Citation
Mechanistic Study of Sorbent Injection for Vanadium Emission Control in Combustion Systems

Material Information

Title:
Mechanistic Study of Sorbent Injection for Vanadium Emission Control in Combustion Systems
Creator:
LEE, SANG-RIN ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Aerosols ( jstor )
Coagulation ( jstor )
Combustion ( jstor )
Condensation ( jstor )
Hydrolysis ( jstor )
Metal particles ( jstor )
Sorbents ( jstor )
Sulfur ( jstor )
Vanadium ( jstor )
Vapors ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Sang-Rin Lee. Permission granted to University of Florida to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
5/1/2005
Resource Identifier:
71230804 ( OCLC )

Downloads

This item has the following downloads:

lee_s ( .pdf )

lee_s_Page_129.txt

lee_s_Page_021.txt

lee_s_Page_006.txt

lee_s_Page_109.txt

lee_s_Page_141.txt

lee_s_Page_016.txt

lee_s_Page_012.txt

lee_s_Page_117.txt

lee_s_Page_142.txt

lee_s_Page_018.txt

lee_s_Page_042.txt

lee_s_Page_033.txt

lee_s_Page_040.txt

lee_s_Page_047.txt

lee_s_Page_014.txt

lee_s_Page_123.txt

lee_s_Page_024.txt

lee_s_Page_048.txt

lee_s_Page_147.txt

lee_s_Page_027.txt

lee_s_Page_084.txt

lee_s_Page_075.txt

lee_s_pdf.txt

lee_s_Page_136.txt

lee_s_Page_146.txt

lee_s_Page_140.txt

lee_s_Page_098.txt

lee_s_Page_058.txt

lee_s_Page_134.txt

lee_s_Page_009.txt

lee_s_Page_002.txt

lee_s_Page_149.txt

lee_s_Page_119.txt

lee_s_Page_086.txt

lee_s_Page_003.txt

lee_s_Page_044.txt

lee_s_Page_035.txt

lee_s_Page_073.txt

lee_s_Page_097.txt

lee_s_Page_128.txt

lee_s_Page_039.txt

lee_s_Page_063.txt

lee_s_Page_148.txt

lee_s_Page_093.txt

lee_s_Page_064.txt

lee_s_Page_125.txt

lee_s_Page_107.txt

lee_s_Page_017.txt

lee_s_Page_030.txt

lee_s_Page_096.txt

lee_s_Page_065.txt

lee_s_Page_090.txt

lee_s_Page_088.txt

lee_s_Page_079.txt

lee_s_Page_121.txt

lee_s_Page_034.txt

lee_s_Page_010.txt

lee_s_Page_070.txt

lee_s_Page_008.txt

lee_s_Page_092.txt

lee_s_Page_038.txt

lee_s_Page_135.txt

lee_s_Page_043.txt

lee_s_Page_037.txt

lee_s_Page_102.txt

lee_s_Page_022.txt

lee_s_Page_050.txt

lee_s_Page_118.txt

lee_s_Page_061.txt

lee_s_Page_056.txt

lee_s_Page_015.txt

lee_s_Page_068.txt

lee_s_Page_112.txt

lee_s_Page_091.txt

lee_s_Page_139.txt

lee_s_Page_143.txt

lee_s_Page_138.txt

lee_s_Page_113.txt

lee_s_Page_099.txt

lee_s_Page_103.txt

lee_s_Page_051.txt

lee_s_Page_110.txt

lee_s_Page_026.txt

lee_s_Page_100.txt

lee_s_Page_020.txt

lee_s_Page_095.txt

lee_s_Page_150.txt

lee_s_Page_054.txt

lee_s_Page_074.txt

lee_s_Page_133.txt

lee_s_Page_081.txt

lee_s_Page_114.txt

lee_s_Page_029.txt

lee_s_Page_059.txt

lee_s_Page_060.txt

lee_s_Page_013.txt

lee_s_Page_137.txt

lee_s_Page_019.txt

lee_s_Page_052.txt

lee_s_Page_087.txt

lee_s_Page_131.txt

lee_s_Page_032.txt

lee_s_Page_005.txt

lee_s_Page_101.txt

lee_s_Page_124.txt

lee_s_Page_036.txt

lee_s_Page_083.txt

lee_s_Page_116.txt

lee_s_Page_049.txt

lee_s_Page_046.txt

lee_s_Page_132.txt

lee_s_Page_111.txt

lee_s_Page_011.txt

lee_s_Page_089.txt

lee_s_Page_053.txt

lee_s_Page_007.txt

lee_s_Page_106.txt

lee_s_Page_126.txt

lee_s_Page_004.txt

lee_s_Page_045.txt

lee_s_Page_078.txt

lee_s_Page_041.txt

lee_s_Page_104.txt

lee_s_Page_055.txt

lee_s_Page_001.txt

lee_s_Page_072.txt

lee_s_Page_085.txt

lee_s_Page_122.txt

lee_s_Page_127.txt

lee_s_Page_062.txt

lee_s_Page_066.txt

lee_s_Page_067.txt

lee_s_Page_130.txt

lee_s_Page_094.txt

lee_s_Page_069.txt

lee_s_Page_076.txt

lee_s_Page_082.txt

lee_s_Page_144.txt

lee_s_Page_057.txt

lee_s_Page_115.txt

lee_s_Page_120.txt

lee_s_Page_025.txt

lee_s_Page_145.txt

lee_s_Page_028.txt

lee_s_Page_071.txt

lee_s_Page_080.txt

lee_s_Page_031.txt

lee_s_Page_023.txt

lee_s_Page_108.txt

lee_s_Page_105.txt

lee_s_Page_077.txt


Full Text












MECHANISTIC STUDY OF SORBENT INJECTION FOR VANADIUM EMISSION
CONTROL IN COMBUSTION SYSTEMS


















By

SANG-RIN LEE


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005


































Copyright 2005

by

Sang-Rin Lee

































This achievement is dedicated to my parents, Jung-Sik Lee and Ho-Suk Bae. My father
died in 1995, but his dedication and love remain in my heart. Without my mother's
encouragement and support, I could not have this achievement.















ACKNOWLEDGMENTS

My sincere gratitude is given to Dr. Chang-Yu Wu for his support and guidance in

this academic endeavor. It was the cornerstone of my life as an aerosol researcher when

he admitted me as his first Ph.D. student. His advice, encouragement and patience have

been invaluable during my graduate career. Special thanks go to Dr. Jean M Andino who

taught me the meaning of atmospheric chemistry and gave me continuous comments and

advice on my research. I sincerely thank Dr. David W. Hahn for allowing me to use his

Raman spectroscopy and for giving me advice on product identification and high-

temperature reactor design. I appreciate Dr. Wolfgang Sigmund's advice and research

comments. I give special thanks to Dr. Dale Lundgrun for allowing me to use his low-

pressure impactor. His abundant experience with and knowledge of aerosols are my

example of expertise.

I also thank the staff in Engineering Research Center and Major Analytic

Instrument Center. They teach and allow me to use many instruments such as ICP,

SEM/EDX, and BET. I would like to thank our past and present group members. Special

thanks go to Nawarat, David, Scott, and Shanna. I also thank the other air group graduate

and undergraduate students.

I appreciate the Korean students in the Environmental Engineering and Sciences

department. They helped to get me settled here in Gainesville. My special thanks go to

Dr. Jae-Hyun Cho and his family who treated me like family. I also appreciate my tennis

friends. They helped me to blow off stress and energize myself to work.









Finally I would like to give my sincere gratitude to my mother, Ho-Sok Bae. Even

though she lost her husband in 1995, she kept encouraging her only son, me, to study

abroad. Without her endless care and support, I could not have achieved this goal. I also

appreciate the support and patience of my four older sisters and their families.
















TABLE OF CONTENTS

Page

ACKNOW LEDGM ENTS ........................................ iv

LIST OF TABLES ................... .................. ......... .................................. ix

LIST O F FIG U RE S ........................................

KEY TO SYMBOLS OR ABBREVIATIONS ........................................... xii

ABSTRACT ................................ xiv

CHAPTER

1 GENERAL INTRODUCTION ....................................................... 1

2 VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS BY
THERMODYNAMIC EQUILIBRIUM ANALYSES .................................................8

Introduction ................. .. ........ ...................8
M methodology .................. .......... ................... .....................10
R results and D discussions ................... .... ... .. ................... ........... .................12
Set I: Baseline Behavior of Vanadium in a Typical Coal Combustion System
w without Sorbents ........................................ .......... .. ........ 12
Set II: Perform ance of Individual Sorbent..................................... ...................13
Set III: Effects of Chlorine and Sulfur on the Performance of Individual
Sorbent ................. .................... .... .. ........ . ...............15
Effects of sulfur on the performance of individual sorbent..........................15
Effects of chlorine on the performance of individual sorbent....................17
Set IV : Com petition am ong Three Sorbents .................................................... 19
Case IV-1: each sorbent is 33% of the stoichiometric amount of sulfur in the
sy stem ...................................... ... .... ..............19
Case IV-2: each sorbent is 40% of the stoichiometric amount of sulfur in the
system .............. .. . .......... .............. .. .. ...............2 1
Case IV-3 and IV-4: each sorbent is 66.7% and 110% of the stoichiometric
amount of sulfur in the system......................................................21
Conclusions....................................... ........ .24










3 SIZE DISTRIBUTION EVOLUTION OF FINE AEROSOLS DUE TO INTER-
COAGULATION WITH COARSE AEROSOLS .............................................26

Introduction .......... ......... ......... ......................... 26
Methodology .......................... .. ........... .........29
Model Description ........................ .......... ........29
Simulation Conditions ............... ..... ........ ...............32
Results and Discussion ............................................... ........32
N um ber Concentration ............................................... ............... 32
G eom etric Standard D aviation: ...................................................35
Mean Size Difference ...................... ........ ................38
Application to Sorbent Injection ....................... .. .......... ........ 41
Conclusions............................. ... ........46

4 MECHANISTIC STUDY OF SORBENT INJECTION TO CONTROL
VANADIUM EMISSION USING AEROSOL REACTOR............... ...............48

Introduction............................ ....................48
Experiments .................................................. ........50
Pot Experiment: Feasibility Study........................... ......... 50
A erosol R eactor Experim ent ........................................................ 51
Coagulation dominant system .......................................... 51
Condensation dom inant system ............. ............... .................... .......... 53
Product characterization.. ..................................... 55
M odel Description ............................................. ...............56
Results and Discussion ........................ ................. .......... 58
Pot Experiment ................ ........ ........ ..........58
A erosol R eactor Experim ent ........................................................ 59
Coagulation dominant ................. ................................59
Condensation dom inant................................ ................... 63
Bimodal Lognormal Model Study......... ..........................76
Case 1: Bimodal coagulation only and unimodal coagulation only..........76
Case 2: Condensation only ................................................. ........ 77
Case 3: 50% instant nucleation .............................................. ......78
Conclusions................................................ .80

5 CONCLUSION AND RECOMMENDATIONS ........................................ ...82

APPENDIX

A FORTRAN CODE FOR BIMODAL LOGNORMAL MODEL.............................86

B MOMENT RELATIONSHIPS FOR THE LOGNORMAL DISTRIBUTION........125

C M ATERIAL SAFETY DATA SHEET ................................................ ...............127









LIST OF REFERENCES ............................................. ......... ........129

B IO G R A PH IC A L SK E T C H ...................................................................................... 135
















LIST OF TABLES


Table Page

2-1. Potential sorbents and corresponding vanadium-sorbent compounds.....................11

2-2. Species used in the equilibrium calculations......................................................11

2-3. Simulation conditions for evaluating the performance of various sorbents in
capturing vanadium .............................. .................12

3-1. Simulation condition for investigation of the effects of inter-coagulation on fine
m ode particle rem oval. .................................................. ............... 32

3-2. Size distribution parameters and removal time for Linak et al. (2003)..................45

3-3. Operating parameters of a FGD system and particle size distribution from a
power plant ........................... ............................46

4-1. Feasibility study experimental conditions...... ............................................ 51

4-2. Coagulation dominant system experimental conditions............... ............ 54

4-3. Condensation dominant system experimental conditions ......................................56

4-4. Summarized simulation conditions ............................... ............... 58
















LIST OF FIGURES


Figure Page

1-1. Aerosol dynamic processes of vanadium in a combustion system vanadium only
and sorbent injection. ................................................ ...... ..2

1-2. Various types of coagulation for bimodally distributed particles ............................6

2-1. Partition of vanadium speciation in a typical coal combustion system .................14

2-2. Partition of vanadium species in a typical coal combustion system with chlorine
and sulfur .......................................................14

2-3. Partition of vanadium species in a coal-air-V-sorbent system.............................16

2-4. Partition of major vanadium species air-V-S02-sorbent system .............................18

2-5. Partition of major sorbent species air-V-S02-sorbent system .............................18

2-6. Partition of major vanadium species in a coal-air-V-HCl-sorbent system............20

2-7. Mole fraction of major vanadium species in the coal combustion system and
partition of sulfur in the coal combustion system ...................................... 23

3-1. Various types of coagulation for bimodally distributed particles .........................28

3-2. Fine mode removal time as a function of fine mode number concentration for
various coarse mode number concentrations. The dashed line connects points
where Nfo/N o= 10. ................................................. ....... .34

3-3. Dimensionless removal time as function of normalized number concentration and
normalized mass or volume concentration........................................................36

3-4. The evolution of fine mode (cgmo=1.4) and coarse mode (Ggco=1.3) particle size
distribution by inter-coagulation to reach removal time (tr=0.106 sec) ................37

3-5. A dimensionless removal time as function of fine mode geometric standard
deviation where gco=1.3. and coarse mode geometric standard deviation where
gfo= 1.3 ................... ..................................................... ......... 3 9

3-6. Dimensionless removal time and half removal time as function of dgc/dgf and fine
mode mean size .....................................................42









3-7. Removal time as function of coarse mode standard deviation and dgc/dgf with same
number concentration (1010/cm3) and with same mass concentration (10[lg/cm3)..43

4-1. Experimental set-up of the aerosol reactor system.......... ...................................52

4-2. Measured reactor temperature profile from bottom to top at 7400C. ....................54

4-3. XRD result for vanadium only at 673 and 873K ......................................60

4-4. XRD result for CaCO3 with vanadium at 673, 873, and 1073K .............................60

4-5. XRD results for Na2CO3 with vanadium at 673K................................ ..........61

4-6. Element PSD of vanadium and calcium for Set I and Set II at 7400C ...................62

4-7. Element PSD of vanadium with and without water vapor at 7400C ........................65

4-8. Morphology of vanadium oxide compound collected on fiber filter by hydrolysis
and thermal decomposition and by thermal decomposition............................66

4-9. Collected vanadium particles on filter with water vapor and without water vapor at
740 C ...................................... ................................... ........ 67

4-10. Element PSD of vanadium when water droplet injected at 740 C ..........................68

4-11. Element PSD of vanadium and calcium at 7400C by mass fraction and surface
area fraction................................ ..... ........ 70

4-12. Morphology of collected particles when Ca-based sorbent was injected and Ca-
based sorbent with VTIPO were injected............................................................. 71

4-13. SEM picture of whole product and EDX mapping of Ca and C) of V ....................72

4-14. SEM picture of single particle and corresponding EDX spectrum .......................72

4-15. Element PSD of vanadium and silica at 740 C ...................... .................................74

4-16. Morphology of product when silica only are injected and when silica with VTIPO
was injected 740 C ................... ...... .................... ......... .............. 75

4-17. The change of total number concentration of the fine mode, total volume
concentration of the fine mode, geometric standard deviation (og) of the fine mode,
MMD of the fine mode, Saturation ratio of V vapor, and MMD of the coarse mode
as function of time.. .. ... ............. .............. .. ............ 79















KEY TO SYMBOLS OR ABBREVIATIONS

0 subscript 0 represents initial condition

bo constant, f(o)
bo=0.633+0.092 02-0.022 a3

co continuum regime

dp particle size (|Jm)

dpg geometric mean size (|Jm)

f c subscriptf represents fine mode, c represents coarse mode

fm free molecule regime

Ko initial inter-coagulation rate

kB Boltzman's constant dynee cm/K)

n aerosol number concentration distribution function (#/cm3cm)

N total number concentration (#/cm3)

11 monomer number concentration at saturation (molecules/cm3)

mi mass of monomer (g)

M3f total volume concentration of free molecule regime (cm3/cm3)

M3c total volume concentration of continuum regime (cm3/cm3)

Mk kth moment of aerosol size distribution

S saturation ratio

t time (s)

T temperature (K)









tI/2 particle half life time (s)

tr particle removal time (s)

vI Volume of monomer (cm3)

p collision frequency function

A mean free path of air (cm)

P gas viscosity (g/cm s)

o-g geometric standard deviation

pp particle density (g/cm3)

Ts fine mode particle scavenge characteristic time (second)
















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

MECHANISTIC STUDY OF SORBENT INJECTION FOR VANADIUM EMISSION
CONTROL IN COMBUSTION SYSTEMS

By

Sang-Rin Lee

May, 2005

Chair: Chang-Yu Wu
Major Department: Environmental Engineering Sciences

Mechanistic study of sorbent injection for vanadium emission control in

combustion systems was conducted experimentally and theoretically. Potential sorbent

material for chemisorption was determined by thermodynamic equilibrium analysis. The

computer code, STANJAN, was used to implement the calculations. Ca-, Na- and Mg-

based sorbents were evaluated for a wide range of combustion temperatures. The strong

affinity between vanadium and these sorbents was identified which implies the great

potential of these sorbents to chemically adsorb vanadium. Sulfur was found to strongly

impair the performance of these sorbents at lower temperatures (<1000K) due to the

formation of sorbent sulfates that depleted the available sorbents in the system.

Bimodal lognormal model was applied to investigate the impact of inter-

coagulation rate on the size distributions of fine-mode aerosols. Fine mode particle

removal time was found to strongly depend on the number concentration of coarse mode

particles but independent on the number concentration of fine mode particles. A 60%









increase of geometric standard deviation of fine mode particles significantly increased the

dimensionless removal time. Fine mode particles ultimately approached monodisperse

when the dominant mechanism was inter-coagulation. Meanwhile, coarse mode particles

approached the asymptotic shape because intra-coagulation was the dominant

mechanism. On a constant mass, monodisperse and 1 |tm mean diameter are the optimal

condition for coarse particles to effectively remove fine particles through inter-

coagulation.

Aerosol reactor was applied for a mechanistic study of sorbent injection.

Condensation was found to be the preferred mechanism for sorbent injection based on

experimental results. CaCO3 sorbent which has strong chemical affinity with vanadium

and silica sorbent which has no chemical affinity but high surface area successfully

reduced vanadium submicron particle formation. Vanadium was highly concentrated

where surface area was high. Surface hydrolysis enhanced physical adsorption while gas

phase hydrolysis reduced the efficiency of sorbent technique by forming nano particles.

Bimodal lognormal modeling based on the experimental condition showed that

condensation was a very effective means to scavenge vanadium oxide vapor. A high

number concentration of fine particles by instant nucleation reduced inter-coagulation

rate and quickly scavenged vanadium oxide vapor. Therefore, enhancing condensation

while suppressing nucleation was shown to be critical to successful removal of vanadium

compound.














CHAPTER 1
GENERAL INTRODUCTION

Heavy metal emissions from combustion sources such as utility boilers and

incinerators are of great concern because of their adverse effects on human health and the

environment. Because of the increased concern, research is being conducted to assess the

actual exposure of human beings to toxic metals (Hogu, 2000), and vanadium is one of

these metals. Vanadium is an abundant metal constituent in coal, heavy oil, and

petroleum coke (Bryers, 1995; Linak and Miller, 2000; Swain, 1991; Yee and

Rosenquist, 1996). The Energy Information Administrator (EIA, 1996) estimated North

America emitted about 497 tons/year of Vanadium into the atmosphere in 2000, and 92%

of it was emitted by coal and oil combustion sources. Thus, power and heat-producing

plants using fossil fuel cause the most widespread discharge of vanadium into the

environment. In ambient fine particle characterization studies, vanadium is often an

excellent marker for oil combustion aerosol (Campen et al., 2001; Divita et al., 1996;

Tolocka et al., 2004).

Once emitted, vanadium can be transported long distances in the atmosphere,

resulting in adverse environmental and health effects (Bylinska, 1996). Vanadium is

known to be more toxic when inhaled and relatively less toxic when ingested (Boyd and

Kustin, 1984). It may also cause cardiovascular diseases, bronchitis, and lung cancer

(Yee and Rosenquist, 1996). There are also several reports of ecological disasters caused

by poorly controlled industrial emissions containing high concentrations of toxic metals

including vanadium (Lin and Chiu, 1995; Pirrone et al., 1999).









Condensation
Nucleation

E-4
E>4 Submicron
nt v vo Particles
Form vanadium Darticles
Coagulation



^~-- ^ *






S2 Finally form
Condensation --- Supermicron
I Sorbent vanadium vapor Particles

( Vanadium particle


Figure 1-1. Aerosol dynamic processes of vanadium in a combustion system
A) Vanadium only. B) Sorbent injection.

Vanadium, like many other metals in the fuel matrix, may enter the gas stream in

combustion systems by vaporization of volatile organic vanadium compounds, or by

entrainment of particles containing vanadium compounds. At high temperatures, it

undergoes various chemical reactions to speciate into different compounds such as

oxides, chlorides or sulfates, depending on the combustion environment and the

composition of the system. As the temperature decreases once the gas stream exits the

combustion zone, various aerosol dynamics such as (nucleation, coagulation and

condensation) proceed, resulting in the transformation of vanadium into the particulate

phase. Many studies have shown that metals undergoing this pathway generally form









aerosols in the submicron regime. Figure 1-1 provides a mechanistic description of such a

process. The resultant particle size distribution depends on the temperature history and

the existing particles in the system (Helble and Sarofim, 1989; Lighty et al., 2000; Wu

and Biswas, 2000). In a study on characterizing the particulate emissions from a large oil

fuel fired power plant, 88 wt% of vanadium was reported to be in the size range of 0.01

to 1.0 |jm (Bacci et al., 1983). Linak and Wendt (1994) summarized that vanadium as a

trace metal was enriched in submicron fly ash, in many coal combustion investigations.

Unfortunately, traditional particulate control devices have their minimum control

efficiencies in this size regime (Biswas and Wu, 1998; Flagan and Seinfeld, 1988; Linak

et al., 1993). The consequence is further illustrated by ambient particulate matter

measurement that showed vanadium enrichment in the submicron regime (Tolocka et al.,

2004). Thus, it is important to develop new techniques to effectively control vanadium

emissions.

Sorbent technique is one promising measure to control submicron particles. In

recent years, various studies have been conducted using mineral sorbents to capture

heavy metals. In most studies, sorbent particles are injected into combustion systems to

chemically adsorb heavy metals on the injected particle surface. As these sorbent

particles are typically in the supermicron range, the metal-sorbent particles can be

collected easily using traditional particulate control devices. Figure 1-lb illustrates the

mechanism of the sorbent injection technique. Shadman and co-workers (Scotto et al.,

1994; Uberoi and Shadman, 1991) used silica, alumina and various naturally available

materials (bauxite, kaolinite, and lime) to capture lead and cadmium. Linak et al. (1995)

used them to capture nickel, lead, and cadmium. Mahuli et al. (1997) tested hydrated









lime, alumina, and silica for arsenic control. Venkatesh et al. (1996) evaluated various

mineral sorbents constituting a spectrum of alumino-silicate compounds and a pulgite

clay for immobilization of several trace metallic species. Biswas and co-workers

(McMillin et al., 1996; Owens and Biswas, 1996; Wu et al., 1998) generated sorbent

particles with very high surface area in-situ to capture lead and mercury. However, no

study using mineral sorbent materials to capture vanadium has been conducted. Chlorine

and sulfur (common constituents of coal and oil) may react with the metal or with the

sorbent, thus reducing the effectiveness of the sorbent technique to remove the metal

(Linak et al., 1995; Linak and Wendt, 1993; Owens et al., 1995; Wu and Biswas, 1993).

Therefore, the impact of chlorine and sulfur on the sorbent process must also be

evaluated.

Figure 1-1 shows the aerosol dynamic processes in a combustion system (including

chemical reaction, nucleation, condensation, and coagulation) resulting in the evolution

of the particle size distribution that ultimately affects the fate of the metals. Consequently

the development of a dynamic aerosol size distribution model considering all these

mechanisms is instrumental to provide insights into the processes and to determine the

key mechanisms of a sorbent injection technique in such a complex system.

There are several aerosol models available to describe aerosol dynamic processes:

moment (Frenklach and Harris, 1987; Whitby, 1979), continuous (Tsang and Brock,

1982), and sectional model (Gelbard and Seinfeld, 1980; Landgrebe and Pratsinis, 1990;

Wu and Biswas, 1998). These models were categorized by their mathematical size

distribution function. Whitby et al. (1991) and Williams and Loyalka (1991) reviewed

these models in more details. Among them, aerosol moment model is one of the most









commonly applied, because of its flexible model structure and low computational cost.

Assuming a uni-modal log-normal aerosol size distribution, Lin and Biswas (1994)

developed a model to study metallic particle formation and growth dynamic during

incineration. Wu and Biswas (2000) also used a uni-modal lognormal size distribution

model to evaluate the effects of chlorine on the evolution of lead aerosol size distribution.

However, such models may not be appropriate for a metal-sorbent system, because of the

extreme differences between their particle sizes. Metals alone in a combustion system

eventually form submicron particles through nucleation, condensation, and coagulation.

The presence of supermicron sorbent particles may suppress the formation of submicron

metal particles through inter-coagulation, and condensation which are the key parameters

for the effectiveness of the sorbent technique (Biswas and Wu, 1998). Consequently, a

bimodal lognormal model will more pertinently represent a metal-sorbent system.

The sorbent processes can be divided into two steps: mass transfer (vanadium

transfer to the surface of sorbent), and followed by surface interaction. The mass transfer

step is condensation, and/or coagulation while the surface interaction is chemical, and/or

physical adsorption. In previous studies (Carey et al., 2000; Linak et al., 1995; Linak et

al.1998; Uberoi and Shadman, 1991), the system was generally condensation favored

condition because the temperature was very high where the metals were in the vapor

state. However, Friedlander et al. (1991) demonstrated that scavenging fine particles by

coarse mode particles through coagulation can be the dominant mechanism. Surface

interaction can be either physical adsorption or chemical adsorption. Previous sorbent

studies (Mahuli et al., 1997; Punjak et al., 1989; Uberoi and Shadman, 1991)

demonstrated that the dominant mechanism of surface interaction was chemical









adsorption, according to their XRD measurement data. Though its amount was much less

than chemical adsorption, physical adsorption was also identified in some sorbent-metal

systems (Chen et al., 2001; Punjak et al., 1989).

Nucleation of vaporized metal compound results in nano-size particles in the fine

mode. Eventually they can grow to the submicron regime by coagulation, and

condensation. However, their typical concentration, and short residence time in a

combustion system do not allow these fine particles to grow to the supermicron regime

(Biswas and Wu, 1998; Friedlander et al., 1991). Thus, coagulation with larger particles

is the only mechanism to remove submicron fine mode particles once they are formed.

Figure 1-2 conceptually depicts the two types of coagulation mechanisms (inter-

coagulation, and intra-coagulation) in a system consisting of bimodally distributed

particles. Intra-coagulation is self growth within the same mode. Meanwhile inter-

coagulation is the process to transfer the fine mode to the coarse mode which plays the

key role in the sorbent technique.





Inter-
Coagulation
a o> Intra-
Intra- coagulation
coagulation




Fine mode Coarse mode


Figure 1-2. Various types of coagulation for bimodally distributed particles









In summary, the sorbent technique is a promising measure to control vanadium

emission. Mechanisms of the sorbent injection technique were investigated theoretically,

and experimentally. Modal Aerosol Dynamic (MAD) model was used for theoretical

study. An aerosol reactor system was applied for experimental study. In Chapter 2, the

potential materials for sorbent injection technique were determined by thermodynamic

equilibrium analysis. The impact of chlorine, and sulfur on the effectiveness of the

process was also assessed. In Chapter 3, the evolution of fine mode particles due to

coagulation with coarse mode particles sorbentt) was investigated using a bimodal

lognormal model. Effects of size distribution parameters (such as number concentration,

standard deviation, and mean diameter on inter-coagulation) were evaluated. In Chapter

4, a feasibility study was conducted to verify the result of Chapter 2. Mechanistic

experiments using an aerosol reactor were then performed. The preferable mechanism

(condensation or coagulation) for the mass transfer process, and the surface interaction

was determined. Finally, the bimodal lognormal model developed in Chapter 3 was

applied to investigate the role of condensation, and nucleation in the sorbent technique. In

Chapter 5, conclusion of this work and recommendations were provided.














CHAPTER 2
VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS BY
THERMODYNAMIC EQUILIBRIUM ANALYSES

Introduction

Toxic metal emissions from combustion sources (such as utility boilers and

incinerators) are of great concerns. Because of the increased concerns, research is being

conducted to assess the actual exposure of human beings to toxic metals (Hogu, 2000),

and vanadium is one of these metals. Vanadium is one of the highly concentrated metals

in certain types of coal (Deluliis, 1993). In addition to coal, heavy oil and petroleum coke

also have been reported to have high concentrations of vanadium (Bryers, 1995; Linak

and Miller, 2000; Swain, 1991; Yee and Rosenquist, 1996). It is also found enriched in

combustion ash and interest has even been developed to recover vanadium from ash

(Alemany et al., 1998; Fang et al., 1998; Tsuboi et al., 1991) due to its high market

values. Once emitted, it can be transported to distance, resulting in adverse environmental

and health effects (Bylinska, 1996). It is known that vanadium may cause cardiovascular

disease, bronchitis, and lung cancer (Yee and Rosenquist, 1996). There are several

reports of ecological disasters caused by poorly controlled industrial emissions containing

high concentrations of toxic metals including vanadium (Lin and Chiu, 1995; Pirrone et

al., 1999). Even in the Arctic, the annual flux of vanadium is estimated to be 474 tonnes

(Akeredolu et al., 1994).

Vanadium, like many other metals in the fuel matrix, may enter the gas stream in

combustion systems by vaporization of volatile organic vanadium compounds or by









entrainment of particles containing vanadium compounds. At high temperatures, it

undergoes various chemical reactions to speciate into different compounds (such as

oxides, chlorides or sulfates) depending on the combustion environment and the

composition in the system. As the temperature decreases once the gas stream exits the

combustion zone, various aerosol dynamics proceed resulting in the transformation of

vanadium into particulate phase. The particle size distribution depends on the temperature

history and the existing particles in the system (Wu and Biswas, 2000; Helble and

Sarofim, 1989; Lighty et al., 2000). Many research studies have shown that metals

undergoing this pathway generally form aerosols in the submicrometer regime. In a study

on characterizing the particulate emissions from a large oil fuel fired power plant, 88

wt% of vanadium is in the size range of 0.01 to 1.0 |tm (Bacci et al., 1983).

Unfortunately, traditional control devices have their minimum control efficiencies in this

size regime (Flagan and Seinfeld, 1988). Thus, it is important to develop new techniques

to effectively control vanadium emissions.

In recent years, various studied have been conducted to use mineral sorbents to

capture heavy metals. Sorbent particles are injected into combustion systems and heavy

metals can be chemically adsorbed on the injected particle surface. As these sorbent

particles are typically in the supermicron range, the metal-sorbent particles can be easily

collected using traditional particulate control devices. Shadman and co-workers (Scotto et

al., 1994; Uberoi and Shadman, 1991) used silica, alumina, and various naturally

available materials (bauxite, kaolinite, and lime) to capture lead and cadmium. Mahuli et

al. (1997) tested hydrated lime, alumina and silica for arsenic control. Venkatesh et al.

(1996) evaluated various mineral sorbents constituting a spectrum of alumino-silicate









compounds and a pulgite clay for immobilization of several heavy metals. Biswas and co-

workers (McMillin et al., 1996; Owens and Biswas, 1996; Wu et al., 1998) generated

sorbents particles with very high surface area in-situ to capture lead and mercury.

However, no study using mineral sorbent materials to capture vanadium has been

conducted.

At high temperature environments, reactions are fast and equilibrium conditions

can very possibly be achieved. Hence thermodynamic equilibrium methods can be

applied to determine the potential sorbent materials (Lee, 1988) that have chemical

affinity with the metal. Equilibrium calculations have been applied to study the behavior

of metals like lead, arsenic and cadmium in combustion systems. Good agreement

between experimental data and theoretical predictions under certain conditions has been

reported, indicating equilibrium calculation to be a good tool for estimating the behavior

of metals in combustion systems (Biswas and Wu, 1998; Owens et al., 1995). The

objective of this study was to use thermodynamic equilibrium calculations to determine

effective materials that can chemically adsorb vanadium. Optimal conditions to achieve

high collection efficiencies were determined. The impact of various common constituents

in combustion systems on the performance was assessed. The most effective material for

controlling vanadium was determined.

Methodology

The computer code STANJAN (Reynolds, 1995) was used to implement the

equilibrium calculations. The principle of STANJAN is to minimize the Gibbs free

energy of the system by using the method of elemental potentials combined with atom

constraints. Thermodynamic data for all relevant species were obtained from the

literature (Barin, 1995; Chase et al., 1986). Table 2-1 lists the potential sorbent materials









(using their element as the representative) and their corresponding vanadium-sorbent

compounds. Table 2-2 lists the other species that were included in the calculations.

