
terms of activity coefficients, therefore, for the adsorbate solid ln γ Ri n ≡ ln γ RGE;i
n n
xi − ln γ RGE;j x0i ð15Þ
solution:
For the adsorbent each of the second terms is dropped. The
GE n Xk
combinatorial and the residual part
¼ nni ln γ ni ð7Þ is given as it is presented by
RT Fredenslund et al. (1977). γ nGE;i xn0i represents the activity coeffi-
i¼0
cient of a binary ASS, that is, the pure adsorbed liquid and the
where the index 0 refers to the adsorbent and the activity adsorbent.
coefficient is computed using the UNIFAC model (Berti et al.,
1999).
Similarly, the definition of the activity coefficients for the 3. Adsorption of radium and barium
surface phase yields their connection with the Gibbs excess free
energy as shown below: In this section, we study the adsorption equilibrium of radium
and barium on different adsorbents. Based on experimental data
G Es X
k
¼ nsi ln γ i ðxsi Þ ð8Þ reported in the literature, we predict the interaction parameters of
RT i¼1 the functional groups that are commonly contained in adsorbents
of radium and barium.
where nsi is the molar quantity in the surface phase for component
i and it is computed using the following equation:
3.1. Problem description of radium removal
nsi ¼ xsi nnT ð1 xn0 Þ ð9Þ
and xsi (fraction of component i in the surface phase) can be The first problem considers the adsorption of radium from
calculated by: aqueous solutions on different minerals. The experimental data
was reported by Ames et al. (1983a, 1983b). In the paper, the
xni authors presented radium adsorption efficiencies as a function of
xsi ¼ ð10Þ
ð1 xn0 Þ temperature determined in a 0.01 M NaCl solution. In the initial
Finally, for the pure component adsorption, the respective solution, when no adsorbent is added, the components involved
activity coefficient γ noi is given by Eq. 11: are: radium (1), sodium chloride (2), and water (3). The equili-
brium is reached after 30 days of mineral solution continuous
contact. At this point, two new phases are presented: bulk phase
1
γ noi ¼ 1 þ ð11Þ and the adsorbate solid solution (see Fig. 2b). The bulk phase
Γ smi Mo
contains part of the radium that was not adsorbed, sodium
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5
chloride, and water while in the ASS phase contains the adsorbed For combinatorial part of component i:
radium and the adsorbent as shown in Fig. 3. !
φi θ φX j
Table A3 of Appendix 3 presents the minerals used by Ames lnðγ i Þ ¼ ln b þ 5qi ln i þ li bi
Cb
xbj lj ð23Þ
et al. (1983a, 1983b) with their properties which used the xi φi xi 1
Freundlich equation (Eq. 2) to describe the adsorption equilibrium.
where
The values of k and n are given by Ames et al. (1983a, 1983b).
These values depend on the temperature and type of adsorbent. li ¼ 5 r i qi ðr i 1Þ ð24Þ
Therefore, for the purpose of this paper we use the information
reported at 25 1C of adsorption temperature for experimental q xbi
results. θi ¼ PMi ð25Þ
j qj xbj
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
6 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
3.4. Problem description for barium removal (i.e. aa112 and aa121). The optimization problem presented in the
main paper was used to find these interaction parameters.
Produced water can also contain barium at high concentration.
