Metal  Partitioning in  Combustion Processes
WILLIAM P. L1NAK
    Air Pollution Prevention and Control Division. MD-65
    National Risk Management Research Laboratory
    U.S. Environmental Protection Agency
    Research Triangle Park, NC 27711 USA
                                                                                                          EPA/600/A-97/085
                                                                             Fourth International Conference on
                                                                             Technologies and Combustion for a Clean Environment
                                                                             July 7-10, 1997
                                                                             Lisboa, Portugal
Abstract -• This article summarizes ongoing research efforts at  the
National Risk Management Research Laboratory of the U.S. Environmental
Protection Agency examining [high temperature) metal behavior within
combustion environments.  The partitioning of non-volatile (Cr and Ni),
semi-volatile (Cd and Pb), and volatile (Hg) metals in combustion systems
was investigated theoretically and experimentally. Theoretical predictions
were based on chemical equilibrium and suggested  that such calculations
can be useful in predicting relative volatility and speciaticm trends, and to
direct experimental efforts.  Equilibrium calculations, however, are not
sufficient to quantitatively predict the behavior of metals even in simple
combustion environments.   Experimental studies employing a 59 kW
laboratory scale combustor examined the behavior (volatility, particle size,
and specialion) of metal  vapors and particles produced by aqueous mclal
solutions sprayed through a swirling natural gas diffusion  flame. These
experiments were designed to study metal transformation mechanisms in a
relatively simple combustion environment without the complex effects of
additional species. Further experiments examined the potential use of
common inorganic sorbents (kaolinite, bauxite, and hydrated lime) to
adsorb metal vapor, offering a potential means of metal emissions control.
                                                         its spcciation and partitioning between the submicron (<1 nm in diameter)
                                                         and supermicron (>1 jim in diameter) aerosol in the exhaust. Of particular
                                                         interest are methods of affecting metal speciation and particle size to
                                                         maximize control and minimize toxicity and risk.
                                                            While much previous  research on  mechanisms governing the
                                                         transformations of metal compounds in combustion systems has focused on
                                                         coal combustion systems (Flagan and Fricdlander, 1978; Nettlelon, 1979;
                                                         Smith, 1980; Haynes rial., 1982; Neville and Sarofim, 1982), relatively
                                                         little is available on the fate of single metal compounds, introduced one at a
                                                         time, in the absence of other major ash constituents, as they might be during
                                                         incineration trial burns.  The coal data suggest that trace metals are often
                                                         enriched in the submicron particle size fraction, and this has been explained
                                                         by mechanisms involving metal vaporization and subsequent condensation
                                                         or surface reaction. There are complicated exceptions, including cases
                                                         where a volatile metal, such as sodium (Na), may be scavenged by alumino-
                                                         silicates displacing  other bound metals. Therefore, experiments involving
                                                         relatively simple combustion environments and pure metal compounds
                                                         (such as those presented here) are useful to test theoretical hypotheses and
                                                         help isolate mechanisms.
 Key Words:
Trace Metals; Partitioning; Volatility; Sorbent Interactions;
Combustion
 INTRODUCTION