Table 2-1. Potential sorbents and corresponding vanadium-sorbent compounds.
Sorbent Potential Vanadium-Sorbent Compounds
Ca- Ca(V03)2, Ca2V207, Ca3(VO4)2
Mg- Mg(V03)2, Mg2V207
Na- NaVO3, Na3VO4, Na4V207

Table 2-2. Species used in the equilibrium calculations
Reactant Phase
Gas V, VCl2, VCl4, VO, VOCl3, VO2
V Condensed V, VCl2, VCl3, VCl4, VO, VO2, VOC13 VCl4, V203,
V204, V205
Gas Ca, CaCl2, CaS
Ca Ca, CaCl2, CaO, CaO2, CaSO3, CaSO4, CaS,
Condensed
Ca(OH)2, CaCO3
Gas Mg, MgCl2, MgOH
Mg Condensed Mg, Mg(OH)2, MgCO3, MgCl2, MgO, MgSO4

Gas Na, NaCl, NaOH, Na202H2, Na2SO4
Na Na, NaCl, NaOH, NaC10O4, NaHCO3, Na202H2,
Na2CO3, Na20, Na202, Na2SO3, Na2SO4
CO, CO2, 02, HC1, HOC1, H2S, NOC1, CIO, N, NO,
Common Gas
NO2, N2, N20, S, SO2, SO3, H2SO4, HNO3, H20
compound
Condensed H20, S, Coal

Calculations were conducted for four scenarios. Simulation conditions are listed in

Table 2-3. The concentrations of the various compounds correspond to the levels found in

a typical coal combustion system burning Eastern bituminous coal (Deluliis, 1993) with

20% excess air at 1 atmosphere. In order to establish the baseline, calculations were first

conducted in Set I to determine the behavior of vanadium in a typical combustion system

with no sorbent in the system. In Set II, the performance of each potential sorbent was

individually investigated for a wide range of temperatures. As chlorine and sulfur in coal

may very well react with vanadium or the sorbents, in Set III, parametric analyses were

conducted to evaluate the impact of these constituents on the performance of the










chemisorptions process. In Set IV, all the sorbents were included in the calculations.

Competition for vanadium among the sorbents was placed by varying their amounts to

determine the best sorbent. All calculations were performed for temperatures ranging

from 400 to 1700K.

Table 2-3. Simulation conditions for evaluating the performance of various sorbents in
capturing vanadium.
Set. V Coal 02 N2 SO, HCI CaCO3 Na20O MgO
1-1 0 0 0 0 0
I-2 0 8.25 x103 0 0 0
II-1 0 0 2.52x10-2 0 0
11-2 0 0 0 3.01x103 0
II-3 0 0 0 0 7.28x103
III-1 0.285 0 2.52x10-2 0 0
III-2 9.64 0.285 0 0 3.01xlO-3 0
111-3 xlO-5 1 29.7 111.7 0.285 0 0 0 7.28x103
III-4 0 8.25 x103 2.52x10-2 0 0
III-5 0 8.25 x103 0 3.01x10 0
III-6 0 8.25 x103 0 0 7.28x10-3
IV-la 9.49x10'2 9.49x10-2 9.49x10-2
IV-l1 3 0.114 0.114 0.114
0.285 8.25 x10
IV-05 0.190 0.190 0.190
IV-ld 0.313 0.313 0.313
a : each sorbent is 33.3% of the stoichiometric amount of sulfur.
b : each sorbent is 40% of the stoichiometric amount of sulfur.
c : each sorbent is 66.7% of the stoichiometric amount of sulfur.
d : each sorbent is 110% of the stoichiometric amount of sulfur.

Results and Discussions

Set I: Baseline Behavior of Vanadium in a Typical Coal Combustion System without
Sorbents

In the first set of calculations, the behavior of vanadium in a typical coal

combustion system was studied, and it served as the baseline for evaluating the

performance of the sorbents under various conditions. The partition of major vanadium

species in such a system is shown Figure 2-1.

As shown, divanadium pentaoxide (V205) is the dominant species in the entire

temperature range studied while V204 becomes increasingly important at high

temperatures. As discussed earlier, metal oxides formed in combustion systems generally









form submicrometer aerosols and hence are not desired (Biswas and Wu, 1998). Chlorine

is well known to have strong affinity for many metals (Owens et al., 1995; Wu and

Biswas, 1993). Therefore, its effect on vanadium speciation was also studied. The results

are shown in Figure 2-2. The result is very similar to that without chlorine (Figure 2-1). It

indicates relatively weak affinity between chlorine and vanadium, compared with earlier

studies on other heavy metals (Owens et al., 1995; Wu and Biswas, 1993). Hence, it is

probable that the presence of chlorine in typical coal combustion systems will not

significantly affect vanadium's speciation. Our finding agrees with the study of metals

from combustion of waste oil conducted by Nerin et al. (1999). Although the chlorine

content in the oil burn was very high, vanadium showed very low affinity with chlorine.

Sulfur has been also reported to possess strong affinity with certain metals (Biswas

and Wu, 1998). However, it was not included in the baseline calculations because there

were no thermodynamic data available for vanadium-sulfur compounds; hence, it was

expected to have no effect on the baseline calculations.

Set II: Performance of Individual Sorbent

In Set II the performance of individual sorbent was studied. The partition of

vanadium is shown in Figure 2-3A to 3C for Ca-, Na-, and Mg- based sorbent systems,

respectively. Since the goal of the sorbent technique is to chemically react vanadium on

sorbent particle surface, the total mole fraction of vanadium-sorbent compounds can also

be interpreted as the capture efficiency of the proposed sorbent process.

The results are very encouraging. As shown, the predominant products are

vanadium-sorbent species in all three systems (although they may be present in different

forms). Undesirable vanadium oxide is present only at extremely high temperatures.












- \/


VIN A\%
1000


/I\ /i\
1200


/1\ /I
1400


1600


Temperature (K)


Figure 2-1.


Partition of vanadium speciation in a typical coal combustion system. (c:
condensed phase, g: gas phase)


400 600 800 1000 1200 1400 1600

Temperature (K)


Figure 2-2 Partition of vanadium species in a typical coal combustion system with
chlorine and sulfur (c: condensed phase, g: gas phase)


100


80

60


40

20


-X- V204 (c)

-0- V205 (c)

-*- V02 (g)




\/ \1/ \L/ i\I \I / \1/ \/


0 0
400


/I\ /I\
600


A/\ ,Il
800


- ---


M


v









The strong affinity between vanadium and the sorbents demonstrates the excellent

potential of the use of these sorbent materials to effectively capture vanadium in

combustion systems. However it should be noted that there are some factors, such as

chlorine and sulfur in the system, that may prohibit the success of this process. Their

impact will be discussed more in the next section.

Set III: Effects of Chlorine and Sulfur on the Performance of Individual Sorbent

Chlorine and sulfur are very common in coal and other fuels. As discussed earlier,

they have been reported to affect the speciation of metals and sorbent materials in the

system. They may react with vanadium to form vanadium chloride or sulfate. In a pilot

study of combustion of residual fuel oil with a high sulfur content, Huffman et al. (2000)

identified vanadyl sulfate to be the major product in the fine particulate matter produced

in the combustion process. Chlorine and sulfur can also react with the sorbent materials

forming, for example, calcium sulfate (gypsum), thus reducing the available amount of

sorbents to react with vanadium. Hence, it is important to investigate their impact.

Calculations were performed for systems with sulfate and with chlorine. The results are

shown in Figures 2-4 and 2-5 for sulfur cases and in Figure 2-6 for chlorine cases.

Effects of sulfur on the performance of individual sorbent

As shown in Figure 2-4A, Ca-based sorbent becomes ineffective to react with vanadium

at lower temperatures (<900K) and V205 becomes the dominant species. The

ineffectiveness is due to the strong affinity between sulfur and calcium, which depletes

the available calcium in the system. This is evidenced by the predominant CaS04 shown

in Figure 2-5A (the partition of Ca compounds). Ca-based sorbent is effective only at

higher temperatures when CaS04 becomes less stable, releasing Ca for reacting with

vanadium.
















































ool O O O O O O O O O


80


60


40


20


800 1000 1200
Temperature (K)


1400 1600


Figure 2-3. Partition of vanadium species in (c: condensed phase, g: gas phase) A) a
coal-air-V-CaCO3 system B) a coal-air-V-Na20 system C) a coal-air-V-
MgO system


C


-e-Mg2V207 (c)
-E V204 (c)
---V205 (c)
-* V02 (g)

- -


II = ~-yu. 1


' *" -- ----









Similar to Ca-based sorbent, Na-based sorbent becomes ineffective at lower

temperatures (<800K) when sulfur is present in the system (Figure 2-4B). The strong

affinity between sulfur and sodium (forming Na2SO4) depletes the available sodium

sorbent in the system for capturing vanadium. The sorption process revives only at higher

temperatures when Na2SO4 becomes less stable (Figure 2-5B).

Compared to the other two sorbents, Mg-based sorbent is the most severely

affected by the presence of sulfur in the system. As shown in Figure 4C, it becomes

ineffective in the entire temperature range studied. This is due to the high affinity

between magnesium and sulfur forming MgS04 (Figure 2-5C). The depletion of Mg-

based sorbent by sulfur in the system results in the absence of magnesium vanadate.

The results discussed above clearly evidenced that the presence of sulfur in the

system significantly affects the performance of the sorption process.

Effects of chlorine on the performance of individual sorbent

When chlorine is present in the system, all sorbents still posses excellent capture

efficiency as shown in Figure 2-6. The results for Ca- (Figure 2-6A) and Mg- (Figure 2-

6C) based sorbents are similar to those with no chlorine. The results imply that neither

magnesium nor calcium has strong affinity to react with chlorine.

However, the products for the Na- case (Figure 2-6B) are different from the no

chlorine case (Figure 2-3B). Na3VO4 and Na4V207 are the dominant species at high

temperatures (above 800K) in the no chlorine case while NaVO3 is the dominant in the

same temperature range (except Na4V207 at approx. 1100K) in the chlorine case. The

shift of the Na/V ratio from 3 (Na3VO4) in the no chlorine case to 1 (NaVO3) in the

chlorine case indicates that there is competition between chlorine and vanadium for

sodium.


















































I 60
45
S40
o
0


400
400


600 800 1000 1200 1400 1600
Temperature (K)


Figure 2-4. Partition of major vanadium specie!
(c: condensed phase g: gas phase)
A) a coal-air-V-S02-CaCO3 system
B) a coal-air-V-SO2-Na2O system
C) a coal-air-V-S02-MgO system


- NaOH (g)

-E3-Na2SO4 (c)


-C- KMgSO4 (c)


400 600 800 1000 1200 1400 1600
Temperature (K)



s Figure 2-5. Partition of major sorbent
species(c: condensed phase g: gas phase)
A) a coal-air-V-S02-CaCO3 system
B) a coal-air-V-SO2-Na2O system
C) a coal-air-V-S02-MgO system


c~e ~ex>exxxxxxx xx3









Earlier studies have also shown the adverse effects of chlorine on alkaline metal (Na, K)-

based sorbents (Wu and Barton, 1999). Hence a higher level of Na- based sorbent in the

system is necessary to ensure the adsorption of both vanadium and chlorine.

Set IV: Competition among Three Sorbents

As discussed, Ca-, Na- and Mg- based sorbents all have demonstrated excellent

potential to capture vanadium in combustion systems. The natural question that comes

next is which of the three is the most effective sorbent. Four conditions were simulated to

determine the most effective sorbent by varying the amount of the sorbents in the system.

In the first case (IV-1), the total amount of all sorbents was the exact amount required to

stoichiometrically react with sulfur. The amount of sorbent was then increased and in the

last case (IV-4) each sorbent itself was enough to completely deplete sulfur in the system.

The results are shown in Figures 2-7A to 7D for the mole fraction of major vanadium

species and Figure 2-7E 7H for the partition of sulfur in the system.

Case IV-1: each sorbent is 33% of the stoichiometric amount of sulfur in the system

Because of the strong affinity of sulfur with all sorbents as discussed earlier, all

sorbents are consumed by sulfur in this case (Figure 2-7E) and no sorbent is available to

capture vanadium (i.e. forming vanadium-sorbent product) at temperatures below 600K

(Figure 2-7A). Thus, V20s appears at low temperatures. When sulfur's affinity with each

sorbent fades, some sorbents are released and start to react with vanadium. NaVO3 is the

dominant in the intermediate temperature range and Ca3(V04)2 is the major compound in

the high temperature range. This can be explained by the relative affinity of these

sorbents with sulfur shown in Figure 2-7E. Na2SO4 starts to gradually decrease from 600

K (though the change is indistinctive in Figure 7E at temperatures below 1200K).











100

80 A
-in- Ca2V207 (c)
S60 -a-Ca3V208 (c)

S 40

20

0

100
B
80
-e- NaV03 (c)
60 Na4V207 (c)
-- V02 (g)
40

20



100 E0 E)E)G

80 C

0 60 -e- Mg2V207 (c)
-o--V205 (c)
4 40
a 40 V02 (g)

20
2 20 -^V204 (c) ,


400 600 800 1000 1200 1400 1600
Temperature (K)

Figure 2-6. Partition of major vanadium species in (c: condensed phase g: gas phase)
A) a coal-air-V-HC1-CaCO3 system B) a coal-air-V-HC1-Na20 system
C) a coal-air-V-HCl-MgO system

As the concentration of vanadium is only a tiny fraction of that of sulfur (less than

0.05%), a slight release of Na-based sorbent would result in a high yield of vanadium-

sorbent product. Thus NaVO3 is the dominant species above 600K. CaS04 starts to

decrease rapidly from 1100 K and disappears at 1300 K. Consequently, a lot more Ca-









based sorbent than Na-based sorbent is available at high temperatures, resulting in the

dominance of calcium vanadate at higher temperatures. On the other hand, MgSO4 is

very stable in the entire temperature range studied. Thus, no magnesium-vanadium

compound is formed.

Case IV-2: each sorbent is 40% of the stoichiometric amount of sulfur in the system

In this case, totally 20% more sorbent was provided. As discussed earlier, the

stronger the affinity with sulfur, the less available the sorbent is. Mg-sorbent apparently

has the strongest affinity with sulfur as it is completely converted into MgSO4 (Figure 2-

7F, mole fraction is 40% at all temperatures). Meanwhile, Ca-based sorbent is shown to

have the weakest affinity with sulfur among the three sorbents. Only half of Ca-based

sorbent (20% out of 40%) forms CaSO4. Consequently, Ca-based sorbent is the most

effective one in most of the temperature range due to its abundance (Figure 2-7B).

However, Na-based sorbent is the most effective one at lower temperatures even though

the amount of Na-based sorbent is much less than that of Ca-based sorbent. Thus, Na-

based sorbent may have a stronger affinity with vanadium than Ca-based sorbent does.

The results imply that the ultimate fate of vanadium depends on the relative affinity and

quantity of these two sorbents with vanadium and sulfur. This will be further manifested

in the following sections.

Case IV-3 and IV-4: each sorbent is 66.7% and 110% of the stoichiometric amount
of sulfur in the system

In case IV-3, while Mg-based sorbent is still completely consumed by sulfur

(Figure 2-7G), more Na- and Ca-based sorbents are released, and the ratio of available

Na-based sorbent/Ca-based sorbent not consumed by sulfur increases. As shown in

Figure 2-7C, the dominant range of Na-based sorbent expands. The result suggests that









Na-based sorbent, if not impeded by sulfur, has a stronger affinity with vanadium in the

entire temperature range then Ca-based sorbent does. This is further evidenced in Case

IV-4 (Figure 2-7D), where the ratio of available Na-/Ca-based sorbent is approaching 1

and Na-V compounds are the dominant in the entire temperature range. The more

available the Na- based sorbent is, the less the calcium-vanadium product is. A field

measurement from an oil fuel fired power plant has also identified vanadium to be

present in the form ofNaVO3 (Bacci et al., 1983).

From the above analyses, it can be concluded that Na-based sorbent has the

strongest affinity to bind with vanadium. However, Na-based sorbent is also more

vulnerable to the sulfur in the system than Ca-based sorbent is. The affinity of the three

sorbents with vanadium and sulfur can be summarized as the following sequence.

Affinity with V: Na > Ca > Mg
Affinity with S: Mg > Na > Ca

It should be emphasized again that the competition of vanadium and sulfur for the

available sorbents in the system ultimately determines the fate of vanadium.

Nevertheless, it should be addressed that sorbent use is dependent upon mixing with the

stream, porosity, surface area and others. Scrubbing of smaller metal particles by larger

sorbent particles in the combustor is also a potential mechanism (e.g. fluidized bed

combustion; Fan et al., 1999). Very often, these factors are dynamically intertwined. For

examples, porosity and surface area change as sulfur is scavenged and temperature goes

up. The system studied is an ideal one where these factors are not considered.















^80

. 60
0
t 40
c,
o
H 20


108.

80

. 60
0
S40

20
S20


-80

60

S40
a)
o
on


400 600 800 1000 1200 1400 1600
Temperature (K)


Figure 2-7.


- CaS04 (c) E

--MgS04 (c)

--Na2SO4 (c)







-0- CaS04 (c)

MgS04 (c) F
--Na2SO4 (c)







CaS04 (c)
--MgS04 (c) G
--Na2S04 (c)













CaS04 (c) H
-a- MgS04 (c)
-A- Na2S04 (c)


400 600


800 1000 1200 1400 1600
Temperature (K)


Mole fraction of major vanadium species (c: condensed phase, g: gas phase)
in the coal combustion system where each sorbent is A) 33.3% B) 40% C)
66.7% D) 110% of the stoichiometric amount of sulfur; Partition of sulfur in
the coal combustion system where each sorbent is E) 33.3% F) 40% G)
66.7% H) 110% of the stoichiometric amount of sulfur









Another important point to be addressed is the leachability of metal from the

generated metal-sorbent product, a major consideration in the selection of a sorbent

material. If the collected ash is to be landfilled, the vanadium-sorbent compound should

have a low leachability. However, if vanadium is to be recovered from ash containing a

high concentration of vanadium (Alemany et al., 1998; Fang et al., 1998; Tsuboi et al.,

1991) soluble products are desired.

Conclusions

Vanadium is concentrated in various fuels and the emission of vanadium from

combustion systems is of concern. Mineral sorbents have been demonstrated to be

effective to control various toxic metals in combustion systems. In this study, equilibrium

calculations were conducted to identify potential sorbent materials to chemically adsorb

vanadium. Na-, Ca-, and Mg- based sorbents were found to be effective in a wide range

of temperatures. However, the presence of sulfur in the system significantly affected the

performance of these sorbents. Sulfur and sorbents were shown to have high affinity

(forming sulfates) at temperatures lower than 1000K. The strong affinity resulted in the

depletion of available sorbents in the system. Sufficient sorbent in excess of sulfur should

be provided in this temperature range to effectively capture vanadium compounds. At

high temperatures (> 1000K), the effectiveness of Na- and Ca-based sorbents to capture

vanadium revived as sorbent sulfates became less stable, releasing available sorbents to

react with vanadium. Meanwhile, Mg- based sorbent still showed very strong affinity

with sulfur in the entire range of temperatures, resulting in the worst performance among

the three sorbents. When sufficient sorbent is available, Na- based sorbent was found to

be the most effective one.






25


This study provides the insight of the reactions between vanadium and sorbents as

well as the impact of various operating conditions. The information obtained is important

for developing a better strategy for managing vanadium emission problems.














CHAPTER 3
SIZE DISTRIBUTION EVOLUTION OF FINE AEROSOLS DUE TO INTER-
COAGULATION WITH COARSE AEROSOLS

Introduction

Heavy metal emission from combustion sources is of great concern because of its

toxicity to human health and the environment. It is well known that metal particles from

combustion are enriched in submicron regime (Linak et al, 1993; Biswas and Wu, 1998).

Unfortunately, submicron particles show poor capture efficiency in traditional particulate

control devices (Flagan and Seinfeld, 1988). Sorbent technique is one promising

technique to control these fine metal emissions. In recent years, various studies have been

conducted that use mineral sorbents to capture heavy metals (Linak et al., 1995;

Venkatesh et al., 1996; Mahuli et al., 1997).

Sorbent particles injected into combustion systems are expected to chemically

adsorb heavy metals on the surface. As these sorbent particles are typically in the

supermicron range, the metal-sorbent product can be easily collected using traditional

particulate control devices. In addition to chemical adsorption, other mechanisms such as

nucleation and coagulation are also present in the system. At combustion source, removal

of metal vapor by condensation so that fine mode particle formation by vapor nucleation

can be suppressed is the preferred mechanism. Rodriguez and Hall (2003) developed an

aerosol dynamic model based on a hybrid sectional model to study condensational

removal of heavy metals from exhaust gases onto sorbent particles. Comparison of their

model results with experimental data (Rodriguez and Hall, 2001) showed good









agreement. Alternatively, after gas stream exits the combustion zone, coagulation with

coarse mode sorbent particles can be a possible mechanism to scavenge fine mode

particles (Fan et al., 1999). Friedlander et al. (1991) studied the scavenge of a coagulating

fine aerosol by a coarse particle mode. They derived a simple analytical criterion for

competing processes such as coagulation (intra-coagulation) and diffusion (inter-

coagulation). Based on the assumption of monodisperse nucleated fine particles and

entrained ash as the coarse particles, the criterion provides a convenient tool for

estimating the importance of these competing processes. In real system, however,

polydisperse fine particles are present and supermicron sorbent particles are injected.

While important, there has been no modeling study about the role of coagulation in

sorbent injection technique. The understanding of the evolution of fine mode particle size

distribution in such a system is of fundamental importance, though it is seldom studied.

Thus, a model to simulate the aerosol dynamics in the system can help understand the

dynamic interactions between metals and sorbents as well as implement sorbent injection

technique to real system.

There are several aerosol models available to describe aerosol dynamic processes

such as moment (Whitby, 1979; Frenklach and Harris, 1987), continuous (Tsang and

Brock, 1982), and sectional model (Gelbard and Seinfeld, 1980; Landgrebe and Pratsinis,

1990; Wu and Biswas, 1998). These models were categorized by their mathematical size

distribution function. Whitby et al. (1991) and Williams and Loyalka (1991) reviewed

these models in more details. Among them, aerosol moment model is one of the most

commonly applied ones because of its flexible model structure and low computational

cost. Assuming a uni-modal log-normal aerosol size distribution, Lin and Biswas (1994)









developed a model to study metallic particle formation and growth dynamic during

incineration. Wu and Biswas (2000) also used a uni-modal lognormal size distribution

model to evaluate the effects of chlorine on the evolution of lead aerosol size distribution.

However, such models may not be appropriate for a metal-sorbent system because of the

extreme differences between their particle sizes. Metals alone in a combustion system

eventually form submicron particles through nucleation, condensation and coagulation.

The presence of supermicron sorbent particles may suppress the formation of submicron

metal particles through inter-coagulation and condensation which are the key parameters

for the effectiveness of the sorbent technique (Biswas and Wu, 1998). Consequently, a

bimodal lognormal model will more pertinently represent a metal-sorbent system.

Figure 3-1 conceptually depicts the two types of coagulation mechanisms, i.e. inter-

coagulation and intra-coagulation, in a system consisting of bimodally distributed

particles. Nucleation of vaporized metal compound result in nano-size particles in the

fine mode. Eventually they can grow to the submicron regime by coagulation and

condensation.


Inter-
Coagulation coagulation


Intra-
coagulation





Fine mode Coarse mode
Figure 3-1. Various types of coagulation for bimodally distributed particles









However, their typical concentration and short residence time in combustion system do

not allow these fine particles to grow to the supermicron regime by intra-coagulation

alone (Biswas and Wu, 1998; Friedlander et al., 1991). Many measurements of metal

compounds emission have confirmed that metals are enriched in the submicron regime

(Bacci et al, 1983; Osan et al, 2000; Lyyranen et al, 1999). When coarse sorbent particles

are introduced into the system, fine mode particles coagulating with the sorbent particles

are then transformed into the supermicron regime, which can easily be collected by

conventional particulate control devices. As shown, inter-coagulation plays the key role

in removing fine mode particles and therefore its effect was the focus of this work.

In this study, the Modal Aerosol Dynamic (MAD) model (Whitby and McMurry,

1997) based on a lognormal moment model that has multi modal structure and low

computation cost was adopted. The evolution of fine mode particles due to coagulation

with coarse mode particles sorbentt) was investigated. The effects of size distribution

parameters such as number concentration, standard deviation and mean diameter on inter-

coagulation were evaluated.

Methodology

Model Description

The development of the MAD model adopted in this study was provided in detail in

Whitby and McMurry (1997), and hence is not repeated. Only inter-coagulation which

was the main mechanism to be investigated is discussed as follows. The inter-coagulation

rate for the fine mode is


(M ) = dp /dpf dp,)nf (dpf ,)n (dp,)ddpfddp, (3-1)
d(kf) -f -1)~f~u~


and for the coarse mode is









a k(k
J-^) p (c +dp')k /3(dpf dp)nf (dpf )n (dp)cddp, ddP
t (3-2)

0 p/k /(d/f c, dpc )f (dpf )n, (dp )ddfdcpc

where subscripts k is order of moment (0, 3 or 6) andf c stand for fine and coarse,

respectively. The definition of parameter is provided in the Nomenclature section. Once

these moments are determined, the key size distribution parameters can then be

determined according to


k2
Mk =Ndpexp(2ln2^) (3-3)


dp, =MkMk2 (3-4)


2 Mkl1\
In 2 = 2 In( )Mk (3-5)
g k1(k k) Mk

where


Mk = Mk
N
r
k2




(k, k,)

With these size distribution parameters, the evolution of fine mode particle size

distribution due to inter-coagulate with coarse mode particles can be investigated. In this

study, harmonic average method (Pratsinis, 1988) was used to calculate the collision

frequency function in the transition regime.









Inter-coagulation to remove fine particles is the key mechanism in this study.

Thus, the time needed to remove fine mode particles is of interest. The fine mode particle

scavenge characteristic time is defined as follows,

Ts No / Ko (3-6)

where Nfo is the initial total number concentration of fine mode particles and K, is the

initial inter-coagulation rate. Once Ko is calculated, the fine mode particle scavenge

characteristic time can be easily estimated with size distribution parameters. The

analytical solution (Whitby and McMurry, 1991) for the inter- coagulation rate of Oth

moment for the continuum regime can be expressed by


Ko co = Nf N(2kT/3fl)[2+1.392( ) 00783 2 exp[( )ln 2 ]+
dp dp^ 2

( )exp[21n2 g]exp[()In2 ug}+1.392( 22 )73 22 {expV[()1n2 gl+
dPgc 2 gdpgc dgc 2
dpgf 272gf d1gc 1 27p12
( )exp[21n2 g]exp[() In2 ugf]}+ + cexp[( )In ] exp[() In2 gC ]}1
dpgc 2 dpgc dp 2 2
(3-7)

and for the free molecule regime by


2 dpgf 1 2a ] \ dp]2
K,=NN 3kB Pb exp[ )n expp[( )ln' "gx[expI + P )I

2( )exp[( )ln 2 ]exp[(A)In2 7])+( ) exp[(9)2 a g]exp[21n2 Ug +
dpgcf 3 2 rdpf 1 u 8

(c )3 exp[21n2 o]exp[( )ln2 ug]+2( )exp[( )ln f]exp[( )ln2 8 ]gC


(3-8)


With emission control in mind, in this work the fine mode particle removal time (tr)

is defined as time to remove 99.99 wt% of fine mode particles. The dimensionless

removal time can then be defined as tr/cs









Simulation Conditions

Inter-coagulation rate depends on the size distributions of both modes. Hence the

effect of number concentration, geometric standard deviation and geometric mean size of

each mode on inter-coagulation rate was investigated. Table 3-1 summarizes all

simulation conditions. In the first set of simulation, the number concentration of each

mode was varied, and the corresponding size distribution and removal time of the fine

mode were determined. In the second set, the effects of the geometric standard deviations

of coarse mode and fine mode were studied. In the third set, the impact of varying the

geometric mean size of each mode on inter-coagulation rate was evaluated. The final set

of simulation was carried out to determine the optimal sorbent particle size distribution

that could enhance inter-coagulation rate, when the mass loading of sorbent particles was

fixed.

Table 3-1. Simulation condition for investigation of the effects of inter-coagulation on
fine mode particle removal.
dgf dgc
gf igc gf Ggc Nfo (#/cc) Neo (#/Cc)

Set I 0.01 10 1.3 1.3 104 10 106 1012

Set II 0.01 10 1.0-1.6 1.0-3.0 108 108

Set III 0.01-0.5 1-500 1.3 1.3 108 108

Set IV 0.01 1-100 1.3 1.0-3.0 106 104-1013

All simulations conducted for T = 7400C


Results and Discussion

Number Concentration

Since coagulation rate strongly depends on number concentration, inter-coagulation

is expected to play a key role in enhancing the fine particle removal when the number









concentration are high. Figure 3-2 shows the simulation results of set I. As shown, for the

same initial fine mode number concentration, the removal time strongly depends on the

coarse mode number concentration. It decreases as the coarse mode number

concentration increases. However, for a fixed coarse mode number concentration,

interestingly, the removal time is not affected by the change of the fine mode number

concentration unless the fine mode number concentration is 10 times more than the

coarse mode number concentration. Careful examination of the coagulation rates (intra

and inter) shows that when the fine mode intra-coagulation rate is much faster than the

inter-coagulation rate, the growth of fine mode particles decreases its number

concentration which subsequently yields a low inter-coagulation rate and therefore a

longer removal time. Thus, the graph can be divided into two regions as shown in Figure

3-2 by the dashed line: inter-coagulation dominant and intra-coagulation dominant. In

short, the coarse mode number concentration clearly plays a major role in determining

fine mode removal time when inter-coagulation is the dominant mechanism; on the other

hand, the fine mode number concentration must be considered to estimate the fine mode

removal time (tr) when intra-coagulation is the dominant mechanism.

These results were further presented by dimensionless removal time (tr/cs) as a

function of normalized number concentration Nfo/Nco in Figure 3-3A. As shown, all data

merge into one line, thus yielding a useful tool for estimating the dimensionless removal

time as


r= 33.7 NflNco < 10 (3-9)
S










t N,,
= 33.7exp(0.0007 )
s, N,,


Nf1/Nco >10


for the geometric standard deviation of the fine mode and the coarse mode of 1.3 and dgf

= 0.01[tm. Hence, as long as Ts is determined for a given system (Eq 3-6), the

corresponding removal time (tr) can be estimated accordingly. Since the prior study

(Friedlander et al., 1991) used half life time (tl/2), also included in the plot are the results

for tl/,/zT. The corresponding equations are


0 0 0 0 0 \0 0

Inter-coagulation
dominant

V V V V V V V


No=106
0 No=108
y NCO=1010
V NC=1012


Intra-coagulation
dominant


10-5 P


10-6 L
103




Figure 3-2.


V V V V V V V V


104 105 106 107 108 109 1010 1011 1012 1013 1014 1015 1016 1017

Nfo (#/cm3)

Fine mode removal time as a function of fine mode number concentration for
various coarse mode number concentrations. The dashed line connects
points where Nfo/Nco=10.


(3-10)


10-2 F










2 1.15 NfONco < 10 (3-11)


and

N
= 1.15exp(0.0003 f) Nf Nco >10 (3-12)
Z- NO

As shown, the same patterns are followed.

Since in practical applications, mass loading is of interest, Figure 3-3B shows

dimensionless removal time as a function of normalized mass/volume. The same trend

observed in Figure 3-3A exists. Dimensionless removal time is independent on

normalized mass or volume until it exceeds 108.

Geometric Standard Deviation:

It is well known that the geometric standard deviation of a log-normally distributed

uni-modal aerosol approaches 1.35 and 1.32 for free molecule regime and continuum

regime, respectively, when coagulation is the dominant mechanism (Pratsinis, 1988).

Parallely, this applies to intra-coagulation for multi-modal aerosols. However, the effects

of inter-coagulation on the geometric standard deviations of multi-mode aerosol have

never been investigated before.

These effects were studied as set II and the results are presented in Figure 3-4 for

the size distributions at various times. In this simulation condition (with sorbent

application in mind), inter-coagulation was dominant for fine mode. As shown, fine mode

became narrower with its mode size shifted toward the bigger size. It should be

emphasized that the increase of the mode size was not due to intra-coagulation. Under

intra-coagulation dominant case, there was no discernible change in the fine mode size

distribution during the simulation time (0.1 s; also marked in Figure 3-4). Rather, it











100


A

S S S S S S S *




10 1


t" t Is
0 tl1/2/



0

1 0 0 0 0 0 0 0 0 0 0 0



10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 101 102 103 104

Nfo/Nco


100






10
U)
Hi

H-


0 .1 1 .
10-18 10-17 10-16 10-15 10-14 10-13 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5

mfo/mo or VfJVo0


Figure 3-3. Dimensionless removal time as function of A) normalized number
concentration B) normalized mass or volume concentration


B S

* *








0
0 0 0 0 0 0 0 0 0 0 0
O


OOOOOOOOOO










resulted from the faster scavenge of the smaller sizes of the fine mode due to inter-

coagulation. Inter-coagulation rate is dependent on collision frequency function, which

strongly depends on particle size difference, and it increases as the size difference

increases. Consequently, the smaller size of the fine mode aerosol is scavenged much

faster then the larger size in the same mode. Therefore, when inter-coagulation is the

dominant mechanism, ogf (fine mode geometric standard deviation) ultimately

approaches monodisperse (i.e. ogf= 1). Meanwhile, intra-coagulation was dominant for

the coarse mode since the addition of the fine mode moments to the coarse mode is rather

insignificant. Although ogc (coarse mode geometric standard deviation) did not change in

the short period (fine mode removal time) shown in the figure, it will finally approach the

asymptotic value (1.32, not shown in Figure 3-4).