Therefore, in order to reach high efficiency of the recovery of
water, the barium must be removed. The experimental data was 4. Estimation of the interaction parameters: an optimization
obtained from two papers: Chavez et al. (2010) and Arafa et al. problem
(1974). In the first paper, the adsorption isotherms were deter-
mined by contacting 0.2 g of two adsorbents: clinoptilolite tuff and As mentioned before, our objective is to find the unknown
montmorillonite clay with 10 ml of a solution that contained values of the UNIFAC interaction parameters (i.e. abnm and aanm ) by
barium chloride (BaCl2) and calcium chloride (CaCl2). Similarly to fitting different adsorption systems and compared them with the
radium adsorption, two phases are presented after reaching information found in the literature. Therefore, we use optimization
equilibrium (See Fig. 4). The second paper by Arafa et al. (1974) techniques to solve the two parameter estimation problems
– 5 g of adsorbent were mechanically shaken with 100 cm3 of a presented in the previous section (i.e. radium and barium adsorp-
solution of barium chloride and sodium chloride (0.01 M) for three tion). The parameter estimation problem deals with solving a
hours. Fig. 4 can also be used to represent the adsorption scheme system of equations based on measured or empirical data so that
process presented by Arafa et al. (1974). However, in the latter the values of the parameters can be estimated by optimizing an
case, the adsorbent used was the β-Mn2O and the bulk phase objective function. We use an error function (Err) (mean square
composition contained barium chloride that did not react, sodium error) as an objective function for this problem. This function
chloride and water. Table A4 in Appendix 3 presents the chemical represents the square root of the difference between the adsorp-
structure of the adsorbents used for adsorption of radium and tion capacity from the experimental values (qexp ) given by the
their molecular weight. adsorption equilibrium and a new calculated qcal . This adsorption
The adsorption equilibrium of Ba þ 2 by Ca-Cln and by Ca-Mnt capacity difference is applied for each adsorbent used. For the
are described by the Langmuir equation shown in Eq. 1. The radium adsorption problem, we considered 4 adsorbents (Kaoli-
constant values of b and K can be found in Chavez et al. (2010). On nite, montmorillinite, glauconite, and clinoptinolite) since they
the other hand, the adsorption equilibrium of barium by β-Mn2O is contain all the functions group needed for the prediction of the
taken from the adsorption isotherm presented in the paper by interaction parameters. For the barium adsorption problem, we
Arafa et al. (1974). Its isotherm was of the Langmuir type and it solved two problems: adsorption of barium onto Ca-Cln and Ca-
was reported at pH 7.5 and 35 1C of temperature. We use the same Mnt and adsorption of barium onto β-MnO2. Then, using the
procedure presented for radium adsorption to estimate the moles interaction parameters as the decision variables, the objective
and mole fractions in the bulk and ASS phase. Table 3 presents the function whose goal is to minimize the sum of this square of
known and unknown values of interaction parameters based on errors is expressed as:
the groups used in each problem in the bulk phase (abnm ). The van Eq h
X
AD X i2
der Waals volume and surface parameters for each functional minErr ¼ min qcal exp
Eq;AD qEq;AD ð33Þ
group can be found in Table A1 of Appendix 1. As it can be seen in 1 1
these tables, there are four unknown interaction parameters in
where AD is the adsorbent used, and Eq is the total number of
each problem, in which all of them are related to barium group.
equation used to solve the problem, we used around 20–40
Finally, Table 4 shows the interaction parameters involved in
experimental data points (i.e. 20–40 equations) depending on
the ASS phase (aanm) for the two adsorption problems, respec-
tively. For problem one (i.e. barium adsorption onto Ca-Cnt and
Ca-Mnt) there are 20 unknowns while for problem two (i.e.
barium adsorption onto β-Mn2O) there are only two unknowns
Table 1
UNIFAC interaction parameters ðabnm Þ for bulk phase.
Table 3
Table 2 UNIFAC interaction parameters ðabnm Þ for bulk phase for adsorption of barium onto
UNIFAC interaction parameters ðaanm Þ for ASS. Ca-Cnt and Ca-Mnt and adsorption onto β-Mn2O.
Groups (g) Ra SiO2 Al2O3 Na2O MgO Fe2O3 K2O CaO Groups (g) Ba2 þ Ca2 þ Cl H2O Na þ
Ra 0 aa12 aa13 aa14 aa15 aa16 aa17 aa18 Ba2 þ 0 ab12 ab13 558.57a ab15
SiO2 aa21 0 0 0 0 0 0 0 Ca2 þ ab21 0 4166.3a 838.2a NAb
Al2O3 aa31 0 0 0 0 0 0 0 Cl ab31 21983a 0 230.2a 14548a
Na2O aa41 0 0 0 0 0 0 0 H2O 374.92a 897.2a 982.5a 0 22.38a
MgO aa51 0 0 0 0 0 0 0 Na þ ab51 NAb 6342.2a 165.0a 0
Fe2O3 aa61 0 0 0 0 0 0 0
a
K2O aa71 0 0 0 0 0 0 0 The known values of abnm for can be found in Kikic and Fermeglia (1991).