 Metal compounds are present in the stack effluents of many combustion
 processes.  As health and environmental studies further identify the scope
 and magnitude of their adverse effects, the release of metals from boilers,
 furnaces, and incinerators into the environment is coming under increasing
 regulatory scrutiny in a number of countries.  In the U.S.,  metal  air
 emissions  from hazardous waste incinerators (HWis) and boilers and
 industrial furnaces (BIFs) which destroy hazardous waste are regulated by
 the Resource Conservation and Recovery Act (RCRA, 1986). Current air
 emission limits are based on risk assessment arguments which limit the
 ground level concentrations that may be inhaled by the "maximum exposed
 individual* Metals regulated by RCRA include a set of carcinogenic metals
 [arsenic (As), beryllium (Be), cadmium (Cd), and chromium (Cr)] and a set
 of noncarcinogemc metals [antimony (Sb), barium (Ba), lead (Pb), mercury
 (Hg), nickel (Ni), selenium (Se), silver (Ag). and thallium (Tl)]. Metal air
 emissions from many other combustion sources are regulated under Title
 ffl. Section 112 of the Clean Air Act Amendments (CAAA, 1990). Metals
 regulated by Section 112 include Sb, As, Be, Cd, Cr, cobalt (Co), Pb,
 manganese (Mn), Hg, Ni, and Se.  While municipal waste incinerators
 (MWIs) are also regulated under Title III, these units are covered by their
 own set of metal emission regulations (Section 129), which are likely to
 place limits on the emissions of Cd, Pb, and Hg.
   This paper, which presents results from theoretical and experimental
 studies conducted by  EPA's National Risk Management  Research
 Laboratory, is concerned with the partitioning of metals in combustion
 processes.  Under well defined  conditions, metals can  be segregated
 according to their relative volatilities. Rizeq et nl (1994) group mcials into
 three volatility classes. Metals, such as Hg and Se, tend to be volatile, even
 at moderate stack temperatures.  Others, such as Sb, As, Cd, Pb,  and Tl,
 are semi-volatile,  and  have  the  potential  of vaporizing at the high
 temperatures in the flame zone. Finally, a third group including Ba, Be, Cr,
 and Ni, are considered to be refractory (non-volatile) over the entire range
 of combustion temperatures usually encountered.  Of interest  arc the
 physical and  chemical transformations from the initial form of the metal to
 Mechanism Overview

 Metals may be contained in solid or liquid fuels, chemically bound to the
 organic fuel matrix (inherent mineral mailer), dispersed within the solid fuel
 as mineral crystallites (included mineral matter), or completely extraneous to
 the fuel particle (excluded mineral matter).  Metals may be chemically bound
 within organomctallic compounds such as chelates or physically mixed, as
 in paints, pigments, and solvents!  They may enter combustion processes
 with other  inorganic clays and soils,  as during the thermal treatment of
 contaminated soils (Eddings and Lijhty, 1992), or they may be contained in
 aqueous  solutions and  sludges.   Metals may be introduced into  a
' combustion environment continuously through atomizers (Bhalia and
 Sirignano, 1991), lances, or screw feeders, or through entrance  chutes in a
 batch mode as solids or contained liquids. They may be introduced as
 single salts or individual compounds, or they may enter as mixtures,  to
 fact, a critical issue in designing test burns for incinerators is whether the
 form in which the metal is introduced during the trial burn is representative
 of the behavior of other forms of that metal that may be introduced during
 routine operation.  The left-hand side of Figure 1 represents some of the
 various forms in which a metal may be introduced into a combuslor, and
 how this physical state can influence the ultimate fate of that metal.
   Upon entry into a combustion environment, the metals contained in a fuel
 or waste stream can be transformed into various physical forms. Dissolved
 metal salts, such as nitrates or pyrites, may form reactive metal compounds
 which may decompose violently at elevated temperatures. They may also
 decompose as viscous melts to form cenospheres  which can burst into
 submicron fragments (Mulholland and Sarofim, 1991). Alternatively, the
 metal compound may be confined within  a porous char matrix. This will
 happen, for example, to inherent (organically bound) metals in heavy oil or
 both inherent and  included mineral matter in pulverized coal.  The metal
 must then be released either within the matrix,  subsequently to diffuse
 through it, or it will enter the gas phase as the char matrix itself is oxidized.
 In the event of the former process, there is the opportunity for the metal to
 react with  included silicates to form stable compounds which fail to
 vaporize. Alternatively, the metal which has been released may diffuse back
 into the remaining char matrix, to react with included  silicates situated there,
 or it may react with excluded silicate particles in the disperse phase.
 Finally, one metal (e.g.,  Na) may displace another metal [e.g., potassium
 (K)J which would otherwise be immobile and bound in a stable mineral