1013
Fine mode
1012 t 0 t=0.1 intra-coagulation only
1011 Coarse mode
t t=0.02
E 1010 t=0.04 t= 0 to0.1
109 t=0.06
1 t=0.08 /
S10 -= / t4 = 0.1 / \

106 =1. I t 0I
g /

106 -1
o =1.242 /
E 104 =1.3


102 =1.1761
101
0.001 0.01 0.1 1 10 100

particle size (jim)

Figure 3-4. The evolution of fine mode (cgfo=1.4) and coarse mode (Ggco=1.3) particle
size distribution by inter-coagulation to reach removal time (tr=0.106 sec)









Figure 3-5A shows the dimensionless removal time and the half life time (note: two

y-axes) as a function ogf, and Figure 3-5B shows the results for varying ogc. When ogf

changed from 1 to 1.6, dimensionless removal time increased drastically from 9.3 to 250

demonstrating the sensitivity to the change of the fine mode distribution. A larger ogf

implies more mass in the larger size part. The corresponding lower inter-coagulation rate

for the larger size resulted in a longer removal time as discussed earlier. Similar to the

previous set, an equation can be obtained as a useful tool for estimating the dimensionless

time due to the change of cgf as


= 0.0058897 exp(6.658o ) (3-13)


tl1 = 0.578 + 0.000276 exp(5.88uo ) (3-14)


In contrast, the dimensionless removal time changed within 3%, almost negligibly,

when ogc was varied from 1 to 3. Clearly demonstrated above is that there is no point of

using coarse mode particles with a wide size distribution to enhance inter-coagulation

rate. The geometric standard deviation of the fine mode particles plays a more important

role. Friedlander et al. (1991) assumed monodisperse nucleated particles (i.e. inter-

coagulation rate was the fastest). In many practical systems, aerosols are present as

polydisperse particles, including metal compounds from combustion system. As revealed

by this study, polydispersity needs to be considered and the assumption of monodisperse

particles underestimates the required time to scavenge polydisperse aerosol.

Mean Size Difference

If particle size difference is large, its collision frequency function is large which

enhances coagulation. Figure 3-6A shows the effects of mean size difference.











400


300 k-


U,

-20
20


10


40




30




20




10




0


Figure 3-5.


* tr/ts
0 tli/'/Tc


0


tl /2/Ts=0.578+0.000276exp(5.880cgf)

0


tr/ts=0.0058897exp(6.658cgyfg) O.. **


1.0 1.1 1.2 1.3 1.4 1.5 1.
Ggf


B









tr/ s
0 t112/2,






900009 0 9 9 9
1.0 1.5 2.0 2.5 3.0
Ggc
A dimensionless removal time as function of A) fine mode geometric
standard deviation where Ggco=1.3. B) coarse mode geometric standard
deviation where ogfo=1.3










As shown, when the mean size difference between the fine mode and the coarse

mode increased, the removal time decreased accordingly. This clearly shows that large

mean size sorbent (with same number concentration) will enhance inter-coagulation rate

due to the large cross-section surface area for coagulation. When the results are presented

in the dimensionless form as shown in Figure 3-6A, even though mean size difference

increased, dimensionless removal time showed only 3-7% difference for two orders of

magnitude of size ratio. The reason for the insignificant difference is due to the increase

of fine mode scavenge characteristic time for increased mean size difference.

In addition to size difference, fine mode mean size also affects the dimensionless

removal time. Their relationship is plotted in Figure 3-6B. A linear relationship in log

graph is observed which can be expressed as

t
r = -9.493 log(d ) + 14.814 (3-15)
S

t
12 =-0.12910og(d )+0.891 (3-16)
S

Considering the combined effects of number concentration, geometric standard

deviation and mean size, the following final form can be used to estimate the removal

time and half life time for Nf/Nco < 10,

t
r = 0.0058897exp(6.658c 9){ -0.28521og(d ) + 0.4296} (3-17)


t
/2 = {0.578 + 0.000276 exp(5.8 80 )} {-0.11261og(d) + 0.7748} (3-18)
S

and for Njo/Nco > 10









t Nf0
--= [0.0058897exp(6.658c ){-0.2852og(d )+0.4296}]exp(0.0007 ) (3-19)
T cO


1s1 N
= [{0.578+0.000276exp(5.880-g)} {-0.11261og(d )+0.7748}]exp(O.0003 fO) (3-20)


Application to Sorbent Injection

For many practical applications such as sorbent injection for toxic metal removal,

minimizing the mass loading while having good removal efficiency is desired as cost

consideration is always important to the feasibility of the technology. With the same total

mass concentration, there are several combinations of size distribution parameters

(number concentration, geometric mean size, and geometric standard deviation). In other

word, the optimal conditions based on mass concentration will be different from those

obtained for number concentration. Figure 3-7A shows the removal time for various

combinations of ogc and dgc/dgf based on the same number concentration. Figure 3-7B, on

the other hand, shows the results based on the same mass concentration. As discussed

earlier, a high number concentration of the coarse mode and a wide mean size difference

allow for a short removal time (shown in Figure 3-7A). However, different patterns were

observed for the results based on the same mass concentration (Figure 3-7B). As shown,

a narrow size deviation and a small coarse mode mean size had the shortest removal time.

Careful examination of the size distribution parameters reveals that under the same mass

concentration, a smaller mean size and a narrower geometric standard deviation are

translated into a higher number concentration, which is favorable for inter-coagulation. In

summary, for practical application sorbent mean size close to 1 |tm, monodisperse and a

number concentration larger than 107 #/cm3 are the most effective.










100


1000
d9c/dgf


40




30



1,
-< 20




10


1000'


0 1-
0.01


A













* tris '8
O tl/2s -
tr
E t1/2

0 i p .. -


Figure 3-6. Dimensionless removal time and half removal time as function of A) dgc/dgf
and B) fine mode mean size


10-2





10-3





10-4a0O





10-5





10-6
0


4




3




2 '




1




0


dgf(Lm)0-1












10-2
4





10-3
C



10U4


* d 9/d f,=100
0 d 9g/d ,=1000
V d g/d g,=10000

O O






V V


10-5 I -
1.0 1.5 2.0 2.5 3


* dgc/dgf=100
0 dgc/dgf=500
V dgc/dgf=1000


cgo 2.0


Removal time as function of coarse mode standard deviation and dgc/dgf A)
with same number concentration (1010/cm3) and B) with same mass
concentration (10,jg/cm3)


00000










V V V V V


10-2


1 0-3


10-4


10-5


10-6
1 .


0


Figure 3-7.









Our finding was also applied to an experimental study conducted by Linak et al.

(2003). Kaolinite sorbent was used to reduce fine particle formation. A 35% reduction of

metal compounds in the fine mode was observed. Condensation was reported to be the

responsible mechanism while coagulation was said to have no effect on the reduction.

The sorbent feed concentration was 680 mg/m3 with a mean size of 1.4 |tm, while the

fine particle feed concentration rate was 83 mg/m3, with a residence time of 4.1 s. Table

3-2 shows the size distribution parameters used in the calculation and the estimated

removal times. Three cases based on the same mass concentration were studied. For case

I, dgf was assumed to be 0.01 |tm and both geometric standard deviations were assumed

to be 1 for the most effective condition. Equations 3-19 and 3-20 were used to calculate

the dimensionless removal time and the half life time, and the fine mode particle

scavenge characteristic time was calculated using Equation 3-7. The resultant removal

time and half life time were 8.1x1085 and 8.6x1034 s, respectively, clearly showing that

coagulation could not work for their system since these values were much greater than

the residence time (4.1 Is). Case II was to simulate sorbent injection at a later stage where

intra-coagulation had lowered the fine mode number concentration (using measured PSD

of fine mode without sorbent that is reported in Linak et al., 2003). If the same amount

of sorbent was injected at this stage, its corresponding removal and half life times were

2,094 and 263 s, both still much longer than the residence time. Other conditions (e.g.

larger og) will only result in a longer removal time. Hence, the operating condition in

their system clearly was ineffective in removing the fine mode aerosol by inter-

coagulation. However, if the sorbent feed was increased two orders higher (68,000

mg/m3, Case III), the removal and the half life times were down to 2.65 and 0.12 s that









were shorter than the residence time. Using the formula developed in this work, the

reason why coagulation was not important in their study was clearly seen and the criteria

for effective removal by inter-coagulation can be clearly established.

Analysis was then carried out for an FGD (Flue Gas Desulfurization) system

(Harris et al., 1993) that uses slurry droplet injection. For this case, particulate matters

which are common components of flue gas also can potentially be removed by inter-

coagulation. Table 3-3 summarizes the operating conditions (Harries et al., 1993) and the

corresponding droplet size distribution. The particle size distribution of particulate matter

(Bacci et al, 1983) is also listed in Table 3. Particulate matter and slurry mass ratio is

about 1.07x10-6. However, the large slurry droplet (bigger than 500 [tm) with

monodisperse assumption results in a very low number concentration (2.3x102/cm3) that

requires an extremely long residence time to effectively remove particulate matter in the

flue gas. If the droplet mean size is 10 |tm instead and the geometric standard deviation is

1.3, its number concentration is 2.1x107 #/cm3 and the corresponding particle removal

time is only 0.5 sec. The analysis demonstrates that fine particulate matter can potentially

be removed using a typical mass loading in an FGD system but the selection of particle

size and number concentration is very critical to accomplishing the goal.

Table 3-2. Size distribution parameters and removal time for Linak et al. (2003).
dgf dge
S gf ogc Nfo (#/cc) Nco (#/cc) tr (sec) ti/2 (sec)


Case I 0.01 1.4 1 1 4.72 x1010 1.80x105 8.1x1085 8.6x1034


Case II 0.07 1.4 1 1 1.38x108 1.80x105 2,094 263


Case III 0.01 1.4 1 1 4.72x1010 1.80x107 2.65 0.12









Table 3-3. Operating parameters of a FGD system and particle size distribution from a
power plant.
FGD system (Harris, 1993) Particulate matter (Bacci et al., 1983)

Diameter: 1.25 m

Operating L/G: 0.015 m3/Nm3
Conditions Gas flow: 2000 Nm3/h N/A

Slurry flow: 30 m3/h
Actual
mo: 17.4x106 mg/m3
Droplet size > 500 |tm mfo: 18.7 mg/m3
ogc: 1 (assumed) MMD: 0.01 |tm (assumed)
Particle Size No: 2.3x102 #/cm3(calculated)
Distribution Dg :1.3 (assumed)
Desired
Droplet size : 10 tm Nfo: 7.8x109 #/cm3 (calculated)
ogc: 1.3 (assumed)
Nco: 2.1x107 #/cm3(calculated)

Conclusions

Sorbent injection technique is one promising method to control the emission of

submicron metal compounds from the combustion system. Under this bimodally

distributed condition, inter-coagulation can be a key mechanism for sorbent injection

technique.

In this work, the effects of inter-coagulation on removing fine mode aerosol were

studied. High number concentration of coarse mode, large mean size difference, and

narrow geometric standard deviation of fine mode were found to favor higher inter-

coagulation rate. The removal time can also be expressed in the dimensionless form. Two

key parameters that affect the dimensionless removal time are the geometric standard

deviation of the fine mode (when the normalized fine mode concentration is less than 10)









and the geometric mean size of the fine mode. The developed formula can be used as a

convenient tool to estimate the removal time once the particle scavenge characteristic

time is determined.

When inter-coagulation is the dominant mechanism for removing fine mode

particles, its geometric standard deviation approaches monodisperse. For coarse mode

particles, the geometric standard deviation approaches the asymptotic value because

intra-coagulation is still the dominant mechanism. With regard to sorbent application

where minimal mass loading is desired, a narrow size deviation and a small mean size of

the coarse mode yield the optimal effectiveness. The reason is due to the corresponding

high number concentration that is the critical parameter to inter-coagulation. Using the

formula developed in this work, the criteria for effective removal of fine mode aerosol

can be clearly established.















CHAPTER 4
MECHANISTIC STUDY OF SORBENT INJECTION TO CONTROL VANADIUM
EMISSION USING AEROSOL REACTOR

Introduction

Vanadium is one of the major trace components in coal and oil (Bryers, 1995;

Linak and Miller, 2000; Swain, 1991; Yee and Rosenquist, 1996). In a study on

characterizing the particulate emissions from a large oil fuel fired power plant, 88 wt% of

vanadium was reported to be in the size range of 0.01 to 1.0 |tm (Bacci et al., 1983).

Ambient particulate matter sampling at urban area used vanadium as a primary marker

for fuel oil combustion (Divita et. al., 1996) and showed it was highly concentrated in the

submicron regime (Tolocka et. al, 2004). Vanadium is known to be more toxic when

inhaled and relatively less so when ingested (Boyd and Kustin, 1984). It may also cause

cardiovascular diseases, bronchitis and lung carcinoma (Yee and Rosenquist, 1996).

Unfortunately, traditional control devices have their minimum collection efficiencies in

the submicron size regime (Biswas and Wu, 1998; Flagan and Seinfeld, 1988; Linak et al,

1993). Thus, it is important to develop new techniques to effectively control vanadium

emissions.

In recent years, various studies (Linak et al., 2003; Mahuli et al., 1997; Scotto et al,

1994; Uberoi and Shadman, 1991) have been conducted to use mineral sorbents to

capture heavy metals. Figure 1-1 illustrates the mechanism of sorbent injection technique.

As shown in Figure 1-1A, vanadium vapor will nucleate and then coagulate and/or

condense to form submicron vanadium oxides particles. When sorbents are injected,









vanadium vapor can be adsorbed on the surface of sorbent and nucleation rate will be

reduced (Figure 1-1B). As these sorbent particles are typically in the supermicron range,

the metal-sorbent particles can be easily collected using traditional particulate control

devices. Shadman and co-workers (Scotto et al, 1994; Uberoi and Shadman, 1991) used

silica, alumina and various naturally available materials (i.e. bauxite, kaolinite and lime)

to capture lead and cadmium. Linak et al. (1995) used them to capture nickel, lead and

cadmium. Mahuli et al. (1997) tested hydrated lime, alumina and silica for arsenic

control. Venkatesh et al. (1996) evaluated various mineral sorbents constituting a

spectrum of alumino-silicate compounds and a pulgite clay for immobilization of several

trace metallic species. Biswas and co-workers (McMillin et al., 1996; Owens and Biswas,

1996; Wu et al., 1998) generated sorbent particles with very high surface area in-situ to

capture lead and mercury. However, there have been few studies that examine the use of

mineral sorbent materials to capture vanadium.

The sorbent process can be divided into two steps: mass transfer (vanadium transfer

to the surface of sorbent) followed by surface interaction. If the metal is in vapor phase,

condensation is the mass transfer process. If sorbents are injected where the metal vapor

has nucleated, coagulation is the mass transfer process. In the previous studies (Carey et

al, 2000; Linak et al., 1995; Linak et al., 1998; Uberoi and Shadman, 1991), the system

was generally condensation favored condition because temperature was very high where

the metal was in the vapor phase. However, Friedlander et al. (1991) demonstrated that

the scavenging of fine particles by coarse mode particles through coagulation can be the

dominant mechanism. Surface interaction can be either physical adsorption or chemical

adsorption. Previous sorbent studies (Mahuli et al., 1997; Punjak et al, 1989; Uberoi and









Shadman, 1991) demonstrated that the dominant mechanism of surface interaction was

chemical adsorption according to their XRD measurement data. Though its amount was

much less than chemical adsorption, physical adsorption was also identified in some

sorbent-metal system (Chen et al, 2001; Punjak et al, 1989; lida et al., 2004). In such a

metal-sorbent system, there are fine mode particles formed by metal compound and

coarse mode particles of injected sorbent. Thus, a bimodal aerosol model which

incorporates various aerosol mechanisms can be a useful tool to investigate the dynamics

of the system. Lee and Wu (2004) has developed a convenient equation to estimate the

removal time (99.99 wt%) for sorbent injection technique using a bimodal lognormal

model when coagulation is the dominant mechanism.

Most studies have focused on demonstrating the ability of various sorbent materials

to remove metal vapors from the gas stream. In this study, a mechanistic study was

conducted to determine the preferable mechanism for sorbent injection technique. First,

sorbent material was selected by feasibility experiments based on their chemical affinity.

Coagulation only and condensation only cases were then conducted and compared.

Finally the effect of surface interaction was determined through the use of different

sorbent materials. Bimodal lognormal modeling was also conducted to gain insights into

the experimental results. With this study, the preferable process for mass transfer and

limiting process for sorbent injection technique was determined.

Experiments

Pot Experiment: Feasibility Study

To verify the thermodynamic equilibrium results (Chapter 2), pot experiments were

conducted. Two sorbent materials, CaCO3 (Fisher, powder, 99%, reagent grade) and

Na2CO3 (Fisher, powder, 99.5%, reagent grade) were tested. The furnace (Thermolyne, F-









A1730) was heated up to the designated temperature. A ceramic pot containing sorbent

and/or vanadium (Fisher, vanadium powder 99.5%) in 2% nitric acid was then placed

into the furnace. The sample was retrieved 10 minutes later. The products were identified

by X Ray Diffraction (XRD) and compared with the result of thermodynamic analyses

(Chapter 2). The experimental conditions are summarized in Table 4-1.

Table 4-1. Feasibility study experimental conditions
Conditions Molar ratio
sorbentt : vanadium) Temperature (OC)

Set I Vanadium only 0 : 1 400, 600

Set II CaCO3 + Vanadium 1.2 :1 400, 600, 800

Set III Na2CO3 + Vanadium 1.2 :1 400,600


Aerosol Reactor Experiment

Coagulation dominant system

Figure 4-1 shows the schematic diagram of the experimental setup of the aerosol

reactor system. Vanadium solution was prepared by dissolving elemental vanadium in

2% nitric acid which was introduced into the system by a Collison nebulizer (BGI,

TN25). The atomized vanadium containing mist was passed through a diffusion dryer to

remove the water content. The aerosol particle size can be controlled by varying the

properties of the atomized droplets according to the following equation (Hinds, 1999) .

d,=dd(Fv)l1/3 (4-1)

where, dp is diameter of final solid aerosol particle, dd is droplet diameter, and Fv is

volume fraction of solid material. Sorbent droplets were fed into the system using either a

Collison nebulizer (BGI, TN25) or ultrasonic nebulizer (Sonaer, 24M) right before the

inlet of the tubular aerosol reactor. The sorbent solution was also prepared by dissolving


































Figure 4-1. Experimental set-up of the aerosol reactor system

sorbent material in 2% nitric acid.An ultrasonic nebulizer can provide a higher number

concentration of sorbent than a Collison nebulizer can and therefore was used to

investigate the effect of number concentration change. The aerosol reactor (diameter:

1.651 cm, height: 45.72 cm, stainless steel) temperature was maintained by a tubular

furnace (Thermolyne 73900). Reactor temperature profile at 740 TC is in Figure 4-2.

Aerosol free dilution air through an ultra fiber filter (MSA, 76876) was introduced at the

exit of the tubular reactor to quench the reaction and aerosol dynamics. A Lundgren low

pressure impactor (LPI) was used downstream to collect and classify particles by their

aerodynamic size. It has 6 stages for atmospheric pressure and 6 stages for low pressure

(under 1 atm) which are designed to collect submicron particles. Glass fiber filter

(Millipore, AP2004700) was placed at the last stage of impactor to collect particles less









than the cutoff size of the 5th stage. Its cutoff diameters are determined by orifice

pressure and temperature (Lundgren and Vanderpool, 1988). When orifice temperature is

29.3oC and pressure is 209.3 mmHg, its cutoff sizes are 11.05, 5.54, 2.91, 1.73, 0.96,

0.50, 0.44, 0.29, 0.18, 0.09 [tm. Its upper limit (30 [tm) and lowest limit (0.01 [tm) of

cutoff size were determined where the mass cumulated curve was 100%. Apiezon L

grease (10% in benzene) was applied on stainless steel substrate to reduce bounce off

effect. To eliminate condensed water vapor, substrates were dried in desiccators for at

least one day before and after they were used. A highly sensitive micro balance

(Sartorius, MC 210S, 105g) was used to measure the sample mass.

Experimental condition

To investigate the effect of size distribution on the sorbent injection technique, the

concentration of vanadium solution was varied and the sorbent was generated by two

methods. Experiments were first conducted for vanadium alone to characterize its particle

size distribution. Experiments were then carried out by adding CaCO3 sorbents to

examine its effect on the size distribution of vanadium particles. In set I, a Collison

nebulizer (BGI, TN-25) was used and an ultrasonic nebulizer (Sonaer, 24M) was used in

set II because the ultra sonic nebulizer generated 10 times more sorbent particles

compared to the Collison nebulizer. Elemental calcium and vanadium mass concentration

in reactor were 10 and 0.21 mg/m3 for Set I, 12.7 and 0.07 mg/m3 for Set II. The

experimental conditions are summarized in Table 4-2.

Condensation dominant system

The experimental setup for the condensation dominant system is similar to that for

the coagulation dominant system except the section of vanadium precursor generation.




























300 400 500 600 700
Temperature C
Figure 4-2. Measured reactor temperature profile from bottom to top at 7400C.

Table 4-2. Coagulation dominant system experimental conditions


Solution concentration
(ppm)
CaCO3 Vanadium

0 500

23,000 500

0 50

23,000 50


Time
(hr)


Temperature Residence time
(0C) (sec)


740

740

740

740


0.34

0.34

0.42

0.42


To provide vanadium in the vapor form, Vanadium Tri-I-Propoxy Oxide (VTIPO,

Strem) which has a relatively high vapor pressure at room temperature was chosen. A

bubbler was used to provide a constant vanadium precursor supply by saturating 2 L/min

dry air with VTIPO vapor.


Set I-1

Set 1-2

Set II-1

Set 11-2









Experimental conditions

Experiments were first carried out for VTIPO. Since hydrolysis would be a key

mechanism for generating vanadium oxide, the effect of temperature and humidity was

investigated by using dry and humid air at room temperature and 740 TC. Finally, surface

hydrolysis effect was tested by injecting water droplet instead of water vapor.

Experiments were then carried out for two types of sorbent. Based on the

thermodynamic equilibrium study and feasibility study, CaCO3 was chosen for chemical

adsorption. Silica sorbent (specific surface area 170 m2/g, amorphous fumed, Alfa Aesar)

with high surface area but no chemical affinity with vanadium was selected for physical

adsorption. CaCO3 was dissolved in 2% nitric acid while silica was mixed and suspended

in 2% nitric acid. The sorbent particles were generated by the same ultrasonic nebulizer.

As baseline, the particle size distributions of CaCO3 and silica sorbent were measured

respectively. The sorbent was then introduced into the reactor together with VTIPO

vapor. Elemental calcium, silica, vanadium mass concentration at reactor were 40, 80,

and 9.25 mg/m3, respectively. The experimental conditions are summarized in Table 4-3.

Product characterization

Element size distribution: To determine the elemental distribution by size, the

collected particles at each impactor stage were dissolved in 2% nitric acid. The

concentration of each element (V and Ca) was measured by Inductively Coupled Plasma

emission spectroscopy (ICP, Perkin-Elmer Plasma 3200). The total amount at each stage

was then determined by multiplying the concentration by the solution volume (50 mL).

Product morphology and elemental mapping: Scanning Electron Microscopy

(SEM, JEOL 6330) / Energy Dispersive X-ray (EDX) was used for surface morphology

and mapping elemental distribution on the surface of particles.












Table 4-3. Condensation dominant system experimental conditions
Vanadium Sorbent in 2% Nitric acid (10 g/L) Temperature
Temperature T,.,
Carrier air (oC)
Carrier air Carrier air feedrate time (sec)
Material feedrate Material (L/min)
(L/m(L/min)
(L/min)
Set I-1 VTIPO 2 dry air 5 room

Set 1-2 VTIPO 2 humid air 5 room 0.3

Set 1-3 VTIPO 2 dry air 5 740 0.3

Set 1-4 VTIPO 2 humid air 5 740 0.3

Set 1-5 VTIPO 2 Water 5 740 0.3

Set II-1 Air 2 CaCO3 5 740 0.3

Set II-2 VTIPO 2 CaCO3 5 740 0.3

Set III-1 Air 2 Silica 5 740 0.3

Set III-2 VTIPO 2 Silica 5 740 0.3


Product speciation: X-Ray Diffraction (XRD, Philips APD 3720) was used to

identify crystalline species of collected particles. Raman spectroscopy (Confocal system,

632-nm excitation) was also used to identify species that were not in crystalline form.

Model Description

The Modal Aerosol Dynamic (MAD) model that was described in detail in Chapter

3 was used for this study to simulate and examine the aerosol dynamics in the system. In

this study, coagulation and condensation were compared to determine the effective

mechanism that was responsible for capturing vanadium.

The condensational volume growth for the fine mode (free molecule regime) is


= B, (S 1)fd2n(d )d(d ) (4-2a)
(t P P P


and that for the coarse mode (continuum regime) is









d3 = B (S -l)fJOdn(d )d(d ) (4-2b)
dt P P P


where, = 36 B and B3 = 9;2 AvI B Definition of the
N 27mn, \ 9 r 7"na

variables can be found in the Nomenclature section. Condensational volume growth is

proportional to saturation ratio and its total surface area for the free molecule regime and

total diameter for the continuum regime, respectively (Prastinis, 1988).

Volume change by inter-coagulation (Whitby and McMurry, 1997) for the fine

mode and the coarse mode are given by


a(M3) = dp 8(dp,,dp)nf (dpf n (dp )ddpddp, (4-3a)


a (M )=J (dpi +dp3) fP(dpf ,dp,)nf (dpf )n (dpc)ddpfddpc
at 0 (4-3b)
dp(dp,, dp)nf (dpf )nc (dp )ddpfddpc

Nucleated particles are in the free molecule regime and sorbent particles are in the

continuum regime. Volume of the fine mode particles is reduced while volume of the

coarse mode is increased by inter-coagulation.

The initial conditions of the simulations were based on the experimental condition.

Metal vapor was assumed as vanadium pentaoxide. Three cases were simulated. First, all

vapor was assumed to have nucleated instantly. Simulations were conducted with sorbent

particles and without sorbent particles. Mechanism for the case with sorbent particles was

bimodal coagulation which includes intra-coagulation and inter-coagulation (case la).

Mechanism for the case without sorbent particles is unimodal coagulation which is intra-

coagulation (case Ib). Second, nucleation was suppressed and condensation was the only

mechanism allowed (case 2). Finally, 50% of vanadium vapor assumed to have instantly










nucleated while the remaining 50% was in the vapor phase. Both condensation and

coagulation were possible mechanisms in such a scenario (case 3). The simulation

conditions are listed in Table 4-4.

Table 4-4. Summarized simulation conditions
M10 dpi M20 dp2 Mechanism
S(#/cm3) () (A) 2 (#/cm3) (m) (mmHg) inpessue estimated

Casela 1 6.210x103 6.285 2.087 3.89x105 1.07 N/A Intra-, inter-
coagulation
Caselb 1 6.210x1013 6.285 N/A N/A N/A N/A Intra-coagulation

Case2 N/A N/A N/A 2.087 3.89x105 1.07 0.006572 Condensation
Intra-, inter-
Case3 1 3.106x101 6.285 2.087 3.89x105 1.07 0.003496 coagulation
Condensation

Results and Discussion

Pot Experiment

Figures 2-1 and 2-2 show the partition of vanadium species in a typical coal

combustion system without and with chlorine and sulfur, according to equilibrium

analysis. Vanadium pentaoxide was the dominant product until the temperature reached

1000 K. Figure 4-3 shows the XRD results for vanadium only case at 673 K and 873 K.

These compounds were identified as vanadium pentaoxide which agreed with the

prediction by the thermodynamic equilibrium analysis.

Figure 4-4 shows the crystalline species identified when CaCO3 sorbent was added

to the pot in addition to vanadium. It clearly showed that different compounds were

formed at different temperatures. Ca2V207 and Ca3(VO4)2 were the major compounds.

Other products such as CaO and Ca(OH)2 were also detected since Ca-based sorbent was

more than the stoichiometric amount of vanadium. Thermodynamic equilibrium









calculations showed that different dominating compounds at different temperatures

(Figure 2-3A) and they well matched the results of the pot experiment.

Figure 4-5 shows the result when Na- based sorbent was added. It also well

matched with the thermodynamic calculation (Figure 2-3B). This figure showed only the

result at 673 K because the material produced at 873 K was extremely sensitive to

ambient air moisture and formed damp material that could not be identified by XRD.

The feasibility study discussed above verified the results of thermodynamic

equilibrium analysis and clearly demonstrated both sorbents' ability to chemically bond

with vanadium. It should be noted that the residence time in a typical combustion system

is much shorter than the residence time in the pot experiment. The ability of the sorbent

to capture vanadium during the short flight time therefore needs to be evaluated in an

aerosol reactor. Due to the difficulty to identify the speciation of Na- based sorbent

system by XRD, Ca-based sorbent was selected for further study in the aerosol reactor.

Aerosol Reactor Experiment

Coagulation dominant

After vanadium vapor has nucleated, coagulation is the only mechanism to remove

these submicron particles. Hence, coagulation dominant case was also investigated by

introducing vanadium in particulate form. The element size distributions of vanadium

only and vanadium with sorbents are shown in Figure 4-6. For both cases, the mass

median diameter (MMD) of vanadium increased slightly (0.55 to 0.63 |jm for Set I and

0.40 to 0.44 for Set II) when Ca sorbent (MMD was 0.74 and 0.97 |tm for Set I and II)

was fed into the system. As discussed in Chapter 3, the number concentrations of fine and

coarse mode particle are the key parameters that determine whether inter-coagulation or
































18


2B


38


48


58 6B r[2e1 78


XRD result for vanadium only at 673 and 873K


10 20 30 40 58 60 ['20]
Figure 4-4. XRD result for CaCO3 with vanadium at 673, 873, and 1073K


600
[counts]

S00


400


380


288


Iee*


a-


673



I ilv~iJy


Figure 4-3.


800-
[counts] -
780-


608-


588


488-


380-


288


100-
10 -


I r 1









[GountsB -
58e


ONaVO3 peaks







I":
3BB -









18 28 30 4B 58 6B 'ZB81 7B
Figure 4-5. XRD results for Na2CO3 with vanadium at 673K

intra-coagulation is the dominant mechanism. The number concentration of vanadium

particles was 2.27x106/cm3 and 2.37x106/cm3 for set I and II, respectively; and the

number concentration of Ca-based sorbent was 5.43x103/cm3 and 6.45x104/cm3 for set I

and II, respectively. Assuming the best condition (monodisperse aerosol for both fine

mode and coarse mode) for inter-coagulation, its fine mode 99.99% removal time and

half removal time following Eqs. 3.19 and 3.20 derived in Chapter 3 were more than 7.6

and 1.8 hours. Since the residence time of these sets of experiments was only 0.3

seconds, the theoretical analysis clearly showed that inter-coagulation could not work as

the major mechanism to remove the fine mode vanadium particles under this

experimental condition. Typically the flue gas residence time in a combustion system is

less than 10 seconds. Thus, the proper number concentration of sorbent particles to

reduce the removal time to less than 10 seconds should be more than 107/cm3. The result

agrees with the observation reported by Linak et al. (2003) who applied kaolinite sorbent

to reduce fine particle formation. Its number concentration of kaolinite sorbent was

1.80x105/cm3 while that of fine mode metal aerosol was 4.72x1010 #/cm3. They reported
















W V at Vonly


I V at Vonly


m V at Ca+V


1.0 I


m Ca at Ca+V


0.0 1
0.0


4 I I 1


0.1
Dp (uam)


1.5
CL

.0
1.0
2


1.5 -
0
0)
o

2


0


m Vat Ca+V

-^


I Ca at Ca+V

l


0n i 1 i 1 1i


0.01


0.1
Dp (uam)


Figure 4-6. Element PSD of vanadium and calcium A) for Set I and B) for Set II at
7400C


II I


1.5 k


1.5 -


11


Tf
JLS









that 35% reduction of fine particles was due to condensation and there was no

coagulation effect on fine particle reduction. To accomplish fine mode removal by

coagulation, the sorbent feed needs to be increased at least 2 orders higher (See Table 3-

2).

Condensation dominant

Vanadium precursor characterization. VTIPO forms vanadium oxide compound

by hydrolysis and/or thermal decomposition. As baseline, VTIPO vapor was introduced

into the reactor with dry air (no hydrolysis) at room temperature (no thermal

decomposition). As expected, there were no particles collected by the LPI. To compare

hydrolysis and thermal decomposition, VTIPO was introduced into the reactor with

humid air and dry air at 7400C respectively. With dry air, only thermal decomposition

occurred, while both thermal decomposition and hydrolysis occurred with the presence of

water vapor. Finally instead of water vapor, water droplets generated by ultra sonic

nebulizer were introduced into the system. In this case, hydrolysis would occur on the

surface of the water droplet.

