A STRATEGIC APPROACH FOR OPTIMIZING THE USE OF THTANIA
PARTICLES FOR MERCURY REMOVAL IN COAL-BURNING UTILITIES
Achariya Suriyawong, Marina Smallwood, Myong-Hwa Lee and Pratim Biswas*
Environmental Engineering Science,
Campus Box 1180, Washington University in St. Louis,
St. Louis, MO, 63130, United States.
Submitted for presentation at
the Air & Waste Management Association’s 98th Annual Conference & Exhibition
June 2005
*To whom correspondence should be addressed:
Tel.: +1-314-935-5482; Fax: +1-314-935-5464
E-mail address: pratim.biswas@seas.wustl.edu
Abstract
Mercury is a toxic metal whose emission control was stipulated under Title III of
Clean Air Act Amendments of 1990. Over 40 percent of anthropogenic mercury emissions in
the United States are from coal-burning utilities. Control technologies for mercury from flue
gas include scrubbing solutions and sorbent injection. Sorbent injection is one of the
promising technology for controlling mercury emission. However, this method relies on
available surface area for adsorption, which depends on time-temperature history in the
system. In this study, the evolution of sorbent morphology, size and number concentration
undergoing chemical formation, collisional growth and sintering are modeled and a
laboratory scale system was constructed to test mercury capture efficiency during Powder
River Basin (PRB) coal combustion.
The results obtained from the simulations showed that when TiO2 was injected after
heat exchanger the highest surface area was obtained, while the surface area obtained from
the injection of TiO2 and TTIP at the combustor and after combustor were 5 and 3 times less.
The sorbent agglomerate size were found largest when TiO2 was injected into the flue gas
after heat exchanger unit, and smallest when TiO2 and TTIP were injected into the
combustor. The results obtained from experiment showed that the best overall reduction was
achieved using the in-situ generated method; a total mercury reduction of 85% was achieved.
When pre-synthesized TiO2 particles was injected in prior to the photo-chemical reactor, a
total of 82% mercury reduction was observed. When commercially TiO2 particles were
introduced to the system with coal, a total of 70% reduction of mercury was observed.
Introduction
Mercury emission from combustion systems has been of great concern because of its
high toxicity, its tendency to bio-accumulate and difficulties in its control. In the United
States, approximately 120 tons of mercury is released to atmosphere each year from
anthropogenic sources; approximately forty percent of this amount comes from coal-burning
utilities1. Even though mercury is a trace metal found in coal in the range of 0.01 to 3.3
ppmw, its emission is significant due to the large quantity of coal burned every year.
Presently, mercury emission regulations for coal-fired power plants are being finalized by the
United States Environmental Protection Agency (USEPA) for implementation in 2007.
Extensive research and development work are also in progress for designing control
methodologies.
There are several methods have been proposed for control of mercury emission from
coal-fired combustors. Sorbent injection method has been shown to be effective for capture
of heavy metal species in combustion systems. This method relies profoundly on the
available surface area for adsorption. Wu et al.2 have demonstrated the effectiveness of the
in-situ generated titanium-based precursor in conjunction with uv irradiation in capturing
mercury; and over 96 percent mercury removal efficiency has been exhibited in their labscale experiments. The purpose of using in-situ generated sorbent is to maximize the
utilization of sorbents as the available surface area for adsorption is the key parameter for the
effectiveness of the sorbent. Hence this study presents a strategic approach for maximizing
mercury removal efficiency as well as to minimizing the amount of sorbent used.
In this study, the evolution of sorbent morphology, size and number concentration
undergoing chemical formation (nucleation), collisional growth and sintering are modeled.
The model equations, based on the work by Kruis et al.3 and Jeong and Choi 4, are modified
to predict the evolution of the sorbent size distribution and the morphology of the sorbent
agglomerates. The variation of flue gas temperatures in a typical full-scale coal fired utility
(137 MW), are used in the model; since temperature is an essential parameter in both
determining the collision kernel and characteristic sintering time. A laboratory scale system
was constructed to test mercury capture efficiency during Powder River Basin (PRB) coal
combustion.