-------
                                                                                                  Submicron Aerosol
               ( Aqueous Solutions  )
                                                                        NucIeatiofVCondensation/
                 Distillate Oil and
                 Organ ometallics
                                                 Dissolved/\
                                                  Inherent,
                                               Reactive Metal
                                                                      X   '
                                                                 Porous^ "^
                                                                                                 Submicron Inclusions,
                                                                                                   Residual Particles.
                                                                                                 Secondary Atomization
Heavy Oil and
    Coal
                                                                        ndMalal \Cenospheric
                                                                                   Particle
               I      Solids and
               ^Contaminated Soils
                                              Inorganic Mixture
                                                                                         Supermicron (Collectable)
                                                                                                  Aerosol
                                     FIGURE 1 Possible controlling mechanisms for particle formation in
                                                           combustion systems.
form, such as illite. Mineral inclusions may also coalesce as the carbon
matrix recedes lo form panicles larger lhan the individual inclusions.  The
physicochemical processes involved in  the release of metals may thus be
quite complex. They may be kinetically controlled, and the overall amount
of a metal released may, under certain circumstances, have little to do with
equilibrium.
  When a metal contained in contaminated soils, sludges, or slurries is
introduced into an incineration  environment, an  inorganic  mixture
containing both the  metal and potential scavenging agents, such as clays and
glasses, is formed.  It is not surprising, therefore, that (upon heating) much
of the metal may react with the clay and only a little will be released to form
a condensable vapor.  This has been experimentally  confirmed for
Pb/montmorillonite clay mixtures  by Eddings and Lighiy (1992).  Their
study showed large  deviations from simple equilibria based on selected pure
condensed phases, during the heating of inorganic mixtures. The work of
Queneau  et at (1991) addresses the  thermodynamics of the vitrified
mixtures likely 10 be formed under these conditions, and  is useful where
inorganic mixtures are formed, as shown in Figun: 1.
  The primary physical forms of the meuls outlined above (reactive metal
compounds, porous chars,  and inorganic mixtures as shown  in Figure 1)
can then undergo further transformations to other physical forms including
metal  vapors,  porous metal ash particles,  cenospheric (hollow) ash
particles, or dense  ash particles. Upon cooling, the supersaturated vapor
may condense on the surfaces of existing particles, or if sufficient surface
area is not available, homogeneously nucleate to form tiny particles. These
particles will subsequently  collide, coagulate,  and  agglomerate.
Alternatively, there is evidence that a metal vapor may react on the surface
of existing particles or sorbents.  In contrast to the first two processes
(heterogeneous condensation and  homogeneous  nucleation), surface
reaction does not require the metals' partial pressures to exceed their vapor
pressures.  These  mechanisms  and processes strongly influence the
chemical and physical form in which the metal under consideration enters
the environment. They are likely to depend strongly on the combustion
environment and temperature history experienced by the metal compound,
as well as on the initial form  in which it is introduced into the combustion
chamber, and on the presence or absence of other species in the mixture.
 Panicle Growth via Coagulation