Figure 4-7 shows the PSDs of vanadium element in dry air and humid air at 7400C,

respectively. These two cases showed very similar distributions. Vanadium mostly

concentrated in the filter stage that is smaller than 0.178 |tm. When vanadium precursor

(VTIPO) undergoes hydrolysis or thermal decomposition, it forms vanadium oxides.

These vanadium oxides quickly nucleate because their saturation vapor pressure is very

low (1.12xl108 mmHg and Saturation ratio: 5868) at 7400C. Instantly, nucleation results

in a burst of extremely high concentration of nanoparticles. These particles will grow

following coagulation and/or condensation. However, the short residence time of the









aerosol reactor (only 0.3 seconds) would only allow the growth to very fine sizes. Figure

4-8 shows the morphology of vanadium oxide compound collected on the filter. When

water vapor was present, vanadium oxide formed well shaped spherical particles

indicating the strong effects of condensation (Figure 4-8A). On the other hand vanadium

oxide formed by only thermal decomposition showed much smaller primary particles

implying the importance of coagulation (Figure 4-8B).

VTIPO is one of metal alkoxides which undergo hydrolysis. The rate of hydrolysis

of a metal alkoxide depends on the characteristics of the metal and those of the alkyl

group. In general, silicon alkoxides are among the slowest to hydrolyze, and for a given

metal alkoxide the hydrolysis rate increases as the length of the alkyl group decreases

(Rahaman, 1995). There are limited studies about the hydrolysis of VTIPO. However,

Titanium TetralsoPropoxide (TTIP), which has the same alkyl group, is a common

material to make TiO2 nanoparticles and there are many studies on hydrolysis and

thermal decomposition of TTIP. The reaction rate constant is 3.96xl05exp(-8479.7/T) for

gas phase thermal decomposition (Okuyama et al., 1990) and 3xlO15exp(-1013.9/T) for

gas phase hydrolysis (Seto et al., 1995). Hydrolysis clearly is much faster than thermal

decomposition. Based on hydrolysis of TTIP, it is likely that hydrolysis of VTIPO is also

faster than thermal decomposition of VTIPO. Supporting this analysis, the total amount

of vanadium element collected in humid air was 5 times more than that in dry air in this

study. Thermodynamic equilibrium analysis study in Chapter 2 showed that the stable

compound for vanadium was vanadium pentaoxide. Assuming that hydrolysis of VTIPO

generates vanadium pentaoxide, VTIPO hydrolysis can be given as follows.

2VO(OC3H7)3+3H20 -> V205 + 6C3H70H










Figure 4-9 shows the collected particles on the filer for both cases. With humid air

the particles collected were dark green, while with dry air yellow particles were observed.

Its final product of vanadium oxide compound could be different based on the

observation of the color of collected particles. These particles were characterized by

XRD although no crystalline species were identified. Unfortunately Raman spectroscopy

also could not identify the species.

The final set of experiment was injecting water droplets to the reactor instead of

feeding water vapor. The mass mean diameter of water droplets generated by ultrasonic

nebulizer was 1.7 |tm (Sonaer, 24M). As shown in Figure 4-10, vanadium PSD with

water droplets was shifted to much larger size. Gas phase hydrolysis generated

nanoparticles. However, the surface of water droplets was also the active site for

hydrolysis once VTIPO vapor diffused to the surface. The surface hydrolysis

consequently captured vanadium and reduced the fine particle formation by gas phase

hydrolysis.


10
0 8 I Vw water vapor
S06






0 8 Z V w/o water vapor
0 6 -

S04
02

00
001 01 1
Dp (im)

Figure 4-7. Element PSD of vanadium with and without water vapor at 7400C


























































Figure 4-8. Morphology of vanadium oxide compound collected on fiber filter A) by
hydrolysis and thermal decomposition and B) by thermal decomposition























Figure 4-9. Collected vanadium particles on filter A) with water vapor B) without water
vapor at 740 TC

Sorbent injection. Since vapor diffusivity is relatively large, the rate limiting step for

effective capture of vanadium vapor for the condensation case will be the surface

interaction. Two sets of experiments were conducted to determine the effective surface

interaction in the system. In the first set Ca-based sorbent, which showed strong chemical

affinity with vanadium in the equilibrium analysis, was used. In the second set Si- based

sorbent which has high surface area but no chemical affinity with vanadium, was studied.

Comparing results of these two experimental conditions helps reveal the preferable

mechanism for surface interaction: chemisorption or physisorption.

Case 1. Ca-based sorbent. Figure 4-10 displays the element PSD of vanadium and

calcium. When only VTIPO was introduced, most vanadium was concentrated in the

filter stage. After CaCO3 sorbent was injected, vanadium PSD was shifted to around 1

|tm size range. Since condensation depends on the surface area of sorbent, the surface

area fraction distribution of the sorbent was also plotted in Figure 4-11. As shown, PSD

of vanadium when sorbent was injected is very similar to the sorbent surface area fraction

distribution, verifying condensation to be the key mechanism in the system.







68



1.0 -

0.8 ----V
0 m Vw water vapor

0) 0.6 -

2 0.4 -

0.2 -

0.0
2.0 -


1.5 I I V w water droplet

0)
<1.0 -


0.5 -


0.0
0.01 0.1 1
Dp (utm)

Figure 4-10. Element PSD of vanadium when water droplet injected at 740 oC

Figure 4-12A shows the morphology of CaCO3 sorbent particles in the range of

1.73-0.96 |jm. Instead of being present as individual particles in that size range, the

collected material appeared to be a big chunk that resulted from merging. It should be

noted that all collected materials were placed in a desiccator before characterization.

However, the strong hygroscopic property of CaCO3 still yielded merging of the material.

Figure 4-12B shows the morphology of particles collected on the same size range when

both vanadium and the sorbent were present. It is interesting to observe individual

spherical particles in this case, totally different morphology from the sorbent only case.

The comparison of these morphologies indicate that when vanadium oxide compounds









were well coated on the surface of sorbent particles, those particles became insensitive to

moisture.

In the latter case, vanadium oxide compound formed by surface hydrolysis will

deposit on the surface of sorbent. Figure 4-13A shows SEM/EDX for the particles when

both CaCO3 and vanadium were present while Figure 4-13B is the Ca mapping and

Figure 4-13C is the V- mapping. As shown in Figures 4-13B and 13C, vanadium was

much widely distributed than calcium was. If surface chemical reaction forming calcium-

vanadium compounds was the main mechanism, these two mapping should show a

similar distribution pattern. Figure 4-14 is a SEM picture of one single particle and its

corresponding EDX spectrum. As shown, the intensity of vanadium is much stronger than

that of calcium. Since calcium and vanadium has similar atomic number, its

corresponding concentration based on intensity for vanadium was higher than calcium.

However, this is relative value. For the precise quantification of measured particles the

calibration of EDX with standard sample are needed. EDX measurement could show that

calcium and vanadium were on the one single particle. However, it could not determine

the surface interaction.

Case 2. Si- based sorbent. Since chemical affinity between vanadium and the sorbent is

not necessary, experiments were carried out using silica to verify if physical adsorption

alone can be a possible mechanism for the surface interaction. Silica is hydrophilic but it

does not dissolve in water. Thus, silica was selected to investigate the hypothesis. It was

well mixed and suspended in 2% nitric acid, then sonicated. Silica sorbent aerosol was

generated using the same ultrasonic nebulizer. As baseline, silica sorbent PSD was







70




1.0 -


0.8 V at V only
.0
0 0.6 ____________


I 0.4 -


0.2 -


0.0
2.0


1.5 I V at Ca+V

0)
1.0 -


0.5 -






o1. Ca surface at Ca + V
S 1.5
0.5
tlO





1.5 Ca at Ca + V





0)
1.0 -







0.5 -















Figure 4-11. Element PSD of vanadium and calcium at 740T by mass fraction and
1.0


0.5


0.0
0.01 0.1 1
Dp (jim)


Figure 4-11. Element PSD of vanadium and calcium at 7400C by mass fraction and
surface area fraction

























































Figure 4-12. Morphology of collected particles when A) Ca-based sorbent was injected
and B) Ca-based sorbent with VTIPO were injected

















Figure 4-13. SEM picture of A) whole product and EDX mapping B) of Ca and C) of V


Full scale = 392 counts


Cursor: 20.1275 keV


S4 F. 10 12 14 1E. 1:3

Figure 4-14. SEM picture of A) single particle and B) corresponding EDX spectrum


B









measured. Experiments were then carried out by injecting sorbent with VTIPO to the

reactor.

Figure 4-15 shows element PSD of silica and vanadium. Silica sorbent was mostly

concentrated on the 1st and 2nd stages (> 5 [tm). When silica sorbent was injected,

interestingly vanadium PSD shifted to the 1st and 2nd stages as well. Since silica has

high inner pore surface area, its fractional surface area PSD is the same as its fractional

mass PSD. Consequently, more vanadium was captured in the 1st and 2nd stages.

Although gas phase hydrolysis with water vapor was still present in this system, the

results clearly indicated hydrolysis at the pores was the dominating process.

Figure 4-16A shows the SEM image of silica only and Figure 4-16B shows the

SEM image of silica with vanadium. As shown, no discernible difference in morphology

can be observed indicating that vanadium was captured inside the pores of silica.

As discussed above, both CaCO3 and Si02 have been tested for their sorption capability.

If the metal capture is limited to chemical adsorption, silica sorbent would not capture

any vanadium. The fact that silica sorbent collected the same amount (2.5 mg for 20

minute) of vanadium as CaCO3 sorbent demonstrates that chemical adsorption to form

metal-sorbent product is not a necessary step for capturing the metal. In this system

studied, surface hydrolysis is the critical mechanism. In a study using silica spheres (lida

et al., 2004) also reported physical condensation to be responsible for manganese vapor

removal. Because of the humid condition in the ultrasonic nebulizer, the carrier air was

saturated. Hence gas phase hydrolysis was inevitable when sorbent particles were

generated by the ultrasonic nebulizer. Once small particles are formed by gas phase

hydrolysis, they can not be adsorbed by surface hydrolysis. Consequently, increasing







74



1.6 -

1.4 vatvony
1 I V at V only

1.2

S1.0 -
O --
S0.8 -

r 0.6 -

0.4 -



0.0
1.6 -

1.4 -
I V at Si+V


1.0 -
1.0 -

S0.8 -
4-
r 0.6 -

0.4 -

0.2 -

0.0
1.6 -

1.4 I Si at Si only

1.2 -
CL
o 1.0 -
0)

2 0.8 -
4-


0.4 -

0.2 -

0.0 --
0.01 0.1 1 10
Dp (jim)


Figure 4-15. Element PSD of vanadium and silica at 7400C























































Figure 4-16. Morphology of product A) when silica only are injected and B) when silica
with VTIPO was injected 740TC









surface hydrolysis and reducing gas phase hydrolysis are the key factor to enhance the

effectiveness of sorbent injection technique in this system.

Bimodal Lognormal Model Study

To gain insight into the effective mechanisms of sorbent injection technique, a

bimodal lognormal model study was conducted. In this model, both coagulation and

condensation were the key mechanisms considered. The reduction of the fine mode

volume and the increase of MMD of the fine mode (more than 1 [tm) are the key

parameters showing the effectiveness of sorbent injection technique.

Case 1: Bimodal coagulation only and unimodal coagulation only

When sorbent particles were injected where vapor had nucleated instantly, the system

undergoes bimodal coagulation. The MMD of the fine mode grows by bimodal

coagulation (inter-coagulation and intra-coagulation), however the volume of the fine

mode can be removed by only inter-coagulation. To evaluate the effects of inter-

coagulation in bimodal coagulation, simulation was also conducted for unimodal

coagulation (i.e. no coarse mode sorbent particles). As shown in Figure 4-17A, the initial

total number concentration was the same for both cases bimodall coagulation, dot line,

and unimodal coagulation, solid line) and they rapidly decreased from 1013 to 1010 in 0.05

seconds. However, the final number concentration of bimodal coagulation was less than

that of unimodal coagulation indicating the effect of inter-coagulation. The fine mode

total volume concentration for bimodal coagulation decreased very slowly (Figure 4-17B,

dotted line) and 6% of the initial volume was removed. Since the total volume is

conserved in intra-coagulation, this reduction was due to inter-coagulation. The

geometric standard deviation of the fine mode for unimodal coagulation approached

1.326 (Figure 4-17C, solid line) which is the asymptotic value for coagulating aerosols in









the transition regime, while for bimodal coagulation it reached 1.172 indicating the effect

of inter-coagulation (Figure 4-17C, dotted line). The MMD of fine mode reached 0.0178

|jm for unimodal coagulation and 0.0203 |jm for bimodal coagulation (Figure 4-17D). As

explained in Chapter 3, the inter-coagulation rate of the small particles in the fine mode

was faster than that of the large particles in the fine mode. Hence the smaller particles of

the fine mode were removed more by inter-coagulation resulting in 20% number

concentration reduction with 6% volume reduction. The larger particles in the fine mode,

however, were not affected much by inter-coagulation. The MMD of the fine mode for

bimodal coagulation increased slightly faster than that of unimodal coagulation (Figure 4-

17D); however, the MMD was still much less than 1 |tm within the given residence time.

The above results demonstrate that coagulation only could not be an effective mechanism

for removing fine mode particles.

Case 2: Condensation only

When condensation was the only mechanism, its vapor consumption rate was very

fast. As shown in Figure 4-17E (dash-dotted line), the vapor was depleted in 0.036

seconds (i.e., its saturation ratio reached 0). Figure 4-17F shows the change of MMD of

the coarse mode as a function of time. MMD of the coarse mode increased rapidly until

0.036 seconds. It then did not change after vapor was depleted. This clearly showed that

condensation was a faster process for mass transfer of the metal to coarse particles

compared to inter-coagulation. It can also be concluded that the preferable condition for

sorbent injection technique is condensation.









Case 3: 50% instant nucleation

Since the nucleation rate of vanadium particles is unknown, simulation was

conducted for a scenario where 50% of the vapor has instantly nucleated. This allowed

for evaluation of the competition of these two mechanisms bimodall coagulation and

condensation) on the evolution of the fine mode aerosol. Interestingly, as shown in Figure

4.17E, the saturation ratio (dash line) went down to 0 rapidly (4.3x10-3 seconds) which

was faster than the condensation only case (dash-dotted line). The high number

concentration of instantly nucleated fine mode particles had a high surface area that was

responsible for the rapid scavenge of the metal vapor in the system. The total volume of

the fine mode particles was increased by condensation in a short period of time (4.3x10-3

seconds) showing that metal vapor condensed on the fine mode particles (Figure 4-17B,

dash line). The increase of the fine mode volume due to condensation was greater than

the reduction due to inter-coagulation. Nonetheless MMD of the fine mode reached only

0.019 |jm (Figure 4-17D) which was much less than the desired MMD (> 1 [tm). The

initial number concentration of the fine mode was half of the bimodal coagulation case.

As shown in Figure 4.17A, nevertheless its number concentration quickly approached

that of the bimodal coagulation case (dash line and dotted line were overlapped).

Although fast condensation on coarse mode particles was desired, metal vapor

condensation on fine mode aerosol had an adverse effect. As shown in Figure 4-17A, the

number concentration of the fine mode was high and intra-coagulation was still the

dominant for the fine mode particles. Consequently, the removal of the fine mode volume

concentration by inter-coagulation in this case was still negligible as shown in Figure 4-

17B. Figure 4-17E shows MMD of the coarse mode as a function of time. In the 50%
















- Unimodal coag only
........ Bimodal coag only
- 50% instant nucleation


E
I 1013
_0
0
E 1012


1011
0

0
0 1010
E
z
109
E




0


0
E


0

0


F-
I- 8x10-9


'4.. ...

S* ......


- Unimodal coag only
........ Bimodal coag only
-- -50% instant nucleation


0.025



0.020


0.015



0.010



0.005


0.000

10000



1000



100


10



1
1.0710

_0




o
61.0706
E

S1.0706
0
0)
1.0704

o l~l


D



..***- *







S Unimodal coag only
S ........ Bimodal coag only
--- 50% instant nucleation


50% instant nucleation
-*-*-* Cond ony


Ii


F


............ Bimodal coag only
- -- 50% instant nucleation
-*-*-* Cond only


.Iu


-I I.-
0.00 0.05 0.10 0.15 0.20 0.25
0.00 0.05 0.10 0.15 0.20 0.25 C


10700
0.00 0.05 0.10 0.15 0.20 0.25 0.30
0.00 0.05 0.10 0.15 0.20 0.25 0.30


ime (S) time (s)

Figure 4-17. The change of A) total number concentration of the fine mode, B) total
volume concentration of the fine mode, C) geometric standard deviation (og)
of the fine mode, D) MMD of the fine mode, E) Saturation ratio of V vapor,
and F) MMD of the coarse mode as function of time.


B











........... Bimodal coag only
- 50% instant nucleation


I I I









instant nucleation case, some vapor also condensed on the coarse mode particles that

grow slightly larger than the bimodal coagulation case. However, the condensation on the

coarse mode was relatively insignificant once a high number of fine mode aerosol was

present. As discussed earlier, the vapor dominantly condensed on the fine mode particles.

Since the removal of the fine mode by inter-coagulation was slow, the contribution to the

coarse mode growth through inter-coagulation was insignificant. These observations

clearly demonstrate that nucleation is not desired for the sorbent injection technique. If

gas phase hydrolysis or thermal decomposition occurs, it forms a high number

concentration of fine mode particles that can not be effectively removed by coarse mode

particles due to slow inter-coagulation. In addition, the fine mode is competing with the

coarse mode for the metal vapor and is actually more effective. Such a competition

renders the sorbent technique less effective.

Conclusions

A mechanistic study for sorbent injection technique was conducted theoretically and

experimentally. The feasibility study verified the results of the thermodynamic

equilibrium analysis and assisted in selecting the sorbent candidate for chemical

adsorption (Ca-based sorbent). Experiments were divided into coagulation dominant and

condensation dominant cases. The results of experiments clearly showed that

condensation was the preferred mechanism for mass transfer process.

In the coagulation only case, element vanadium PSD showed only a slight shift to

larger particles. At least 107/cm3 number concentration of sorbent particles was needed to

improve the efficiency of sorbent injection technique by coagulation in the system

studied that had a residence time of a few seconds. In the condensation case, the

experimental results clearly showed that elemental vanadium PSD was shifted to the









larger sizes of sorbent particles. CaCO3 sorbent which has high affinity with vanadium

and silica sorbent which has no affinity but high surface area both exhibited the similar

capture efficiency. Gas phase hydrolysis resulted in the formation of fine particles while

the surface hydrolysis enhanced physical adsorption. Thus, surface hydrolysis was the

key mechanism to effectively capture the metal.

The bimodal lognormal model study showed that complete vapor consumption was

achieved within 0.03 seconds, proving surface interaction was the rate limiting process.

The reduction of the fine mode volume concentration and number concentration for

coagulation only case were 6% and 20% in 0.3 second residence time. Smaller particles

in the fine mode were removed more than bigger particles in fine mode. However, slow

inter-coagulation rate by high number concentration of fine mode particles supported that

condensation was the preferable mechanism for sorbent technique. Instant nucleation

formed high surface area of fine particles resulting in fast metal vapor condensation on

fine mode particles. Its condensed volume was more than reduced volume by inter-

coagulation. These results clearly demonstrated that nucleation was not desired for the

sorbent injection technique.

In conclusion, the preferable mechanism for mass transfer is condensation with fast

surface interaction such as surface hydrolysis. Alternatively, other scenarios that can

result in fast surface interaction are also desired, e.g. maintaining metals in the vapor

state at high temperature so that nucleation can be avoided. If in a system where it is

impossible to suppress nucleation, a sufficiently high number concentration of sorbent

particles must be provided to achieve a good collection efficiency by inter-coagulation.














CHAPTER 5
CONCLUSION AND RECOMMENDATIONS

Vanadium is one of the potential hazardous metal compounds from combustion

sources. It is concentrated in the submicron regime and sorbent injection is one promising

measure to control submicron vanadium particles. This study was carried out

theoretically and experimentally to evaluate sorbent injection technique for the control of

vanadium emission from combustion sources.

Theoretical study includes thermodynamic equilibrium analysis and bimodal

lognormal model study. Potential material for sorbent injection technique was determined

by thermodynamic equilibrium analysis, and bimodal lognormal model study used to gain

the insight of aerosol dynamic as the mechanisms of sorbent injection technique.

Experimental study includes pot experiment and aerosol reactor experiments. Pot

experiment verified the results of thermodynamic analysis. Aerosol reactor experiment

determined preferable mechanism and investigated the surface interaction for sorbent

injection technique.

By thermodynamic equilibrium analysis, Na-, Ca-, and Mg-based sorbents were

found to be effective in a wide range of temperatures. However, sulfur and sorbents were

shown to have high affinity (forming sulfates) at temperatures lower than 1000 K. The

strong affinity resulted in the depletion of available sorbents in the system. Sufficient

sorbent in excess of sulfur should be provided in this temperature range to effectively

capture vanadium compounds. At high temperatures (> 1000 K), the effectiveness of Na-









and Ca-based sorbents to capture vanadium revived as sorbent sulfates became less

stable, releasing available sorbents to react with vanadium.

When vanadium formed submicron particles by nucleation, inter-coagulation was

the only mechanism to remove these particles by sorbent particles. Hence, the effects of

inter-coagulation on removing fine mode aerosol were studied using a bimodal lognormal

model. Number concentration of coarse mode, geometric standard deviation of fine

mode, and MMD of fine mode were found to be important parameters for inter-

coagulation rate. Interestingly when inter-coagulation was the dominant mechanism for

removing fine mode particles, its geometric standard deviation approached monodisperse.

For coarse mode particles, the geometric standard deviation approached the asymptotic

value because intra-coagulation was still the dominant mechanism. The removal time

could also be expressed in the dimensionless form and two formulas were developed for

inter-coagulation dominant and intra-coagulation dominant respectively. The developed

formula could be used as a convenient tool to estimate the removal time once the particle

scavenge characteristic time was determined.

Ca-based sorbent and Na-based sorbent showed their chemical affinity with

vanadium in the pot experiments. Aerosol reactor experiments were divided into

coagulation dominant and condensation dominant cases. The results of experiments

clearly showed that condensation is the preferred mechanism for mass transfer process. In

the coagulation only case, element vanadium PSD showed only a slight shift to

supermicron. Formulas developed by inter-coagulation study showed that at least 107/cm3

number concentration of sorbent particles were needed to improve the efficiency of

sorbent injection technique by coagulation for a typical combustion system with a









residence time of a few seconds. In the condensation case, the experimental results

clearly showed that element vanadium PSD was shifted to the larger sizes of sorbent

particles. CaCO3 sorbent which had high affinity with vanadium and silica sorbent which

had no affinity but high surface area both exhibited the similar capture efficiency. Gas

phase hydrolysis resulted in the formation of fine particles while the surface hydrolysis

enhanced physical adsorption. Silica sorbent which had no chemical affinity with

vanadium also successfully captured vanadium vapor. Vanadium PSD shifted to the size

where the surface area of sorbent was high. This was the strong evidence for physical

adsorption. In real system, atomic vanadium vapor can be possibly condensed on sorbent

particles. Condensation rate is not limiting process, however its fast nucleation rate may

affect on the efficiency of sorbent injection.

A bimodal lognormal modeling study based on the initial condition of experiments

showed that condensation was the preferable mechanism. When condensation (based on

vanadium pentaoxide vapor) was the only mechanism, it showed the fastest and complete

collection of vanadium vapor. If nucleation occurred, condensation on the instantly

nucleated fine particles could yield undesired results because of the slow inter-

coagulation rate that would not be effective in removing them. Meanwhile, the amount

condensed on the fine mode was insufficient to increase MMD of fine mode to greater

than 1 |tm.

In conclusion, the preferable mechanism for mass transfer was condensation with

the fast surface interaction such as surface hydrolysis. Alternatively, other scenarios that

could result in fast surface interaction were also desired, e.g. maintaining metals in the

vapor state at high temperature so that nucleation could be avoided. If in a system where









it is impossible to suppress nucleation, a sufficiently high number concentration of

sorbent particles must be provided to achieve a good collection efficiency.

Based on the conclusions presented above and the experience gained in performing

this research, recommendations can be made that should further the understanding and

application of sorbent technique to the control of vanadium.

1. According to thermodynamic equilibrium analysis, most stable compound in
combustion temperature was vanadium pentaoxide. However, if atomic vanadium
vapor is formed, its nucleation rate is faster than vanadium pentaoxide at same
temperature because of its higher saturation ratio. It consequently results adverse
effect on sorbent injection. Hence, formed chemical species analysis and nucleation
rate study need to be performed.

2. There will be the relationship among number concentration of nucleated particles,
condensable vapor and PSD of sorbent particles to suppress nucleation. The
bimodal lognormal model which is used in this study could be an excellent tool for
this study.

3. Collison nebulizer and ultrasonic nebulizer used in this study generated
polydisperse particles. They also generated undesired submicron particles.
Theoretical study performed by this work showed that monodisperse articles
around 1 |jm particles was the most effective PSD of sorbent. Experimental studies
are needed to verify the theoretical finding.

4. Surface hydrolysis enhanced sorbent injection technique. However, its final product
on the surface of sorbent was unknown in this study. Based on the thermodynamic
equilibrium study and the study of TTIP hydrolysis which has similar structure, the
final product might be vanadium pentaoxide which is an excellent catalyst for NOx
removal. Hence, performing further study of the surface hydrolysis can determine
the surface interaction. This process can also be used to make micron particles
coated by nanosize vanadium pentaoxides.

5. Physical adsorption should be further studied. In this study, silica particles captured
vanadium vapor successfully which is the evidence of the importance of physical
adsorption. However, there was the surface hydrolysis supporting physical
adsorption. Studies should be carried out to examine if other surface interaction in
addition to hydrolysis can also help physical condensation. Dry silica particles with
different particle generation methods such as dust feeder and screw feeder can be
used for this type of study.




Full Text

PAGE 1

MECHANISTIC STUDY OF SORBENT INJECTION FOR VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS By SANG-RIN LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Sang-Rin Lee

PAGE 3

This achievement is dedicated to my parent s, Jung-Sik Lee and Ho-Suk Bae. My father died in 1995, but his dedication and love re main in my heart. Without my mother's encouragement and support, I co uld not have this achievement.

PAGE 4

iv ACKNOWLEDGMENTS My sincere gratitude is give n to Dr. Chang-Yu Wu for his support and guidance in this academic endeavor. It was the cornerstone of my life as an aerosol researcher when he admitted me as his first Ph.D. student. His advice, encouragement and patience have been invaluable during my graduate career. Special thanks go to Dr. Jean M Andino who taught me the meaning of atmospheric chem istry and gave me continuous comments and advice on my research. I sincerely thank Dr. David W. Hahn for allowing me to use his Raman spectroscopy and for giving me a dvice on product identification and hightemperature reactor design. I appreciate Dr Wolfgang Sigmund's advice and research comments. I give special thanks to Dr. Da le Lundgrun for allowing me to use his lowpressure impactor. His abundant experien ce with and knowledge of aerosols are my example of expertise. I also thank the staff in Engineering Research Center and Major Analytic Instrument Center. They teach and allow me to use many instruments such as ICP, SEM/EDX, and BET. I would like to thank our past and present group members. Special thanks go to Nawarat, David, Scott, and Sha nna. I also thank the other air group graduate and undergraduate students. I appreciate the Korean students in the Environmental Engineering and Sciences department. They helped to get me settled he re in Gainesville. My special thanks go to Dr. Jae-Hyun Cho and his family who treated me like family. I also appreciate my tennis friends. They helped me to blow off stress and energize myself to work.

PAGE 5

v Finally I would like to give my sincere gr atitude to my mother, Ho-Sok Bae. Even though she lost her husband in 1995, she kept encouraging her only son, me, to study abroad. Without her endless care and support, I could not have achieved this goal. I also appreciate the support and pa tience of my four older si sters and their families.

PAGE 6

vi TABLE OF CONTENTS Page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES.............................................................................................................ix LIST OF FIGURES.............................................................................................................x KEY TO SYMBOLS OR ABBREVIATIONS................................................................xii ABSTRACT.....................................................................................................................xi v CHAPTER 1 GENERAL INTRODUCTION....................................................................................1 2 VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS BY THERMODYNAMIC EQUILIBRIUM ANALYSES.................................................8 Introduction................................................................................................................... 8 Methodology...............................................................................................................10 Results and Discussions..............................................................................................12 Set I: Baseline Behavior of Vanadium in a Typical Coal Combustion System without Sorbents..............................................................................................12 Set II: Performance of Individual Sorbent...........................................................13 Set III: Effects of Chlorine and Sulf ur on the Performance of Individual Sorbent.............................................................................................................15 Effects of sulfur on the perfor mance of individual sorbent..........................15 Effects of chlorine on the perfor mance of individual sorbent......................17 Set IV: Competition among Three Sorbents.......................................................19 Case IV-1: each sorbent is 33% of the stoichiometric amount of sulfur in the system.....................................................................................................19 Case IV-2: each sorbent is 40% of the stoichiometric amount of sulfur in the system.....................................................................................................21 Case IV-3 and IV-4: each sorbent is 66.7% and 110% of the stoichiometric amount of sulfur in the system................................................................21 Conclusions.................................................................................................................24

PAGE 7

vii 3 SIZE DISTRIBUTION EVOLUTION OF FINE AEROSOLS DUE TO INTERCOAGULATION WITH COARSE AEROSOLS.....................................................26 Introduction.................................................................................................................26 Methodology...............................................................................................................29 Model Description...............................................................................................29 Simulation Conditions.........................................................................................32 Results and Discussion...............................................................................................32 Number Concentration........................................................................................32 Geometric Standard Deviation:...........................................................................35 Mean Size Difference..........................................................................................38 Application to Sorbent Injection.........................................................................41 Conclusions.................................................................................................................46 4 MECHANISTIC STUDY OF SORBENT INJECTION TO CONTROL VANADIUM EMISSION USI NG AEROSOL REACTOR......................................48 Introduction.................................................................................................................48 Experiments................................................................................................................50 Pot Experiment: Feasibility Study.......................................................................50 Aerosol Reactor Experiment...............................................................................51 Coagulation dominant system......................................................................51 Condensation dominant system....................................................................53 Product characterization...............................................................................55 Model Description......................................................................................................56 Results and Discussion...............................................................................................58 Pot Experiment....................................................................................................58 Aerosol Reactor Experiment...............................................................................59 Coagulation dominant..................................................................................59 Condensation dominant................................................................................63 Bimodal Lognormal Model Study.......................................................................76 Case 1: Bimodal coagulation onl y and unimodal coagulation only.............76 Case 2: Condensation only...........................................................................77 Case 3: 50% instant nucleation....................................................................78 Conclusions.................................................................................................................80 5 CONCLUSION AND RECOMMENDATIONS.......................................................82 APPENDIX A FORTRAN CODE FOR BIMODAL LOGNORMAL MODEL................................86 B MOMENT RELATIONSHIPS FOR TH E LOGNORMAL DISTRIBUTION........125 C MATERIAL SAFETY DATA SHEET....................................................................127

PAGE 8

viii LIST OF REFERENCES.................................................................................................129 BIOGRAPHICAL SKETCH...........................................................................................135

PAGE 9

ix LIST OF TABLES Table Page 2-1. Potential sorbents and corresp onding vanadium-sorbent compounds.......................11 2-2. Species used in the equilibrium calculations.............................................................11 2-3. Simulation conditions for evaluating the performance of various sorbents in capturing vanadium..................................................................................................12 3-1. Simulation condition for investigation of the effects of inter-coagulation on fine mode particle removal..............................................................................................32 3-2. Size distribution parameters and re moval time for Linak et al. (2003).....................45 3-3. Operating parameters of a FGD system and particle size di stribution from a power plant...............................................................................................................46 4-1. Feasibility study ex perimental conditions.................................................................51 4-2. Coagulation dominant syst em experimental conditions............................................54 4-3. Condensation dominant system experimental conditions.........................................56 4-4. Summarized simulation conditions...........................................................................58

PAGE 10

x LIST OF FIGURES Figure Page 1-1. Aerosol dynamic processes of vanadium in a combustion system vanadium only and sorbent injection..................................................................................................2 1-2. Various types of coagulation fo r bimodally distributed particles..............................6 2-1. Partition of vanadium speciation in a typical coal combustion system ...................14 2-2. Partition of vanadium speci es in a typical coal combusti on system with chlorine and sulfur .................................................................................................................14 2-3. Partition of vanadium species in a coal-air-V-s orbent system.................................16 2-4. Partition of major vanadium species air-V-SO2-sorbent system .............................18 2-5. Partition of major sorbent species air-V-SO2-sorbent system .................................18 2-6. Partition of major vanadium species in a coal-air-V-HCl-sorbent system...............20 2-7. Mole fraction of major vanadium speci es in the coal combustion system and partition of sulfur in the coal combustion system ...................................................23 3-1. Various types of coagulation fo r bimodally distributed particles............................28 3-2. Fine mode removal time as a function of fine mode number concentration for various coarse mode number concentrati ons. The dashed line connects points where Nf0/Nc0=10.....................................................................................................34 3-3. Dimensionless removal time as functi on of normalized number concentration and normalized mass or volume concentration...............................................................36 3-4. The evolution of fine mode ( gf0=1.4) and coarse mode ( gc0=1.3) particle size distribution by inter-coagulati on to reach removal time (tr=0.106 sec)...................37 3-5. A dimensionless removal time as func tion of fine mode geometric standard deviation where gc0=1.3. and coarse mode geometric standard deviation where gf0=1.3.....................................................................................................................39 3-6. Dimensionless removal time and half removal time as function of dgc/dgf and fine mode mean size........................................................................................................42

