Chemical Engineering Journal 343 (2018) 244–257

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Reaction kinetic study of elemental mercury vapor oxidation with CuCl2 T

⁎
Vishnu Sriram, Can Li, Zhouyang Liu, Mina Jafari, Joo-Youp Lee
Chemical Engineering Program, Department of Chemical and Environmental Engineering, University of Cincinnati, Cincinnati, OH 45221-0012, United States

H I G H L I G H T S G RA P H I C A L AB S T R A C T

• Kinetics for heterogeneous Hg(0) oxi-

dation with CuCl was studied for the
2
first time.
• CuCl2 over α-Al2O3 was not com-
pletely used for Hg(0) oxidation due to
agglomeration/sintering.
• Grain model under agglomeration/
sintering effect was used to determine
rate constant.
• Activation energy was much lower
than those for other Hg(0) oxidation
catalysts.
• CuCl2 significantly enhances Hg(0)
oxidation by reduction of Cu(2+) to
Cu(1+).

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, the reaction kinetics for a heterogeneous oxidation reaction of elemental mercury (Hg(0)) vapor
Reaction kinetics with CuCl2 was studied in a fixed-bed reactor using 2%(wt) CuCl2/α-Al2O3 between 100 and 180 °C for Hg(0)
Heterogeneous elemental mercury oxidation oxidation after air preheater at a typical coal-fired power plant. The reaction rate expression was first order with
reaction respect to Hg(0). However, between 100 and 180 °C, CuCl2 over α-Al2O3 agglomerates and sinters. This sintering
Cupric chloride
effect added significant mass-transfer resistance to the diffusion of Hg(0) vapor, and thus made the conversion of
Mercury emissions control
Grain model
CuCl2 incomplete. Therefore, a grain model was formulated to determine the rate constant by taking into ac-
count the mass-transfer resistance. The model constituted a two parameter estimation problem for the de-
termination of the rate constant and product layer diffusivity. The model predictions with the two optimum
parameters were in good agreement with the experimental data. The activation energy value determined from
the rate constant values was significantly lower than those for other Hg(0) oxidation catalysts under HCl and O2
gases reported in the literature. This result corroborates that CuCl2 can enhance Hg(0) oxidation by lowering the
activation energy barrier with the reduction of Cu(2+) to Cu(1+) and supplying thermally stable surface Cl sites
following a Mars-Maessen mechanism. CuCl2-based catalyst has potential to be applied after air preheater for Hg
(0) oxidation followed by the separation of Hg(2+) in wet flue gas desulfurization (FGD) system or by activated
carbon (AC) injection.

1. Introduction and Air Toxics Standards (MATS) rule for coal- and oil-fired electric
utility generating units (EGUs) [1]. It is important to design mercury
Mercury, a hazardous air pollutant (HAP), is regulated under the separation technologies using homogeneous and heterogeneous Hg(0)
United States Environmental Protection Agency (U.S. EPA)’s Mercury oxidation reactions for subsequent capture of the resultant and existing

⁎
Corresponding author.
E-mail address: joo.lee@uc.edu (J.-Y. Lee).

https://doi.org/10.1016/j.cej.2018.02.127
Received 22 December 2017; Received in revised form 5 January 2018; Accepted 28 February 2018
Available online 01 March 2018
1385-8947/ © 2018 Elsevier B.V. All rights reserved.
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Nomenclature product layer, m

‴
−rHg Hg(0) oxidation rate based on particle cluster volume used
CHg,b Hg(0) concentration in the bulk gas phase, gmol Hg(0)/m3 in general model, gmol Hg(0) reacted/(m3 particle clus-
in ter⋅s)
CHg ,b inlet Hg(0) concentration in the bulk gas phase, gmol Hg
(0)/m3 ″
−rHg Hg(0) oxidation rate based on grain surface area used in
CHg,p Hg(0) concentration in the particle cluster phase used in grain model, gmol Hg(0) reacted/(m2 grain⋅s)
general model, gmol Hg(0)/m3 ‴ ,obs
−rHg observed Hg(0) oxidation rate based on particle cluster
CHg,pdt Hg(0) concentration inside the product layer, gmol Hg(0)/ volume used in general model, gmol Hg(0) reacted/(m3
m3 particle cluster⋅s)
CCuCl2 CuCl2 concentration inside particle cluster phase used in ″ ,obs
−rHg observed Hg(0) oxidation rate based on grain surface area
general model, gmol CuCl2/m3 used in grain model, gmol Hg(0) reacted/(m2 grain⋅s)
CCuCl2,0 initial CuCl2 concentration inside particle cluster phase, CuCl2 reaction rate based on grain surface area used in
gmol CuCl2/m3 grain model, gmol CuCl2 reacted/(m2 grain⋅s)
D dispersion of CuCl2 over α-Al2O3, % ‴ 2
−rCuCl CuCl2 reaction rate based on particle cluster volume used
De effective pore diffusion coefficient for Hg(0) vapor be- in general model, gmol CuCl2 reacted/(m3 particle clus-
tween the pores of the particles inside the cluster, m2/s ter⋅s)
Deq limiting dispersion at equilibrium, % t time, s
Dm molecular diffusion coefficient for Hg(0) vapor, m2/s uz superficial gas velocity, m/s
D0 initial dispersion, % XCuCl2 conversion of CuCl2
DS product layer diffusion coefficient for Hg(0) vapor used in z axial distance inside fixed-bed reactor, m
grain model, m2/s
k‴ oxidation rate constant based on particle cluster volume Greek letters
used in general model, m3/(g·s)
k″ oxidation rate constant based on grain surface used in εb bed porosity, –
grain model, m/s εp particle porosity, –
kg gas-phase mass-transfer coefficient, m/s ρCuCl2 CuCl2 solid density, gmol/m3
ks sintering rate constant, 1/s δ dimensionless dispersion, see Eq. (18)
L height of fixed-bed reactor, m δeq dimensionless dispersion at infinite time, see Eq. (18)
m sintering order ξ dimensionless radial distance inside particle, see Eq. (18)
Rep Reynolds number ξC dimensionless radial distance inside product layer, see Eq.
Rg radius of CuCl2 grain, m (18)
Rp radius of CuCl2/α-Al2O3 particle cluster, m γ dimensionless Hg(0) concentration in product layer, see
Sc Schmidt number Eq. (18)
Sh Sherwood number γB dimensionless Hg(0) bulk concentration, see Eq. (18)
r radial distance inside particle cluster, m ζ dimensionless axial distance inside fixed-bed reactor, see
rC reaction interface between unreacted CuCl2 core and CuCl Eq. (18)

oxidized gaseous mercury species (Hg(2+)) in order to comply with the kinetics over CuCl2/α-Al2O3 was studied using gas-solid reaction
MATS rule. Presently, Powdered Activated Carbon (PAC) sorbent in- models. A temperature window between 100 and 180 °C was used for
jection and Selective Catalytic Reduction (SCR) catalysts followed by potential applications after an air preheater at a typical coal-fired
wet Flue Gas Desulfurization (FGD) system are used to reduce mercury power plant. Both the general model and grain model were used to
emissions in the post-combustion window of the U.S. coal-fired power describe the reaction between Hg(0) vapor and CuCl2 for the determi-
plants [2–4]. nation of a reaction kinetic expression.
Several models have been proposed to describe various hetero- In our previous studies, cupric chloride (CuCl2)/α-Al2O3 demon-
geneous gas–solid reactions with different reaction mechanisms in- strated excellent Hg(0) oxidation capability, and its reaction me-
cluding the general model, unreacted shrinking core model, and grain chanism was studied [13]. Low surface area α-Al2O3 was used because
model [5–8]. The general model developed by Wen consists of a simple it did not change the speciation of CuCl2 [13]. It was found that CuCl2
model that takes into account external and internal mass-transfer re- reacts with Hg(0) via a Mars-Maessen mechanism to form CuCl and
sistances and chemical reaction [6,9]. The unreacted shrinking core HgCl2, respectively, as a result of the reaction as shown in Eq. (1) [13].
model is extensively used for the reactions with solids of low porosity, However, a kinetic study of the Hg(0) oxidation reaction with CuCl2 has
thereby forming ash layer or product layer around the unreacted core of not been reported in the literature.
solid reactant. As the reaction proceeds at a sharp interface between the
Hg(0) + 2CuCl2 → HgCl 2 + 2CuCl (1)
reactant gas diffused through the product layer and the unreacted solid
core, the unreacted core shrinks and the product layer thickness in- Before the kinetic study, the equilibrium conversion of Hg(0) was
creases. External mass-transfer resistance around the solid, internal calculated at the reaction temperatures (i.e. 20, 100, 140, 180 °C) using
mass-transfer resistance inside the solid and the chemical reaction are Eq. (2).
considered in this model. On the other hand, grain model, developed by
ΔG ∘ K eq,1 ΔH ∘ 1 1
Sohn and Szekely, is used for porous solid [8,10]. The solid is con- lnK eq = − RT ; ln K =− R
(T −T )
eq,2 1 2
sidered to contain small non-porous grains at a microscopic level. The where
shrinking core mechanism is applied to the grain phase forming the y HgCl ,eq
2 ( n HgCl2,0 + ξν HgCl2 )
solid. Another predominant phenomenon taking place in gas-solid re- K eq = yHg (0),eq
= (nHg (0),0 − ξνHg (0) )
actions is thermal sintering leading to deactivation of solids [11]. A loss nHg (0),0 − (nHg (0),0 − ξνHg (0) )
Equilibrium conversion of Hg(0) =
in activity due to a decrease in active surface sites is common for re- nHg (0),0 (2)
actions at high temperatures [12]. In this study, the Hg(0) oxidation
where ξ is the extent of reaction; yHg,eq is the equilibrium mole fraction