 In addition to the heterogeneous condensation of vapor-phase species,
 particle growth occurs through panicle coagulation and agglomeration.
 Figure 2 illustrates the predicted evolution of an aerosol due to coagulation
 only.  The MAEROS code (Gelbard and Seinfeld, 1980) was used in which
 the particle size domain [0.001 to 20.0 )lm particle diameter (Dp)] was
 divided into 13 geometrically equal sections or bins. Coagulation was the
 only  mechanism considered; all other mechanisms (condensation,
 nucleation, deposition, etc.) were disabled. At time zero, an initial mass of
           was assigned to section 2 (0.0021 - 0.0046 \im diameter) to
                                                       simulate the nucleation of a submicron fume. This aerosol was assumed to
                                                       have the properties of Pb-oxide, producing a number concentration of
                                                       4.0x10" /mj.  An additional 500 rng/m' of mass was assigned to section
                                                       11 (2.0-4.4 urn diameter) to simulate the presence of flyash or sorbent,
                                                       System pressure and temperature were maintained at l.OlxlO5 Pa (1 aun)
                                                       and 810 K (1000 °F) to simulate post-flame conditions. Following the
                                                       initial distribution. Figure 2 presents six mass distributions which follow
                                                       the evolving aerosol through six orders in lime (t=0.1, 1.0, 10, 100, 1000,
                                                       10000 s).  Note that coagulation does not change the total aerosol mass and
                                                       that the areas  under all seven curves represent  HX50 mg/mA  Number
                                                       concentrations, however, are greatly affected, and Figure 2 shows that at
                                                       0.1 s the number concentration has fallen approximately  2 orders of
                                                       magnitude (2.3x10" /m>), and the average nuclei size has grown to
                                                       approximately 0.03 fim. At 1.0 and 10.0 s the distributions have grown
                                                       only slightly fanner into the 0.01 to 1.0 fim diameter range.  This is
                                                       important, as these times represent a range of typical residence times within
                                                       combustion/incineration systems.  In fact, even after 10000 s (2.S hours)
                                                       the average Dp is only approximately 1.0 pm with a number concentration
                                                       of 3.3x10" /m}.  Thus, as can be seen from Figure 2, aerosol nuclei tend
                                                       to coagulate very  quickly  at small  times, due  to the large  number
                                                       concentrations of nuclei particles, and then, at  larger limes, as number
                                                       concentrations fall, coagulation slows considerably causing the aerosol to
                                                       accumulate into a mode approximately between 0.1 and 1.5 jlm diameter.
                                                       Panicles in this size range, posessing neither high diffusivities nor high
                                                       momentum, exhibit minimum collection efficiencies in most air pollution
                                                       control devices. This characteristic distribution of a coagulating aerosol has
                                                       been termed the accumulation mode. Note that the coagulation mechanism
                                                       does not include the effect of differing fractal properties of the agglomerate
                                                       formed, as developed by Matsoukas and Fricdlander (1991).  It has been
                                                       assumed here that only spheres result from the coagulation process.
                                                          Also evident from Figure 2-is that coagulation between the evolving
                                                       nuclei and the sorbent is very slow, and the small particles grow as if the
                                                       sorbent were not present. Again, this is due in part to the large differences
                                                       in number concentrations. At time zero, nuclei are present in concentrations
                                                       of approximately 4.0xl018 /m'.   Sorbent  particles  are  present  in
                                                       concentrations of 4.7x109 /m'. This difference encourages nuclei-nuclei
                                                       coagulation.  Thus, it would seem that the use of a sorbent to scavenge
                                                       submicron metal particles through coagulation is not possible in the times
                                                       available. However, as has been described previously, mechanisms other
                                                       than coagulation may allow sorbents to be utilized to remove aerosol mass
                                                       from the submicron fraction.  Again, these results assume a fractal
                                                       dimension of 3 (i.e., spheres) for the resulting agglomerates.  Even though
                                                       the simulation presented in Figure 2  shows very little nuelei-sorbent
                                                       interaction, fine particle coagulation and diffusion to the coarse mode are
                                                       competitive processes. Friedlander el at, (1991) reason that, since diffusion
                                                       and coagulation are first and second order with respect to particle number
                                                       concentration, respectively, one should expect the following qualitative
                                                       behavior. If, as presented in Figure 2, particle nucleation results in a large
                                                       initial number concentration, then coaRulalion will dominate, causing the