PAGE 11

xi 3-7. Removal time as function of coar se mode standard deviation and dgc/dgf with same number concentration (1010/cm3) and with same mass concentration (10 g/cm3)..43 4-1. Experimental set-up of the aerosol reactor system...................................................52 4-2. Measured reactor temperature pr ofile from bottom to top at 740oC........................54 4-3. XRD result for vanadium only at 673 and 873K.....................................................60 4-4. XRD result for CaCO3 with vanadium at 673, 873, and 1073K..............................60 4-5. XRD results for Na2CO3 with vanadium at 673K....................................................61 4-6. Element PSD of vanadium and calcium for Set I and Set II at 740oC.....................62 4-7. Element PSD of vanadium with and without water vapor at 740oC........................65 4-8. Morphology of vanadium oxide compound collected on fiber filter by hydrolysis and thermal decomposition and by thermal decomposition.....................................66 4-9. Collected vanadium particles on filter w ith water vapor and without water vapor at 740 oC.......................................................................................................................67 4-10. Element PSD of vanadium when water droplet injected at 740 oC..........................68 4-11. Element PSD of vanadium and calcium at 740oC by mass fraction and surface area fraction..............................................................................................................70 4-12. Morphology of collected particles when Ca-based sorbent was injected and Cabased sorbent with VTIPO were injected.................................................................71 4-13. SEM picture of whole product and EDX mapping of Ca and C) of V....................72 4-14. SEM picture of single particle and corresponding EDX spectrum..........................72 4-15. Element PSD of vanadium and silica at 740 oC.......................................................74 4-16. Morphology of product when silica only ar e injected and when silica with VTIPO was injected 740 oC..................................................................................................75 4-17. The change of total number concentr ation of the fine mode, total volume concentration of the fine mode, geometric standard deviation ( g) of the fine mode, MMD of the fine mode, Saturation ratio of V vapor, and MMD of the coarse mode as function of time....................................................................................................79

PAGE 12

xii KEY TO SYMBOLS OR ABBREVIATIONS 0 subscript 0 represents in itial condition b0 constant, f( ) b0=0.633+0.092 2-0.022 3 co continuum regime dp particle size ( m) dpg geometric mean size ( m) f, c subscript f represents fine mode, c represents coarse mode fm free molecule regime K0 initial inter-coagulation rate kB Boltzman's constant (dyne cm/K) n aerosol number concentration distribution function (#/cm3cm) N total number concentration (#/cm3) ns monomer number concentration at saturation (molecules/cm3) m1 mass of monomer (g) M3f total volume concentration of free molecule regime (cm3/cm3) M3c total volume concentration of continuum regime (cm3/cm3) Mk kth moment of aerosol size distribution S saturation ratio t time (s) T temperature (K)

PAGE 13

xiii t1/2 particle half life time (s) tr particle removal time (s) v1 Volume of monomer (cm3) collision frequency function mean free path of air (cm) gas viscosity (g/cm s) g geometric standard deviation p particle density (g/cm3) s fine mode particle scavenge characteristic time (second)

PAGE 14

xiv Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MECHANISTIC STUDY OF SORBENT INJECTION FOR VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS By Sang-Rin Lee May, 2005 Chair: Chang-Yu Wu Major Department: Environmental Engineering Sciences Mechanistic study of sorbent injectio n for vanadium emission control in combustion systems was conducted experimental ly and theoretically. Potential sorbent material for chemisorption was determined by thermodynamic equilibrium analysis. The computer code, STANJAN, was used to impl ement the calculations. Ca-, Naand Mgbased sorbents were evaluated for a wide range of combustion temperatures. The strong affinity between vanadium and these sorbents was identified which implies the great potential of these sorbents to chemically ad sorb vanadium. Sulfur was found to strongly impair the performance of these sorbents at lower temperatures (<1000K) due to the formation of sorbent sulfates that depleted the available sorbents in the system. Bimodal lognormal model was applied to investigate the impact of intercoagulation rate on the size distributions of fine-mode aerosols. Fine mode particle removal time was found to strongly depend on th e number concentration of coarse mode particles but independent on the number concen tration of fine mode particles. A 60%

PAGE 15

xv increase of geometric standard deviation of fine mode particles si gnificantly increased the dimensionless removal time. Fine mode pa rticles ultimately approached monodisperse when the dominant mechanism was inter-coagu lation. Meanwhile, coarse mode particles approached the asymptotic shape becau se intra-coagulation was the dominant mechanism. On a constant mass, monodisperse and 1 m mean diameter are the optimal condition for coarse particles to effectiv ely remove fine particles through intercoagulation. Aerosol reactor was applied for a mech anistic study of sorbent injection. Condensation was found to be the preferred mechanism for sorbent injection based on experimental results. CaCO3 sorbent which has strong chemical affinity with vanadium and silica sorbent which has no chemical a ffinity but high surface area successfully reduced vanadium submicron particle form ation. Vanadium was highly concentrated where surface area was high. Surface hydrolysis enhanced physical adsorption while gas phase hydrolysis reduced the efficiency of so rbent technique by forming nano particles. Bimodal lognormal modeling based on the experimental condition showed that condensation was a very effective means to scavenge vanadium oxide vapor. A high number concentration of fine particles by instant nucleation reduced inter-coagulation rate and quickly scavenged vanadium oxide vapor. Therefore, enhancing condensation while suppressing nucleation was shown to be cr itical to successful removal of vanadium compound.

PAGE 16

1 CHAPTER 1 GENERAL INTRODUCTION Heavy metal emissions from combustion sources such as ut ility boilers and incinerators are of great con cern because of their adverse effects on human health and the environment. Because of the increased concer n, research is being conducted to assess the actual exposure of human beings to toxic metals (Hogu, 2000), and vanadium is one of these metals. Vanadium is an abundant metal constituent in coal, heavy oil, and petroleum coke (Bryers, 1995; Linak and Miller, 2000; Swain, 1991; Yee and Rosenquist, 1996). The Energy Information Ad ministrator (EIA, 1996) estimated North America emitted about 497 tons/year of Vana dium into the atmosphere in 2000, and 92% of it was emitted by coal and oil combusti on sources. Thus, power and heat-producing plants using fossil fuel cause the most wi despread discharge of vanadium into the environment. In ambient fine particle char acterization studies, vanadium is often an excellent marker for oil combustion aerosol (Campen et al., 2001; Divita et al., 1996; Tolocka et al., 2004). Once emitted, vanadium can be transporte d long distances in the atmosphere, resulting in adverse environmental and hea lth effects (Bylinska, 1996). Vanadium is known to be more toxic when inhaled and re latively less toxic when ingested (Boyd and Kustin, 1984). It may also cause cardiovasc ular diseases, bronchitis, and lung cancer (Yee and Rosenquist, 1996). There are also seve ral reports of ecologi cal disasters caused by poorly controlled industrial emissions cont aining high concentrations of toxic metals including vanadium (Lin and Chiu, 1995; Pirrone et al., 1999).

PAGE 17

2 Figure 1-1. Aerosol dynamic processes of vanadium in a combustion system A) Vanadium only. B) Sorbent injection. Vanadium, like many other metals in the fu el matrix, may enter the gas stream in combustion systems by vaporization of vola tile organic vanadium compounds, or by entrainment of particles containing vana dium compounds. At high temperatures, it undergoes various chemical reactions to spec iate into different compounds such as oxides, chlorides or sulfates, dependi ng on the combustion environment and the composition of the system. As the temperatur e decreases once the gas stream exits the combustion zone, various aerosol dynamics such as (nucleation, coagulation and condensation) proceed, resulting in the transf ormation of vanadium into the particulate phase. Many studies have shown that meta ls undergoing this pathway generally form NucleationA Condensation Form vanadiu m p articles Coa g ulation S S u u b b m m i i c c r r o o n n P P a a r r t t i i c c l l e e s s B N N u u c c l l e e a a t t i i o o n n r r e e d d u u c c e e d d Condensation Coa g ulation Finally form Supermicron Particles Sorbent vanadium vapor Vanadium particle

PAGE 18

3 aerosols in the submicron regime. Figure 1-1 pr ovides a mechanistic description of such a process. The resultant particle size distri bution depends on the te mperature history and the existing particles in the system (Helbl e and Sarofim, 1989; Lighty et al., 2000; Wu and Biswas, 2000). In a study on characterizing th e particulate emissions from a large oil fuel fired power plant, 88 wt% of vanadium was reported to be in the size range of 0.01 to 1.0 m (Bacci et al., 1983). Linak and Wendt ( 1994) summarized that vanadium as a trace metal was enriched in submicron fly ash, in many coal combustion investigations. Unfortunately, traditional particulate cont rol devices have their minimum control efficiencies in this size regime (Biswas and Wu, 1998; Fl agan and Seinfeld, 1988; Linak et al., 1993). The consequence is further illustrated by ambien t particulate matter measurement that showed vanadium enrichment in the submicron regime (Tolocka et al., 2004). Thus, it is important to develop new t echniques to effectively control vanadium emissions. Sorbent technique is one pr omising measure to control submicron particles. In recent years, various studies have been c onducted using mineral sorbents to capture heavy metals. In most studies, sorbent partic les are injected into combustion systems to chemically adsorb heavy metals on the inje cted particle surface. As these sorbent particles are typically in the supermicron range, the metal-sorbent particles can be collected easily using traditional particulat e control devices. Figure 1-1b illustrates the mechanism of the sorbent inj ection technique. Shadman and co -workers (Scotto et al., 1994; Uberoi and Shadman, 1991) used silica, alumina and various naturally available materials (bauxite, kaolinite, and lime) to ca pture lead and cadmium. Linak et al. (1995) used them to capture nickel, lead, and cad mium. Mahuli et al. (1997) tested hydrated

PAGE 19

4 lime, alumina, and silica for arsenic control. Venkatesh et al. ( 1996) evaluated various mineral sorbents constituti ng a spectrum of alumino-sili cate compounds and a pulgite clay for immobilization of several trace metallic species. Biswas and co-workers (McMillin et al., 1996; Owens a nd Biswas, 1996; Wu et al., 1998) generated sorbent particles with very high surface area in-situ to capture lead and mercury. However, no study using mineral sorbent mate rials to capture vanadium ha s been conducted. Chlorine and sulfur (common constituents of coal and oil) may react with the metal or with the sorbent, thus reducing the e ffectiveness of the sorbent t echnique to remove the metal (Linak et al., 1995; Linak and Wendt, 1993; Owens et al., 1995; Wu and Biswas, 1993). Therefore, the impact of chlorine and sulfur on the sorbent process must also be evaluated. Figure 1-1 shows the aerosol dynamic processes in a combustion system (including chemical reaction, nucleation, condensation, and coagulation) resulti ng in the evolution of the particle size distribution that ultimately affects the fate of the metals. Consequently the development of a dynamic aerosol size distribution model considering all these mechanisms is instrumental to provide insigh ts into the processes and to determine the key mechanisms of a sorbent injection technique in such a complex system. There are several aerosol models availabl e to describe aerosol dynamic processes: moment (Frenklach and Harris, 1987; Wh itby, 1979), continuous (Tsang and Brock, 1982), and sectional model (Gelbard and Sein feld, 1980; Landgrebe a nd Pratsinis, 1990; Wu and Biswas, 1998). These models were categorized by their mathematical size distribution function. Whitby et al. (1991) and Williams and Loyalka (1991) reviewed these models in more details. Among them, aerosol moment model is one of the most

PAGE 20

5 commonly applied, because of its flexible m odel structure and low computational cost. Assuming a uni-modal log-normal aerosol si ze distribution, Lin and Biswas (1994) developed a model to study metallic particle formation and growth dynamic during incineration. Wu and Biswas (2000) also used a uni-modal lognormal size distribution model to evaluate the effects of chlorine on the evolution of lead aerosol size distribution. However, such models may not be appropriate for a metal-sorbent system, because of the extreme differences between th eir particle sizes. Metals alone in a combustion system eventually form submicron particles through nucleation, c ondensation, and coagulation. The presence of supermicron sorbent partic les may suppress the formation of submicron metal particles through inter-coagulation, and condensation which are the key parameters for the effectiveness of th e sorbent technique (Biswas and Wu, 1998). Consequently, a bimodal lognormal model will more pertinen tly represent a metal-sorbent system. The sorbent processes can be divided in to two steps: mass transfer (vanadium transfer to the surface of sorbent), and fo llowed by surface interaction. The mass transfer step is condensation, and/or co agulation while the surface inte raction is chemical, and/or physical adsorption. In previous studies (Carey et al., 2000; Linak et al., 1995; Linak et al.1998; Uberoi and Shadman, 1991), the syst em was generally condensation favored condition because the temperature was very hi gh where the metals were in the vapor state. However, Friedlander et al. (1991) demonstrated that scavenging fine particles by coarse mode particles through coagulati on can be the dominant mechanism. Surface interaction can be either physical adsorpti on or chemical adsorption. Previous sorbent studies (Mahuli et al., 1997; Punjak et al., 1989; Uberoi and Shadman, 1991) demonstrated that the dominant mechan ism of surface interaction was chemical

PAGE 21

6 adsorption, according to their XRD measurem ent data. Though its amount was much less than chemical adsorption, physical adsorption wa s also identified in some sorbent-metal systems (Chen et al., 2001; Punjak et al., 1989). Nucleation of vaporized metal compound resu lts in nano-size particles in the fine mode. Eventually they can grow to th e submicron regime by coagulation, and condensation. However, their typical concen tration, and short residence time in a combustion system do not allow these fine particles to grow to the supermicron regime (Biswas and Wu, 1998; Friedlander et al., 1991). Thus, coagulation with larger particles is the only mechanism to remove submicron fine mode particles once they are formed. Figure 1-2 conceptually depi cts the two types of coagul ation mechanisms (intercoagulation, and intra-coagul ation) in a system consisti ng of bimodally distributed particles. Intra-coagulation is self growth within the same mode. Meanwhile intercoagulation is the process to transfer the fine mode to th e coarse mode which plays the key role in the sorbent technique. Fine m ode Coarse mode Figure 1-2. Various types of coagulati on for bimodally distributed particles Intracoagulation InterCoagulation Intracoagu lation

PAGE 22

7 In summary, the sorbent technique is a promising measure to control vanadium emission. Mechanisms of the sorbent injection technique were investigated theoretically, and experimentally. Modal Aerosol Dynamic (MAD) model was used for theoretical study. An aerosol reactor system was applie d for experimental study. In Chapter 2, the potential materials for sorbent injection technique were determined by thermodynamic equilibrium analysis. The impact of chlori ne, and sulfur on the effectiveness of the process was also assessed. In Chapter 3, the evolution of fine mode particles due to coagulation with coarse mode particles (s orbent) was investigated using a bimodal lognormal model. Effects of size distribution parameters (such as number concentration, standard deviation, and mean diameter on inte r-coagulation) were evaluated. In Chapter 4, a feasibility study was conducted to verify the result of Chapter 2. Mechanistic experiments using an aerosol reactor were then performed. The preferable mechanism (condensation or coagulation) for the mass transfer process, and the surface interaction was determined. Finally, the bimodal lognormal model developed in Chapter 3 was applied to investigate the role of condensation, and nucleation in the sorbent technique. In Chapter 5, conclusion of this work and recommendations were provided.

PAGE 23

8 CHAPTER 2 VANADIUM EMISSION CONTROL IN COMBUSTION SYSTEMS BY THERMODYNAMIC EQUILIBRIUM ANALYSES Introduction Toxic metal emissions from combustion sources (such as ut ility boilers and incinerators) are of great concerns. Because of the increased concerns research is being conducted to assess the actual exposure of human beings to toxic metals (Hogu, 2000), and vanadium is one of these metals. Vanadium is one of the highly concentrated metals in certain types of coal (DeIul iis, 1993). In addition to coal heavy oil and petroleum coke also have been reported to have high concen trations of vanadium (Bryers, 1995; Linak and Miller, 2000; Swain, 1991; Yee and Rosenqui st, 1996). It is also found enriched in combustion ash and interest has even been developed to recover vanadium from ash (Alemany et al., 1998; Fang et al., 1998; Ts uboi et al., 1991) due to its high market values. Once emitted, it can be transported to distance, resulting in adverse environmental and health effects (Bylinska, 1996). It is known that vanadium ma y cause cardiovascular disease, bronchitis, and lung cancer (Yee and Rosenquist, 1996). There are several reports of ecological disaster s caused by poorly controlled in dustrial emissions containing high concentrations of toxic metals including vanadium (Lin and Chiu, 1995; Pirrone et al., 1999). Even in the Arctic, the annual flux of vanadium is estimated to be 474 tonnes (Akeredolu et al., 1994). Vanadium, like many other metals in the fu el matrix, may enter the gas stream in combustion systems by vaporization of vola tile organic vanadium compounds or by

PAGE 24

9 entrainment of particles containing vana dium compounds. At high temperatures, it undergoes various chemical reactions to spec iate into different compounds (such as oxides, chlorides or sulfates) dependi ng on the combustion environment and the composition in the system. As the temperatur e decreases once the gas stream exits the combustion zone, various aerosol dynamics proceed resulting in the transformation of vanadium into particulate phase. The particle size distribution depe nds on the temperature history and the existing part icles in the system (Wu and Biswas, 2000; Helble and Sarofim, 1989; Lighty et al ., 2000). Many research studies have shown that metals undergoing this pathway generally form aerosol s in the submicrometer regime. In a study on characterizing the particulate emissions fr om a large oil fuel fired power plant, 88 wt% of vanadium is in the size range of 0.01 to 1.0 m (Bacci et al., 1983). Unfortunately, traditional control devices have their minimum control efficiencies in this size regime (Flagan and Seinfeld, 1988). Thus, it is important to develop new techniques to effectively control vanadium emissions. In recent years, various st udied have been conducted to use mineral sorbents to capture heavy metals. Sorbent particles are injected into combustion systems and heavy metals can be chemically adsorbed on the injected particle surface. As these sorbent particles are typically in the supermicron range, the metal-sorb ent particles can be easily collected using traditional particulate control devices. Shadman and co-workers (Scotto et al., 1994; Uberoi and Shadman, 1991) used silica, alumina, and various naturally available materials (bauxite, kaolinite, and lim e) to capture lead and cadmium. Mahuli et al. (1997) tested hydrated lime, alumina and s ilica for arsenic control. Venkatesh et al. (1996) evaluated various mineral sorbents c onstituting a spectrum of alumino-silicate

PAGE 25

10 compounds and a pulgite clay for immobilizatio n of several heavy metals. Biswas and coworkers (McMillin et al., 1996; Owens and Biswas, 1996; Wu et al., 1998) generated sorbents particles with very high surface area in-situ to capture lead and mercury. However, no study using mineral sorbent ma terials to capture vanadium has been conducted. At high temperature environments, reacti ons are fast and equilibrium conditions can very possibly be achieved. Hence th ermodynamic equilibrium methods can be applied to determine the potential sorbent materials (Lee, 1988) that have chemical affinity with the metal. Equilibrium calculatio ns have been applied to study the behavior of metals like lead, arsenic and cadmium in combustion systems. Good agreement between experimental data and theoretical predictions under certain conditions has been reported, indicating equilibrium calculation to be a good tool for estimating the behavior of metals in combustion systems (Biswa s and Wu, 1998; Owens et al., 1995). The objective of this study was to use thermodyna mic equilibrium calculations to determine effective materials that can chemically adsorb vanadium. Optimal conditions to achieve high collection efficiencies were determine d. The impact of various common constituents in combustion systems on the performance was assessed. The most effective material for controlling vanadium was determined. Methodology The computer code STANJAN (Reynolds, 1995) was used to implement the equilibrium calculations. The principle of STANJAN is to minimize the Gibbs free energy of the system by using the method of elemental potentials combined with atom constraints. Thermodynamic data for all re levant species were obtained from the literature (Barin, 1995; Chase et al., 1986). Table 2-1 lists th e potential sorbent materials

PAGE 26

11 (using their element as the representative) and their corresponding vanadium-sorbent compounds. Table 2-2 lists the other species th at were included in the calculations. Table 2-1. Potential sorbents and co rresponding vanadium-sorbent compounds. Table 2-2. Species used in the equilibrium calculations Reactant Phase Gas V, VCl2, VCl4, VO, VOCl3, VO2 V Condensed V, VCl2, VCl3, VCl4, VO, VO2, VOCl3, VCl4, V2O3, V2O4, V2O5 Gas Ca, CaCl2, CaS Ca Condensed Ca, CaCl2, CaO, CaO2, CaSO3, CaSO4, CaS, Ca(OH)2, CaCO3 Gas Mg, MgCl2, MgOH Mg Condensed Mg, Mg(OH)2, MgCO3, MgCl2, MgO, MgSO4 Gas Na, NaCl, NaOH, Na2O2H2, Na2SO4 Na Condensed Na, NaCl, NaOH, NaClO4, NaHCO3, Na2O2H2, Na2CO3, Na2O, Na2O2, Na2SO3, Na2SO4 Gas CO, CO2, O2, HCl, HOCl, H2S, NOCl, ClO, N, NO, NO2, N2, N2O, S, SO2, SO3, H2SO4, HNO3, H2O Common compound Condensed H2O, S, Coal Calculations were conducted for four scen arios. Simulation conditions are listed in Table 2-3. The concentrations of the various compounds correspond to the levels found in a typical coal combustion system burning Ea stern bituminous coal (DeIuliis, 1993) with 20% excess air at 1 atmosphere. In order to es tablish the baseline, calculations were first conducted in Set I to determine the behavior of vanadium in a typical combustion system with no sorbent in the system. In Set II, th e performance of each potential sorbent was individually investigated for a wide range of temperatures. As chlorine and sulfur in coal may very well react with vanadium or the sorb ents, in Set III, parametric analyses were conducted to evaluate the impact of thes e constituents on the performance of the Sorbent Potential Vanadium-Sorbent Compounds CaCa(VO3)2, Ca2V2O7, Ca3(VO4)2 MgMg(VO3)2, Mg2V2O7 NaNaVO3, Na3VO4, Na4V2O7

PAGE 27

12 chemisorptions process. In Se t IV, all the sorbents were included in the calculations. Competition for vanadium among the sorbents was placed by varying their amounts to determine the best sorbent. All calculations were performed for temperatures ranging from 400 to 1700K. Table 2-3. Simulation conditions for evaluati ng the performance of various sorbents in capturing vanadium. a : each sorbent is 33.3% of the stoichiometric amount of sulfur. b : each sorbent is 40% of the stoichiometric amount of sulfur. c : each sorbent is 66.7% of the stoichiometric amount of sulfur. d : each sorbent is 110% of the stoichiometric amount of sulfur. Results and Discussions Set I: Baseline Behavior of Vanadium in a Typical Coal Combustion System without Sorbents In the first set of calcula tions, the behavior of vana dium in a typical coal combustion system was studied, and it serv ed as the baseline for evaluating the performance of the sorbents under various c onditions. The partition of major vanadium species in such a system is shown Figure 2-1. As shown, divanadium pentaoxide (V2O5) is the dominant species in the entire temperature range studied while V2O4 becomes increasingly important at high temperatures. As discussed earlier, metal oxi des formed in combustion systems generally Set. V Coal O2 N2 SO2 HCl CaCO3 Na2O MgO I-1 I-2 0 0 0 8.25 x10-3 0 0 0 0 0 0 II-1 II-2 II-3 0 0 0 0 0 0 2.52x10-2 0 0 0 3.01x10-3 0 0 0 7.28x10-3 III-1 III-2 III-3 III-4 III-5 III-6 0.285 0.285 0.285 0 0 0 0 0 0 8.25 x10-3 8.25 x10-3 8.25 x10-3 2.52x10-2 0 0 2.52x10-2 0 0 0 3.01x10-3 0 0 3.01x10-3 0 0 0 7.28x10-3 0 0 7.28x10-3 IV-1a IV-1b IV-1c IV-1d 9.64 x10-5 1 29.7 111.7 0.285 8.25 x10-3 9.49x10-2 0.114 0.190 0.313 9.49x10-2 0.114 0.190 0.313 9.49x10-2 0.114 0.190 0.313

PAGE 28

13 form submicrometer aerosols and hence are no t desired (Biswas and Wu, 1998). Chlorine is well known to have strong affinity for many metals (Owens et al., 1995; Wu and Biswas, 1993). Therefore, its effect on vanadi um speciation was also studied. The results are shown in Figure 2-2. The result is very sim ilar to that without chlo rine (Figure 2-1). It indicates relatively weak affinity between ch lorine and vanadium, compared with earlier studies on other heavy metals (Owens et al., 1995; Wu and Biswas, 1993). Hence, it is probable that the presence of chlorine in typical coal combustion systems will not significantly affect vanadiumÂ’s speciation. Ou r finding agrees with the study of metals from combustion of waste oil conducted by Nerin et al. (1999). Although the chlorine content in the oil burn was ve ry high, vanadium showed very low affinity with chlorine. Sulfur has been also reported to possess st rong affinity with certain metals (Biswas and Wu, 1998). However, it wa s not included in the baseline calculations because there were no thermodynamic data available for vanadium-sulfur compounds; hence, it was expected to have no effect on the baseline calculations. Set II: Performance of Individual Sorbent In Set II the performance of individua l sorbent was studie d. The partition of vanadium is shown in Figure 2-3A to 3C fo r Ca-, Na-, and Mgbased sorbent systems, respectively. Since the goal of the sorbent t echnique is to chemically react vanadium on sorbent particle surface, the total mole fr action of vanadium-sorbent compounds can also be interpreted as the capture efficien cy of the proposed sorbent process. The results are very encouraging. As shown, the predominant products are vanadium-sorbent species in all three system s (although they may be present in different forms). Undesirable vanadium oxide is pres ent only at extremely high temperatures.

PAGE 29

14 0 20 40 60 80 100 4006008001000120014001600 Tem p erature ( K ) M o l e f ract i on (%) V2O4 (c) V2O5 (c) VO2 (g) Figure 2-1. Partition of vana dium speciation in a typical coal combustion system. (c: condensed phase, g: gas phase) 0 20 40 60 80 100 4006008001000120014001600 Temperature (K)Mole fraction (%) V2O4 (c) V2O5 (c) VO2 (g) Figure 2-2 Partition of vana dium species in a typical co al combustion system with chlorine and sulfur (c: conde nsed phase, g: gas phase)

PAGE 30

15 The strong affinity between vanadium and the sorbents demonstrates the excellent potential of the use of these sorbent mate rials to effectively capture vanadium in combustion systems. However it should be noted that there are some factors, such as chlorine and sulfur in the system, that ma y prohibit the success of this process. Their impact will be discussed more in the next section. Set III: Effects of Chlorine and Sulfur on the Performance of Individual Sorbent Chlorine and sulfur are very common in co al and other fuels. As discussed earlier, they have been reported to affect the specia tion of metals and sorbent materials in the system. They may react with vanadium to form vanadium chloride or sulfate. In a pilot study of combustion of residual fuel oil with a high sulfur content, Huffman et al. (2000) identified vanadyl sulfate to be the major product in the fine particulate matter produced in the combustion process. Chlorine and sulfur can also react with the sorbent materials forming, for example, calcium sulfate (gypsum ), thus reducing the available amount of sorbents to react with vanadium. Hence, it is important to investigate their impact. Calculations were performed fo r systems with sulfate and with chlorine. The results are shown in Figures 2-4 and 2-5 for sulfur cas es and in Figure 2-6 for chlorine cases. Effects of sulfur on the performance of individual sorbent As shown in Figure 2-4A, Ca-based sorbent b ecomes ineffective to react with vanadium at lower temperatures (<900K) and V2O5 becomes the dominant species. The ineffectiveness is due to the strong affinity between sulfur and calcium, which depletes the available calcium in the system. This is evidenced by the predominant CaSO4 shown in Figure 2-5A (the partition of Ca compounds ). Ca-based sorbent is effective only at higher temperatures when CaSO4 becomes less stable, releasing Ca for reacting with vanadium.

PAGE 31

16 Figure 2-3. Partition of vanadi um species in (c: condensed phase, g: gas phase) A) a coal-air-V-CaCO3 system B) a coal-air-V-Na2O system C) a coal-air-VMgO system 0 20 40 60 80 100Mole fraction (%) Ca(VO3)2 (c) Ca2V2O7 (c) Ca3V2O8 (c) 0 20 40 60 80 100Mole fraction (%) NaVO3 (c) Na3VO4 (c) Na4V2O7 (c) VO2 (g) 0 20 40 60 80 100 4006008001000120014001600Temperature (K)Mole fraction (%) Mg2V2O7 (c) V2O4 (c) V2O5 (c) VO2 (g)A B C

PAGE 32

17 Similar to Ca-based sorbent, Na-based sorbent becomes ineffective at lower temperatures (<800K) when sulfur is presen t in the system (Figure 2-4B). The strong affinity between sulfur and sodium (forming Na2SO4) depletes the available sodium sorbent in the system for capturing vanadium. The sorption process revives only at higher temperatures when Na2SO4 becomes less stable (Figure 2-5B). Compared to the other two sorbents, Mg -based sorbent is the most severely affected by the presence of sulfur in the system. As shown in Figure 4C, it becomes ineffective in the entire temperature range studied. This is due to the high affinity between magnesium and sulfur forming MgSO4 (Figure 2-5C). The depletion of Mgbased sorbent by sulfur in the system results in the absence of magnesium vanadate. The results discussed above clearly evidenced that the presence of sulfur in the system significantly affects the perf ormance of the sorption process. Effects of chlorine on the performance of individual sorbent When chlorine is present in the system, all sorbents still posses excellent capture efficiency as shown in Figure 2-6. The resu lts for Ca(Figure 2-6A) and Mg(Figure 26C) based sorbents are similar to those with no chlorine. The results imply that neither magnesium nor calcium has strong a ffinity to react with chlorine. However, the products for the Nacase (Figure 2-6B) are different from the no chlorine case (Figure 2-3B). Na3VO4 and Na4V2O7 are the dominant species at high temperatures (above 800K) in the no chlorine case while NaVO3 is the dominant in the same temperature range (except Na4V2O7 at approx. 1100K) in the chlorine case. The shift of the Na/V ratio from 3 (Na3VO4) in the no chlorine case to 1 (NaVO3) in the chlorine case indicates that there is competition between chlorine and vanadium for sodium.

PAGE 33

18 Figure 2-4. Partition of major va nadium species Figure 2-5. Part ition of major sorbent (c: condensed phase g: gas phase) species(c: condensed phase g: gas phase) A) a coal-air-V-SO2-CaCO3 system A) a coal-air-V-SO2-CaCO3 system B) a coal-air-V-SO2-Na2O system B) a coal-air-V-SO2-Na2O system C) a coal-air-V-SO2-MgO system C) a coal-air-V-SO2-MgO system 0 20 40 60 80 100Mole fraction (%) Ca(VO3)2 (c) Ca2V2O7 (c) Ca3V2O8 (c) V2O5 (c) 0 20 40 60 80 100Mole fraction (%) NaVO3 (c) Na4V2O7 (c) V2O5 (c) VO2 (g) 0 20 40 60 80 100 4006008001000120014001600 Temperature (K)Mole fraction (%) V2O4 (c) V2O5 (c) VO2 (g)A B C CaO (c) CaSO4 (c) NaOH (g) Na2SO4 (c)A B C 4006008001000120014001600 Temperature (K) MgSO4 (c)

PAGE 34

19 Earlier studies have also shown the adverse e ffects of chlorine on alkaline metal (Na, K)based sorbents (Wu and Barton, 1999). Hence a higher level of Nabased sorbent in the system is necessary to ensure the adso rption of both vanadium and chlorine. Set IV: Competition among Three Sorbents As discussed, Ca-, Naand Mgbased sorb ents all have demonstrated excellent potential to capture vanadium in combustion systems. The natural question that comes next is which of the three is the most effec tive sorbent. Four conditions were simulated to determine the most effective so rbent by varying the amount of the sorbents in the system. In the first case (IV-1), the total amount of all sorbents was the exact amount required to stoichiometrically react with sulfur. The amount of sorbent wa s then increased and in the last case (IV-4) each sorbent itself was enough to completely deplete sulfur in the system. The results are shown in Figures 2-7A to 7D for the mole fraction of major vanadium species and Figure 2-7E – 7H for the partition of sulfur in the system. Case IV-1: each sorbent is 33% of the stoichiometr ic amount of sulfur in the system Because of the strong affinity of sulfur w ith all sorbents as discussed earlier, all sorbents are consumed by sulfur in this cas e (Figure 2-7E) and no so rbent is available to capture vanadium (i.e. forming vanadium-sor bent product) at temp eratures below 600K (Figure 2-7A). Thus, V2O5 appears at low temperatures. When sulfur’s affinity with each sorbent fades, some sorbents are released and start to react with vanadium. NaVO3 is the dominant in the intermedia te temperature range and Ca3(VO4)2 is the major compound in the high temperature range. This can be e xplained by the relative affinity of these sorbents with sulfur shown in Figure 2-7E. Na2SO4 starts to gradually decrease from 600 K (though the change is indis tinctive in Figure 7E at te mperatures below 1200K).