b
CaO aa81 0 0 0 0 0 0 0 There is not interaction parameter between Na and Ca since each of them
belongs to different adsorption problem.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7
Table 4
UNIFAC interaction parameters ðabnm Þ for ASS phase onto Ca-Cln and onto Ca-Mnt and onto β-Mn2O.
Groups (g) Ba2 þ SiO2 Al2O3 (AlO2) FeO MgO CaO Na2O K2O Ca2 þ Kþ β-MnO2
Ba2 þ 0 aa12 aa13 aa14 aa15 aa16 aa17 aa18 aa19 aa110 aa111 aa112
SiO2 aa21 0 0 0 0 0 0 0 0 0 0 NAa
Al2O3 aa31 0 0 0 0 0 0 0 0 0 0 NAa
(AlO2) aa41 0 0 0 0 0 0 0 0 0 0 NAa
FeO aa51 0 0 0 0 0 0 0 0 0 0 NAa
MgO aa61 0 0 0 0 0 0 0 0 0 0 NAa
CaO aa71 0 0 0 0 0 0 0 0 0 0 NAa
Na2O aa81 0 0 0 0 0 0 0 0 0 0 NAa
K2O aa91 0 0 0 0 0 0 0 0 0 0 NAa
Ca2 þ aa101 0 0 0 0 0 0 0 0 0 0 NAa
Kþ aa111b 0 0 0 0 0 0 0 0 0 0 NAa
β-MnO2 aa121 NAa NAa NAa NAa NAa NAa NAa NAa NAa NAa 0
a
There is not interaction parameter between β-MnO2 and the other groups since β-MnO2 belongs to the second problem by Arafa et al. (1974).
the problem. To calculate qcal , we use Eq. (34): 5. Results and discussion
n n
x 1 nT
qcal ¼ ð34Þ Table 5 presents the values of the 20 UNIFAC interaction
m0
parameters determined between all possible functional group that
To find nnT , we use the equations of the adsorption equilibrium could be part of the new adsorbent for radium adsorption. We can
formulation presented in the previous sections which include the observed the strong interaction between radium and different
equilibrium equation (Eq. (5)) and the difference of the chemical groups such as water and sodium oxide since these parameters
potential of the adsorbent (Eq. (6)). This last equation involves the represent the energy of the interaction between the functional
calculation of the Gibbs excess free energy in three parts: adsor- groups.
bate solid solution (Eq. (35)), surface phase (Eq. (36)), and for the As mentioned before, these parameters were determined by
pure component (Eq. (37)) fitting adsorption systems of four different adsorbents: Kaolinite,
clinoptinolite, montmorillonite, and glauconite. The isotherms
GEn ¼ RT nn0 ln γ n0 þ nn1 ln γ n1 ¼ RT xn0 nn ln γ n0 þ xn1 nn ln γ n1 ð35Þ (calculated and experimental) of these systems are presented in
Fig. 5. In this figure, we compared the capacity of adsorption
GEs ¼ RT ns1 ln γ 1 ðxs1 Þ ¼ RT ns xs1 ln γ 1 ðxs1 Þ ð36Þ (qexp ) from the experimental results and the calculated adsorp-
tion capacity (qcal ) with the interaction parameter shown in
!
1 Table 5. It can be seen that the experimental and the estimated
GE0in ¼ RTnn0i ln γ n0i ¼ RTxn0i nn ln 1 þ ð37Þ adsorption capacity values of q are in good agreement. These
Γ smi M0
results are also reflected in the error function value obtained (i.e.
where xn0i ¼xn1 since there only one adsorbate (either radium or 1.86e 4) which it corresponds to the value of the objective
barium) in the ASS phase. GEs is the Gibbs excess free energy of the function shown in Eq. (33).
surface phase and in this case is equal to zero since only involves On the other hand, Table 6 reports the values of UNIFAC
the activity coefficient for pure component (xs1 ¼1), therefore, interaction parameters between all possible functional group that
replacing Eqs. 34, 35 and 36 into Eq. (6), we obtain an expression could be part of the new adsorbents for barium adsorption. In this
to calculate nnT : case, the energy interaction of barium is very high especially with
the anion chloride and the potassium oxide.