245
V. Sriram et al.
of Hg(0) vapor; y HgCl2,eq is the equilibrium mole fraction of HgCl2; K eq is
the equilibrium constant; ΔH is the enthalpy of reaction; and R is the
universal gas constant; n HgCl2,0 is the number of moles of HgCl2 initially
present in the system; nHg (0),0 is the number of moles of Hg(0) initially
present in the system; ν HgCl2 and νHg are the stoichiometric coefficients
(i.e. 1). Table 1 shows the calculated thermodynamic values. Standard
free energy values were obtained from NIST-JANAF tables [14]. The
equilibrium conversion was calculated to be 100% using Eq. (2), sug-
gesting complete CuCl2 conversion at equilibrium in the temperature
range.
2. Experiment
2.1. Preparation of CuCl2/α-Al2O3
In this study, α-Al2O3 was used as a support since it neither changes
the CuCl2 speciation nor contains potential impurities that can react
with Hg(0) vapor [13]. The use of the α-Al2O3 support helped disperse
CuCl2 and thus reduce its amount required to study the reaction ki-
netics. Low surface area α-Al2O3 pellets were purchased from Alfa
Aesar (Alfa Aesar aluminum oxide-43862, 1/8″ pellets, measured BET
surface area = 0.79 m2/g). The pellets were crushed and passed
through a 40-μm sieve. The powdered α-Al2O3 was impregnated with
CuCl2⋅2H2O (Sigma-Aldrich, > 98% purity) using wetness incipient
method by adding a known amount of aqueous solution of CuCl2⋅2H2O.
The loading used on the support was set to 2%(wt) CuCl2. After the
impregnation, the sample was left to dry at room temperature over-
night, and was subsequently used for different characterizations and
performance tests.
2.2. CuCl2/α-Al2O3 characterizations
Thermal gravimetric analysis-mass spectrometry (TGA, TA
Instruments TGA Q5000IR and MS, Pfeiffer-Vacuum Thermostar) was
used to check the thermal stability of the sample. The 2%(wt) CuCl2/α-
Al2O3 sample was heated from room temperature to 180 °C at a ramp
rate of 1 °C/min in 25 mL/min of N2 flow (99.999% UHP, Wright
Brothers, Inc.). After the temperature reached 180 °C, the sample was
kept isothermally for 4 h. The amount of sample used was ∼60 mg.
In order to examine the sample morphology, scanning electron
microscopy (SEM, FEI, XL30) was used. The samples were pretreated by
heating them at the respective reaction temperatures and removing
them from the furnace at different times in order to view the mor-
phological changes with respect to temperature and time. Before
viewing these samples under the SEM, the samples were sputter coated
with a thin layer of gold. The samples were viewed in back-scattering
mode in order to obtain atomic number contrast for the differentiation
between CuCl2 grains and α-Al2O3 particles. Focused ion beam scan-
ning electron microscopy (FIB/SEM, FEI) was also performed on fresh
and spent 5%(wt) CuCl2/α-Al2O3 to investigate the porosity change in
the samples. Gallium ion was used to mill the cross-section of the
sample. A thin layer of platinum was deposited on top of the samples
before it was milled, to protect it from gallium ions.
Separate chemisorption tests were carried out in order to determine
the dispersion kinetics for CuCl2 over the α-Al2O3 support with respect
to temperature [15]. The samples were kept at different reaction tem-
peratures (i.e. 100, 140, and 180 °C) in N2 flow. Then ∼200 mg of a
sample were taken out for chemisorption tests. The chemisorption tests
were performed by static-volumetric carbon monoxide (CO) adsorption
measurements using Micromeritics ASAP 2020. CO was found to have
excellent affinity to Cu(I) sites. It was reported that Cu(I)–CO bonds are
stronger than Cu(II)–CO interactions [15]. Therefore, it was necessary
to reduce CuCl2 to CuCl in order to determine its dispersion over the α-
Al2O3 support. Ethylene was used to reduce CuCl2 to CuCl as used for
ethylene oxychlorination reactions in the literature [16]. The samples
were heated to 110 °C for 15 min to remove any potential moisture
246
Chemical Engineering Journal 343 (2018) 244–257
before the tests. Then the samples were reduced using ethylene gas at
200 mmHg for 30 min at 75 °C followed by evacuation for 15 min. This
reduction step was repeated three times to ensure that CuCl2 was
completely reduced to CuCl. Then an evacuation process was performed
by pulling a vacuum at 10 µm Hg to desorb ethylene molecules po-
tentially adsorbed onto Cu sites at 35 °C for 15 min followed by CO
chemisorption at 35 °C. Volume adsorbed by the sample was measured
at a pressure range of 50–500 mmHg. Some physically adsorbed CO gas
molecules were removed by evacuating the sample after the first ana-
lysis. This was followed by repeating the CO adsorption. Thus, the first
analysis gave the volume adsorbed derived from both chemical and
physical adsorption while the repeated analysis gave the volume ad-
sorbed as a result of physical adsorption only. The dispersion values
were determined by calculating the volume intercept derived from the
differences between the first and repeated analysis obtained from the
isotherms.
X-ray photoelectron spectroscopy (XPS) depth profiling was carried
out using a Kratos Axis Ultra XPS spectrometer using monochromatic Al
Kα radiation equipped with a monoatomic argon ion beam for etching
at the Ohio State University. In order to use argon gas cluster ion beam
for etching, another XPS depth profiling experiment was performed on
a PHI Versa Probe III XPS system (ULVAC-PHI) using a monochromatic
Al Kα radiation at the University of Minnesota. Carbon 1s peak at
284.5 eV was used for calibration.
2.3. Hg(0) breakthrough curves in fixed-bed system
For the Hg(0) oxidation kinetic study, Hg(0) breakthrough curves
were obtained using a fixed-bed system shown in Fig. 1. Different inlet
Hg(0) concentrations (5, 10, and 20 ppbv) were used to test the sample
at different reaction temperatures of 20, 100, 140, and 180 °C. Ele-
mental mercury vapor was generated by flowing 1 L/min of N2
(> 99.999% ultra-high purity, Wright Brothers, Inc.) gas through an Hg
(0) permeation tube (Dynacal, VICI Metronics) immersed in an oil bath.
A temperature controller (Fisher Scientific Isotemp 2100) was used to
control the temperature of the oil bath within ± 0.1 °C of a set point
temperature. For each run, 25 mg of the CuCl2/α-Al2O3 sample mixed
with 4 g of silica sand was added to the fixed-bed reactor made of
quartz. The reaction zone was measured to be ∼20 mm. At the outlet of
the reactor, the concentrations of Hg(0) and Hg(2+) were measured
using the Ontario Hydro method [17]. A combination of 1 M KCl fol-
lowed by 4%(w/v) KMnO4/10%(v/v) H2SO4 impinger solutions were
used to capture the Hg(2+) and Hg(0), respectively. Then, both of the
mercury species were quantified using a cold vapor atomic absorption
spectrophotometer (Model 400A, Buck Scientific, Inc.). For each ex-
perimental run, a typical total mercury mass balance was ∼85% at the
start of the reaction while it increased to 95–100% as the reaction
proceeded.
2.4. CuCl2 conversion
Since the amount of Hg(0) reacted can be determined from the
breakthrough curves, the amount of CuCl2 reacted can also be calcu-
lated based on the reaction stoichiometry in Eq. (1) at a given
Table 1
Calculated thermodynamic values for reaction between CuCl2 and Hg(0) at different re-
action temperatures.
Reaction ΔHr (kJ/ ΔG (kJ/gmol) Equilibrium Equilibrium
temperature gmol) constant (Keq) conversion (%)
(°C)
20 −75.19 −94.87 8.03 × 1016 100
100 −72.26 −101.05 1.40 × 1014 100
140 −72.25 −104.14 1.47 × 1013 100
180 −72.09 −107.24 2.30 × 1012 100
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Fixed-bed reactor
0
Hg
permeation H CuCl2/α-Al2O3
g
tube