PAGE 35

20 Figure 2-6. Partition of major vanadi um species in (c: condensed phase g: gas phase) A) a coal-air-V-HCl-CaCO3 system B) a coal-air-V-HCl-Na2O system C) a coal-air-V-HCl-MgO system As the concentration of vanadium is only a tiny fraction of that of sulfur (less than 0.05%), a slight release of Na-based sorbent would result in a high yield of vanadiumsorbent product. Thus NaVO3 is the dominant sp ecies above 600K. CaSO4 starts to decrease rapidly from 1100 K and disappears at 1300 K. Consequently, a lot more Ca0 20 40 60 80 100Mole fraction (%) Ca2V2O7 (c) Ca3V2O8 (c) 0 20 40 60 80 100Mole fraction (%) NaVO3 (c) Na4V2O7 (c) VO2 (g) 0 20 40 60 80 100 4006008001000120014001600 Temperature (K)Mole fration (%) Mg2V2O7 (c) V2O5 (c) VO2 (g) V2O4 (c)A B C

PAGE 36

21 based sorbent than Na-based sorbent is ava ilable at high temperatures, resulting in the dominance of calcium vanadate at highe r temperatures. On the other hand, MgSO4 is very stable in the entire temperature range studied. Thus, no magnesium-vanadium compound is formed. Case IV-2: each sorbent is 40% of the stoichiometr ic amount of sulfur in the system In this case, totally 20% more sorben t was provided. As discussed earlier, the stronger the affinity with sulf ur, the less available the sorben t is. Mg-sorbent apparently has the strongest affinity with sulfur as it is completely converted into MgSO4 (Figure 27F, mole fraction is 40% at all temperatures ). Meanwhile, Ca-based sorbent is shown to have the weakest affinity with sulfur among the three sorbents. Only half of Ca-based sorbent (20% out of 40%) forms CaSO4. Consequently, Ca-based sorbent is the most effective one in most of the temperature range due to its abun dance (Figure 2-7B). However, Na-based sorbent is the most effective one at lower temperatures even though the amount of Na-based sorbent is much less than that of Ca-based sorbent. Thus, Nabased sorbent may have a stronger affinity w ith vanadium than Ca-based sorbent does. The reslts imply that the ultimate fate of vanadium depends on the relateive affinity and quantity of these two sorbents with vanadium and sulfur. This will be further manifested in the following sections. Case IV-3 and IV-4: each sorbent is 66.7 % and 110% of the stoichiometric amount of sulfur in the system In case IV-3, while Mg-based sorbent is still completely consumed by sulfur (Figure 2-7G), more Naand Ca-based sorben ts are released, and the ratio of available Na-based sorbent/Ca-based sorbent not c onsumed by sulfur increases. As shown in Figure 2-7C, the dominant range of Na-based sorbent expands. The result suggests that

PAGE 37

22 Na-based sorbent, if not impeded by sulfur, ha s a stronger affinity with vanadium in the entire temperature range then Ca-based sorben t does. This is further evidenced in Case IV-4 (Figure 2-7D), where the ratio of ava ilable Na-/Ca-based sorbent is approaching 1 and Na-V compounds are the dominant in th e entire temperature range. The more available the Nabased sorbent is, the le ss the calcium-vanadium product is. A field measurement from an oil fuel fired power pl ant has also identified vanadium to be present in the form of NaVO3 (Bacci et al., 1983). From the above analyses, it can be conc luded that Na-based sorbent has the strongest affinity to bind with vanadium. However, Na-based sorbent is also more vulnerable to the sulfur in the system than Ca -based sorbent is. The affinity of the three sorbents with vanadium and sulfur can be summarized as the following sequence. Affinity with V: Na > Ca > Mg Affinity with S: Mg > Na > Ca It should be emphasized again that the competition of vanadium and sulfur for the available sorbents in the system ultima tely determines the fate of vanadium. Nevertheless, it should be addressed that sorbent use is depend ent upon mixing with the stream, porosity, surface area and others. Scrubb ing of smaller metal particles by larger sorbent particles in the combustor is also a potential mechanism (e.g. fluidized bed combustion; Fan et al., 1999). Very often, th ese factors are dynami cally interwined. For examples, porosity and surface area change as sulfur is scavenged and temperature goes up. The system studied is an ideal one where these factors are not considered.

PAGE 38

23 Figure 2-7. Mole fraction of major vanadium species (c: condensed phase, g: gas phase) in the coal combustion system where each sorbent is A) 33.3% B) 40% C) 66.7% D) 110% of the stoichiometric amount of sulfur; Partition of sulfur in the coal combustion system where ea ch sorbent is E) 33.3% F) 40% G) 66.7% H) 110% of the stoichio metric amount of sulfur 0 20 40 60 80 100Mole fraction (%) Ca3V2O8 (c) NaVO3 (c) V2O5 (c) 0 20 40 60 80 100Mole fraction (%) Ca2V2O7 (c) Ca3(VO4)2 (c) NaVO3 (c) 0 20 40 60 80 100Mole fraction (%) Ca3(VO4)2 (c) NaVO3 (c) Na3VO4 (c)A B C D CaSO4 (c) MgSO4 (c) Na2SO4 (c) CaSO4 (c) MgSO4 (c) Na2SO4 (c) CaSO4 (c) MgSO4 (c) Na2SO4 (c)E F G H 0 20 40 60 80 100 4006008001000120014001600 Temperature (K)Mole fraction (%) Ca3(VO4)2 (c) Na3VO4 (c) NaVO3 (c) 4006008001000120014001600 Temperature (K) CaSO4 (c) MgSO4 (c) Na2SO4 (c)

PAGE 39

24 Another important point to be addressed is the leachability of metal from the generated metal-sorbent product, a major cons ideration in the sele ction of a sorbent material. If the collected ash is to be landfilled, the vanadium-sorbent compound should have a low leachability. However, if vanadium is to be recovered from ash containing a high concentration of vanadium (Alemany et al., 1998; Fang et al., 1998; Tsuboi et al., 1991) soluble products are desired. Conclusions Vanadium is concentrated in various fu els and the emission of vanadium from combustion systems is of concern. Mineral so rbents have been demonstrated to be effective to control various toxic metals in combustion systems. In this study, equilibrium calculations were conducted to id entify potential sorbent materials to chemically adsorb vanadium. Na-, Ca-, and Mgbased sorbents were found to be effective in a wide range of temperatures. However, the presence of sulf ur in the system significantly affected the performance of these sorbents. Sulfur and so rbents were shown to have high affinity (forming sulfates) at temperatures lower than 1000K. The strong affinity resulted in the depletion of available sorbents in the system. Sufficient sorb ent in excess of sulfur should be provided in this temperature range to effectively capture vanadium compounds. At high temperatures (> 1000K), the effectiveness of Naand Ca-based sorbents to capture vanadium revived as sorbent sulfates became less stable, releasing available sorbents to react with vanadium. Meanwhile, Mgbased so rbent still showed very strong affinity with sulfur in the entire range of temper atures, resulting in the worst performance among the three sorbents. When suffi cient sorbent is available, Nabased sorbent was found to be the most effective one.

PAGE 40

25 This study provides the insight of the reac tions between vanadium and sorbents as well as the impact of various operating conditi ons. The information obtained is important for developing a better strategy for managing vanadium emission problems.

PAGE 41

26 CHAPTER 3 SIZE DISTRIBUTION EVOLUTION OF FINE AEROSOLS DUE TO INTERCOAGULATION WITH COARSE AEROSOLS Introduction Heavy metal emission from combustion sour ces is of great concern because of its toxicity to human health and the environment. It is well known that metal particles from combustion are enriched in submicron regime (Linak et al, 1993; Biswas and Wu, 1998). Unfortunately, submicron particles show poor cap ture efficiency in traditional particulate control devices (Flagan and Seinfeld, 1988) Sorbent technique is one promising technique to control thes e fine metal emissions. In recent years, various studies have been conducted that use mineral sorbents to cap ture heavy metals (Linak et al., 1995; Venkatesh et al., 1996; Mahuli et al., 1997). Sorbent particles injected into combus tion systems are expected to chemically adsorb heavy metals on the surface. As thes e sorbent particles ar e typically in the supermicron range, the metal-sorbent product can be easily collected using traditional particulate control devices. In addition to ch emical adsorption, other mechanisms such as nucleation and coagulation are also present in the system. At combustion source, removal of metal vapor by condensation so that fine mode particle formation by vapor nucleation can be suppressed is the preferred mechanis m. Rodriguez and Hall (2003) developed an aerosol dynamic model based on a hybrid sectional model to study condensational removal of heavy metals from exhaust gases onto sorbent particles. Comparison of their model results with experimental data (Rodriguez and Hall, 2001) showed good

PAGE 42

27 agreement. Alternatively, after gas stream exits the combustion zone, coagulation with coarse mode sorbent particles can be a po ssible mechanism to scavenge fine mode particles (Fan et al., 1999). Frie dlander et al. (1991) studied the scavenge of a coagulating fine aerosol by a coarse partic le mode. They derived a simple analytical criterion for competing processes such as coagulation (intra-coagulation) and diffusion (intercoagulation). Based on the assumption of m onodisperse nucleated fine particles and entrained ash as the coarse particles, the criterion provides a convenient tool for estimating the importance of these competi ng processes. In real system, however, polydisperse fine particles are present and supermicron sorbent particles are injected. While important, there has been no modeling study about the role of coagulation in sorbent injection technique. The understanding of the evolution of fine mode particle size distribution in such a system is of fundam ental importance, though it is seldom studied. Thus, a model to simulate the aerosol dynami cs in the system can help understand the dynamic interactions between me tals and sorbents as well as implement sorbent injection technique to real system. There are several aerosol models availabl e to describe aerosol dynamic processes such as moment (Whitby, 1979; Frenklach and Harris, 1987), continuous (Tsang and Brock, 1982), and sectional model (Gelbard an d Seinfeld, 1980; Landgr ebe and Pratsinis, 1990; Wu and Biswas, 1998). These models were categorized by their mathematical size distribution function. Whitby et al. (1991) and Williams and Loyalka (1991) reviewed these models in more details. Among them, aerosol moment model is one of the most commonly applied ones because of its flexib le model structure and low computational cost. Assuming a uni-modal log-normal aerosol size distribution, Li n and Biswas (1994)

PAGE 43

28 developed a model to study metallic particle formation and growth dynamic during incineration. Wu and Biswas (2000) also used a uni-modal lognormal size distribution model to evaluate the effects of chlorine on the evolution of lead aerosol size distribution. However, such models may not be appropriate for a metal-sorbent system because of the extreme differences between th eir particle sizes. Metals alone in a combustion system eventually form submicron particles through nucleation, c ondensation and coagulation. The presence of supermicron sorbent partic les may suppress the formation of submicron metal particles through inter-coagulation and condensation which are the key parameters for the effectiveness of th e sorbent technique (Biswas and Wu, 1998). Consequently, a bimodal lognormal model will more pertinen tly represent a metal-sorbent system. Figure 3-1 conceptually depict s the two types of coagula tion mechanisms, i.e. intercoagulation and intra-coagul ation, in a system consisting of bimodally distributed particles. Nucleation of vapor ized metal compound result in nano-size particles in the fine mode. Eventually they can grow to the submicron regime by coagulation and condensation. Fine m ode Coarse mode Figure 3-1. Various types of coagulati on for bimodally distributed particles Intracoagulation InterCoagulation Intracoagu lation

PAGE 44

29 However, their typical concentration and s hort residence time in combustion system do not allow these fine particles to grow to the supermicron regime by intra-coagulation alone (Biswas and Wu, 1998; Friedlander et al., 1991). Many measurements of metal compounds emission have confirmed that metals are enriched in the submicron regime (Bacci et al, 1983; Osan et al, 2000; Lyyranen et al, 1999). When coarse sorbent particles are introduced into the system, fine mode par ticles coagulating with the sorbent particles are then transformed into the supermicron regime, which can easily be collected by conventional particulate contro l devices. As shown, inter-coa gulation plays the key role in removing fine mode particles and therefor e its effect was the focus of this work. In this study, the Modal Aerosol Dynamic (MAD) model (Whitby and McMurry, 1997) based on a lognormal moment model th at has multi modal structure and low computation cost was adopted. The evolution of fine mode particles due to coagulation with coarse mode particles (sorbent) was i nvestigated. The effects of size distribution parameters such as number concentration, sta ndard deviation and mean diameter on intercoagulation were evaluated. Methodology Model Description The development of the MAD model adopted in this study was provided in detail in Whitby and McMurry (1997), and hence is no t repeated. Only inter-coagulation which was the main mechanism to be investigated is discussed as follows. The inter-coagulation rate for the fine mode is 00) ( ) ( ) ( ) (c f c c f f c f k f kfddp ddp dp n dp n dp dp dp M t (3-1) and for the coarse mode is

PAGE 45

30 c f c c f f c f k c c f c c f f c f k c f kcddp ddp dp n dp n dp dp dp ddp ddp dp n dp n dp dp dp dp M t ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (00 3 / 00 3 3 (3-2) where subscripts k is order of moment (0, 3 or 6) and f, c stand for fine and coarse, respectively. The definition of parameter is provided in the Nomenclature section. Once these moments are determined, the key si ze distribution parameters can then be determined according to ) ln 2 exp(2 2 g k g kk Ndp M (3-3) 2 2 1 1ˆ ˆ k k k k gM M dp (3-4) ) ln( ) ( 2 ln2 12 1 1 2 r k k gM M k k k (3-5) where ) ( ˆ ) ( 1 ˆ2 1 2 1 2 1 2 1k k r k k k r k k k r N M Mk k With these size distribution parameters, th e evolution of fine mode particle size distribution due to inter-coagulat e with coarse mode particles can be investigated. In this study, harmonic average method (Pratsinis, 1988 ) was used to calculate the collision frequency function in the transition regime.

PAGE 46

31 Inter-coagulation to remove fine particle s is the key mechanism in this study. Thus, the time needed to remove fine mode par ticles is of interest. The fine mode particle scavenge characteristic time is defined as follows, s = Nf0 / Ko (3-6) where Nf0 is the initial total number concentration of fine mode particles and Ko is the initial inter-coagulation rate. Once K0 is calculated, the fine mode particle scavenge characteristic time can be easily estimated with size distribution parameters. The analytical solution (Whitby and McMurry, 1991 ) for the intercoag ulation rate of 0th moment for the continuum re gime can be expressed by ]}] ln ) 2 1 exp[( ] ln ) 2 1 }{exp[( { ]} ln ) 2 1 exp[( ] ln 2 exp[ ) ( ] ln ) 2 1 {exp[( 2 ) 2 ( 392 1 ]} ln ) 2 1 exp[( ] ln 2 exp[ ) ( ] ln ) 2 1 {exp[( 2 ) 2 ( 392 1 2 )[ 3 / 2 (2 2 2 2 2 0783 0 2 2 2 0783 0 gc gf gf gc gc gf gf gc gc gf gc gc gc gc gf gc gf gf gf gf B c f co odp dp dp dp dp dp dp dp dp dp dp dp T k N N K (3-7) and for the free molecule regime by ]} ln ) 8 1 exp[( ] ln ) 2 1 exp[( ) ( 2 ] ln ) 8 9 exp[( ] ln 2 exp[ ) ( ] ln 2 exp[ ] ln ) 8 9 exp[( ) ( ] ln ) 2 1 exp[( ] ln ) 8 1 exp[( ) ( 2 ] ln ) 8 1 exp[( ] ln ) 8 1 {exp[( / 32 2 2 2 3 2 2 4 2 2 2 2 0 gc gf gc gf gc gf gf gc gc gf gf gc gc gf gf gc gc gf gc gf gf P B c f fm odp dp dp dp dp dp dp dp dp dp dp b T k N N K (3-8) With emission control in mind, in this work the fine mode particle removal time (tr) is defined as time to remove 99.99 wt% of fine mode particles. The dimensionless removal time can then be defined as tr/ s

PAGE 47

32 Simulation Conditions Inter-coagulation rate depends on the size distributions of both modes. Hence the effect of number concentrati on, geometric standard deviation and geometric mean size of each mode on inter-coagulation rate was investigated. Table 3-1 summarizes all simulation conditions. In the first set of si mulation, the number concentration of each mode was varied, and the corr esponding size distribution and removal time of the fine mode were determined. In the second set, the effects of the geometric standard deviations of coarse mode and fine mode were studied. In the third set, the impact of varying the geometric mean size of each mode on inter-c oagulation rate was evaluated. The final set of simulation was carried out to determine the optimal sorbent particle size distribution that could enhance inter-coagulation rate, wh en the mass loading of sorbent particles was fixed. Table 3-1. Simulation condition for investigat ion of the effects of inter-coagulation on fine mode particle removal. dgf dgc ( m) gf gc Nf0 (#/cc) Nc0 (#/cc) Set I 0.01 10 1.3 1.3 104 – 1015 106 1012 Set II 0.01 10 1.0–1.6 1.0-3.0 108 108 Set III 0.01-0.5 1-500 1.3 1.3 108 108 Set IV 0.01 1-100 1.3 1.0-3.0 106 104-1013 All simulations conducted for T = 740oC Results and Discussion Number Concentration Since coagulation rate strongly depends on number concentrati on, inter-coagulation is expected to play a key role in enhanci ng the fine particle removal when the number

PAGE 48

33 concentration are high. Figure 32 shows the simulation results of set I. As shown, for the same initial fine mode number concentration, the removal time strongly depends on the coarse mode number concentration. It decreases as the coarse mode number concentration increases. However, for a fi xed coarse mode number concentration, interestingly, the removal time is not affect ed by the change of the fine mode number concentration unless the fine mode number concentration is 10 times more than the coarse mode number concentration. Careful examination of the coagulation rates (intra and inter) shows that when the fine mode intr a-coagulation rate is much faster than the inter-coagulation rate, the gr owth of fine mode partic les decreases its number concentration which subsequently yields a low inter-coagulation rate and therefore a longer removal time. Thus, the graph can be di vided into two regions as shown in Figure 3-2 by the dashed line: inter-coagulation dom inant and intra-coagul ation dominant. In short, the coarse mode number concentration clearly plays a major role in determining fine mode removal time when inter-coagulation is the dominant mechanism; on the other hand, the fine mode number concentration must be considered to estimate the fine mode removal time (tr) when intra-coagulation is the dominant mechanism. These results were further presented by dimensionless removal time (tr/ s) as a function of normalized number concentration Nf0/Nc0 in Figure 3-3A. As shown, all data merge into one line, thus yielding a useful tool for estimating the dimensionless removal time as 7 33 s rt Nf0/Nc0 < 10 (3-9) and

PAGE 49

34 ) 0007 0 exp( 7 330 0c f s rN N t Nf0/Nc0 10 (3-10) for the geometric standard deviation of the fi ne mode and the coarse mode of 1.3 and dgf = 0.01 m. Hence, as long as s is determined for a given system (Eq 3-6), the corresponding removal time (tr) can be estimated accordingly. Since the prior study (Friedlander et al., 1991) used half life time (t1/2), also included in the plot are the results for t1/2/ s. The corresponding equations are Nf0 (#/cm3) 10310410510610710810910101011101210131014101510161017 tr (sec) 10-610-510-410-310-210-1100101102 Nc0=106 Nc0=108 Nc0 =1010 Nc0 =1012 Intra-coagulation dominant Inter-coagulation dominant Figure 3-2. Fine mode removal time as a f unction of fine mode number concentration for various coarse mode number concen trations. The dashed line connects points where Nf0/Nc0=10.

PAGE 50

35 15 12 / 1st Nf0/Nc0 < 10 (3-11) and ) 0003 0 exp( 15 10 0 2 / 1 c f sN N t Nf0/Nc0 10 (3-12) As shown, the same patterns are followed. Since in practical applications, mass loading is of interest, Figure 3-3B shows dimensionless removal time as a function of normalized mass/volume. The same trend observed in Figure 3-3A exists. Dimens ionless removal time is independent on normalized mass or volume until it exceeds 10-8. Geometric Standard Deviation: It is well known that the geometric standa rd deviation of a log-normally distributed uni-modal aerosol approaches 1.35 and 1.32 for free molecule regime and continuum regime, respectively, when coagulation is the dominant mechanis m (Pratsinis, 1988). Parallely, this applies to intra-coagulation for multi-modal aerosols. However, the effects of inter-coagulation on the geometric standard deviations of multi-mode aerosol have never been investigated before. These effects were studied as set II and th e results are presented in Figure 3-4 for the size distributions at various times. In this simulation condition (with sorbent application in mind), inter-coagul ation was dominant for fine mode. As shown, fine mode became narrower with its mode size shifted toward the bigger size. It should be emphasized that the increase of the mode size was not due to intra-coagulation. Under intra-coagulation dominant case, there was no discernible change in the fine mode size distribution during the simulation time (0.1 s; also marked in Figure 3-4). Rather, it

PAGE 51

36 Nf0/Nc0 10-910-810-710-610-510-410-310-210-1100101102103104 tr/ s & t1/2/ s 1 10 100 tr/ s t1/2/ s A mf0/mc0 or Vf0/Vc0 10-1810-1710-1610-1510-1410-1310-1210-1110-1010-910-810-710-610-5 t r / s ts 0.1 1 10 100 t r / s t1/2/ s B Figure 3-3. Dimensionless removal time as function of A) normalized number concentration B) normalized mass or volume concentration

PAGE 52

37 resulted from the faster scavenge of the sm aller sizes of the fine mode due to intercoagulation. Inter-coagulati on rate is dependent on collision frequency function, which strongly depends on particle size difference, and it incr eases as the size difference increases. Consequently, the smaller size of the fine mode aerosol is scavenged much faster then the larger size in the same mode. Therefore, when inter-coagulation is the dominant mechanism, gf (fine mode geometric sta ndard deviation) ultimately approaches monodisperse (i.e. gf = 1). Meanwhile, intra-co agulation was dominant for the coarse mode since the addition of the fine mode moments to the coarse mode is rather insignificant. Although gc (coarse mode geometric standard deviation) did not change in the short period (fine mode removal time) shown in the figure, it will finally approach the asymptotic value (1.32, not shown in Figure 3-4). particle size ( m) 0.0010.010.1110100 number concentration (#/cm3) 1011021031041051061071081091010101110121013 Coarse mode Fine mode g =1.4 g =1.242 g =1.207 g =1.188 g =1.176 g =1.167 g =1.3 t = 0 t = 0.02 t = 0.04 t = 0.06 t = 0.08 t = 0.1 t = 0 to 0.1 t=0.1 intra-coagulation only Figure 3-4. The evolution of fine mode ( gf0=1.4) and coarse mode ( gc0=1.3) particle size distribution by inter-coagul ation to reach removal time (tr=0.106 sec)

PAGE 53

38 Figure 3-5A shows the dimensionless remova l time and the half life time (note: two y-axes) as a function gf, and Figure 3-5B shows the results for varying gc. When gf changed from 1 to 1.6, dimensionless remova l time increased drastically from 9.3 to 250 demonstrating the sensitivity to the change of the fine mode di stribution. A larger gf implies more mass in the larger size part. The corresponding lower inter-coagulation rate for the larger size resulted in a longer removal time as discussed earlier. Similar to the previous set, an equation can be obtained as a useful tool for estimating the dimensionless time due to the change of gf as ) 658 6 exp( 0058897 0gf s rt (3-13) ) 88 5 exp( 000276 0 578 02 / 1 gf st (3-14) In contrast, the dimensionless removal time changed within 3%, almost negligibly, when gc was varied from 1 to 3. Clearly demonstr ated above is that there is no point of using coarse mode particles with a wide size distribution to enhance inter-coagulation rate. The geometric standard deviation of the fine mode particles plays a more important role. Friedlander et al. (1 991) assumed monodisperse nucle ated particles (i.e. intercoagulation rate was the fastest). In many practical systems, aerosols are present as polydisperse particles, including metal compoun ds from combustion system. As revealed by this study, polydispersity needs to be c onsidered and the assumption of monodisperse particles underestimates the required time to scavenge polydisperse aerosol. Mean Size Difference If particle size difference is large, its collision frequency f unction is large which enhances coagulation. Figure 3-6A show s the effects of mean size difference.

PAGE 54

39 gf 1.01.11.21.31.41.51.6 t r / s 0 100 200 300 400 t1/2/ s 0 1 2 3 4 t r / s t 1/2 / s t r / s =0.0058897exp(6.658 gf ) t 1/2 / s =0.578+0.000276exp(5.88 gf ) gc 1.01.52.02.53.0 t r / st1/2/ s 0 10 20 30 40 t r / s t1/2/ s Figure 3-5. A dimensionless removal time as function of A) fine mode geometric standard deviation where gc0=1.3. B) coarse mode geometric standard deviation where gf0=1.3 A B

PAGE 55

40 As shown, when the mean size difference between the fine mode and the coarse mode increased, the removal time decreased accordingly. This clearly shows that large mean size sorbent (with same number concentr ation) will enhance inter-coagulation rate due to the large cross-section surface area for coagulation. Wh en the results are presented in the dimensionless form as shown in Figure 3-6A, even though mean size difference increased, dimensionless removal time showed only 3-7% difference for two orders of magnitude of size ratio. The reason for the insi gnificant difference is due to the increase of fine mode scavenge characteristic time for increased mean size difference In addition to size difference, fine mode mean size also affects the dimensionless removal time. Their relationship is plotted in Figure 3-6B. A linear relationship in log graph is observed which can be expressed as 814 14 ) log( 493 9 gf d s r t (3-15) 891 0 ) log( 129 02 / 1 gf d s t (3-16) Considering the combined effects of numb er concentration, geometric standard deviation and mean size, the following final form can be used to estimate the removal time and half life time for Nf0/Nc0 < 10, } 4296 0 ) log( 2852 0 ){ 658 6 exp( 0058897 0 gf d gf s r t (3-17) } 7748 0 ) log( 1126 0 )}{ 88 5 exp( 000276 0 578 0 {2 / 1 gf d gf s t (3-18) and for Nf0/Nc0 10

PAGE 56

41 ) 0007 0 exp( }] 4296 0 ) log( 2852 0 ){ 658 6 exp( 0058897 0 [0 0 c fN N gf d gf s r t (3-19) ) 0003 0 exp( }] 7748 0 ) log( 1126 0 )}{ 88 5 exp( 000276 0 578 0 [{0 0 2 / 1 c f gfN N gf d s t (3-20) Application to Sorbent Injection For many practical applications such as so rbent injection for toxic metal removal, minimizing the mass loading while having good removal efficiency is desired as cost consideration is always importa nt to the feasibility of the technology. With the same total mass concentration, there are several comb inations of size distribution parameters (number concentration, geometric mean size, and geometric standard deviation). In other word, the optimal conditions based on mass c oncentration will be different from those obtained for number concentration. Figure 3-7A shows the rem oval time for various combinations of gc and dgc/dgf based on the same number concentration. Figure 3-7B, on the other hand, shows the results based on th e same mass concentration. As discussed earlier, a high number concentration of the coarse mode and a wide mean size difference allow for a short removal time (shown in Figure 3-7A). However, diffe rent patterns were observed for the results based on the same ma ss concentration (Figure 3-7B). As shown, a narrow size deviation and a small coarse m ode mean size had the shortest removal time. Careful examination of the size distribution parameters reveals that under the same mass concentration, a smaller mean size and a narrower geometric standard deviation are translated into a higher number concentration, which is favorable for inter-coagulation. In summary, for practical applicati on sorbent mean size close to 1 m, monodisperse and a number concentration larger than 107 #/cm3 are the most effective.

PAGE 57

42 dgc/dgf 100100010000 t r / s & t1/2/ s 1 10 100 t r & t1/2 10-610-510-410-310-2 tr/ s t1/2/ s tr t1/2 d gf m) 0.010.1 t r / s 0 10 20 30 40 t1/2/ s 0 1 2 3 4 t r / s t1/2/ s Figure 3-6. Dimensionless removal time a nd half removal time as function of A) dgc/dgf and B) fine mode mean size A B

PAGE 58

43 gc 1.01.52.02.53.0 t r (sec) 10-510-410-310-2 dgc/dgf=100 dgc/dgf=1000 dgc/dgf=10000 gc 1.01.52.02.53.0 t r (sec) 10-610-510-410-310-210-1100 d gc /d gf =100 d gc /d gf =500 d gc /d gf =1000 Figure 3-7. Removal time as function of coarse mode standard deviation and dgc/dgf A) with same number concentration (1010/cm3) and B) with same mass concentration (10 g/cm3) A B

PAGE 59

44 Our finding was also applied to an expe rimental study conducted by Linak et al. (2003). Kaolinite sorbent was used to reduce fi ne particle formation. A 35% reduction of metal compounds in the fine mode was observe d. Condensation was reported to be the responsible mechanism while coagulation was said to have no effect on the reduction. The sorbent feed concentration was 680 mg/m3 with a mean size of 1.4 m, while the fine particle feed concentration rate was 83 mg/m3, with a residence time of 4.1 s. Table 3-2 shows the size distribution parameters us ed in the calculation and the estimated removal times. Three cases based on the same mass concentration were studied. For case I, dgf was assumed to be 0.01 m and both geometric standard deviations were assumed to be 1 for the most effective condition. E quations 3-19 and 3-20 were used to calculate the dimensionless removal time and the half life time, and the fine mode particle scavenge characteristic time was calculate d using Equation 3-7. The resultant removal time and half life time were 8.1x1085 and 8.6x1034 s, respectively, clearly showing that coagulation could not work for their system since these values were much greater than the residence time (4.1s). Case II was to simulate sorbent injection at a later stage where intra-coagulation had lowered the fine mode number concentrati on (using measured PSD of fine mode without sorbent that is reporte d in Linak et al., 2003) If the same amount of sorbent was injected at this stage, its corresponding removal and half life times were 2,094 and 263 s, both still much longer than the residence time. Other conditions (e.g. larger g) will only result in a longer removal tim e. Hence, the operating condition in their system clearly was in effective in removing the fine mode aerosol by intercoagulation. However, if the sorbent feed was increas ed two orders higher (68,000 mg/m3, Case III), the removal and the half lif e times were down to 2.65 and 0.12 s that

PAGE 60

45 were shorter than the residence time. Usi ng the formula developed in this work, the reason why coagulation was not important in their study was clearly seen and the criteria for effective removal by inter-coagula tion can be clearly established. Analysis was then carried out for an FGD (Flue Gas Desulfurization) system (Harris et al., 1993) that uses slurry droplet injection. For this case, particulate matters which are common components of flue gas al so can potentially be removed by intercoagulation. Table 3-3 summarizes the opera ting conditions (Harries et al., 1993) and the corresponding droplet size distribution. The part icle size distribution of particulate matter (Bacci et al, 1983) is also listed in Table 3. Particulate matter and slurry mass ratio is about 1.07x10-6. However, the large slurry droplet (bigger than 500 m) with monodisperse assumption results in a ve ry low number concentration (2.3x102/cm3) that requires an extremely long residence time to effectively remove particulate matter in the flue gas. If the droplet mean size is 10 m instead and the geometri c standard deviation is 1.3, its number concentration is 2.1x107 #/cm3 and the corresponding particle removal time is only 0.5 sec. The analys is demonstrates that fine particulate matter can potentially be removed using a typical mass loading in an FGD system but the selection of particle size and number concentration is very critical to accomplishing the goal. Table 3-2. Size distributi on parameters and removal ti me for Linak et al. (2003). dgf dgc ( m) gf gc Nf0 (#/cc) Nc0 (#/cc) tr (sec) t1/2 (sec) Case I 0.01 1.4 1 1 4.72 x1010 1.80x105 8.1x1085 8.6x1034 Case II 0.07 1.4 1 1 1.38x108 1.80x105 2,094 263 Case III 0.01 1.4 1 1 4.72x1010 1.80x107 2.65 0.12

PAGE 61

46 Table 3-3. Operating parameters of a FGD system and particle size distribution from a power plant. FGD system (Harris, 1993) Partic ulate matter (Bacci et al., 1983) Operating Conditions Diameter: 1.25 m L/G: 0.015 m3/Nm3 Gas flow: 2000 Nm3/h Slurry flow: 30 m3/h N/A Actual mc0: 17.4 106 mg/m3 Droplet size > 500 m gc: 1 (assumed) Nc0: 2.3x102 #/cm3(calculated) Particle Size Distribution Desired Droplet size : 10 m gc: 1.3 (assumed) Nc0: 2.1x107 #/cm3(calculated) mf0 : 18.7 mg/m3 MMD: 0.01 m (assumed) gf :1.3 (assumed) Nf0 : 7.8x109 #/cm3 (calculated) Conclusions Sorbent injection technique is one prom ising method to control the emission of submicron metal compounds from the combustion system. Under this bimodally distributed condition, inter-coagulation can be a key mechanism for sorbent injection technique. In this work, the effects of inter-coagul ation on removing fine mode aerosol were studied. High number concentration of coar se mode, large mean size difference, and narrow geometric standard deviation of fi ne mode were found to favor higher intercoagulation rate. The removal time can also be expressed in the dimensionless form. Two key parameters that affect the dimensionless removal time are the geometric standard deviation of the fine mode (when the normalized fine mode concentration is less than 10)

PAGE 62

47 and the geometric mean size of the fine mode The developed formula can be used as a convenient tool to estimate the removal time once the particle scavenge characteristic time is determined. When inter-coagulation is the dominant mechanism for removing fine mode particles, its geometric sta ndard deviation approaches monodisperse. For coarse mode particles, the geometric standard deviati on approaches the asymptotic value because intra-coagulation is still th e dominant mechanism. With regard to sorbent application where minimal mass loading is desired, a na rrow size deviation and a small mean size of the coarse mode yield the op timal effectiveness. The reas on is due to the corresponding high number concentration that is the critical parameter to inter-coagulation. Using the formula developed in this work, the criteria for effective removal of fine mode aerosol can be clearly established.