ϕn ϕnoi m0 Similarity, we compared the capacity of adsorption for barium
nT ¼
n
ð38Þ from the experimental ðqexp Þ and estimated ðqcal Þ values and the
RT xn0 ln γ n0 þ xn1 ln γ n1 xn01 ln 1 þ Γ s 1M
m1 o results can be shown in the isotherms of Figs. 6 and 7. The results
are for the 3 adsorbents used: Ca-exchange clinoptilolite tuff,
The surface phase capacity Γ is given by Eq. (4). The pore
s
mi
Ca-exchange montmorillonite clay and βMnO2. There is an excel-
volume values of the adsorbents used are listed in Table A2 of
lent agreement between of the experimental and the estimated
Appendix 2. Finally, to compute ðϕ ϕoi Þ, if Eq. (5) is used, then:
n n
! values of q with error values are 5.21e 10 for the first problem
xb γ b and 5.44e 12 for adsorption onto β-MnO2 problem. Appendix 5
φn φnoi ¼ R T Γ smi ln in in ð39Þ
xi γ i presents the Confidence interval calculation for the interaction
parameters.
The nonlinear programming (NLP) optimization solution methods
are used to determine the UNIFAC interaction parameters using
Matlab software. As an example, we present the calculation steps to 6. Conclusion
compute the adsorption capacity for Kaolinite in Appendix 4
(Tables A5 and A6). This is a first paper in the series of two papers to find an
We used the nonlinear least square fit for this purpose. We took optimal adsorbent following computer-aided molecular design
41 experimental points and converted it into a single objective formulation. In this paper, we developed group contribution
function of sum error square between these experimental values methods for barium and radium adsorption. The UNIFAC interac-
and calculated values. We applied nonlinear optimization techni- tion parameters for the groups are computed using the available
que to obtain parameters. As the number of experimental points experimental isotherms for various adsorbents for barium and
are not small and the data was not too noisy (as can be seen from radium and an NLP optimization technique. These parameters are
the errors), we do not have the problem of over fitting here. used to estimate the adsorption capacity at different
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
8 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Table 5
Values of interaction parameter in the adsorption of radium.
Bulk phase
ASS Phase
Note: The values of interaction parameters are presented within their confidence interval CI. The value of the parameter is shown in bold. In some cases there was is no
confidence interval, meaning that the results are very good and have tight bounds.
a
The CI is not estimated since the objective function was insensitive to those parameters.
Fig. 5. Isotherm for adsorption of radium using Kaolinite, Clinoptinolite, Montmorrillonite and Glauconite. Comparison of the experimental adsorption capacity found in the
literature and the calculated adsorption capacity.
concentrations and these parameter values are found to be in very optimization algorithms to find optimal adsorbents for the
good agreement with experimental results. In part 2 of the paper, removal of barium and radium from produced waters from
we discuss the detail of the CAMD formulation and the fracking.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 9
Table 6
Values of interaction parameter in the adsorption of barium onto Ca-Cln and Ca-Mnt and onto β-MnO2.
Bulk phase
ASS Phase
Note: The values of interaction parameters are presented within their confidence interval CI. The value of the parameter is shown in bold. In some cases there was is no
confidence interval, meaning that the results are very good and have tight bounds.
a
The CI is not estimated since the objective function was insensitive to those parameters.