N2: 100 ml/min

Temperature-
controlled furnace
N2: 900 ml/min Hood

On-line Hg
analyzer
Bypass
(AAS)

1 M KCl
4%(w/v) KMnO4/
Solution
10%(v/v) H2SO4
Solution

4%(w/v) KMnO4/
10%(v/v) H2SO4
Solution
(Impingers are used for Hg speciation)

Fig. 1. Schematic of experimental set-up for Hg(0) vapor oxidation reaction.

temperature and an inlet Hg(0) concentration. The amount of Hg(0) CuCl2 reacted, g mol
CuCl2 conversion = ⎜⎛ ⎞
⎟ × 100
reacted with CuCl2 was calculated by taking the difference between a CuCl initally placed in reactor, g mol ⎠
⎝ 2
total amount of Hg(0) introduced to the reactor and the area under the
(3)
breakthrough curve. CuCl2 conversion was determined at a given re-
action temperature and an inlet Hg(0) concentration by calculating the
amount of Hg(0) reacted with a known amount of CuCl2 placed in the 3. Model for Hg(0) oxidation reaction kinetics
fixed-bed reactor based on the molar ratio 2 of CuCl2 to Hg(0). Then the
molar conversion of CuCl2 during the reaction was determined as 3.1. Model theory and assumptions
shown in Eq. (3).
In this study, two models were used to describe the reaction kinetics

Fig. 2. Schematic of (a) general model and (b) grain model.

247
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

for the reaction between Hg(0) and CuCl2 at low and high tempera- ∂CHg,b ∂CHg,b
εb = −uz −(1−εb)(−r‴
Hg,obs)
tures. It is based on our experimental observations that the agglom- ∂t ∂z (6)
eration of CuCl2 grains is not noticeable at 20 °C but becomes sig-
where −rHg‴ ,obs is the observed Hg(0) oxidation rate based on particle
nificant between 100 and 180 °C. More detailed explanations are given
cluster volume (gmol Hg(0) reacted/(m3 particle cluster⋅s)); CHg,b is the
in the Results and Discussion section. Two schematics for the different
Hg(0) concentration in the bulk gas phase (gmol Hg(0)/m3); uz is the
models are shown in Fig. 2. For the Hg(0) oxidation reaction at 20 °C,
superficial gas velocity (m/s); and εb is the bed porosity. The observed
the general model assuming no sintering of CuCl2 grains was used to
reaction rate was taken by considering a volume integral of Hg(0)
describe the reaction [6,9]. External and internal mass-transfer re-
oxidation rate over the entire CuCl2/α-Al2O3 particle cluster phase and
sistances were taken into consideration in this model. On the other
is given as
hand, the Hg(0) oxidation reaction at higher temperatures (i.e. 100,
140 and 180 °C) illustrates the combined effects of simultaneous Hg(0) ∫V (−rHg
‴ ,int ) dV 3 ∂CHg,p
‴ ,obs ) =
(−rHg = De
oxidation reaction with CuCl2 and sintering of CuCl2 grains. The mass-
transfer resistance related to the product layer diffusion (i.e. Hg(0)
(
V = 3 πRp3
4
) Rp ∂r (z ,r = Rp,t )
(7)
diffusion through reacted CuCl/α-Al2O3 layer) and the chemical reac- ‴ ,obs ) derived from the local
By coupling the observed reaction rate (−rHg
tion at the reaction interface was taken into account. It is assumed that Hg(0) concentration inside the pores (CHg,p ) with Eq. (6), the con-
each grain follows a shrinking core mechanism [10,18,19]. In this centration of Hg(0) in the bulk phase (CHg,b ) can be determined.
study, the Hg(0) oxidation reaction with CuCl2 in the fixed-bed reactor A set of initial conditions (ICs) and boundary conditions (BCs) for
was modeled using the following assumptions: Eqs. (4)-(6) are given as
ICs:
1. The gas flow in the fixed-bed system is considered to be plug flow.
CHg,b (z ,t = 0) = 0; CHg,p (z ,r ,t = 0) = 0; CCuCl2 (t = 0) = CCuCl2,0 (8)
2. Axial diffusion along the fixed-bed reactor is negligible.
3. The α-Al2O3 particles and CuCl2 grains are considered to be sphe- BCs:
rical and uniform in size.
∂CHg ,p
4. The CuCl2/α-Al2O3 particles are uniformly dispersed in the fixed- in
CHg,b (z = 0,t ) = CHg ,b; De ∂r
= 0;
z ,r = 0,t
bed reactor.
∂CHg ,p
5. The pressure drop across the fixed-bed system is negligible. De ∂r
= k g (CHg,b (z ,t )−CHg,p (z ,Rp,t ))
6. The reaction is considered to be isothermal.
z ,r = Rp,t (9)
It is worth mentioning that the external mass transfer between the
3.2. Model equations particle and bulk gas phases was taken into account in one of the BCs.

3.2.1. General model 3.2.2. Grain model

As shown in Fig. 2(a), CuCl2 was assumed to be uniformly coated
over the α-Al2O3 substrate in the general model. The general model is a
lumped model formulated to describe the CuCl2/α-Al2O3 particle
cluster phase comprising multiple CuCl2/α-Al2O3 particles. Inter-par-
ticle diffusion between the pores of CuCl2/α-Al2O3 particles inside a
cluster (please see Fig. 2(a)) was taken into account. Based on the re-
action chemistry shown in Eq. (1), the reaction kinetics was further
assumed to be first order with respect to Hg(0) and CuCl2 as shown in
Eq. (4). Therefore, the reaction rate expression for Hg(0) oxidation is
given as
dCHg,p 1
‴)=−
(−rHg = k‴CHg,p CCuCl2 = ‴ 2)
× (−rCuCl
dt 2 (4)
where −rHg‴ is the Hg(0) oxidation rate based on particle cluster volume
(gmol Hg(0) reacted/(m3 particle cluster⋅s)); −rCuCl
‴ 2 is the reaction rate
based on particle cluster volume (gmol CuCl2 reacted/(m3 particle
cluster⋅s)); k‴ is the oxidation rate constant based on particle cluster
volume (m3/(g·s)); CHg,p is the Hg(0) concentration inside the particle
cluster phase (gmol Hg(0)/m3); and CCuCl2 is the CuCl2 concentration
inside the particle cluster phase (gmol CuCl2/m3).
To describe the inter-particle diffusion of Hg(0) between the pores
of the CuCl2/α-Al2O3 particles inside the cluster and the reaction of Hg
(0) with CuCl2 over the support (refer to schematic in Fig. 2(a)), a shell
mole balance for Hg(0) was formulated over the entire CuCl2/α-Al2O3
particle cluster phase as shown in Eq. (5).
∂CHg,p 1 ∂ ⎛ 2 ∂CHg,p ⎞
εp = ⎜ r De ‴)
−(−rHg
⎟ 2
∂t r 2 ∂r ⎝ ∂r ⎠ (5)
where De is the effective pore diffusion coefficient for Hg(0) vapor
between the pores of the CuCl2/α-Al2O3 particles inside the cluster
(m2/s); and εp is the particle porosity.
Another shell mole balance for Hg(0) was used to describe the
change in Hg(0) concentrations as a result of the reaction along the
length of the reactor, and is given in Eq. (6).
The grain model was formulated so that the effects of CuCl2 sin-
tering and Hg(0) oxidation reaction were taken into consideration. As
shown in Fig. 2(b), each CuCl2 grain forms a product layer around itself
as Hg(0) vapor diffuses and reacts with CuCl2, thereby forming CuCl
between 100 and 180 °C. Then diffusing Hg(0) reacts with CuCl2 at the
interface between the product layer comprising CuCl and the unreacted
CuCl2 core (at r = rc). At the beginning of the Hg(0) reaction with pure
CuCl2, the reaction interface (rc) is the same as the grain radius (Rg). As
the reaction proceeds, the reaction interface retreats toward the center
of the grain (i.e. rc < Rg) throughout the reaction. Therefore, this
model constitutes a moving boundary problem. Thus, as the reaction
proceeds, the mass-transfer resistance to Hg(0) diffusion through the
product layer increases with an increase in the thickness of the product
layer.
An Hg(0) oxidation kinetic expression is written in first order with
respect to Hg(0) as shown in Eq. (10). In the grain model, CuCl2 is
assumed to be available only at the reaction interface, and thus the
reaction kinetics is also dependent on CuCl2 available only at the re-
action interface. This dependence can be shown by taking a shell mole
balance for CuCl2 around the unreacted CuCl2 shrinking core as shown
in Eq. (10). These two rate expressions are stoichiometrically related,
and thus the reaction kinetics for Hg(0) shows the dependence of CuCl2
in terms of the radius of the unreacted CuCl2 shrinking core. It is also
worthwhile to mention here that this kinetic expression should be in-
cluded as one of the boundary conditions (BCs).
1 1 dr
″ ) = k ″CHg,pdt |r = rC =
(−rHg ″ 2 ) = − ρCuCl2 C
(−rCuCl
2 dt (10)
where CHg,pdt is the concentration of Hg(0) inside the product layer
(gmol Hg(0)/m3); ρCuCl2 is the density of CuCl2 (gmol/m3); k″ is the
reaction rate constant (m/s); and rC is the reaction interface between
the unreacted CuCl2 core and the CuCl product layer (m).
The sintering kinetics for grain sintering and agglomeration (details
are given in the Results and Discussion section) is expressed by CuCl2
dispersion, and is given by a general power law expression in Eq. (11)