-------
      600
      400-

      200-
mitial distribulion
t = 0s
number/m3 = 4.0x1018
                     t = 0.1 s
                     number/m3 = 2.3x1016
                     t= 1.0s
                     number/m3= 1.5x1015
                     t=10s
                     number/m3? 1.2x1014
                     t=100s
                     number/m3 = 1.2x1013
                     I= 1000 s
                     number/m3 = 1.7x1012
                     t = 10000 S
                     number/m3 = 3.3x1011
0.001     0.01      0,1        1
                                                           100
                                Dp (urn)
FIGURE 2  Predicted evolution of Pb-oxidc nuclei panicles in the presence
    of a simulated sorhenl via coagulation in a post-flame comhusiion
                           envifonmenl.

nuclei lo grow inlo panicles with low dift'usivities and little possibility to be
scavenged by the coarse mode panicles. If. on the other hand, nucleation
results in a small initial  nuclei  number concentration,  nuclei-nuclei
coagulation rales will be small, the nuclei mode will not grow substantially
but, provided sufficient lesidence time is available, will diffuse to, and be
scavenged by, the coarse mode particles.  Thus, using ihe self preserving
size distribution theory, Friedlander et al. (1991) argue that significant
scavenging of nuclei by coarse mode (sorbent) particles is possible if: the
diffusion time (see Hagan  and Friedlander,  (1978); Friedlander et al,
(1991)3  *s small compared to the residence lime; and the mass of the line
mode is  below a threshold value for a given coarse mode aerosol.
EXPERIMENTAL APPROACH AND PROCEDURE
Laboratory Swirl Flame Combuxtor

Experiments were performed using the laboratory scale 59 kW (actual), 82
kW (maximum rated) horizontal tunnel comhustor presented in Figure 3.
This rttfraciory-lined research cornbustor was designed to simulate the
time/lernperature and mixing characteristics of practical industrial liquid and
gas combustion systems. Natural gas fuel, aqueous metal  solutions, gas
dopants, and combustion air were  introduced into the burner section
through an International Flame Research Foundation (1FRF) moveable-
block variable-air swirl burner. This burner incorporates an interchangeable
injector positioned along its center axis.  Swirling air passes  through the
annuliis around ihe fuel injector promoting flame stability and attachment to
the water-cooled quarl.  A high swirl (IFRF type 2) flame with internal
                                                             recirculaiion (Swirl No.=1.48) was used. Gaseous and aerosol samples
                                                             were taken from a stack location 5,9 m from the burner quarl.  The
                                                             temperature at this location was approximately 670 K (745 °F).  Further
                                                             details regarding die experimental combuslor can he found elsewhere (Unak
                                                             eta!., 1994, 1995, 1996, 1997).
                                                 Metal Syslenis Investigated

                                                 To dale, experimental investigations examining four metals (Cd, Cr, Pb,
                                                 and Ni) have been conducted and published (Linak et al.,  1994,  1995,
                                                 1996, 1997). Additional studies focusing on Hg, vanadium (V), and zinc
                                                 (Zn) are on-going.  Typically, water soluble metal nitrates or oxides were
                                                 introduced as aqueous solutions through a special fuel/waste injector which
                                                 incorporated a small air atomizing system down the center of a standard
                                                 natural gas injector. The resulting droplet size distribution was relatively
                                                 narrow wiih z mean droplet diameter of approximately 50-80 Jim (Linak et
                                                 al., 1994). Diatomic chlorine (C\i'l or sulfur dioxide (SOj) dopants were
                                                 introduced, separately  from the metal solutions, with the (secondary)
                                                 combustion air. Thus,  the metal and chlorine or sulfur, were not mixed
                                                 prior to their introduction into the cornbustor. All interactions between the
                                                 components were dependent upon normal turbulent mixing patterns.
                                                   Aqueous solutions typically containing  1.5% metal (by weight) were
                                                 used.  Solution flow rates were maintained so as to produce stack gas
                                                 concentrations of approximately 100 ppm metal (by volume).  Metal feed
                                                 rales corresponded to constant molar feed rates of approximately 0.005 g-
                                                 moles/min.  Experimental programs included evaluating Ihe effects of
                                                 chlorine and sulfur on metal partitioning.  For these experiments, Cl| or
                                                 SOj were introduced at different  molar ratios of chlorine or sulfur to metal.
                                                 These feed rates and resulting  stack concentrations varied from
                                                 substoichiometric with respect to  the metal  concentration lo excess
                                                 concentrations approaching 10,000 ppm (stack). Typically, chlorine and
                                                 sulfur slack concentrations of 10(X) ppm were examined.  Excess air was
                                                 maintained at 20%. No air preheat was employed.
                                                 Aerosol Particle Size Distribution: Sampling ami Analysis