Methods
Model
The evolution of sorbent size distribution and agglomerate structure is predicted by a
bimodal model developed by Jeong and Choi4. This model accounts for the morphology of
the aggregates in a dynamic manner, which would give more accurate prediction of the
surface area and the collision rates of growth, in contrast to more traditional models that
assume spherical particles. The model dynamics are described by the following equations:
Mode 1 (nucleation mode):
dN 1
1
 r 
   11 N 12 
(1),
   12 N 1 N 2  I
dt
2
 r  1
where A1 ( N1a0 ) is a total surface area of the agglomerates in mode 1, V1 ( N1v0 ) is the total
volume, N1 is the particle number concentration in mode 1; 11 is the collision kernel between
particles in mode 1; 12 is collision kernel between particles in mode 1 and mode 2; a0 and v0
are surface area and volume of a primary particle.
Mode 2 (accumulation mode):
dN 2 1
 1  1
2
  11 N 12 
   22 N 2 .
dt
2
 r  1 2
dA2 1
1
 r 
  11 N 12 
a 0   12 N 1 N 2 a 0   A2  N 2 a 2 s 
dt 2

 r  1
dV2 1
 r 
  11 N 12 
v 0   12 N 1 N 2 v 0
dt 2
 r  1
da 2 1
N2  r 
1
  11 1 
a 0  12 N 1 a 0  a 2  a 2 s 
dt 2
N 2  r  1

(2)
(3)
(4)
(5)
dv 2 1
N 12  r 
 11

v0  12 N 1v0
dt 2
N 2  r  1
where r 
(6)
v2
is the ratio of the size v2 to v1; a2 is surface area of an aggregate in mode 2; and
v1
v2 is volume of an aggregate in mode 2. The total number concentration, area and volume of
the particles are given by N t  N1  N 2 ; At  A1  A2 and Vt  V1  V2 ; where N2 is the
particle number concentration in mode 2, A2 is the total surface area of particle in mode 2 , V2
is the total volume of particle in mode 2, and the other parameters as described earlier. The
second mode, is the larger mode and accounts for the agglomerate characteristics, and the
appropriate collision coefficients, which are also a function of fractal dimension. The
collision diameter will be estimated using dc2 = d p 2 n p 2 D f where Df is the fractal
1
dimension of the aggregate3.
Since temperature is an essential parameter in both determining the collision kernel as
well as characteristic sintering time, the variation of flue gas temperatures are taken into
account. The temperature profile used in this simulations, shown in Figure 1, is obtained a
full-scale coal-fired utility (137 MW). Based on the plant structure, three sorbent injection
locations are selected: (1) at the combustor, (2) after combustor, and (3) after heat exchanger.
The temperatures at these locations are 1600, 1200, and 150 oC, respectively. In this study,
sorbents are introduced into the system in two ways: as sorbent precursor, and as bulk
sorbents. In the former approach, the sorbent precursor undergoes chemical formation
(nucleation), collisional growth and sintering. And in the latter approach, the bulk sorbent
undergoes only collisional growth and sintering. One of the most important parameters in
determining sorbent morphology is the intitial size of the sorbent. For the precursor feed, the
initial size of the sorbent is their monomer size, and for the bulk sorbent the initial size is the
size of commercially produced sorbents. The values of the important parameters used in the
model are shown in Table 1.
Injection Location 1: in the combustor
2000
Injection Location 2: after the combustor
Temperature (K)
1500
Injection Location 3: after heat exchanger
1000
500
0
0
10
20
30
40
50
60
time (sec)
Figure 1: Temperature Profile of a Typical Coal-fired Power Plant.
Table 1: Values of important parameters used in the model
Selected Sorbent
Titanium Dioxide
Amount of sorbent feed as precursor
1 mole TTIP/m3
4.0
for sorbent precursor feed
Initial Diameter of the sorbent (nm)
21
for bulk sorbent feed
Fractal Dimension
1.8
Experimental
The experimental set up to study Hg capture using sorbents in a coal combustor is
depicted in Figure 2. A detailed description of the flow reactor system, in situ generated
TiO2, and the experimental procedure is similar to the one developed by Wu et al.2 and
Zhuang and Biswas5 . In this work, the sorbent was injected into a laboratory scale coal
combustion system. Titanium (IV) isopropoxide (Ti[OCH(CH3)2]4; 97% Aldrich) was used
as precursor for TiO2. Coal feed rate was maintained at 0.9 g/hr. The sorbent precursor was
pre-heated to 80 oC; and the carrier gas (N2) flow rate was 0.2 lpm. Furnace temperature was
maintained at 1100 oC with excess air (1.75 lpm air feed rate) to ensure complete combustion.