248
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

[11,12]. model, most mass-transfer resistance would exist in the product layer
where sintering takes place throughout the reaction. Thus the external
dD
= −kS (D−Deq)m mass-transfer resistance between the grain and bulk gas phases was
dt (11)
neglected as shown in the BCs. The profiles for rC(t), D(t), CHg,pdt(z, r, t)
where D is dispersion (%); kS is the sintering rate constant (1/s); m is and CHg,b(z, t) are obtained by solving eqs. (10)-(13) with ICs and BCs
the sintering order; and Deq is the limiting dispersion at equilibrium in eqs. (15) and (16) and the two unknown parameters of k″ and Ds.
(%). Therefore, this model constitutes a two parameter optimization pro-
A shell mole balance for Hg(0) is taken to describe the diffusion of blem solving for k″ and Ds by comparing the model predictions with the
Hg(0) through the product layer of the grain phase for the Hg(0) re- experimental breakthrough data. Upon the completion of an iterative
action with CuCl2 taking place at the interface between the CuCl2 solution process for k″ and Ds, a temporal profile for the reaction in-
shrinking core and the CuCl product layer as shown in Eq. (12). terface (rc(t)) is also determined.
The conversion of the solid-phase CuCl2 (XCuCl2) is calculated by
∂CHg,pdt 1 ∂ ⎛ 2 ⎛ D−Deq ⎞ ∂CHg,pdt ⎞
= r DS ⎜ taking the volume integral of the local conversion inside the grain over
r 2 ∂r ⎜ ⎟
⎟
∂t ⎝ D0−Deq ⎠ ∂r ⎠ (12) the entire reactor length (L) as shown in Eq. (17).
⎝
where DS is the product layer diffusion coefficient (m2/s); and D0 is the 3
1 L ⎛ ⎛ rc ⎞ ⎞
initial dispersion (%). The sintering kinetics for CuCl2 grains in Eq. (11) XCuCl2 =
L
∫0 ⎜1−⎜ Rg ⎟ ⎟ dz
is incorporated into Eq. (12) to describe the increasing diffusional re- ⎝ ⎝ ⎠ ⎠ (17)
sistance generated by the product layer in the grain phase as the re-
action proceeds.
Another shell mole balance is formulated to describe the Hg(0) 3.3. Normalization
concentration profile along the length of the reactor as shown in Eq.
(13). The set of the governing equations, initial and boundary conditions
given in eqs. (10)-(13), (15) and (16) have been normalized using the
∂CHg,b ∂CHg,b A following set of dimensionless variables.
εb = −uz ″ ,obs )
−(1−εb)(−rHg
∂t ∂z V (13)
rC D Deq
ξC = , δ= , δeq = ,
where −rHg ″ ,obs is the observed Hg(0) oxidation rate based on grain Rg D0 D0

surface area (gmol Hg(0) reacted/(m2 grain⋅s)); uz is the superficial gas ξ=

r
, γ=
CHg,pdt
, ζ=
z
, γB =
CHg ,b
Rg in
CHg L in
CHg
velocity (m/s); A is the surface area of grain; V is the volume of grain; ,b ,b (18)
and εb is the bed porosity. Taking a volume average over the local Hg(0)
A set of the normalized governing eqs. is given as:
oxidation rate (−rHg ‴ ) over the entire grain volume gives the observed
oxidation rate (−rHg ‴ ,obs ) at the external surface of the grain. ∂ξC in
″ ) = 2(−rCuCl
(−rHg ″ 2 ) = ρB Rg = −2k ″γ|ξ = ξC CHg ,b
∂t (19)
A ∫ (−rHg
‴ ) dV
″ ,obs )
(−rHg ‴ ,obs ) = V
= (−rHg
V 4
V = 3 πRg3 ( ) ∂δ
= −kS (δ −δeq)m
(20)
∂t
3 D−Deq ⎞ ∂CHg,pdt
= DS ⎛⎜ ⎟
Rg ⎝ D0−Deq ⎠ ∂r (z ,r = Rg ,t ) (14) ∂γ ∂ ⎛ 2 (δ −δeq) ∂γ ⎞
Rg2 ξ 2 = ⎜ξ DS ⎟
∂t ∂ξ ⎝ (1−δeq) ∂ξ ⎠ (21)
where Rg is the radius of each grain; −rHg ‴ ,obs is the observed Hg(0)
oxidation rate based on grain volume (gmol Hg(0) reacted/(m3
∂γB u ∂γ 3D (δ −δeq) ∂γ
grain⋅s)). The grain radius (Rg) was assumed to be constant since the εB = − Z B −(1−εB ) 2S
∂t L ∂ζ Rg (1−δeq) ∂ξ (22)
grain size did not considerably change when SEM images were taken (ζ ,ξ = 1,t )

after 1 h and 24 h as shown in Fig. 5. The observed oxidation rate given

ICs:
in Eq. (14) is coupled with the mole balance given in Eq. (13) to obtain
the bulk-phase Hg(0) concentration profile along the reactor (CHg,b ). ξC (t = 0) = 1, δ (t = 0) = 1, γ (ζ ,ξ ,t = 0) = 0, γB (ζ ,t = 0) = 0 (23)
The eqs. (10)-(13) have the following set of initial (ICs) and
BCs:
boundary conditions (BCs).
ICs: DS ∂γ
γ (ζ ,ξ = 1,t ) = γB (ζ ,t ), = −k ″γ|ξ = ξC ,γB (ζ = 0,t ) = 1
rC (t = 0) = Rg ; D (t = 0) = D0; Rg ∂ξ (ζ ,ξ = ξC ,t ) (24)
CHg,pdt (z ,r ,t = 0) = 0; CHg,b (z ,t = 0) = 0 (15) The normalized conversion of CuCl2 is given in Eq. (25).
BCs: 1
XB = ∫0 (1−ξC3) dζ (25)
CHg,pdt |z,r = Rg,t = CHg,b (z ,t );
∂CHg,pdt
DS ″ )|z,r = rc,t = k ″CHg,pdt |z,r = rc,t ;
= (−rHg
∂r
z ,r = rc,t 3.4. Parameters used in this model
in
CHg,b (z = 0,t ) = CHg ,b (16)
COMSOL Multiphysics (version 4.4) was used to simulate the set of
In this study, 25 mg of the CuCl2/α-Al2O3 sample mixed with 4 g of the coupled equations in the above Section 3.3 [21]. The bulk-phase
silica sand were added to the fixed-bed reactor made of quartz as de- concentration of Hg(0) at the outlet of the reactor (CHg,b (z = L) ) and the
scribed in Section 2.3. The bed porosity filled with 4 g of silica sand was conversion of the solid-phase CuCl2 (XCuCl2) were used to assess the
determined as 0.38. However, since the silica sand does not adsorb Hg performance of the catalyst at different reaction temperatures and inlet
(0), the bed porosity should include both the sand volume and the void Hg(0) concentrations. A summary of the parameters used in this model
volume between the sand particles. Then the bed porosity filled only is given in Table 2. The mass-transfer coefficient, used in the general
with 25 mg of the sample in the fixed bed was found to be 0.98. This model, was calculated by using the Ranz-Marshall correlation given in
technique was used in our previous publication [20]. In the grain Eq. (26) [22,23].