                                                 Particle size distribution (PSD) measurements were taken from the slack
                                                 localion using three  techniques.  Extractive samples were taken  for
                                                 collection by inertia! impaction and electrical mobility analyses using an
                                                 isokinetic aerosol sampling system based on the modified designs of Scotio
                                                 etal. (1992) and Linak aaL (1994). In order to minimize in-probe gas and
                                                 aerosol kinelics.  the sampling system dilutes and cools the aerosol sample
                                                 using filtered nitrogen and air immediately after sampling. Dilution ratios
                                                 arc measured directly for each experiment and verified independently by the
                                                 measurement of a nitric oxide tracer gas.
                                                    Extracted samples were directed 10 an Andersen Inc. eighi-stage, 28.3
                                                 L/min. (I W/min.), atmospheric pressure cascade impaetor and a Thermo-
                                                 Systems Inc. scanning mobility particle sizer (SMPS). The cascade
                                                 impaclor is designed to collect physical samples less than approximately 10
                                                 Jim diameter (for subsequent gravimetric and/or chemical analysis) on nine
                                                 stages (including the after filler). The SMPS classifies and counts particles
                                                 within a working range of 0.01 lo 1.0  pm  diameter using principles of
                                                 charged  particle mobility through an electric field.  The SMPS, used for
                                                 later experiments, is an upgraded version of the differential mobility particle
                                                 sizer (DMPS), used during early experiments. The SMPS upgrade allows
                                                 for improved PSD resolution and shorter sampling limes.  The SMPS and
                                                 DMPS were configured to yield 54 and 27 channels, respectively, evenly
                                                 spaced (logarithmically) over the O.U! to  1.0 (am diameter range.
                                                    In addition to the inertia! impaction and electrical mobility devices which
                                                 require an extracted sample, in-situ  light scattering PSDs were taken using
                                                 an Insitec  Inc. laser doppler voloeimeter.  This instrument determines
                                                 particle size by measuring the light scatiering intensity of particles which
                                                 pass through a sampling volume- established within the cornbustor stack by
                                                 a laser focused through a set of quartz optical access ports.  The working
                                                 range of this device was approximately  0.3 lt> 30 jim diameter which
                                                 slightly overlapped and extended tins PSD data collected by the SMPS.
                                                    In addition to the three PSD instruments, samples were collected on
                                                 silver fillers  and analyzed using  a field  emission  scanning  electron
                                                 microscope (SEM) equipped with an  energy dispersive x-ray (EDX)
                                                 spectrometer.  This  provided  morphological .information as  well as
                                                 qualitative chemical analysis of individual particles.
                                                  NON-VOLATILE METALS

                                                  Volatile and semi-volatile metals have been of particular interest because
                                                  they remain as vapors within combustion and Hue gas cleaning systems,
                                                  resulting in poor removal efficiencies, or tend to form fumes of submicron
                                                  particles resulting from nuclealion and condensation of metal vapor. These
                                                  mechanisms lead to substantial enrichment of these metals on submicron
                                                  particles which are often difficult to collect in paniculate pollution control
                                                  equipment. It has been shown (Scotto et til., 1992; Wendt, 1994; Linak et
                                                  al., 1995) that some meial vapors can also be reacted at high temperatures
                                                  above their dewpoints with sorbent substrates to form environmentally
                                                  benign water  insoluble products.  Thus, the emissions of some of these
                                                  volatile and semi-volatile metals can be managed because they form a vapor
                                                  which can be scavenged and reacted.   However, several researchers