Fig. 6. Isotherm for adsorption of barium onto Ca-Cln and Ca-Mnt. Comparison of the experimental adsorption capacity found in the literature and the calculated adsorption
capacity for barium adsorption.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
10 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
ri parameter for Van der Waal volume calculation Appendix A1. Calculation of van der Waals volume and surface
rw Van der Waals radius ðÅÞ
r1 ; r2 : Var der Waals radii for both atoms ðÅÞ The constants representing the group sizes and surfaces areas
Rg Van der Waals group volumes of group K (m3) Rg and Q g are obtained from atomic and molecular structural data,
R ideal gas constant (8.314 J/mol K) the van der Waals group volume V g , and surface areas Ag , as it is
T temperature of adsorption (K) shown in Eqs. (A1) and (A2) (Fredenslund et al., 1977).
vp pore volume of the adsorbent (cm3/g)
Rg ¼ V g =15:17 ðA1Þ
voi molar volume of the respective fluid (cm3/mol)
Vg Van der Waals volume (cm3/mol)
Q g ¼ Ag =2:5e9 ðA2Þ
V volume (L)
x molar fraction Where V g and Ag are calculated from experimental data, namely,
the so-called van der Waals radius (r w ) of each atom in the
Greek letters molecule representation
Radium (Ra). In the case of radium, a single spherical atom, the
γ activity coefficient van der Waals volume is given by:
γ nGE;i activity coefficient of a binary adsorbate-solid solution 4
φn chemical potential of the wetted solid adsorption of the V g ¼ π r 3w nN A ðA3Þ
3
pure component i (J/g) _ equivalent to 2.3e 8 cm, and N A is the
Where r w ¼ 2:3 A
φnoi chemical potential of the wetted solid adsorption of the
Avogadro number equals 6.023e23moles/molecule, then V g ¼
pure component i (J/g)
30:696 cm3 =mole. For the surface area value, Eq. (A4) is used:
φi volume fraction for UNIFAC calculation
θi surface area fraction for UNIFAC calculation Ag ¼ π r 2w nN A ðA4Þ
Γ smi surface phase capacity of component i (mol/g)
ρ density (g/ml) Then, Ag ¼ 1e9 cm =moles.
2
Superscripts
exp experimental
Table A1
Van der Waals volume and surface values.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 11
(A5) shows the van der Waals volume of diatomic molecule. Appendix A3. Structure and properties of the adsorbents for
radium and barium removal
V g ¼ N A V 11 þV 2 ΔV 2 1 n10 24 cm3 =mole ðA5Þ
Table A3 presents the minerals used by Ames et al. (1983a,
where V 11 , V 2 and ΔV 2 1 are computed by Eqs. (A6)–(A8) and r is 1983b) with their chemical composition and molecular weight in
given in Angstrom units, then: the adsorption of radium. From this table, we can see the number
of functional groups that are contained in each adsorbent. The
h1 first four adsorbents belong to the clay mineral group also known
V 11 ¼ π h1 r 1
2
ðA6Þ
3 as hydrous aluminum phyllosilicates. The word “clays” was
assigned early to fine grained material in geological formations
4
V 2 ¼ π r 32 ðA7Þ or soils with size less than 2 μm (Velve and Meunier, 2008). The
3 clinoptilolite is one of the most abundant natural zeolite. It has
been used for radioactive waste disposal and ammonia recovery
h2 from sewage effluents (Yang, 2003). The last adsorbent is silica
ΔV 2 1 ¼ π h22 r2 ðA8Þ
3 gel which is most-used as a desiccant or adsorbent for moisture
(Yang, 2003). For the chemical analysis of these minerals see
To compute the height of sphere segments, Eqs. (A9) and (A10) Ames et al. (1983a, 1983b). The calculation of the molecular
are used: weight for all clays was based on the chemical structure general
h1 ¼ r 1 þ 1 ma ðA9Þ formula (Deer et al., 1992).