249
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Table 2 parameter estimation. Two least-square objective functions defined in

Test conditions for fixed-bed runs. Eq. (27) were used to determine the two parameters by comparing the
model predictions with the experimental breakthrough and solid con-
Parameter Value
version results.
Flow mode Down-flow tf
Flow rate 1 L/min at 20 °C and 1 atm cal exp
Operating temperature 20, 100, 140, 180 °C
Min
subject to Ds , k ″
∑ (CHg ,b (z = L,t )−CHg ,b (z = L,t ))
2

Gas Nitrogen t=o

tf
Superficial gas velocity 0.15 m/s exp
cal
Reactor diameter 12 mm Min ∑ (XCuCl2( )
t −XCuCl2
(t ))2
subject to Ds , k′ ′ t = o (27)
Fixed-bed height 20 mm
Sorbent amount 25 mg with 4 g silica sand
Inlet Hg(0) concentration 5, 10, 20 ppbv
Gas hourly space velocity 26,515 h−1 4. Results and Discussion
Dopant loading 2%(wt) CuCl2/α-Al2O3
Bed porosity 0.98
Particle porosity 0.90 4.1. CuCl2 conversion
Effective diffusivity 2.71 × 10−7 m2/s
Fig. 3 shows the final CuCl2 conversion calculated from each ex-
perimental run at the end of the reaction, as a function of reaction
temperature and inlet Hg(0) concentration. It can be seen that the CuCl2
conversion increased with a decrease in reaction temperature while it
was not influenced by inlet Hg(0) concentration. At 180 °C, CuCl2
conversion was ∼15%(wt) while at 20 °C, all of the CuCl2 was reacted.
When the reaction temperature increases from 100 to 180 °C, the
average conversion values of CuCl2 decreased ∼65% to ∼15%. These
results clearly indicate that the CuCl2 conversion decreased as tem-
perature increased.

4.2. TGA-MS analysis

TGA-MS experiments were performed in order to check the thermal

Fig. 3. Final CuCl2 conversion at the end of each experiment as a function of inlet Hg(0)
concentration (5, 10, and 20 ppbv) and different reaction temperatures (20, 100, 140, and
180 °C). Note: average conversion values at room temperature, 100, 140 and 180 °C are
96, 66, 36, and 15%. Conversion of CuCl2 was calculated by from the breakthrough curve
and initial amount of CuCl2 added to the reactor as shown in Eq. (3).
stability of the 2%(wt) CuCl2/α-Al2O3 sample. The results are shown in
Fig. 4. The early weight loss up to 100 °C in the sample was close to
∼0.7%. An expected weight loss derived from the dihydrate (2H2O)
loss in the 2%(wt) CuCl2⋅2H2O/α-Al2O3 was 0.4% [13]. When the
sample was heated to 180 °C, ∼0.8% of total weight loss was observed.
This weight loss was attributed to the moisture from dihydrate (2H2O)
and α-Al2O3 substrate as confirmed by MS. As expected, the sample did
not exhibit any chloride ion signals from CuCl2 at a mass to charge ratio
(m/e) of 35 when it was heated up to 180 °C. This result suggests that
the stoichiometric ratios for the incomplete conversions cannot be at-
tributed to the loss of activity in CuCl2 due to the evolution of chlorine
from the samples.

4.3. SEM analysis of CuCl2/α-Al2O3

Fig. 5 shows the temporal SEM analysis of 2%(wt) CuCl2/α-Al2O3

samples, used in this study. The images were taken at different tem-
peratures and sampling times in back scattering mode to distinguish
between CuCl2 grains and α-Al2O3 particles, based on their different
atomic numbers. As shown in Fig. 5(a) and (b), CuCl2 grains were not
visible at 20 °C. It seems that the CuCl2 grains were uniform over the α-
Al2O3 substrate and do not agglomerates or form clusters. Therefore,
the general model was used for the reaction at 20 °C. However, at
higher temperatures such as 100 and 180 °C as shown in Fig. 5(c)–(f),
CuCl2 grain agglomerates were clearly visible. The grain sizes did not
change much from the samples obtained after 1 h and 24 h over >
∼50-h tests. It seems that agglomeration/sintering took place during
the initial phase of the reaction. Therefore, to simplify the shrinking
core model, we assumed that the grain size remains the same during the
Fig. 4. TGA-MS plot for 2%(wt) CuCl2/α-Al2O3 at a temperature of 180 °C. reaction.
In order to further examine the morphological changes during the
2Rg k g agglomeration, FIB/SEM was performed on fresh CuCl2/α-Al2O3 sam-
Sh = = 2 + 0.6(Rep)1/2 (Sc )1/3 ples before and after heating at 180 °C for 12 h and on spent CuCl2/α-
Dm (26)
Al2O3 sample at 180 °C for 12 h with an inlet Hg(0) concentration of 20
In this model, the two unknown parameters of product layer diffu- ppbv as shown in Fig. 6. In order to prevent morphological changes of
sion coefficient (DS) and kinetic constant (k″) were determined through the sample, a thin platinum coating was deposited on fresh CuCl2/α-

250
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Fig. 5. Temporal SEM pictures of 2%(wt) CuCl2/α-Al2O3 at 20 °C (a, b), 100 °C (c, d), and 180 °C (e, f).

(a) Pt deposition (c)

CuCl2 CuCl2 phase

phase

(b) Pt deposition (d)

CuCl2 phase

CuCl/CuCl2
phase

Fig. 6. FIB/SEM images of fresh CuCl2/α-Al2O3 sample (a) before heating with thin platinum layer and (b) after heating at 180 °C for 12 h with thin platinum layer, (c) after heating
without thin platinum layer and (d) spent CuCl2/α-Al2O3 sample after reaction with 20 ppbv inlet Hg(0) concentration at 180 °C for 12 h.

Al2O3 samples. Since no morphological changes were found as shown in
Fig. 6(b) and (c), platinum deposition was not applied to the spent
sample (Fig. 6(d)). After the sample was milled using gallium ions, SEM
was taken on the cross-sectional area of each sample. Porous structure
was clearly observed from fresh 5%(wt) CuCl2/α-Al2O3 sample before
heating. After heating, the fresh sample became much less porous, but a
pore network was visible. An SEM image on the spent sample after the
reaction showed that pore structure became almost non-porous. This
non-porous phenomenon could be confirmed at multiple milled spots.
The morphological changes seem to be derived from physical heating
combined with the chemical reaction, and lead to significant mass-
transfer resistance for incomplete CuCl2 conversion.
Although it is difficult to validate from the images whether the grain
agglomerates continue to grow with time, it is reasonable to speculate
that the incomplete CuCl2 conversion in this temperature range is at-
tributed to the agglomeration phenomenon. It is considered that it is
difficult for Hg(0) vapor to diffuse into the sintered CuCl2 agglomerates
and react with the unreacted CuCl2 grains. This sintering effect is more
pronounced at higher temperatures, and the diffusion of Hg(0) vapor is
limited due to sintering. The effect of sintering or agglomeration of

251
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Fig. 7. XPS depth profiles for fresh and spent 10%(wt) CuCl2/α-Al2O3 using monoatomic Ar ion gun and pure CuCl2 samples using Ar gas cluster ion beam. Note: (f) and (g) data were
obtained without using the depth profile technique.

CuCl2 grains is reported to take place when the reaction temperature is

higher than the Tamman temperature [24]. Tamman temperature is
generally considered to be approximately one half of the absolute
melting point in Kelvin (i.e. 771 K) of the CuCl2 grains, which is ne-
cessary for them to be mobilized on the α-Al2O3 substrate. Then the
Tamman temperature was estimated to be 385 K (112 °C) for CuCl2.
This is why CuCl2 grains are not clearly visible in the 20 °C samples in
Fig. 5(a) and (b) while CuCl2 grain agglomerates are clearly visible from
the 100 and 180 °C samples.