-------
                          J To air pollution control system

                                     Gas sample port
                                      In-situ light scattering particle sizer

                                    -Aerosol sample port
                                      (electrical mobility & impaction particle sizers)
                                   - Cooling water
                                                                                             Moveable-block
                                                                                             burner
                                                                                                  Natural gas
                                                                                               •  ^    Aqueous metal
                                                                                                   I   solutions

                                                                                                   I—Atomizing air

                                                                                               	CI2 or S02

                                                                                               	Combustion air
                                                                                  Transport air
                                                 FIGURE 3 EPA horizontal lunnd combustor.
(Davisonera/.. 1974; Klein et aL, 1975; Markowski etal.. 1980; Quann
and Sarofim, 1982) have also noted enrichment of the refractory metals Ni
and Cr in submicron  panicles, even though it is not obvious  that these
metals necessarily vaporize in the combustion process. The mechanisms by
which this occurs are not readily apparent.  It is appropriate, therefore, to
examine refractory metals, and to determine if their partitioning among
various particle sizes can be predicted and controlled.
   Among possible refractory metals we focused on Ni and Cr.  Ni is only
slightly less volatile than Cr, and appears in both oil and coal tlyash.  Ni
and Cr (metal) have boiling points  of 3003 and 2945 K. respectively.  Cr is
of particular interest  because it commonly exists  in two forms in  the
environment (Goyer,  1991; Seigneur and Constantinou, 1995), as either
trivalent Cr(Ill) or hexavalent Cr(Vl), as in the chromate artion CrOv2(Vl)
or the compound CrOj(VI).  CrfVT) has the lowest risk specific dose for all
carcinogenic metals (0.00083  ng/m') (Fed. Reg., 1991). while Cr(lII) is
not considered particularly hazardous.  However, from a  regulatory
viewpoint, all Cr must be  considered to be Cr(Vl) unless difficult silo
specific speciation is performed. Therefore, it is important to determine  not
only how Cr is chemically partitioned among valency states in the exhaust.
but also on how this can be manipulated through the addition of sulfur and
chlorine compounds, and how the physical partitioning of Cr among
various particle sizes occurs in a practical combustion configuration.
Equilibrium Predictions

Multicomponent equilibrium calculations can provide insight into which
species are thermodynamically stable at flame and Hue gas temperatures.
However, the accuracy of equilibrium results depends on the accuracy of
the ihermodynamic data available, and on the availability of thermodynamic
data for all  important species containing the elements in question.  In
addition, equilibrium calculations do not take into account kinetic or mixing
limitations and, therefore, represent an idealized solution that may not be
realized in practical systems. Thermochemical predictions were determined
using  the CET89  computer code  for  calculating complex  chemical
equilibrium  compositions (Gordon and McBride, 1986). Twenty-six Ni
species and physical states and 48 Cr  species and physical slates  are
considered in these calculations (see Linak etal. 1996, 1997). Of interest is
the thermodynamic partitioning between vapor and condensed phases, as
well as the partitioning between various species. Also, of special interest
for Cr is the partitioning between Cr(Vl) and other Cr valent species. For
both Ni and Cr, the influences of chlorine, sulfur, and  both chlorine and
sulfur were investigated.
   Equilibrium predictions (Figure 4a)   indicate that both metals  are
refractory, with dewpoinls of 2000 and 1900 K for 100 ppm Ni and Cr,
respectively. Sulfur addition (Figure 4b) has little effect on either metals'
dewpoint.   Chlorine  addition (Figure 4c) lowers  the Ni dewpoint by
approximately 200 K and moves the Ni curve from the right to the left of the
Cr curve, which shews no significant effect of chlorine. The tact that
chlorine is more likely to devolatilize Ni than Cr, at high temperatures,
suggests that, in the presence of chlorine, high temperature sorbents might
be able to capture Ni, but are less likely to capture Cr.
   At low temperatures, Ni and chlorine are predicted to form a condensed
Ni-chloride salt, thus rendering the residue  water soluble. However,
chlcrine's predicted effect on Cr at low temperatures is profound.  Not only
is Cr predicted to form vapor-phase Cr-chlorides, but it is also predicted to
form the hexavalent Cr-chlorides, CrOCl4(VI) and CiCl6(Vl), as shown on
Figure 4c. The presence of sulfur (Figure 4d) completely eliminates the
chlorine enhanced formation of low temperature Cr(Vl) species, but has
           (d) Cl = 2500 ppm. S = 2500 ppm
      0.0
        300     600     900     1200     1500
                           Temperature (K)
1800    2100
  RGURE 4  Ni and Cr equilibrium predictions for four conditions: (a) 0
ppm chlorine, 0 ppm sulfur; (b) 0 ppm chlorine, 2500 ppm sulfur; (c) 2500
 ppm chlorine, 0 ppm sulfur; and (d) 25(K) ppm chlorine, 2500 ppm sulfur.