On the other hand, Table A4, presents the chemical structure
and molecular weight of the adsorbents used for adsorption of
h2 ¼ r 2 ma ðA10Þ barium. Clinoptinolite tuff (Ca-Cln) is constituted by clinoptino-
lite–heulandite and mordenite with minor plagioclase, feldspar,
Finally, Eq. (A11) is used to calculate the auxiliary parameter
cristobalite, quartz, and glass. The montmorillonite clay (Ca-Mnt),
ma:
on the other hand, contains 85% of montmorillonite associated
2 with plagioclase, feldspar, glass, opal, and calcite. Beta-manganese
r 22 r 21 þ l
ma ¼ ðA11Þ oxide was also studied for the adsorption of barium. For more
21
information about the adsorbent's composition please refer to
On the other hand, the van der Waals surface area is computed Chavez et al. (2010) and Arafa et al. (1974) The molecular weight is
using Eq. (A12): calculated based on the chemical composition presented by the
authors.
Ag ¼ 2π N A ðr 1 h1 þ2r 22 r 2 h2 Þ ðA12Þ
Once again, after V g and Ag are calculated from the previous Appendix A4. Calculation steps to compute the adsorption
equations, these values are replaced into Eqs. (A1) and (A2), capacity for Kaolinite
respectively. Table A2 summaries the result of these values for
the groups of adsorbent used here.
Table A5 summarizes the important calculations followed to
compute the adsorption capacity (qcal ) for Kaolinite. The table
Table A2
presents the expression used, its value and the equation used to
Information of functional groups of absorbents for calculation of Rg and Q g .
demonstrate one point of Fig. 5 (i.e. when concentration (C) in the
Groups (g) l ðÅÞ r 2 ðÅÞ r 1 ðÅÞ V g (cm3/mol) Ag (cm2/mol) Rg Qg equilibrium is equal to 0.018 mol/L). We used the interaction
parameters for Kaolinite shown in Table A6, which are also
SiO2 1.61 2.1 1.52 21.067 3.167e9 1.388 1.268 presented in Table 5 for all the interaction parameters involved
Al2O3 1.935 1.84 1.52 16.829 2.759e9 1.109 1.104
in the adsorption of radium.
MgO 2.11 1.73 1.52 15.368 2.602e9 1.013 1.041
Fe2O3 2.098 1.94 1.52 18.528 2.918e9 1.221 1.167
K2O 2.86 2.75 1.52 41.198 4.609e9 2.716 1.844
Na2O 2.41 2.27 1.52 25.600 3.517e9 1.688 1.407 Appendix A5. Confidence interval calculation of the
β-MnO2 1.922 1.26 1.52 10.731 2.036e9 0.707 0.815
interaction parameters
(AlO2) 1.641 1.84 1.52 16.673 2.753e9 1.099 1.101
CaO 2.4 2.31 1.52 26.576 3.594e9 1.752 1.438
FeO 2.17 1.94 1.52 18.580 2.920e9 1.225 1.168 Reliability of the parameters is obtained by finding the con-
fidence interval (Rasmuson et al., 2014). The approximate
Table A3
List of secondary mineral used in adsorption of radium and their molecular weight and pore volume values.
Kaolinite (Al2O3)(SiO2)2(H20) 0.278 222.127 Clay-mica group/phyllosilicate Diamond, (1970) and Benco et al. (2001)
Montmorillonite (Na2O) (MgO)2 (Fe3O2) (Al2O3)2 (SiO2)4 2H2O 0.172 786.374 Clay-mica group/phyllosilicate Diamond (1970)
Nontronite (Na2O)(Fe2O3)(Al2O3)2(SiO2)4 2H2O 0.255 665.920 Clay-mica group/phyllosilicate Diamond (1970)
Glauconite (K2O)(MgO)2(Fe2O3)(Al2O3)2(SiO2)4 0.164 778.736 Iron potassium phyllosilicate Hassan and El Shall (2004)
Clinoptilolite (CaO)(Na2O)(K2O)(Al2O3)3(SiO2)30.24H2O 0.186 2320.62 Natural Zeolite Yang (2003)
Silica gel (SiO2) 0.43 60.083 Silica Yang (2003) and Ruthven (1984)
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
12 P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Table A4
Information about the adsorbents used in adsorption of barium.