4.4. XPS analysis of CuCl2/α-Al2O3

The standards for Cu 2p3/2 binding energies were obtained at

934.6 eV for Cu(2+) in CuCl2 (Fig. 7(a)) and 932.2 eV for Cu(1+) in
CuCl (Fig. 7(b)). XPS depth profiling was performed on 10%(wt) CuCl2/
α-Al2O3 sample using monoatomic Ar ion gun and on pure CuCl2
sample using Ar gas cluster ion beam in order to validate the use of the
shrinking core model for the conversion of CuCl2 to CuCl inside the
grain phase. The fresh 10%(wt) CuCl2/α-Al2O3 sample showed clear Cu
(2+) peak before etching (Fig. 7(c)). However, CuCl2 of the fresh
Fig. 8. Relative CuCl2 dispersion on α-Al2O3 as a function of time at different reaction sample was reduced to CuCl during the depth profiling process using a
temperatures (100, 140, and 180 °C). monoatomic Ar ion gun even after 10-s etching (Fig. 7(d)). On the other
hand, it was reported that gas cluster ion beam could reduce the effect
Table 3 of reduction during the etching process [25,26]. However, when gas
Optimum dispersion kinetic parameters used in Eq. (11) as a function of different reaction cluster ion beam was used for pure CuCl2, the sample was also reduced
temperature.
as shown in Fig. 7(e).
Temperature (°C) Deq ks (1/s) m In summary, unfortunately, the XPS depth profiling technique could
not prevent the reduction of CuCl2 during the etching process, and thus
100 0.68 0.04 0.85 could not be used to validate the existence of the CuCl product layer
140 0.56 0.36 1.43
and the CuCl2 unreacted core. However, XPS analysis without using any
180 0.37 0.91 1.30
type of etching technique on two spent CuCl2/α-Al2O3 samples after 30
and 60 h (Fig. 7(f) and (g)) clearly showed the transition of Cu(2+)
peak into Cu(1+) peak as the reaction of CuCl2 with Hg(0) in N2

252
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Fig. 9. Hg(0) breakthrough data and simulated results for 2%(wt) CuCl2/α-Al2O3 as a function of different inlet Hg(0) concentration and reaction temperature. Note: symbols indicate
experimental data, and lines indicate model predictions.

proceeded. This implies that as a result of the reaction, CuCl is very 4.6. Model predictions for Hg(0) reaction, diffusion, and CuCl2 conversion
likely formed on the outer layer of the CuCl2/α-Al2O3 sample with
a ∼ 1 μm depth of CuCl2 since XPS can probe the sample depth less than Figs. 9 and 10 show the experimental breakthrough curves at the
10 nm. This result suggests that a shrinking-core type reaction take outlet of the reactor and the calculated CuCl2 conversion obtained using
place during the reaction and a CuCl product layer be formed from the Eq. (17) under different reaction temperatures (20, 100, 140 and
outer surface of CuCl2. 180 °C) and three different inlet Hg(0) concentrations (5, 10, and 20
ppbv), respectively. Experiments were conducted at 20 °C so that dif-
fusional limitations, as a result of agglomeration of CuCl2 were negli-
4.5. Dispersion kinetics for CuCl2 over α-Al2O3
gible for Hg(0) oxidation reaction. This was corroborated by SEM
images as no agglomeration was visualized at 20 °C. Overall, model
The effect of sintering was quantified by separately performing
predictions for all the temperatures and inlet Hg(0) concentrations are
chemisorption tests on the samples to determine the dispersion of CuCl2
in good agreement with the experimental breakthrough curve and
over α-Al2O3 support at different reaction temperatures as shown in
conversion results. Among all the breakthrough data, the breakthrough
Fig. 8. Then the dispersion kinetics was determined with an empirical
time at 20 °C for 20 ppbv inlet Hg(0) concentration was longest, and its
equation given in Eq. (11) [11,12]. The optimum sintering rate constant
CuCl2 conversion was almost complete. On the other hand, as the
(ks) and sintering order (m) for different reaction temperatures are
temperature between 100 and 180 °C went higher, the breakthrough
summarized in Table 3. The value of limiting dispersion (Deq) decreases
times became shorter and the final conversion was lower. These results
as the reaction temperature increases. It shows that the dispersion of
are attributed to the sintering and agglomeration phenomenon of the
CuCl2 grains quantified by CO chemisorption decreases to a larger ex-
CuCl2 grain phase at the high temperatures. At all the reaction tem-
tent at higher temperatures. The optimum sintering kinetic constant
peratures, the breakthrough time was longer as inlet Hg(0) concentra-
(ks) increased from 0.04 (1/s) at 100 °C to 0.91 (1/s) at 180 °C. The
tion was lower as seen in Fig. 9. It is also interesting to note that the
early loss of the dispersion of CuCl2 at high temperature accounts for
final CuCl2 conversion reached almost the same value at a given
the low CuCl2 conversion at high temperature shown in Fig. 3.

253
V. Sriram et al.
Fig. 10. CuCl2 conversion data and simulated results for 2%(wt) CuCl2/α-Al2O3 as a function of different inlet Hg(0) concentration and reaction temperature. Note: symbols show CuCl2
conversion calculated from Eq. (3) based on the breakthrough curve in Fig. 9 and the initial amount of CuCl2 added to the reactor. Lines are model predictions.
temperature regardless of inlet Hg(0) concentration as shown in Fig. 10.
This is most likely because the conversion is limited by sintering, which
is a function of temperature. The time required to reach a final con-
version increased with a decrease in inlet Hg(0) concentration.A higher
inlet Hg(0) concentration results in a higher diffusional flux in the
product layer inside the grain defined as ( D − Deq
D0 − Deq ) Rg ∂CHg,pdt
in
CHg ,b
∂r
r = Rg
. As
shown in Fig. 11(a), this leads to a higher Hg(0) oxidation rate at the
reaction interface between the unreacted CuCl2 grains and the product
layer, leading to a shorter breakthrough time. Fig. 11(b) shows the
bulk-phase Hg(0) concentration profiles along the length of the reactor
for an inlet Hg(0) concentration of 10 ppbv and a reaction time of 8 h.
As the reaction temperature increases, the outlet Hg(0) concentration
increases, indicating low reactivity. This is more pronounced at 180 °C,
suggesting that the reaction rate greatly slows down from 140 to 180 °C.
This can also be clearly seen from the reaction rates plotted as a
function of the reactor length in Fig. 11(c) and (d). The observed and
Hg(0) oxidation rates calculated by Eqs. (14) and (10), respectively,
along the reactor length at different temperatures for an inlet Hg(0)
concentration of 10 ppbv and a reaction time of 1 h are shown in
Fig. 11(c). Both the observed and reaction rates are close and decrease
along the reactor length at a short reaction time of 1 h since most Hg(0)
vapor reacts with freshly available CuCl2 close to the reactor inlet. As a
result, both of the reaction rates significantly decrease with a significant
254
Chemical Engineering Journal 343 (2018) 244–257
decrease in Hg(0) concentrations towards the reactor outlet. At higher
reaction temperatures, both of the reaction rates are relatively higher
due to the temperature dependence of the rate constant. However, after
8 h of the reaction shown in Fig. 11(d), a difference in the observed and
Hg(0) reaction rates becomes larger but decreases along the reactor
length. The reaction rates at 140 °C are higher than those at 100 °C.
However, at 180 °C, both of the Hg(0) oxidation rates near the reactor
inlet are even lower than those at 100 and 140 °C although both rates at
180 °C gradually decrease along the reactor length. This is because the
product of DS and (D−Deq)/(D0−Deq) significantly decreases with an
increase in the reaction temperature and time since the loss of disper-
sion is particularly significant at 180 °C (Table 4). A change in the re-
action rates significantly slows down along the reactor length since the
dispersion of CuCl2 starts to reach a limiting dispersion value (i.e. Deq)
at 180 °C. This can be corroborated by low CuCl2 conversion values at
high temperatures.The Hg(0) concentration profile along the product
layer thickness is shown in Fig. 11(e). As the reaction temperature in-
creases, the diffusional resistance derived from the combined effects of
sintering and the increased product layer thickness also increases. This
can be easily observed from the slope of the Hg(0) concentration in the
product layer. At the reactor inlet, the slope becomes steeper with an
increase in the reaction temperature, indicating that the diffusional
resistance becomes larger. At the reactor outlet, this effect becomes
particularly pronounced at 180 °C. The Hg(0) concentration profile
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Fig. 11. (a) Change in diffusional flux as a function of inlet Hg(0) concentration along the reactor length (t = 8 h, T = 100 °C) , (b) Bulk-phase Hg(0) concentration profile along the length
in = 10 ppbv, t = 8 h) , (c) The observed and Hg(0) oxidation rate along the length of the reactor at different reaction temperatures
of the reactor at different reaction temperatures (CHg ,b
in = 10 ppbv, t = 1 h) , (d) The observed and Hg(0) oxidation rate along the length of the reactor at different reaction temperatures (C in = 10 ppbv, t = 8 h) , (e) Hg(0) concentration
(CHg ,b Hg ,b
in = 10 ppbv, t = 8 h) , (f) The rate of change in the radius of
profiles inside the product layer at the inlet and outlet of the reactor (z / L = 0 & 1) and at different reaction temperatures (CHg ,b
in = 10 ppbv ) . Note: dimensionless diffusional
the unreacted CuCl2 core at the inlet and outlet of the reactor (z / L = 0 & 1) and at different reaction temperatures (CHg ,b
D−D eq ⎞ R g ∂CHg ,pdt
flux = ⎛ in
.
⎝ D0 − Deq ⎠ CHg,b ∂r
r = Rg

255
V. Sriram et al. Chemical Engineering Journal 343 (2018) 244–257

Table 4
Optimum reaction rate constant (k″) and product layer diffusivity (Ds) as a function of different reaction temperatures and inlet Hg(0) concentrations.