PAGE 63

48 CHAPTER 4 MECHANISTIC STUDY OF SORBENT INJECTION TO CONTROL VANADIUM EMISSION USING AE ROSOL REACTOR Introduction Vanadium is one of the major trace com ponents in coal and oil (Bryers, 1995; Linak and Miller, 2000; Swain, 1991; Yee and Rosenquist, 1996). In a study on characterizing the particulate em issions from a large oil fuel fired power plant, 88 wt% of vanadium was reported to be in the size range of 0.01 to 1.0 m (Bacci et al., 1983). Ambient particulate matter sampling at urban area used vanadium as a primary marker for fuel oil combustion (Divita et. al., 1996) and showed it was highly concentrated in the submicron regime (Tolocka et. al, 2004). Va nadium is known to be more toxic when inhaled and relatively less so when ingested (Boyd and Kustin, 1984). It may also cause cardiovascular diseases, bronchitis a nd lung carcinoma (Yee and Rosenquist, 1996). Unfortunately, traditional control devices have their minimum collection efficiencies in the submicron size regime (Biswas and Wu, 19 98; Flagan and Seinfe ld, 1988; Linak et al, 1993). Thus, it is important to develop new t echniques to effectively control vanadium emissions. In recent years, various studi es (Linak et al., 2003; Mahul i et al., 1997; Scotto et al, 1994; Uberoi and Shadman, 1991) have been conducted to use mineral sorbents to capture heavy metals. Figure 1-1 illustrates th e mechanism of sorben t injection technique. As shown in Figure 1-1A, vanadium vapor will nucleate and then coagulate and/or condense to form submicron vanadium oxides particles. When sorbents are injected,

PAGE 64

49 vanadium vapor can be adsorbed on the surf ace of sorbent and nucleation rate will be reduced (Figure 1-1B). As these sorbent par ticles are typically in the supermicron range, the metal-sorbent particles can be easily collected using traditional particulate control devices. Shadman and co-workers (Scotto et al, 1994; Uberoi and Shadman, 1991) used silica, alumina and various naturally availabl e materials (i.e. bauxite, kaolinite and lime) to capture lead and cadmium. Linak et al. ( 1995) used them to capture nickel, lead and cadmium. Mahuli et al. (1997) tested hydrat ed lime, alumina and silica for arsenic control. Venkatesh et al. (1996) evaluate d various mineral sorb ents constituting a spectrum of alumino-silicate compounds and a pu lgite clay for immobilization of several trace metallic species. Biswas and co-workers (McMillin et al., 1996; Owens and Biswas, 1996; Wu et al., 1998) generated sorbent particles with very high surface area in-situ to capture lead and mercury. However, there have been few studies that examine the use of mineral sorbent materials to capture vanadium. The sorbent process can be divided into tw o steps: mass transfer (vanadium transfer to the surface of sorbent) fo llowed by surface interaction. If the metal is in vapor phase, condensation is the mass transfer process. If sorbents are injected where the metal vapor has nucleated, coagulation is the mass transfer process. In the previous studies (Carey et al, 2000; Linak et al., 1995; Linak et al., 1998; Uberoi and Shadman, 1991), the system was generally condensation favored condition because temperature was very high where the metal was in the vapor phase. However, Friedlander et al (1991) demonstrated that the scavenging of fine partic les by coarse mode particles through coagulation can be the dominant mechanism. Surface interaction can be either physical adsorption or chemical adsorption. Previous sorbent st udies (Mahuli et al., 1997; Punj ak et al, 1989; Uberoi and

PAGE 65

50 Shadman, 1991) demonstrated that the domin ant mechanism of su rface interaction was chemical adsorption according to their XR D measurement data. Though its amount was much less than chemical adsorption, physical adsorption was also identified in some sorbent-metal system (Chen et al, 2001; Punjak et al, 1989; Iida et al., 2004). In such a metal-sorbent system, there are fine mode particles formed by metal compound and coarse mode particles of injected sorb ent. Thus, a bimodal aerosol model which incorporates various aerosol mechanisms can be a useful tool to investigate the dynamics of the system. Lee and Wu (2004) has developed a convenient equation to estimate the removal time (99.99 wt%) for sorbent inj ection technique using a bimodal lognormal model when coagulation is the dominant mechanism. Most studies have focused on demonstrating the ability of various sorbent materials to remove metal vapors from the gas str eam. In this study, a mechanistic study was conducted to determine the preferable mechan ism for sorbent injection technique. First, sorbent material was selected by feasibility experiments based on their chemical affinity. Coagulation only and condensation only cas es were then conducted and compared. Finally the effect of surface interaction was determined through the use of different sorbent materials. Bimodal lognormal modeling was also conducted to gain insights into the experimental results. With this study, th e preferable process for mass transfer and limiting process for sorbent injection technique was determined. Experiments Pot Experiment: Feasibility Study To verify the thermodynamic equilibrium re sults (Chapter 2), pot experiments were conducted. Two sorbent materials, CaCO3 (Fisher, powder, 99%, reagent grade) and Na2CO3 (Fisher, powder, 99.5%, reagent grade) were tested. The furnace (Thermolyne, F-

PAGE 66

51 A1730) was heated up to the designated temp erature. A ceramic pot containing sorbent and/or vanadium (Fisher, vanadium powder 99. 5%) in 2% nitric acid was then placed into the furnace. The sample was retrieved 10 minutes later. The products were identified by X Ray Diffraction (XRD) and compared with the result of thermodynamic analyses (Chapter 2). The experime ntal conditions are summarized in Table 4-1. Table 4-1. Feasibility study experimental conditions Conditions Molar ratio (sorbent : vanadium) Temperature (oC) Set I Vanadium only 0 : 1 400, 600 Set II CaCO3 + Vanadium 1.2 : 1 400, 600, 800 Set III Na2CO3 + Vanadium 1.2 : 1 400,600 Aerosol Reactor Experiment Coagulation dominant system Figure 4-1 shows the schematic diagram of the experimental setup of the aerosol reactor system. Vanadium solution was prep ared by dissolving elemental vanadium in 2% nitric acid which was introduced into the system by a Collison nebulizer (BGI, TN25). The atomized vanadium containing mi st was passed through a diffusion dryer to remove the water content. The aerosol par ticle size can be controlled by varying the properties of the atomized droplets accordi ng to the following equa tion (Hinds, 1999) dp=dd( Fv )1/3 (4-1) where, dp is diameter of final solid aerosol particle, dd is droplet diameter, and Fv is volume fraction of solid material Sorbent droplets were fed into the system using either a Collison nebulizer (BGI, TN25) or ultrasonic nebulizer (Sonaer, 24M) right before the inlet of the tubular aerosol reactor. The so rbent solution was also prepared by dissolving

PAGE 67

52 Figure 4-1. Experimental set-up of the aerosol reactor system sorbent material in 2% nitric acid.An u ltrasonic nebulizer can provide a higher number concentration of sorbent than a Collison nebulizer can and therefore was used to investigate the effect of number concentra tion change. The aerosol reactor (diameter: 1.651 cm, height: 45.72 cm, stainless steel) temperature was maintained by a tubular furnace (Thermolyne 73900). Reactor temperature profile at 740 oC is in Figure 4-2. Aerosol free dilution air through an ultra fiber filter (MSA, 76876) was introduced at the exit of the tubular r eactor to quench the reaction and aerosol dynamics. A Lundgren low pressure impactor (LPI) was used downstream to collect and classify particles by their aerodynamic size. It has 6 stages for atmospheri c pressure and 6 stages for low pressure (under 1 atm) which are designe d to collect submicron part icles. Glass fiber filter (Millipore, AP2004700) was placed at the last st age of impactor to collect particles less

PAGE 68

53 than the cutoff size of the 5th stage. Its cutoff diameters are determined by orifice pressure and temperature (L undgren and Vanderpool, 1988). When orifice temperature is 29.3oC and pressure is 209.3 mmHg, its cutoff sizes are 11.05, 5.54, 2.91, 1.73, 0.96, 0.50, 0.44, 0.29, 0.18, 0.09 m. Its upper limit (30 m) and lowest limit (0.01 m) of cutoff size were determined where the mass cumulated curve was 100%. Apiezon L grease (10% in benzene) was applied on st ainless steel substrat e to reduce bounce off effect. To eliminate condensed water vapor, su bstrates were dried in desiccators for at least one day before and after they were used. A highly sensitive micro balance (Sartorius, MC 210S, 10-5g) was used to measure the sample mass. Experimental condition To investigate the effect of size distribu tion on the sorbent injection technique, the concentration of vanadium solution was va ried and the sorbent was generated by two methods. Experiments were first conducted for vanadium alone to characterize its particle size distribution. Experiments were then carried out by adding CaCO3 sorbents to examine its effect on the size distribution of vanadium particles. In set I, a Collison nebulizer (BGI, TN-25) was used and an ultr asonic nebulizer (Sonaer, 24M) was used in set II because the ultra sonic nebulizer generated 10 times more sorbent particles compared to the Collison nebulizer. Elementa l calcium and vanadium mass concentration in reactor were 10 and 0.21 mg/m3 for Set I, 12.7 and 0.07 mg/m3 for Set II. The experimental conditions are su mmarized in Table 4-2. Condensation dominant system The experimental setup for the condensation dominant system is similar to that for the coagulation dominant system except the s ection of vanadium precursor generation.

PAGE 69

54 Figure 4-2. Measured reactor temperat ure profile from bottom to top at 740oC. Table 4-2. Coagulation dominant system experimental conditions Solution concentration (ppm) CaCO3 Vanadium Time (hr) Temperature (oC) Residence time (sec) Set I-1 0 500 2 740 0.34 Set I-2 23,000 500 2 740 0.34 Set II-1 0 50 2 740 0.42 Set II-2 23,000 50 2 740 0.42 To provide vanadium in the vapor fo rm, Vanadium Tri-I-Propoxy Oxide (VTIPO, Strem) which has a relatively high vapor pre ssure at room temperature was chosen. A bubbler was used to provide a constant vana dium precursor supply by saturating 2 L/min dry air with VTIPO vapor.

PAGE 70

55 Experimental conditions Experiments were first carried out for VTIPO. Since hydrolysis would be a key mechanism for generating vanadium oxide, the effect of temperat ure and humidity was investigated by using dry and humid air at room temperature and 740 oC. Finally, surface hydrolysis effect was tested by injecting water droplet instead of water vapor. Experiments were then carried out fo r two types of sorbent. Based on the thermodynamic equilibrium study and feasibility study, CaCO3 was chosen for chemical adsorption. Silica sorbent (specific surface area 170 m2/g, amorphous fumed, Alfa Aesar) with high surface area but no chemical affinity with vanadium was selected for physical adsorption. CaCO3 was dissolved in 2% nitric acid while silica was mixed and suspended in 2% nitric acid. The sorbent particles we re generated by the same ultrasonic nebulizer. As baseline, the particle size distributions of CaCO3 and silica sorbent were measured respectively. The sorbent was then introduced into the reactor together with VTIPO vapor. Elemental calcium, silica, vanadium ma ss concentration at reactor were 40, 80, and 9.25 mg/m3, respectively. The experimental conditi ons are summarized in Table 4-3. Product characterization Element size distribution: To determine the elemental distribution by size, the collected particles at each impactor stag e were dissolved in 2% nitric acid. The concentration of each element (V and Ca) was measured by Inductively Coupled Plasma emission spectroscopy (ICP, Perkin-Elmer Plas ma 3200). The total amount at each stage was then determined by multiplying the con centration by the solution volume (50 mL). Product morphology and elemental mapping: Scanning Electron Microscopy (SEM, JEOL 6330) / Energy Dispersive X -ray (EDX) was used for surface morphology and mapping elemental distribution on the surface of particles.

PAGE 71

56 Table 4-3. Condensation dominant system experimental conditions Vanadium Sorbent in 2% Nitric acid (10 g/L) Material Carrier air feedrate (L/min) Material Carrier air feedrate (L/min) Temperature (oC) Residence time (sec) Set I-1 VTIPO 2 dry air 5 room 0.3 Set I-2 VTIPO 2 humid air 5 room 0.3 Set I-3 VTIPO 2 dry air 5 740 0.3 Set I-4 VTIPO 2 humid air 5 740 0.3 Set I-5 VTIPO 2 Water 5 740 0.3 Set II-1 Air 2 CaCO3 5 740 0.3 Set II-2 VTIPO 2 CaCO3 5 740 0.3 Set III-1 Air 2 Silica 5 740 0.3 Set III-2 VTIPO 2 Silica 5 740 0.3 Product speciation: X-Ray Diffraction (XRD, Philips APD 3720 ) was used to identify crystalline species of collected particles. Rama n spectroscopy (Confocal system, 632-nm excitation) was also used to identify sp ecies that were not in crystalline form. Model Description The Modal Aerosol Dynamic (MAD) model that was described in detail in Chapter 3 was used for this study to simulate and ex amine the aerosol dynamics in the system. In this study, coagulation and condensation we re compared to determine the effective mechanism that was responsible for capturing vanadium. The condensational volume growth for the fine mode (free molecule regime) is 0 ) ( ) ( 2 ) 1 (1 3p d d p d n p d S B t Mf (4-2a) and that for the coarse mode (continuum regime) is

PAGE 72

57 0 ) ( ) ( 1 ) 1 (3 3p d d p d n p d S B dt dMc (4-2b) where, 1 1 3 / 1 12 36 m T k n v BB s and 1 1 3 / 1 2 38 9 16 m T k n v BB s Definition of the variables can be found in the Nomenclature section. Condensational volume growth is proportional to saturation ratio and its total surface area for the free molecule regime and total diameter for the continuum re gime, respectively (Prastinis, 1988). Volume change by inter-coagulation (W hitby and McMurry, 1997) for the fine mode and the coarse mode are given by 00 3 3) ( ) ( ) ( ) (c f c c f f c f f fddp ddp dp n dp n dp dp dp M t (4-3a) c f c c f f c f c c f c c f f c f c f cddp ddp dp n dp n dp dp dp ddp ddp dp n dp n dp dp dp dp M t ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (00 3 00 3 3 3 (4-3b) Nucleated particles are in the free molecule regime and sorbent particles are in the continuum regime. Volume of the fine mode particles is reduced while volume of the coarse mode is increased by inter-coagulation. The initial conditions of th e simulations were based on the experimental condition. Metal vapor was assumed as vanadium pentaoxi de. Three cases were simulated. First, all vapor was assumed to have nucleated instan tly. Simulations were conducted with sorbent particles and without sorbent pa rticles. Mechanism for the cas e with sorbent particles was bimodal coagulation which includes intra-co agulation and inter-coa gulation (case 1a). Mechanism for the case without sorbent particles is unimodal coagulation which is intracoagulation (case 1b). Second, nucleation wa s suppressed and condensation was the only mechanism allowed (case 2). Finally, 50% of vanadium vapor assumed to have instantly

PAGE 73

58 nucleated while the remaining 50% was in the vapor phase. Both condensation and coagulation were possible mechanisms in such a scenario (case 3). The simulation conditions are listed in Table 4-4. Table 4-4. Summarized simulation conditions g1 M10 (#/cm3) dp1 (A) g2 M20 (#/cm3) dp2 ( m) Vapor pressure (mmHg) Mechanism investigated Case1a 1 6.210x1013 6.285 2.087 3.89x105 1.07 N/A Intra-, intercoagulation Case1b 1 6.210x1013 6.285 N/A N/A N/A N/ A Intra-coagulation Case2 N/A N/A N/A 2.087 3.89x105 1.07 0.006572 Condensation Case3 1 3.106x1013 6.285 2.087 3.89x105 1.07 0.003496 Intra-, intercoagulation Condensation Results and Discussion Pot Experiment Figures 2-1 and 2-2 show the partition of vanadium species in a typical coal combustion system without and with chlori ne and sulfur, according to equilibrium analysis. Vanadium pentaoxide was the dom inant product until the temperature reached 1000 K. Figure 4-3 shows the XRD results fo r vanadium only case at 673 K and 873 K. These compounds were identified as vanadi um pentaoxide which agreed with the prediction by the thermodynamic equilibrium analysis. Figure 4-4 shows the crystalline species identified when CaCO3 sorbent was added to the pot in addition to vanadium. It cl early showed that different compounds were formed at different temperatures. Ca2V2O7 and Ca3(VO4)2 were the major compounds. Other products such as CaO and Ca(OH)2 were also detected since Ca-based sorbent was more than the stoichiometric amount of vanadium. Thermodynamic equilibrium

PAGE 74

59 calculations showed that di fferent dominating compounds at different temperatures (Figure 2-3A) and they well matched the results of th e pot experiment. Figure 4-5 shows the result when Naba sed sorbent was added. It also well matched with the thermodynamic calculation (F igure 2-3B). This figure showed only the result at 673 K because the material pr oduced at 873 K was extremely sensitive to ambient air moisture and formed damp mate rial that could not be identified by XRD. The feasibility study discussed above ve rified the results of thermodynamic equilibrium analysis and clearly demonstrated both sorbentsÂ’ abil ity to chemically bond with vanadium. It should be noted that the re sidence time in a typical combustion system is much shorter than the residence time in th e pot experiment. The ability of the sorbent to capture vanadium during the short flight ti me therefore needs to be evaluated in an aerosol reactor. Due to the difficulty to id entify the speciation of Nabased sorbent system by XRD, Ca-based sorbent was selected for further study in the aerosol reactor. Aerosol Reactor Experiment Coagulation dominant After vanadium vapor has nucleated, coagul ation is the only mechanism to remove these submicron particles. Hence, coagulat ion dominant case was also investigated by introducing vanadium in particulate form. Th e element size distributions of vanadium only and vanadium with sorbents are show n in Figure 4-6. For both cases, the mass median diameter (MMD) of vanadium increased slightly (0.55 to 0.63 m for Set I and 0.40 to 0.44 for Set II) when Ca sorbent (MMD was 0.74 and 0.97 m for Set I and II) was fed into the system. As discussed in Chap ter 3, the number concentrations of fine and coarse mode particle are the key parameters that determine whether inter-coagulation or

PAGE 75

60 Figure 4-3. XRD result for va nadium only at 673 and 873K Figure 4-4. XRD result for CaCO3 with vanadium at 673, 873, and 1073K

PAGE 76

61 Figure 4-5. XRD results for Na2CO3 with vanadium at 673K intra-coagulation is the dominant mechanis m. The number concentration of vanadium particles was 2.27x106/cm3 and 2.37x106/cm3 for set I and II, respectively; and the number concentration of Ca-based sorbent was 5.43x103/cm3 and 6.45x104/cm3 for set I and II, respectively. Assuming the best condition (monodisperse aerosol for both fine mode and coarse mode) for inter-coagulation, its fine mode 99.99% removal time and half removal time following Eqs. 3.19 and 3.20 derived in Chapter 3 were more than 7.6 and 1.8 hours. Since the resi dence time of these sets of experiments was only 0.3 seconds, the theoretical analysis clearly showed that inter-co agulation could not work as the major mechanism to remove the fine mode vanadium particles under this experimental condition. Typical ly the flue gas residence time in a combustion system is less than 10 seconds. Thus, th e proper number concentratio n of sorbent particles to reduce the removal time to less than 10 seconds should be more than 107/cm3. The result agrees with the observation reported by Linak et al. (2003) who app lied kaolinite sorbent to reduce fine particle formation. Its numb er concentration of kaolinite sorbent was 1.80x105/cm3 while that of fine m ode metal aerosol was 4.72x1010 #/cm3. They reported

PAGE 77

62 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V at V only Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V at Ca+V Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V at V only Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V at Ca+V Dp ( m) 0.010.11 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 Ca at Ca+V Dp ( m) 0.010.11 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 Ca at Ca+V Figure 4-6. Element PSD of vanadium and cal cium A) for Set I and B) for Set II at 740oC A B A A B B

PAGE 78

63 that 35% reduction of fine particles was due to condensation and there was no coagulation effect on fine particle reduc tion. To accomplish fine mode removal by coagulation, the sorbent feed needs to be in creased at least 2 orders higher (See Table 32). Condensation dominant Vanadium precursor characterization. VTIPO forms vanadium oxide compound by hydrolysis and/or thermal decomposition. As baseline, VTIPO vapor was introduced into the reactor with dry air (no hydrolysis) at room temperature (no thermal decomposition). As expected, there were no pa rticles collected by the LPI. To compare hydrolysis and thermal decomposition, VTIPO was introduced into the reactor with humid air and dry air at 740oC respectively. With dry air, only thermal decomposition occurred, while both thermal decomposition and hydrolysis occu rred with the presence of water vapor. Finally instead of water va por, water droplets generated by ultra sonic nebulizer were introduced into the system. In this case, hydrolysis would occur on the surface of the water droplet. Figure 4-7 shows the PSDs of vanadium el ement in dry air and humid air at 740oC, respectively. These two case s showed very similar dist ributions. Vanadium mostly concentrated in the filter st age that is smaller than 0.178 m. When vanadium precursor (VTIPO) undergoes hydrolysis or thermal decomposition, it forms vanadium oxides. These vanadium oxides quickly nucleate because their saturation vapor pressure is very low (1.12x10-8 mmHg and Saturation ratio: 5868) at 740oC. Instantly, nucleation results in a burst of extremely high concentration of nanoparticles. These particles will grow following coagulation and/or condensation. However, the short re sidence time of the

PAGE 79

64 aerosol reactor (only 0.3 seconds ) would only allow the growth to very fine sizes. Figure 4-8 shows the morphology of vanadium oxide compound collected on the filter. When water vapor was present, vanadium oxide formed well shaped spherical particles indicating the strong effects of condensation (Figure 4-8A). On the other hand vanadium oxide formed by only thermal decomposition showed much smaller primary particles implying the importance of coagulation (Figure 4-8B). VTIPO is one of metal alkoxides which unde rgo hydrolysis. The ra te of hydrolysis of a metal alkoxide depends on the characteri stics of the metal and those of the alkyl group. In general, silicon alkoxides are among the slowest to hydrolyze, and for a given metal alkoxide the hydrolysis rate increases as the length of the alkyl group decreases (Rahaman, 1995). There are limited studies about the hydrolysis of VTIPO. However, Titanium TetraIsoPropoxide (TTIP), whic h has the same alkyl group, is a common material to make TiO2 nanoparticles and there are many studies on hydrolysis and thermal decomposition of TTIP. The reaction rate constant is 3.96x105exp(-8479.7/T) for gas phase thermal decomposition (Okuyama et al., 1990) and 3x1015exp(-1013.9/T) for gas phase hydrolysis (Seto et al., 1995). Hydrol ysis clearly is much faster than thermal decomposition. Based on hydrolysis of TTIP, it is likely that hydrolysis of VTIPO is also faster than thermal decomposition of VTIPO. Supporting this analys is, the total amount of vanadium element collected in humid air was 5 times more than that in dry air in this study. Thermodynamic equilibrium analysis stu dy in Chapter 2 showed that the stable compound for vanadium was vanadium pentaoxi de. Assuming that hydrolysis of VTIPO generates vanadium pentaoxide, VTIPO hydrolysis can be given as follows. 2VO(OC3H7)3+3H2O -> V2O5 + 6C3H7OH

PAGE 80

65 Figure 4-9 shows the collected particles on the filer for both cases. With humid air the particles collected were dark green, while with dry air yellow particles were observed. Its final product of vanadium oxide co mpound could be different based on the observation of the color of co llected particles. These par ticles were characterized by XRD although no crystalline species were identified. Unfortunately Raman spectroscopy also could not identify the species. The final set of experiment was injecting water droplets to the reactor instead of feeding water vapor. The mass mean diameter of water droplets ge nerated by ultrasonic nebulizer was 1.7 m (Sonaer, 24M). As shown in Figure 4-10, vanadium PSD with water droplets was shifted to much larg er size. Gas phase hydrolysis generated nanoparticles. However, the surface of wate r droplets was also the active site for hydrolysis once VTIPO vapor diffused to the surface. The surface hydrolysis consequently captured vanadium and reduced the fine particle formation by gas phase hydrolysis. Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 V w water vapor Dp ( m) 0.010.11 Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 V w/o water vapor Figure 4-7. Element PSD of vanadium with and without water vapor at 740oC

PAGE 81

66 Figure 4-8. Morphology of vanadium oxide compound collected on fiber filter A) by hydrolysis and thermal decomposition and B) by thermal decomposition A B

PAGE 82

67 Figure 4-9. Collected vanadium particles on filter A) with water vapor B) without water vapor at 740 oC Sorbent injection. Since vapor diffusivity is relatively large, the rate limiting step for effective capture of vanadium vapor fo r the condensation case will be the surface interaction. Two sets of experiments were conducted to determine the effective surface interaction in the system. In the first set Ca-based sorbent, which showed strong chemical affinity with vanadium in the equilibrium an alysis, was used. In the second set Sibased sorbent which has high surface ar ea but no chemical affinity with vanadium, was studied. Comparing results of these two experimental conditions helps reveal the preferable mechanism for surface interaction: chemisorption or physisorption. Case 1. Ca-based sorbent. Figure 4-10 displays the element PSD of vanadium and calcium. When only VTIPO was introduced, mo st vanadium was concentrated in the filter stage. After CaCO3 sorbent was injected, vanadium PSD was shifted to around 1 m size range. Since condensation depends on the surface area of sorbent, the surface area fraction distribution of the sorbent was al so plotted in Figure 4-11. As shown, PSD of vanadium when sorbent was injected is ve ry similar to the sorb ent surface area fraction distribution, verifying conde nsation to be the key mechanism in the system. A B

PAGE 83

68 Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 V w water vapor Dp ( m) 0.010.11 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V w water droplet Figure 4-10. Element PSD of vanadium wh en water droplet injected at 740 oC Figure 4-12A shows the morphology of CaCO3 sorbent particles in the range of 1.73-0.96 m. Instead of being present as indivi dual particles in that size range, the collected material appeared to be a big chunk that resulted from merging. It should be noted that all collected materi als were placed in a desiccat or before characterization. However, the strong hygroscopic property of CaCO3 still yielded merging of the material. Figure 4-12B shows the morphology of particle s collected on the same size range when both vanadium and the sorbent were present. It is interesting to observe individual spherical particles in this cas e, totally different morphology from the sorbent only case. The comparison of these morphologies indica te that when vanadium oxide compounds

PAGE 84

69 were well coated on the surface of sorbent par ticles, those particles became insensitive to moisture. In the latter case, vanadium oxide compound formed by surface hydrolysis will deposit on the surface of sorbent. Figure 4-13A shows SEM/EDX for the particles when both CaCO3 and vanadium were present while Figure 4-13B is the Ca mapping and Figure 4-13C is the Vmapping. As shown in Figures 4-13B and 13C, vanadium was much widely distributed than calcium was. If surface chemical r eaction forming calciumvanadium compounds was the main mechan ism, these two mapping should show a similar distribution pattern. Figure 4-14 is a SEM picture of one single particle and its corresponding EDX spectrum. As shown, the inte nsity of vanadium is much stronger than that of calcium. Since calcium and vanadium has similar atomic number, its corresponding concentration ba sed on intensity for vanadium was higher than calcium. However, this is relative value. For the prec ise quantification of measured particles the calibration of EDX with standard sample ar e needed. EDX measurement could show that calcium and vanadium were on the one single particle. However, it could not determine the surface interaction. Case 2. Sibased sorbent. Since chemical affinity between vanadium and the sorbent is not necessary, experiments were carried out using silica to verify if physical adsorption alone can be a possible mechanism for the surf ace interaction. Silica is hydrophilic but it does not dissolve in water. Thus, silica was se lected to investigate the hypothesis. It was well mixed and suspended in 2% nitric acid, then sonicated. Silica sorbent aerosol was generated using the same ultrasonic nebuliz er. As baseline, silica sorbent PSD was

PAGE 85

70 2D Graph 4 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 V at Ca+V Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 V at V only Surface fraction/ logDp 0.0 0.5 1.0 1.5 2.0 Ca surface at Ca + V Dp ( m) 0.010.11 Mf/ logDp 0.0 0.5 1.0 1.5 2.0 Ca at Ca + V Figure 4-11. Element PSD of vanadium and calcium at 740oC by mass fraction and surface area fraction

PAGE 86

71 Figure 4-12. Morphology of collected particle s when A) Ca-based sorbent was injected and B) Ca-based sorbent with VTIPO were injected A B

PAGE 87

72 Figure 4-13. SEM picture of A) whole product and EDX mappi ng B) of Ca and C) of V Figure 4-14. SEM picture of A) single particle and B) corresponding EDX spectrum C B A A B

PAGE 88

73 measured. Experiments were then carried ou t by injecting sorbent with VTIPO to the reactor. Figure 4-15 shows element PSD of silica and vanadium. Silica sorbent was mostly concentrated on the 1st and 2nd stages (> 5 m). When silica sorbent was injected, interestingly vanadium PSD shifted to the 1st and 2nd stages as well. Since silica has high inner pore surface area, its fractional su rface area PSD is the sa me as its fractional mass PSD. Consequently, more vanadium was captured in the 1st and 2nd stages. Although gas phase hydrolysis with water vapo r was still present in this system, the results clearly indicated hydr olysis at the pores was the dominating process. Figure 4-16A shows the SEM image of s ilica only and Figure 4-16B shows the SEM image of silica with vanadium. As s hown, no discernible di fference in morphology can be observed indicating that vanadium was captured inside the pores of silica. As discussed above, both CaCO3 and SiO2 have been tested for their sorption capability. If the metal capture is limited to chemical adsorption, silica sorbent would not capture any vanadium. The fact that silica sorbent collected the same amount (2.5 mg for 20 minute) of vanadium as CaCO3 sorbent demonstrates that chemical adsorption to form metal-sorbent product is not a necessary step for capturing th e metal. In this system studied, surface hydrolysis is the critical mech anism. In a study using silica spheres (Iida et al., 2004) also reported phys ical condensation to be res ponsible for manganese vapor removal. Because of the humid condition in th e ultrasonic nebulizer, the carrier air was saturated. Hence gas phase hydrolysis was inevitable when sorbent particles were generated by the ultrasonic nebulizer. On ce small particles are formed by gas phase hydrolysis, they can not be adsorbed by su rface hydrolysis. Consequently, increasing

PAGE 89

74 Dp ( m) 0.010.1110 Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 Si at Si only Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 V at Si+V Mf/ logDp 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 V at V only Figure 4-15. Element PSD of vanadium and silica at 740oC

PAGE 90

75 Figure 4-16. Morphology of product A) when silica only are injected and B) when silica with VTIPO was injected 740oC A B

PAGE 91

76 surface hydrolysis and reducing gas phase hydrol ysis are the key factor to enhance the effectiveness of sorbent injection technique in this system. Bimodal Lognormal Model Study To gain insight into the effective mech anisms of sorbent injection technique, a bimodal lognormal model study was conducted. In this model, both coagulation and condensation were the key mechanisms cons idered. The reduction of the fine mode volume and the increase of MMD of the fine mode (more than 1 m) are the key parameters showing the effectivenes s of sorbent injection technique. Case 1: Bimodal coagulation only and unimodal coagulation only When sorbent particles were injected where vapor had nucleated in stantly, the system undergoes bimodal coagulation. The MMD of the fine mode grows by bimodal coagulation (inter-coagulation and intra-coagulation), however the volume of the fine mode can be removed by only inter-coagulat ion. To evaluate the effects of intercoagulation in bimodal coagulation, si mulation was also conducted for unimodal coagulation (i.e. no coarse mode sorbent par ticles). As shown in Fi gure 4-17A, the initial total number concentration was the same for both cases (bimodal coagulation, dot line, and unimodal coagulation, solid line) and they rapidly decreased from 1013 to 1010 in 0.05 seconds. However, the final number concentr ation of bimodal coagulation was less than that of unimodal coagulation indicating the e ffect of inter-coagulation. The fine mode total volume concentration for bimodal coagul ation decreased very slowly (Figure 4-17B, dotted line) and 6% of the initial volume was removed. Since the total volume is conserved in intra-coagulation, this re duction was due to inter-coagulation. The geometric standard deviation of the fine mode for unimodal coagulation approached 1.326 (Figure 4-17C, solid line) wh ich is the asympto tic value for coagulating aerosols in

PAGE 92

77 the transition regime, while for bimodal coagul ation it reached 1.172 indicating the effect of inter-coagulation (Figure 4-17C, dotted line). The MMD of fine mode reached 0.0178 m for unimodal coagulation and 0.0203 m for bimodal coagulation (Figure 4-17D). As explained in Chapter 3, the inter-coagulation ra te of the small particles in the fine mode was faster than that of the large particles in the fine mode. Hence the smaller particles of the fine mode were removed more by in ter-coagulation resulting in 20% number concentration reduction with 6% volume reduction. The larger particles in the fine mode, however, were not affected much by inter-coa gulation. The MMD of the fine mode for bimodal coagulation increased slightly faster than that of unimodal coagulation (Figure 417D); however, the MMD was still much less than 1 m within the given residence time. The above results demonstrate that coagulati on only could not be an effective mechanism for removing fine mode particles. Case 2: Condensation only When condensation was the only mechanism, its vapor consumption rate was very fast. As shown in Figure 4-17E (dash-dotte d line), the vapor was depleted in 0.036 seconds (i.e., its saturation ratio reached 0) Figure 4-17F shows the change of MMD of the coarse mode as a function of time. MMD of the coarse mode increased rapidly until 0.036 seconds. It then did not change after va por was depleted. This clearly showed that condensation was a faster process for mass tr ansfer of the metal to coarse particles compared to inter-coagulation. It can also be concluded that the preferable condition for sorbent injection techni que is condensation.