Clinoptilolite tuff (Ca-Cln) (K þ )(Ca þ 2)(K2O)(Na2O)(CaO) (AlO2)2 (Al2O3)3.1(SiO2)37.9(H2O)24 187.207 Natural zeolite Chavez et al. (2010)
Montmorillonite (Ca-Mnt) (Na2O) (MgO)2 (FeO) (Al2O3)2 (SiO2)4 2H2O 399.104 Clay Chavez et al. (2010)
Beta-Manganese oxide β-Mn2O 86.937 Oxides Arafa et al. (1974)
Table A5
Calculation steps to compute the adsorption capacity for Kaolinite.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i
P.T. Benavides, U. Diwekar / Chemical Engineering Science ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 13
The approximate numerical estimation of the Jacobian is given by Deer, W., Howie, R., Zussman, J., 1992. An Introduction to the Rock-forming
Eq. (A15): Minerals, 2nd ed. Prentice Hall, England.
Diamond, S., 1970. Pore size distributions in clays. Clays Clay Miner. 18, 7–23.
f xn ; θ^ f xn ; θ^ p;1%
Fredenslund, A., Gmehling, J., Rasmussen, P., 1977. Vapor–liquid Equlibria Using
UNIFAC A Group Contribution Method. Elsevier Scientific Publishing Company,
J n;p ¼ ðA15Þ
Δθp Amsterdam.
Hassan, M., El Shall, H., 2004. Glauconitic clay of El Gidida Egypt: evaluation and
The 1 α confidence interval for θ^ p is determined by Eq. (A16) surface modification. Appl. Clay Sci. 27, 219–222.
Joback, K., Reid, R., 1987. Estimation of pure-component properties from group-
θ^ p 7 se θ^ p t n p; α=2 ðA16Þ contributions. Chem. Eng. Commun. 57, 233–243.
Kikic, I., Fermeglia, M., 1991. UNIFAC prediction of vapor–liquid equilibria in mixed
solvent-salt systems. Chem. Eng. Sci. 46, 2775–2780.
where t n p; α=2 is the Student's
t distribution with n p Lydersen., 1955. Estimation of Critical Properties of Organic Compounds. College
degrees of freedom and se θ^ p is the standard error for parameter Enineering University Wisconsin. Engineering Esperimental Station Report 3.
calculated using Eq. (A17) Madison.
Lustgarten., 2009. Frack Fluid Spill in Dirmock Contaminantes Stream, Killinf Fish.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
Propublica. 〈www.propublica.org/article/frack-fluid-spill-in-dimock-contaminantes-
se θ^ p ¼ s JT J ðA17Þ stream-killing-fish-921〉.
pp Marcellus Shale., 2011. Understanding naturally occurring radioactive material in
the Marcellus Shale. Issue 4 〈http://www.museumoftheearth.org/files/marcel
where s is the standard deviation obtained from the objective lus/Marcellus_issue4.pdf〉.
Marrero, J., Gani, R., 2001. Group-contribution based estimation of pure component
function SSE
properties. Fluid Phase Equilibria 183, 183–208.
sffiffiffiffiffiffiffiffiffiffiffi Purdey, M., 2004. Chronic barium intoxication disrupts sulphated proteoglycan
SSE synthesis: a hypotesis for the origins of multiple sclerosis. Med. Hypotheses 62,
s¼ ðA18Þ
np 746–754.
Rasmuson, A., Andersson, B., Olsson, L., Andersson, R., 2014. Mathematical Model-
For some cases the confidence interval is very tight due to very ing in Chemical Engineering. Cambridge University Press, United Kingdom.
Replica Engineering INC., 2014. 〈http://www.replicaeng.com/Replica-Engineering/
small hessian. For cases where the Hessian is very large, the Hydraulic-Fracturing/Marcellus-Shale-Formation.htm〉.
confidence interval is very large and the objective function is Rowan, E., Engle, M., Kirby, C., Kraemer, T., 2011. Radium Content of Oil- and Gas-
insensitive to the parameter changes. Field Produced Waters in the Northern Appalachian Basin (USA)-Summary and
Discussion of Data. Scientific investigations Report. U.S. Geological Survey,
Reston, Virginia.