Dilution air was added at the exit stream to obtain 0.5 μm cutoff size using cascade impactor.
A UV lamp and electrostatic precipitator were used as photochemical reactors. The real time
differential mobility analyzer (DMA) and condensation particle counter (CPC) were used to
obtain particle size distributions.
Mercury capture was studied using three different generation and injection strategies.
The first injection, TiO2 precursor (TTIP) was introduced into the combustor for in-situ
generation of titania particles. In the second injection strategy, commercially available titania
particles were introduced into the system with coal. And the last injection strategy
investigated was pre-synthesized TiO2 particles injected in prior to the photochemical reactor.
Figure 2. Schematic diagram of the experimental system for coal combustion studies
Results
Model Results
The key parameters used to determined the best injection strategies are surface area
and agglomerate size. The sorbent agglomerates with high specific surface area would retain
their active sites for mercury capture, and the large agglomerate size would result in the most
effective capture in an existing particle control device, such as ESP. From the simulation,
results shown in Figure 3, the maximum surface area is obtained when bulk sorbent is
injected after heat exchanger unit. At this location, flue gas temperature is low enough to
impede the effect of sintering. When sorbent precursor is injected at this location, low flue
gas temperature lowers the formation of the sorbent by chemical transformation of the
precursor; therefore, total surface area at the exit of the stack of sorbent precursor feed is less
than that of the bulk sorbent feed. When sorbents are injected in the combustor or after, total
surface area at the exit are the same for both precursor feed and bulk sorbent feed. This is due
to at these injection locations, flue gas temperatures are high enough to allow instantaneous
chemical transformation of sorbent precursor to the sorbents. However, excessive sintering
resulting in substantial reduction of surface area is also observed for both injection locations.
The simulation results for diameter of primary particle as well as number of primary particle
in an agglomerate also confirm this observation.
Figure 4 shows the evolution of diameter of primary particle in an aggregate. The
largest primary particle diameter is obtained when sorbent is injected in the combustor
resulting in less number of primary particle in an agglomerate although volume of the
agglomerate is high. As the size of primary particle increases, the effect of sintering
decreases. This is due to characteristic sintering time, the time required for solid particles in
contact to sinter to a compact particle, is proprosional to fourth power of radius of primary
particle. The large primary particle would required longer time to sinter.
Another important criteria in determining the best injection strategy is agglomerate
size or collision raduis (Rc). Results of the simulations are illustrated in Figure 5. This
parameter is determined by raduis of primary particle, number of primary particle in the
agglomerate, and fractal dimension in the agglomerate. Since monodisperse distribution is
assumed for the size distribution in nucleation mode and accumulation mode, collision radius
will be the same for all sorbent agglomerates. The largest collision radius (200 µm) is again
obtained when bulk sorbent is injected in after heat exchanger unit. For sorbent precursor
feed, the largest collision radius (120µm) is obtained when the precursor is injected into flue
gas after heat exchanger. This difference results from the influence of flue gas temperature on
the chemical transformation of sorbent precursor, as previously discussed.
7.0x106
Total surface area (m2/m3)
6.0x106
5.0x106
In the combustor
After combustor
After heat exchanger
4.0x106
3.0x106
2.0x106
1.0x106
(A)
0.0
0
10
20
30
40
50
60
70
3.0x106
Total surface area (m2/m3)
2.5x106
2.0x106
1.5x106
1.0x106
(B)
500.0x103
0.0
0
10
20
30
40
50
60
70
time (sec)
Figure 3. Predicted total surface are of sorbents: (A) sorbent precursor feed
and (B) sorbent feed.
350x10-9
Diameter of primary particle (m)
300x10-9
250x10-9
in the combustor
after the combustor
after heat exchanger
200x10-9
150x10-9
(A)
100x10-9
50x10-9
0
0
10
20
30
40
50
60
70
350x10-9
Diameter of primary particle (m)
300x10-9
250x10-9
200x10-9
150x10-9
100x10-9
50x10-9
(B)
0
0
10
20
30
40
50
60
70
time (sec)
Figure 4. Predicted primary particle size diamete: (A) sorbent precursor feed
and (B) sorbent feed.