PAGE 93

78 Case 3: 50% instant nucleation Since the nucleation rate of vanadium particles is unknown, simulation was conducted for a scenario where 50% of the va por has instantly nucleated. This allowed for evaluation of the competition of thes e two mechanisms (bimodal coagulation and condensation) on the evolution of the fine mode aerosol. Intere stingly, as shown in Figure 4.17E, the saturation ratio (dash lin e) went down to 0 rapidly (4.3x10-3 seconds) which was faster than the conde nsation only case (dash-do tted line). The high number concentration of instantly nucleated fine mode particles had a high su rface area that was responsible for the rapid scavenge of the meta l vapor in the system. The total volume of the fine mode particles was increased by c ondensation in a short period of time (4.3x10-3 seconds) showing that metal vapor condensed on the fine mode particles (Figure 4-17B, dash line). The increase of the fine mode volume due to condensation was greater than the reduction due to inter-coagulation. Noneth eless MMD of the fine mode reached only 0.019 m (Figure 4-17D) which was much less than the desired MMD (> 1 m). The initial number concentration of the fine mode was half of the bimodal coagulation case. As shown in Figure 4.17A, nevertheless it s number concentration quickly approached that of the bimodal coagulation case (das h line and dotted line were overlapped). Although fast condensation on coarse mode particles was desired, metal vapor condensation on fine mode aerosol had an a dverse effect. As shown in Figure 4-17A, the number concentration of the fine mode wa s high and intra-coagulation was still the dominant for the fine mode particles. Conse quently, the removal of the fine mode volume concentration by inter-coagulation in this case was still negligible as shown in Figure 417B. Figure 4-17E shows MMD of the coarse m ode as a function of time. In the 50%

PAGE 94

79 Num conc. of the fine mode (#/cm 3 ) 1091010101110121013 Unimodal coag only Bimodal coag only 50% instant nucleation Saturation ratio 1 10 100 1000 10000 50% instant nucleation Cond ony Total volume of the fine mode (cm 3 /cm 3 ) 4x10-98x10-912x10-916x10-9 Bimodal coag only 50% instant nucleation time (s) 0.000.050.100.150.200.250.30 g of the fine mode 1.0 1.1 1.2 1.3 Unimodal coag only Bimodal coag only 50% instant nucleation MMD of fine mode ( m) 0.000 0.005 0.010 0.015 0.020 0.025 Unimodal coag only Bimodal coag only 50% instant nucleation A B C D E time (s) 0.000.050.100.150.200.250.30 MMD of the coarse mode 1.0700 1.0702 1.0704 1.0706 1.0708 1.0710 Bimodal coag only 50% instant nucleation Cond only F Figure 4-17. The change of A) total number concentration of the fine mode, B) total volume concentration of the fine mode C) geometric standard deviation ( g) of the fine mode, D) MMD of the fine mode, E) Saturation ratio of V vapor, and F) MMD of the coarse mode as function of time.

PAGE 95

80 instant nucleation case, some vapor also c ondensed on the coarse mode particles that grow slightly larger than the bimodal coagul ation case. However, the condensation on the coarse mode was relatively insignificant on ce a high number of fi ne mode aerosol was present. As discussed earlier, the vapor domin antly condensed on the fine mode particles. Since the removal of the fine mode by inter-coagulation was slow, the contri bution to the coarse mode growth through inter-coagulati on was insignificant. These observations clearly demonstrate that nucleation is not de sired for the sorbent injection technique. If gas phase hydrolysis or thermal decomposition occurs, it forms a high number concentration of fine mode particles that can not be effectively removed by coarse mode particles due to slow inter-coagulation. In addition, the fine mode is competing with the coarse mode for the metal vapor and is actually more effective. Such a competition renders the sorbent technique less effective. Conclusions A mechanistic study for sorbent injection technique was conducted theoretically and experimentally. The feasib ility study verified the re sults of the thermodynamic equilibrium analysis and assisted in sele cting the sorbent candidate for chemical adsorption (Ca-based sorbent). Experiments we re divided into coa gulation dominant and condensation dominant cases. The results of experiments clearly showed that condensation was the preferred mechan ism for mass transfer process. In the coagulation only case, element vana dium PSD showed only a slight shift to larger particles. At least 107/cm3 number concentration of sorb ent particles was needed to improve the efficiency of sorbent injecti on technique by coagulat ion in the system studied that had a residence time of a fe w seconds. In the condensation case, the experimental results clearly showed that elemental vanadium PSD was shifted to the

PAGE 96

81 larger sizes of sorbent particles. CaCO3 sorbent which has high affinity with vanadium and silica sorbent which has no affinity but high surface area both exhibited the similar capture efficiency. Gas phase hydrolysis resulted in the formation of fine particles while the surface hydrolysis enhanced physical ad sorption. Thus, surface hydrolysis was the key mechanism to effectively capture the metal. The bimodal lognormal model study showed that complete vapor consumption was achieved within 0.03 seconds, proving surface interaction was the rate limiting process. The reduction of the fine mode volume c oncentration and number concentration for coagulation only case were 6% and 20% in 0. 3 second residence time. Smaller particles in the fine mode were removed more than bi gger particles in fine mode. However, slow inter-coagulation rate by high number concentration of fine mode particles supported that condensation was the preferable mechanism for sorbent technique Instant nucleation formed high surface area of fine particles re sulting in fast metal vapor condensation on fine mode particles. Its condensed volu me was more than reduced volume by intercoagulation. These results cl early demonstrated that nucl eation was not desired for the sorbent injection technique. In conclusion, the preferable mechanism fo r mass transfer is c ondensation with fast surface interaction such as surface hydrolysis Alternatively, othe r scenarios that can result in fast surface interaction are also de sired, e.g. maintaining metals in the vapor state at high temperature so th at nucleation can be avoided. If in a system where it is impossible to suppress nucleation, a suffici ently high number concentration of sorbent particles must be provided to achieve a good collection effi ciency by inter-coagulation.

PAGE 97

82 CHAPTER 5 CONCLUSION AND RECOMMENDATIONS Vanadium is one of the potential ha zardous metal compounds from combustion sources. It is concentrated in the submicron regime and sorbent injection is one promising measure to control submicron vanadium particles. This study was carried out theoretically and experimentally to evaluate sorbent injection technique for the control of vanadium emission from combustion sources. Theoretical study includes thermodynami c equilibrium analysis and bimodal lognormal model study. Potential material for sorbent injection tec hnique was determined by thermodynamic equilibrium analysis, and bi modal lognormal model study used to gain the insight of aerosol dynamic as the m echanisms of sorbent injection technique. Experimental study includes pot experiment and aerosol reactor experiments. Pot experiment verified the resu lts of thermodynamic analysis. Aerosol reactor experiment determined preferable mechanism and invest igated the surface in teraction for sorbent injection technique. By thermodynamic equilibrium analysis, Na-, Ca-, and Mg-based sorbents were found to be effective in a wide range of temp eratures. However, sulfur and sorbents were shown to have high affinity (forming sulfat es) at temperatures lower than 1000 K. The strong affinity resulted in the depletion of available sorbents in the system. Sufficient sorbent in excess of sulfur should be provide d in this temperature range to effectively capture vanadium compounds. At high temper atures (> 1000 K), the effectiveness of Na-

PAGE 98

83 and Ca-based sorbents to capture vanadium revived as sorbent sulfates became less stable, releasing available sorbents to react with vanadium. When vanadium formed submicron partic les by nucleation, in ter-coagulation was the only mechanism to remove these particles by sorbent particles. He nce, the effects of inter-coagulation on removing fine mode aerosol were studied using a bimodal lognormal model. Number concentration of coarse m ode, geometric standard deviation of fine mode, and MMD of fine mode were found to be important parameters for intercoagulation rate. Interestingly when intercoagulation was the do minant mechanism for removing fine mode particles, its geometric standard deviation approached monodisperse. For coarse mode particles, the geometric st andard deviation approached the asymptotic value because intra-coagulation was still the dominant mechanis m. The removal time could also be expressed in the dimensionles s form and two formulas were developed for inter-coagulation dominant and intra-coagula tion dominant respectively. The developed formula could be used as a convenient tool to estimate the removal time once the particle scavenge characteristic time was determined. Ca-based sorbent and Na-based sorbent showed their chemical affinity with vanadium in the pot experiments. Aeroso l reactor experiments were divided into coagulation dominant and c ondensation dominant cases. The results of experiments clearly showed that condensation is the prefer red mechanism for mass transfer process. In the coagulation only case, element vanadi um PSD showed only a slight shift to supermicron. Formulas developed by inter-c oagulation study showed that at least 107/cm3 number concentration of sorbent particles we re needed to improve the efficiency of sorbent injection technique by coagulation for a typical combustion system with a

PAGE 99

84 residence time of a few seconds. In the c ondensation case, the experimental results clearly showed that element vanadium PSD wa s shifted to the larger sizes of sorbent particles. CaCO3 sorbent which had high affinity with vanadium and silica sorbent which had no affinity but high surface area both exhibited the similar capture efficiency. Gas phase hydrolysis resulted in the formation of fine particles while the surface hydrolysis enhanced physical adsorption. Silica sorb ent which had no chemical affinity with vanadium also successfully cap tured vanadium vapor. Vanadium PSD shifted to the size where the surface area of sorbent was high. This was the strong evidence for physical adsorption. In real system, atomic vanadium vapor can be possibly condensed on sorbent particles. Condensation rate is not limiting pr ocess, however its fast nucleation rate may affect on the efficiency of sorbent injection. A bimodal lognormal modeling study based on the initial condition of experiments showed that condensation was the preferable mechanism. When condensation (based on vanadium pentaoxide vapor) was the only mechanism, it showed the fastest and complete collection of vanadium vapor. If nucleati on occurred, condensa tion on the instantly nucleated fine particles c ould yield undesired results be cause of the slow intercoagulation rate that would not be effective in removing them. Meanwhile, the amount condensed on the fine mode was insufficient to increase MMD of fine mode to greater than 1 m. In conclusion, the preferable mechanism for mass transfer was condensation with the fast surface interaction such as surface hydr olysis. Alternatively, other scenarios that could result in fast surface interaction were al so desired, e.g. maintaining metals in the vapor state at high temperature so that nuclea tion could be avoided. If in a system where

PAGE 100

85 it is impossible to suppress nucleation, a sufficiently high number concentration of sorbent particles must be provided to achieve a good collection efficiency. Based on the conclusions pres ented above and the experien ce gained in performing this research, recommendations can be made that should further the understanding and application of sorbent technique to the control of vanadium. 1. According to thermodynamic equilibrium analysis, most stable compound in combustion temperature was vanadium pent aoxide. However, if atomic vanadium vapor is formed, its nucleation rate is fa ster than vanadium pentaoxide at same temperature because of its higher saturati on ratio. It consequently results adverse effect on sorbent injection. Hence, formed chemical species analysis and nucleation rate study need to be performed. 2. There will be the relationship among number concentration of nuc leated particles, condensable vapor and PSD of sorbent particles to suppress nucleation. The bimodal lognormal model which is used in th is study could be an excellent tool for this study. 3. Collison nebulizer and ultrasonic nebulizer used in this study generated polydisperse particles. They also ge nerated undesired submicron particles. Theoretical study performed by this work showed that monodisperse articles around 1 m particles was the most effective PS D of sorbent. Experimental studies are needed to verify the theoretical finding. 4. Surface hydrolysis enhanced sorbent injec tion technique. However, its final product on the surface of sorbent was unknown in this study. Based on the thermodynamic equilibrium study and the study of TTIP hydrol ysis which has similar structure, the final product might be vanadium pentaoxide which is an excellent catalyst for NOx removal. Hence, performing further study of the surface hydrolysis can determine the surface interaction. This process can also be used to make micron particles coated by nanosize vanadium pentaoxides. 5. Physical adsorption should be further studi ed. In this study, silica particles captured vanadium vapor successfully which is th e evidence of the im portance of physical adsorption. However, there was the surface hydrolysis supporting physical adsorption. Studies should be carried out to examine if other surface interaction in addition to hydrolysis can also help physic al condensation. Dry silica particles with different particle generation methods such as dust feeder and screw feeder can be used for this type of study.

PAGE 101

86 APPENDIX A FORTRAN CODE FOR BIMODAL LOGNORMAL MODEL

PAGE 102

87

PAGE 103

88

PAGE 104

89

PAGE 105

90

PAGE 106

91

PAGE 107

92

PAGE 108

93

PAGE 109

94

PAGE 110

95

PAGE 111

96

PAGE 112

97

PAGE 113

98

PAGE 114

99

PAGE 115

100

PAGE 116

101

PAGE 117

102

PAGE 118

103

PAGE 119

104

PAGE 120

105

PAGE 121

106

PAGE 122

107

PAGE 123

108

PAGE 124

109

PAGE 125

110

PAGE 126

111

PAGE 127

112

PAGE 128

113

PAGE 129

114

PAGE 130

115

PAGE 131

116

PAGE 132

117

PAGE 133

118

PAGE 134

119

PAGE 135

120

PAGE 136

121

PAGE 137

122

PAGE 138

123

PAGE 139

124

PAGE 140

125 APPENDIX B MOMENT RELATIONSHIPS FOR TH E LOGNORMAL DISTRIBUTION The integral moments of a distribution are defined as, 0 ) ( ) ( p d d p d n k p d k M (b-1) which can also be expressed in terms of ln(dp) as ) (ln ) (ln p d d p d n k p d k M (b-2) The lognormal number di stribution function is g gn D p d g N p d d dN p d n 2 ln ) ln (ln 5 0 exp ln 2 ) (ln ) (ln (b-3) It is convenient to make a change of variables as g p g gn pd d dx x D d g gn D p d x ln / ) (ln ) ln exp( ln ) ln (ln Using this change of variables and subs tituting eq. (b-2) into eq. (b-3) yields dx kx x ND Mg gn k) ln 5 0 exp( 22 (b-4) Expanding the exponential term, comple ting the square, and substituting ) ln (gk x y into eq. (b-4) yields dy y k ND Mg gn k) 5 0 exp( ) ln 5 0 exp( 22 2 2 (b-5a)

PAGE 141

126 ) ln 2 / exp(2 2 g k gnk ND (b-5b) By defining average moments and 1 kMas 2 kM as ) ln 2 / exp(2 2 1 1 1 g k gn kk D M (b-6a) ) ln 2 / exp(2 2 2 2 2 g k gn kk D M (b-6b) and solving for Dgn and g the conversion equations between Dgn and g ; 1 kMand 2 kM are ) 2 1 /( 2 )] 1 2 ( /[ 1 1 k k r k k k r k gnM M D (b-7a) ) / ln( ) 1 2 ( 1 2 ln2 1 2 r k k gM M k k k (b-7b) where 2 / 1 k k r Substituting eq. (b-7a) and (b-7b) into eq. (b -5b) yields an expression for other moments of the distribution in terms of the moments 1 kM and 2 kM 2 2 1 1 k k k k kM M N M (b-8) where ) 1 /( ] ) 2 / ( ) 2 / ( [ 2 ) 1 /( )] 1 / ( ) 1 / ( [ 12 2 r k k k k r k r k k k k r k

PAGE 142

127 APPENDIX C MATERIAL SAFETY DATA SHEET Section 1. Product Identification Chemical name: Vanadium (V) tri-i-propoxy oxide, 98+% Product number: 23-5000 CAS resister number: 5588-84-1 Formula: VO(OC3H7)3 Einecs number: 226-997-4 Synonym: Isopropylorthovanadate, Tri-iso-prop ylvanadate, Vanadyl isopropylate Section 2. Hazardous Identification Emergency overview: Flammable liquid and vapor. Inhalation may cause dizziness and headache. Irritating to eyes, skin and mucous membranes. Ingestion may cause nausea and vomiting. Primary routes of exposure: Eye, skin and inhalation Eye contact: Cause mild to severe irritation of the eyes Skin contact: Cause slight to mild irritation of the skin. Prolonged contact may dry the skin and lead to rashes or more severe irritation. Inhalation: irritating to mouth throat and st omach. May cause dizziness, nausea, vomiting, pain and stomach upset. Acute health effects: irritating to skin, eyes and mu cous membranes. More severe effects (ingestion and inhalation) are dizziness, nausea, vomiting. Chronic health effects: No information available on long term chromic effects. Section 3. Handling and Storage Handling and storage: Store in a cool, dry, well-ve ntilated area away from heat and direct sunlight. Keep c ontainers tightly sealed. This product will react with

PAGE 143

128 atmospheric moisture and chemically degr ade. Transfer material under an inert atmosphere of nitrogen argon or very dry air Section 4. Exposure Controls and Personal Protection Eye protection: Always wear safety glasses. If pouring or transferring this substance wear a face shield for additional protection. Skin protection: Wear protective clothing and groves. Consult with glove manufacturer to determine the proper type of glove. Ventilation: Work with this product in a well-ve ntilated area, preferably a fume hood. Respirator: If ventilation is not available a respirator s hould be worn. The use of respirators requires a Respirator Protecti on Program to be in compliance with 29 CFR 1910.134 Additional protection: No additional protection required Section 5. Physical and Chemical Properties Color and form: Light yellow to light green liquid. Molecular weight: 244.20 Melting point (oC): Not determined Boiling point (oC): 60-61o/0.5 mm Vapor pressure: No data Odor: none Solubility in water: reacts with water

PAGE 144

129 LIST OF REFERENCES Akeredolu, F. A., Barrie, L. A., Olson, M. P., Oikawa, K. K., Pacyna, J. M., and Keeler, G. J., 1994, Flux of anthropogenic trace metals into the Arctic from the midlatitudes in 1979/80, Atmospheric Environment, v 28(8), p 1557-1572. Alemany, L. J., Larrubia, M. A., and Blasco J. M., 1998, Catalytic activity in partial oxidation of methane and physico-chemical characterization of a VPO system obtained from boiler ash, Appl. Catal. B:Environ., v 16(2), p 139-147. Bacci, P., Monte, M. D., Longhetto, A., Pia no, A., Prodi, F. and Redaelli, P. (1983). Characterization of the particulate emissi on by a large oil fuel fired power plant, J. Aerosol Sci. v 14, p 557-572 Barin, I., 3rd Ed., 1995, Thermochemical Data of Pure Substances VCH, New York Biswas, P., and Wu, C. Y., 1998, Control of to xic metal emissions from combustors using sorbents: A review J. Air & Waste Manage. Assoc. v 48, p 113-127. Boyd, D. W., and Kustin, K., 1984, Vandium: A versatile biochemical effector with an elusive biological function. In C. L. Eichhorn and L. G. Marzilli (Eds.), Advances in Inorganic Biochemistry v 6, p 311-363 Bryers, R. W., 1995, Utilization of petroleum coke and petroleum coke/coal blends as a means of steam raising, Fuel Processing Technology, v 44(1-3), p 121-141. Bylinska, E., 1996., Bioindication of titaniu m, vanadium and lanthanum coming from long distance emission in Sudety mountain monitoring, In Proceeding of simulation and control international conference on air pollution—computational Mechanics Inc., Billerica, MA, USA. p 753-759. Campen, M. J., Nolan, J. P., Schladweiler, M. C. J., Kodavanti, U. P., Evansky, P. A., and Costa, D.L., 2001, Cardi ovascular and thermoregulator y effects of inhaled PM associated transition metals: a potential interaction between ni ckel and vanadium sulfate, Toxicology Sci. v 64, p 243-252 Carey, T. R., Richardson, C. F., Chang, R., Meserole, F. B., Rostam-Abadi, M., and Chen, S., 2000, Assessing sorbent injection mercury control eff ectiveness in flue gas streams, Environ. Prog. v 19(3), p 167-173

PAGE 145

130 Chase, M. W., Davies, C. A., Downey, J. R., Frurip, D. J., McDonald R. A. and Syverud, A. N., 3rd Edition, 1986, JANAF Thermochemical tables, American Chemical Society and American Institute of Physics for the National Bureau of Standards, Midland, MI. Chen, J. C., Wey, M. Y., and Liu, Z. S., 2001, Adsorption mechanism of heavy metals on sorbents during incineration, J. Envron. Env. Eng. v 127, p 63-69 Divita, F., Ondov, J.M., and Suarez, A. E., 1996, Size spectra and at mospheric growth of V-containing aerosol in washington DC., Aerosol Sci. Tech. v 25, p 256-273 Energy Information Administration, 1996, Washing D.C., International Energy Annual Report Fan, L. S., Jiang, P., Agnihotri, R., Mahuli, S. K., Zhang, J., Chauk, S., Ghosh-Dastidar, A., 1999, Dispersion and ultra-fast reacti on of calcium-based sorbent powders for SO2 and air toxics remova l in coal combustion, Chemical Engineering Sci. v 54, p 5585-5597 Fang, M. X., Luo, Z. Y., Li, X. T., Wang, Q. H., Ni, M. J., and Chen, K. F., 1998, Multiproduct cogeneration system using combin ed coal gasification and combustion, Energy (Oxford) v 23(3), p 203-212. Flagan, R. C. and Seinfeld, J. H., 1988, Fundamentals of Air Pollution Engineering Prentice Hall, Englewood Cliffs, NJ, p 433-471. Frenklach, M., and Harris, S.J., 1987, Aero sol dynamics modeling using the method of moments, J. Colloid Interf. Sci. v 118, p 252-261 Friedlander, S. K., Koch, W., and Main, H. H., 1991, Scavenging of a Coagulating fine aerosol by a coarse particle mode, J. Aerosol Sci. v 22, p 1-8. Gelbard, E., and Seinfeld, J. H., 1980, Si mulation of multicomponent aerosol dynamics, J. Colloid Interface Sci. v 78, p 485-501 Harries, R. R., 1993, Process modeling for we t limestone flue gas desulphurization. ICHEME sym. Ser. 131, p 167-81 Helble, J. J., and Sarofim, A. F., 1989, Fact ors determining the primary particle size of flame-generated inorganic aerosols, J. Coll. Int. Sci ., v 128 (2), p 348-362. Hind, W. C., 1999, Aerosol Technology John Wiley & Sons, Inc., NY, p 428-434 Hogu, C., 2000, Body burdens of pollutants, Chem. Eng. News v 78 (31), p 28 – 29.

PAGE 146

131 Huffman, G. P., Huggins, F. E., Shah, N. R., Hu ggins, W. P., Linak, C. A., Miller, R. J., Pugmire, H. L. C., Meuzelaar, M., and Seehra, A., 2000, Characterization of Fine Particulate Matter Produced by Co mbustion of Residual Fuel Oil, J. Air & Waste Management Assoc. v 50, p 1106-1114 Iida, K., Wu, C. Y., and Hahn, D. W., 2004, In situ analysis of the interaction of manganese vapor and silica microspheres in a laboratory-scale combustor, Combust. Sci. and Technol. v 176 (3), p 453-480 Landgrebe, J. D., and Pratsini s, S. E., 1990, A discrete-sectional model for particulate production by gas-phase chemical-reaction and aerosol coagulation in the freemolecular regime, Colloid Interf. Sci. v 139, p 63-86 Lee, C. C., 1988, A model analysis of metal partitioning in a hazardous waste incineration system, JAPCA v 38(7), p 941-945. Lee, S. R., and Wu, C. Y., 2002, Study of vanadium emission control in combustion systems by thermodynamic equilibrium analyses, Advances in Environ. Research v 7, p 1-10 Lee, S. R., and Wu, C. Y., 2004, Size distribution e volution of fine aerosols due to intercoagulation with coarse aerosols, submitted to Aerosol Sci & Tech. (in process) Lighty, I. S., Veranth, J. M., and Sarofim, A. F., 2000, 2000 critical review-combustion aerosols: Factors governing their size and composition and implications to human health., J. Air Wastes Mang. Assoc. v 50, p 1565-1618. Lin, C.Y., and Chiu, C. H., 1995, Emissions of industrial furnaces burning vanadiumcontained heavy oils, Environ. Sci. Health, Part A: Environ. Sci. Eng. Toxic Hazard. Substance v 30(1), p 133-142. Lin, W. Y., and Biwas, P., 1994, Metallic pa rticle formation and growth dynamics during incineration, Combust. Sci. Technol. v 101, p 29-43 Linak, W. P., and Wendt, J. O. L., 1998, Part itioning of the refractory metals, nickel and chromium, in combustion systems, Combustion Sci. and Tech v 134, p. 291-314 Linak, W. P., Miller, C. A., Santoianni, D. A., King, C. J ., Shinagawa, T., Wendt, J. O. L., Yoo, J. I., and Seo, Y. C., 2003, Forma tion of fine particles from residual oil combustion: Reducing nuclei through the addition of inorganic sorbent, Korean J. Chem. Eng. v 20(4), p 664-669 Linak, W. P., Srivastava, R. K., and Wendt, J. O. L., 1995, Sorbent capture of nickel, lead, and cadmium in a laboratory swirl flame incinerator, Combustion and flame v 100, p 241-250 Linak, W. P., and Wendt, J.O.L., 1993, T oxic metal emissions from incineration: Mechanisms and control, Prog. Energy Combust. Sci. v 19, p 145-185

PAGE 147

132 Linak, W. P., and Wendt, J.O.L., 1994, Tr ace metal transformation mechanisms during coal combustion, Fuel Process. Technol. v 39 (2), p 173-198 Linak, W. P., and Miller, C. A., 2000, Comparison of par ticle size distributions and elemental partitioning from the combustion of pulverized coal and residual fuel oil, Air & Waste Mange. Assoc ., v 50, p. 1532-1544. Lundgren, D. A., and Vanderpool, R. W., 1988, Design and calibration of an in-stack low pressure impactor, US air fo rce, Contract No. F08635-83-C0136 Lyyranen, J., Jokiniemi, J., Kauppinen, E. I., and Joutsensaari, J., 1999, Aerosol characterization in medium-speed diesel engines operating with heavy fuel oils, J. Aerosol Sci. v 30, p 771-784 Mahuli, S., Agnihotri, R., Chauk, S. G hosh-Dastidar, A., and Fan, L. S., 1997, Mechanism of arsenic sorption by hydrated lime, Environ. Sci. Technol. v 31(11), p 3226-3231. McMillin, B. K., Biswas, P. and Zachariah, M., 1996, In Situ Diagnostics of vapor phase growth of iron oxide-sili ca nanocomposites Part I: 2D planar laser-induced fluorescence and mie imaging, J. Mater. Res. v 11(6), p 1552-1561. Nerin, C., Domeno, C., and Superior, C. P., 1999, Distribution of Pb, V, Cr, Ni, Cd, Cu and Fe in particles formed from the combustion of waste oils, Chemosphere ,v 38, p 1533-1540 Osan, J., Torok, S., Fekete, J., and Rindby, A ., 2000, Case study of the emissions from a heavy-oil-fueled Hungarian power plant, Energy & Fuels v 14, p 986-993 Owens, T. M., and Biswas, P., 1996, Reactions between vapor phase lead compounds and In Situ generated silica particles at various l ead-silicon feed ratios: applications to toxic metal capture in combustors, J. Air Waste Manage. Assoc. v 46, p 530-538. Owens, T. M., Wu, C. Y., and Biswas, P., 1995, An equilibrium analysis for reactions of metal compounds with sorbents in high temperature systems, Chemical Engineering Communications v 133, p 31-52. Pirrone, N., Costa, P., and Pacyna, J. M.; 1999, Past, current and projected atmospheric emissions of trace elements in the Mediterranean region, Water Sci. Technol. v 39(12), p 1-7. Prastinis, S. E., 1988, Simultane ous nucleation, condensation and coagulation in aerosol reactors, J. Colloid Interface Sci. v 124(2), p 416-427 Punjak, W. A., Uberoi, M., and Shadman F., 1989, High-temperature adsorption of alkali vapors on solid sorbents, AIChE Journal v 35, p 1186-1194

PAGE 148

133 Rahaman, M. N.,1995, Ceramic processing and sintering Marcl Dekker Inc., Madison Ave., NY, p 201-263. Reynolds, W. C., 1995, STANJAN-Interactiv e computer programs for chemical equilibrium analysis, Dept. Mechanical E ngineering, Stanford Univ., Standford, CA Rodriguez, A., and Hall, M. J., 2001, Removal of an airborne low volatility metal under fuel-rich and fuel-lean c ondition through condensation on to soot and/or sorbent particles. Waste management v 21(7), p 589-607 Rodriguez, A. and Hall, M. J., 2003, The si mulation of condensation removal of a heavy metal from exhaust gases onto sorbent particles, Waste Management v 23(6), p 493-502 Scotto, M. V., Uberoi, M., Peterson, T. W ., Shadman, F., and Wendt, J. O. L., 1994, Metal capture by sorbents in combustion processes, Fuel Proc. Technol. v 39(1/3), p 357-372. Swain, E. J., 1991, Power, cement industries shape coke future, Oil and Gas, v 89(20), p 49-52. Tolocka M. P., Lake D. A., Johnston M. V., and Wexler A.S., 2004, Number concentrations of fine and ultrafine particles containing metals, v 38(20), p 32633273 Tsang, T. H., and Brock, J. R. (1982). Aero sol coagulation in the plume from a crosswind line source, Atmos. Envrion. 16: 2229-2235 Tsuboi, I. K, Shigetami, K., Eiichi, K., and Isao, K., 1991, Recovery of gallium and vanadium from coal fly ash, J. Chem. Eng. Jpn ., v 24(1), p 15-20. Uberoi, M., and Shadman, F., 1991, Simulta neous condensation and reaction of metal compound vapors in porous solids, Ind. Eng. Chem. Res. v 30, p 624-631. Venkatesh, S., Fournier, D. J. Jr., Waterland, L. R. and Carro ll, G. J., 1996, Evaluation of Mineral-Based Additives as Sorbents for Hazardous Trace Metal Capture and Immobilization in Inci neration Processes, Haz. Waste. Haz. Mat. v 13(1), p 73-94. Whitby, E. R., McMurry, P. H., Binkowski, F., and Shankar, U., 1991, Modal aerosol dynamics modeling. EPA report for contract No. 68-01-7365. Whitby, E. R., and McMurry, P. H., 1997, Modal aerosol dynamics modeling, Aerosol Sci. and Tech. v 27(6), p 673-688

PAGE 149

134 Whitby, K. T., 1979, Lumped mode aeros ol growth model. Particle Technology Laboratory Publication #395, Mechanical E ngineering Department, University of Minnesota, Minneapolis, MN.Wiliams, M. M. R. and Loyalka, S. K., 1991, Aerosol science: Theory and practice Pergamon Press. Wu, C. Y., and Barton, T., 2001, A thermodynami c equilibrium analysis to determine the potential sorbent material for the control of arsenic emissions, Environ. Eng. Sci. v 18(3), p 177-190 Wu, C. Y., and Biswas, P., 1993, An equilibrium analysis to determine the speciation of metals in an incinerator, Combustion and Flame v 92, p 31-40. Wu, C. Y., and Biswas, P., 1998, Study of nu merical diffusion in a discrete-sectional model and its application to aerosol dynamics simulation, Aerosol Sci. and Tech. v 29 (5), p 359-378 Wu, C. Y., and Biswas, P., 2000, Lead sp ecies aerosol formation and growth in multicomponent high temperature environments, Env. Eng. Sci. v 17(1), p 41-60. Wu, C. Y., Lee, T. G., Arar, E., Tyree, G ., and Biswas, P., 1998, Capture of mercury in combustion environments by In-Situ generated titania partic les with UV radiation, Env. Eng. Sci. v 15(2), p 137-148. Yee, B. H., and Rosenquist, W. A., 1996, Petrol eum coke as an alternative fuel for new generation and repowering, In Proceedings of the American Power Conference v 2., Illinois Institute of Technology, Chicago, IL, USA. p 838-841.

PAGE 150

135 BIOGRAPHICAL SKETCH Sang-Rin Lee was born in Pusan, Korea on July 4, 1969. After graduation from Pusan High School, in 1988, he moved to Incheo n to attend Inha university. He applied to the Electrical Engineering department but he was admitted to the Environmental Engineering department which became the cornerstone of his lif e. After receiving Bachelor of Science degree from the Envir onmental Engineering department in 1993, he continued his graduate study at that same university. His special topic was acid rain modeling. In addition to this model work, he al so got involved with lo ts of projects such as PM10 sampling and air quali ty consulting. He worked for LG-Product Engineering Research Center for 1 year af ter receiving his Master of Sc ience degree in 1995. He then moved to Korea Power Engineering Company. He worked as a process engineer for development and construction of the Korean type flue gas desulfurization system at Young-Dong Power Plant (200 MW and 150 MW). He was involved in the design of an Flue Gas Desulfurization system for 2 year s and worked at Young-Dong Power Plant for construction and pre-operation for 1 year. With 4 years of valuable e xperience in the real world he decided to continue his studies in the U.S. He was admitted with a research assistantship from Dr. Chang-Yu Wu in the Environmental Engineering Sciences Department at the University of Florida. He studied aerosol technology and sorbent technology. He was awarded a Ph. D. in May 2005.