References Ruthven, D., 1984. Principles of Adsorption and Adsorption Processes. John Wiley &
Son, Canada.
Silva, J., Matis, H., Kostedt, W., Watkins, V., 2012. Produce Water Preatrement for
Ames, L., McGarrah, J., Walker, B., 1983a. Sorption of trace constituents from Water Recovery and Salt Production. General Electric Global Research Center,
aqueous solutions onto secundary minerals. Radium ll. Clays Clay Miner. 31, Niskayuna, New York.
335–342. Smith, T., 2012. Environmental considerations of shale gas development. Chem.
Ames, L., Mc.Garrah, J., Walker, B., 1983b. Sorption of trace constitutes from Eng. Process. 108, 53–58.
aqueous solutions onto secondary minerals. l uranium. Clays Clay Miner. 31, Suzuki, M., 1990. Adsorption Engineering. Elsevier Science Publishing Company,
321–334. INC, New York.
Arafa, M., Yousef, A., Malati, M., 1974. Adsorption of barium ions by Beta_Manga- U.S. Department of Energy Website, 2014a. E. I. US Energy Information Adminis-
nese dioxide and its activation in oleate flotation. Int. J. Miner. Process. 1, tration. DOE-EIA. http://www.eia.gov/naturalgas/.
267–275. U.S. Department of Energy Website, 2014b. E. I. Annual Energy Outlook 2014, Early
Benco, L., Tunega, D., Hafner, J., Li, H., 2001. Upper limit of the O–H–O hydrogen Release Overviwe. Retrieved from U.S. Department of Energy, Energy Informa-
bond. Ab initio study of the Kaolinite structure. J. Phys. Chem. 105, tion Administrative. 〈http://www.eia.gov/forecasts/aeo/er/pdf/0383er(2014).
10812–10817. pdf〉.
Benavides, P., Diwekar, U., Gebreslassie, B., 2015. Optimal design of adsorbents for U.S.EPA., 2014. Environmental Protection Agency, EPA. 〈http://www.epa.gov/safe
NORM removal from produced water in natural gas fracking. Part 2: CAMD for water/agua/estandares.html〉.
adsorption of radium and barium Submitted for publication. Chem. Eng. Sci.. Velve, B., Meunier, A., 2008. The Origin of Clay Minerals and Soils and Weathered
Berti, C., Ulbig, P., Burdorf, J., Seippel, J., Schulz, S., 1999. Correlation and prediction Rocks. Springer-Verlag, Berlin Heidelberg, Paris.
of liquidi-phase adsorption on zeolites using group contribution methods Warner, N., Christie, C., Jackson, R.B., Vengosh, A., 2013. Impacts of shale gas
based on the adsorbate-solid solution theory. Langmuir 15, 6035–6042. wastewater disposal on water quality in western Pennsylvania. Environ. Sci.
Berti, C., Ulbig, P., Schulz, S., 2000. Correlation and prediction of adsorption from Technol. 47, 11849–11857.
liquid mixtures on solids by use of GE-Models. Adsorption 6, 79–91. Yang, P., 2003. Adsorbents: Fundamentals and Applications. John Wiley & Sons,
Bondi, A., 1968. Physical Properties of Molecular Crystals, Liquids, and Glass. John Hoboken, New Jersey.
Wiley & Sons, New York. Zielinski, R., Otton, J., 1999. Naturally Ocurring Radioactive Materials (NORM) in
Chavez, M., Pablo, L., Garcia, T., 2010. Adsorption of Ba þ 2 by Ca-exchange produce water and oil-field equiment- an issue for the energy industry. USGS,
clinoptilolite tuff and montmorillonite clay. J. Hazard. Mater. 175, 216–223. science for a changing world. Geological Survey. Denver, CO.
Please cite this article as: Benavides, P.T., Diwekar, U., Optimal design of adsorbents for NORM removal from produced water in natural
gas fracking. Part 1: Group contribution method for adsorption. Chem. Eng. Sci. (2015), http://dx.doi.org/10.1016/j.ces.2015.07.012i