140x10-6
in the combustor
after the combustor
after heat exchanger
Collision Radius (m)
120x10-6
100x10-6
80x10-6
60x10-6
40x10-6
20x10-6
0
0
10
20
30
40
50
60
70
0
10
20
30
40
50
60
70
250x10-6
Collision Radius (m)
200x10-6
150x10-6
100x10-6
50x10-6
0
time (sec)
Figure 5: Predicted agglomerate size: (A) sorbent precursor feed and (B)
sorbent feed.
All the numbers reported in this study are system specific and condition specific. The
evolution of sorbent size distribution and agglomerate structure depends on time-temperature
history at which the agglomerates have undergone under specific combustion system. In
addition, initial condition also plays an important role on agglomerate size and surface area,
which can be increased or decreased by changing initial number concentration.
Experimental Results
Figure 6 summarizes the speciated and total mercury concentration in the exhaust for
the various tests performed. The best overall reduction was achieved using the in-situ
generated method, injection strategy 1; a total mercury reduction of 85% was achieved. When
pre-synthesized TiO2 particles was injected in prior to the photo-chemical reactor, injection
strategy 3, a total of 82% mercury reduction was observed. When commercially TiO2
particles were introduced to the system with coal, a total of 70% reduction of mercury was
observed. The tests were also performed with the injection strategy 1, but with uv lamp off
and ESP on, results of mercury reduction were comparable to those obtained when uv light
was on. Thus, it can be deduced that the uv radiation generated in the ESP was strong enough
to activate titania particles and to oxidize mercury in the system.
3.5
Hg total
Hg2+
Hg0
Gas phase Hg concentration at
combustor exit [g/m3]
3.0
2.5
2.0
69%
1.5
82%
1.0
85%
75%
86%
0.5
0.0
coal
coal + TTIP
at inlet
manifold
coal + TTIP coal + TTIP coal + TiO2
at inlet
(TTIP
mixed in
manifold
introduced
impinger
(UV off)
to small
(0.995 mg
furnace)
of TiO2)
coal + TiO2
mixed in
impinger
(150 mg
of TiO2)
Figure 6: Summary of key experimental results for mercury capture from coal
combustion exhaust using titania sorbent introduced by various strategies.
Conclusions
The evolution of sorbent size distribution and agglomerate structure was modeled by a
bimodal model developed by Jeong and Choi4. The simulation results of TiO2 and TiO2
precursor (TTIP) injections were discussed. The TiO2 injection after heat exchanger provided
the highest surface area, while the surface area obtained from the injection of TiO 2 and TTIP
at the combustor and after combustor were 5 and 3 times less. The sorbent agglomerate size
were found to be the largest when TiO2 was injected into the flue gas after the heat exchanger
unit, and smallest when TiO2 and TTIP were injected into the combustor. This shows that
sintering process previal under high temperatures, while coagulation process dominated at
lower temperatures. In addition flue gas temperature plays an important role in chemical
transformation of TiO2 precursor. The precursor was not completely transformed when it is
injected in after heat exchanger unit. Hence, TiO2 should be injected at the location after heat
exchanger unit, and TiO2 precursor should be injected at the location after the combustor.
Titania sorbents were shown to be effective in capturing gas phase mercury species,
oxidized and elemental, released during the combustion of PRB coal. The most controlled
and reliable tests were for tests performed with in-situ generated sorbets. Similar percent
reductions were observed when the uv reactor was turned on and when it was turned off. It
was concluded that the uv present in the ESP was sufficient for the effective capture of
mercury.
Reference:
(1)
EPA "1999 National Emission Inventory for Hazardous Air Pollutants," Office
of Air Quality Planning and Standards, U.S. Environmental Protection Agency, 1999.
(2)
Wu, C. Y.; Lee, T. G.; Tyree, G.; Arar, E.; Biswas, P. Environ Eng Sci 1998, 15,
137-148.
(3)
Kruis, F. E.; Kusters, K. A.; Pratsinis, S. E. Aerosol Science and Technology
1993, 19, 514-526.
(4)
Jeong, I. J.; Choi, M. Aerosol Science 2003, 34, 965-976.
(5)
Zhuang, Y.; Biswas, P. Energ Fuel 2001, 15, 510-516.