Reaction temperature (°C) Inlet Hg(0) concentration (ppbv) k″ (m/s) Average k″ (m/s) DS (10−8 m2/s) DS’=DS (D − Deq)/(1 − Deq) (10−8 m2/s)

@ t = 0.1 h @ t = tf

20 20 0.13 0.13 – – –
100 5 0.17 0.18 ± 0.03 22.5 22.5 1.41
10 0.16 5.9 5.9 0.37
20 0.21 2.5 2.5 0.16
140 5 0.21 0.26 ± 0.07 24.8 24.8 0.07
10 0.24 7.78 7.78 0.02
20 0.34 2.63 2.63 0.01
180 5 0.21 0.28 ± 0.07 37.8 37.8 0.004
10 0.28 7.98 7.98 0.0008
20 0.34 2.7 2.7 0.0003

4.7. Hg(0) reaction and diffusion kinetics

Table 4 shows the optimum values determined for the product layer
diffusion coefficient (Ds) and reaction rate constant (k ″) using the
double optimization method used in Eq. (27) that minimizes the sum of
the least square values between the experimental and simulation data
for both the breakthrough and conversion curves. It can be seen that
both Ds and k″ increase with an increase in reaction temperature. Using
the general model, the reaction rate constant was estimated to be
0.13 m/s at 20 °C while the rate constant obtained using the grain
model increased to 0.28 m/s at 180 °C. As expected, the rate constant
increased as the reaction temperature increased from 20 to 180 °C. The
temperature dependence of the rate constant was plotted in
Fig. 12.Assuming Arrhenius-type temperature dependence in Eq. (28),
the activation energy (Ea) and the pre-exponential factor (k 0″) were
determined.
−E
k ″ = k 0″exp ⎛ a ⎞
⎝ RT ⎠ (28)

Fig. 12. Temperature dependence of reaction rate constant (k″) for reaction between
CuCl2/α-Al2O3 and Hg(0).
inside the product layer at 180 °C is even higher than those at 100 and
140 °C. The Hg(0) concentrations are kept relatively high inside the
product layer at 180 °C due to the low reactivity derived from the
combined effects of sintering and product layer resulting in increased
diffusional resistance. Fig. 11(f) shows the temporal evolution of the
reaction interface between the unreacted core and the product layer at
different temperatures. As shown in Figs. 3 and 8, the final conversion
of CuCl2 at a temperature is almost constant while the conversion de-
creases with an increase in the reaction temperature. This decrease in
conversion is indicated in Fig. 11(f) with the location of the reaction
interface (rc). At the final reaction time, the rc value at 100 °C is much
smaller than that at 180 °C, indicating that the reaction interface moves
much closer to the center of the grain at 100 °C. This shows that the
conversion at a given inlet Hg(0) concentration increases with a de-
crease in the reaction temperature.
The incomplete conversion of the CuCl2 grains are attributed to the
sintering and agglomeration of CuCl2 grains. As discussed in the pre-
vious section, sintering is predominant at the reaction temperature
between 100 and 180 °C since this temperature range is higher than the
Tamman temperature, above which CuCl2 grains become mobile and
form clusters and agglomerates [24]. Since the sintering kinetics for
CuCl2 grains is faster at higher temperatures, the diffusion kinetics for
Hg(0) vapor through the sintered CuCl product layer significantly de-
creases at higher temperatures. This leads to lower CuCl2 conversion at
the high temperatures as shown in Fig. 3.
The activation energy was determined to be 5.8 ± 3.0 kJ/gmol with a
90% confidence interval, and the Arrhenius pre-exponential factor (k 0″)
was determined to be 1.3 m/s. A t-test was performed on the rate
constant (k″) with a 90% confidence interval, and the true value for k″
from the experimental data was estimated to be within 0.19 and
0.27 m/s. The activation energy for Hg(0) oxidation with CuCl2 was
compared with other activation energy values reported in the literature
for homogeneous and heterogeneous Hg(0) reaction systems using ha-
logen gases and different catalysts. It is seen that the activation energy
value for the Hg(0) oxidation using CuCl2 is significantly lower than the
other reported values ranging between 20 and 120 kJ/gmol for other
heterogeneous catalysts [27–33]. Niksa et al. reported an activation
energy value of 142 kJ/gmol for a homogenous Hg(0) oxidation reac-
tion involving Cl2 gas while the homogeneous gas-phase reaction be-
tween Hg(0) and HCl, was predicted to have an activation energy of
332 kJ/gmol [34]. These reported values are much higher than that for
the heterogeneous Hg(0) reaction with CuCl2 found in this study, sug-
gesting that CuCl2 works as an excellent redox catalyst for Hg(0) oxi-
dation. In conjunction with our previous mechanistic study of the re-
action of Hg(0) vapor with CuCl2, this study result also validates that
CuCl2 greatly enhances the Hg(0) oxidation reaction by not only pro-
viding Cl for Hg(0) oxidation but also lowering the activation energy
barrier for the reaction by reducing Cu(2+) to Cu(1+) [13].The pro-
duct layer diffusion coefficient (DS) values in Table 4 show that DS
increases with an increase in reaction temperature. It is interesting to
note that the value of DS is a function of inlet Hg(0) concentration. This
trend has been reported for the sulfation of calcium oxide [35,36]. It
was reported that the crystal growth associated with the product for-
mation leads to a high porous product layer at low inlet gas con-
centration. This leads to a high product-layer diffusion coefficient

256
V. Sriram et al.
inside the product layer. These reported observations are consistent
with the results obtained in this study where the value of DS is high at
low inlet Hg(0) concentration. It is also noteworthy from Table 4 that a
product of DS and (D−Deq)/(D0−Deq) significantly decreases with reac-
tion time, which adds significant mass-transfer resistance to Hg(0)
diffusion through the product layer.
An analysis was performed to estimate the inter-parameter corre-
lation between k″ and DS. A covariance value between k″ and DS was
obtained to be −0.45. Then this value was normalized between −1 and
1 by calculating the correlation coefficient. The normalized covariance
value was determined to be -0.60. This negative relationship between
k″ and Ds. indicates that as the reaction temperature increases, the Hg
(0) oxidation reaction rate increases and the product layer diffusion
rate decreases. At high temperatures, mass-transfer resistance increases
as a result of sintering/agglomeration of grains, resulting in low DS
values.
5. Conclusions
The reaction kinetics for the heterogeneous Hg(0) oxidation reac-
tion with CuCl2 was studied using CuCl2/α-Al2O3 under N2 atmosphere.
The general and grain models were used to determine the reaction rate
constant and product layer diffusion coefficient using a fixed-bed
system at 20 °C and 100–180 °C, respectively. Both model predictions
were generally in good agreement with the experimental data. It was
found that the sintering of CuCl2 over α-Al2O3 support at 100–180 °C
significantly increased the mass-transfer resistance to Hg(0) diffusion
inside the reacted CuCl product layer leading to an incomplete con-
version of the CuCl2 grain phase. Nevertheless, the reaction rate con-
stant for the Hg(0) oxidation reaction was successfully determined by
taking into account the mass-transfer resistance. The activation energy
value determined from the Arrhenius temperature dependence of the
rate constants was much lower than other activation energy values
reported for other catalysts evaluated for heterogeneous Hg(0) oxida-
tion reactions with halogen gases in the literature. In conjunction with
our previous study of the reaction mechanism of Hg(0) vapor with
CuCl2, this study result strongly supports that the activation energy
barrier for Hg(0) oxidation reaction can greatly be lowered with CuCl2
following a Mar-Maessen mechanism by the simultaneous reduction of
Cu(2+) to Cu(1+) and readily available surface Cl sites.
Acknowledgements
This study was supported by the National Science Foundation, NSF
CAREER Grant # 1151017, and Ohio Development Services Agency,
Grant Agreement # OER-CDO-D-14-21. The authors appreciate their
financial support. The XPS characterization was carried out in the
Characterization Facility, University of Minnesota, which receives
partial support from NSF through the MRSEC program and was also
performed at the Ohio State University.
References
[1] National Emission Standards for Hazardous Air Pollutants From Coal and Oil-Fired
Electric Utility Steam Generating Units and Standards of Performance for Fossil-
Fuel-Fired Electric Utility, Industrial-Commercial Institutional, and Small Industrial
Commercial Institutional Steam Generating Units – 40 CFR Parts 60 and 63, in: U.S.
EPA (Ed.), 2016.
[2] R.K. Srivastava, N. Hutson, B. Martin, F. Princiotta, J. Staudt, Control of mercury
emissions from coal-fired electric utility boilers, Environ. Sci. Technol. 40 (2006)
1385–1393.
[3] A.P. Jones, J.W. Hoffmann, D.N. Smith, T.J. Feeley, J.T. Murphy, DOE/NETL's
phase II mercury control technology field testing program: preliminary economic
analysis of activated carbon injection, Environ. Sci. Technol. 41 (2007) 1365–1371.
[4] J.H. Pavlish, E.A. Sondreal, M.D. Mann, E.S. Olson, K.C. Galbreath, D.L. Laudal,
S.A. Benson, Status review of mercury control options for coal-fired power plants,
257
Chemical Engineering Journal 343 (2018) 244–257
Fuel Process. Technol. 82 (2003) 89–165.
[5] G.F. Froment, K.B. Bischoff, J. De Wilde, Chemical Reactor Analysis and Design,
Wiley, New York, 1990.
[6] M. Ishida, C. Wen, Effectiveness factors and instability in solid-gas reactions, Chem.
Eng. Sci. 23 (1968) 125–137.
[7] O. Levenspiel, Chemical Reaction Engineering, John Wiley & Sons, New York, 1999.
[8] J. Szekely, Gas-Solid Reactions, Elsevier, 2012.
[9] M. Ishida, C. Wen, Comparison of kinetic and diffusional models for solid-gas re-
actions, AIChE J. 14 (1968) 311–317.
[10] M. Hartman, R.W. Coughlin, Reaction of sulfur dioxide with limestone and the grain
model, AIChE J. 22 (1976) 490–498.
[11] C.H. Bartholomew, Sintering kinetics of supported metals: new perspectives from a
unifying GPLE treatment, Appl. Catal. A Gen. 107 (1993) 1–57.
[12] C.H. Bartholomew, Mechanisms of catalyst deactivation, Appl. Catal. A Gen. 212
(2001) 17–60.
[13] X. Li, Z. Liu, J. Kim, J.-Y. Lee, Heterogeneous catalytic reaction of elemental mer-
cury vapor over cupric chloride for mercury emissions control, Appl. Catal. B
Environ. 132 (2013) 401–407.
[14] M. Chase Jr, C. Davies, J. Downey Jr, D. Frurip, R. McDonald, A. Syverud, JANAF
thermochemical tables, J. Phys. Chem. Ref. Data 14 (1856).
[15] N. Muddada, U. Olsbye, G. Leofanti, D. Gianolio, F. Bonino, S. Bordiga, T. Fuglerud,
S. Vidotto, A. Marsella, C. Lamberti, Quantification of copper phases, their re-
ducibility and dispersion in doped-CuCl2/Al2O3 catalysts for ethylene oxychlor-
ination, Dalton Trans. 39 (2010) 8437–8449.
[16] G. Leofanti, A. Marsella, B. Cremaschi, M. Garilli, A. Zecchina, G. Spoto, S. Bordiga,
P. Fisicaro, C. Prestipino, F. Villain, Alumina-supported copper chloride: 4. effect of
exposure to O2 and HCl, J. Catal. 205 (2002) 375–381.
[17] Standard Test Method for Elemental, Oxidized, Particle-Bound, and Total Mercury
in Flue Gas Generated from Coal-Fired Stationary Sources (Ontarlo-Hydro),
D6784–02, Standard, ASTM, 2008.
[18] C. Georgakis, C. Chang, J. Szekely, A changing grain size model for gas—solid re-
actions, Chem. Eng. Sci. 34 (1979) 1072–1075.
[19] A. Ghosh-Dastidar, S. Mahuli, R. Agnihotri, L.-S. Fan, Ultrafast calcination and
sintering of Ca (OH) 2 powder: experimental and modeling, Chem. Eng. Sci. 50
(1995) 2029–2040.
[20] X. Li, Z. Liu, J.-Y. Lee, Adsorption kinetic and equilibrium study for removal of
mercuric chloride by CuCl 2-impregnated activated carbon sorbent, J. Hazard.
Mater. 252 (2013) 419–427.
[21] C. Multiphysics, 4.4; COMSOL, Inc.: Burlington, MA, USA, 2014.
[22] W. Ranz, W. Marshall, Evaporation from drops-Part I, Chem. Eng. Prog. 48 (1952)
141–146.
[23] W. Ranz, W. Marshall, Evaporation from Drops-Part II, Chem. Eng. Prog. 48 (1952)
173–l180.
[24] J.A. Moulijn, A.E. van Diepen, F. Kapteijn, Catalyst deactivation: is it predictable?:
What to do? Appl. Catal. A Gen. 212 (2001) 3–16.
[25] A.G. Shard, R. Havelund, M.P. Seah, S.J. Spencer, I.S. Gilmore, N. Winograd,
D. Mao, T. Miyayama, E. Niehuis, D. Rading, Argon cluster ion beams for organic
depth profiling: results from a VAMAS interlaboratory study, Anal. Chem. 84 (2012)
7865–7873.
[26] R. Steinberger, J. Walter, T. Greunz, J. Duchoslav, M. Arndt, S. Molodtsov,
D. Meyer, D. Stifter, XPS study of the effects of long-term Ar+ ion and Ar cluster
sputtering on the chemical degradation of hydrozincite and iron oxide, Corrosion
Sci. 99 (2015) 66–75.
[27] S. Eswaran, H.G. Stenger, Understanding mercury conversion in selective catalytic
reduction (SCR) catalysts, Energy Fuels 19 (2005) 2328–2334.
[28] W. Gao, Q. Liu, C.-Y. Wu, H. Li, Y. Li, J. Yang, G. Wu, Kinetics of mercury oxidation
in the presence of hydrochloric acid and oxygen over a commercial SCR catalyst,
Chem. Eng. J. 220 (2013) 53–60.
[29] A.A. Presto, E.J. Granite, Noble metal catalysts for mercury oxidation in utility flue
gas, Platin. Met. Rev. 52 (2008) 144–154.
[30] A.A. Presto, E.J. Granite, A. Karash, R.A. Hargis, W.J. O'Dow, H.W. Pennline, A
kinetic approach to the catalytic oxidation of mercury in flue gas, Energy Fuels 20
(2006) 1941–1945.
[31] A. Suarez Negreira, J. Wilcox, Uncertainty analysis of the mercury oxidation over a
standard SCR catalyst through a lab-scale kinetic study, Energy Fuels 29 (2014)
369–376.
[32] N. Usberti, S.A. Clave, M. Nash, A. Beretta, Kinetics of Hg° oxidation over a V2O5/
MoO3/TiO2 catalyst: experimental and modelling study under DeNO X inactive
conditions, Appl. Catal. B Environ. 193 (2016) 121–132.
[33] Y. Zhao, M.D. Mann, J.H. Pavlish, B.A. Mibeck, G.E. Dunham, E.S. Olson,
Application of gold catalyst for mercury oxidation by chlorine, Environ. Sci.
Technol. 40 (2006) 1603–1608.
[34] S. Niksa, J.J. Helble, N. Fujiwara, Kinetic modeling of homogeneous mercury oxi-
dation: the importance of NO and H2O in predicting oxidation in coal-derived
systems, Environ. Sci. Technol. 35 (2001) 3701–3706.
[35] W. Duo, K. Laursen, J. Lim, J. Grace, Crystallization and fracture: formation of
product layers in sulfation of calcined limestone, Powder Technol. 111 (2000)
154–167.
[36] W. Duo, K. Laursen, J. Lim, J. Grace, Crystallization and fracture: product layer
diffusion in sulfation of calcined limestone, Ind. Eng. Chem. Res. 43 (2004)
5653–5662.