-------
little effect on the high temperature Cr(VI) species (not shown on Figure 4,
see Linak eiaL, 1996).  This is because sulfur lies up Cr to form trivalent
Cr-sulfate, but only at low temperatures.
   The effect of sulfur on the Ni/chlorine mixture is predictable.  It has no
effect at high temperatures on  the devolatilization of Ni, sinee Ni-sulfales
arc unstable there, while its effect at low temperatures is merely to replace a
solid Ni-chloride by a solid Ni-sulfale. Additional calculations exploring
the equilibrium effects of calcium addition showed that calcium, even at
concentrations in excess wilh respect to sulfur, displaced neither the Ni- nor
the Cr-sulfate and. therefore, had no  appreciable effect  on Ni or Cr
partitioning.
Particle Sue Distributions without Chlorine

Figure 5 presents the PSDs for Ni [injected as Ni(NOj)j], for Cr(IlI)
[injected as Cr(N03)j(III)], and forCr(Vl) [injected as CrOj(VI)]. These
data (open symbols) were obtained using the SMPS for particles in the 0.01
to 1.0 urn diameter size  range and the in-siiu light scattering panicle sizer
for particles the 0.3 to >10 |am diameter size range. Each panel also shows
the effect of chlorine  addition (solid symbols), where the chlorine was
added as  Clj  gas,  with the secondary  combustion  air.   While
complementary impactor samples were taken, those data are not presented
here but may be found elsewhere (Linak et al., 1997).
   It should be noted that all three methods of panicle collection and sizing
produced consistent results that supported each other.  Data from the in-situ
light scattering panicle sizer slightly overlapped and extended the range of
the SMPS for the sampled panicles.  This suggests that the isokinelic
dilution sampling procedure used maintained aerosol size integrity.
   The Ni volume PSD without chlorine (Figure 5a) showed a maximum at
about 0.3 to 0.4 Jim diameter. ForCr(lII) without chlorine, a single panicle
size mode peaking between 1.0 and 10 |am diameter is shown by both the
in-siiu light scattering particle sizer data (Figure 5h) and the impactor data
(data  not shown, see Linak et aL, 1997).  For Cr(Vl), two mode;, are
apparent from the SMPS  and the in-situ light scattering panicle sizer, wilh a
dominant mode peaking  at about 0.1 to 0.3 |Jm diameter (Figure 5c).  This
is verified by the impaetor results  (dala noi shown), which also show a
dominant mode at ahoul  0.2 lo 0.3 |Jm diameter. These results suggest that
CrOj(VI) vaporized, while Cr(NOj).i(IIl) did not. The difference in
volatilization behavior of the iwo Cr compounds is, of course, in conlrasl to
    10s.

    10V

    10".

    10*.

    10

    10°.

    10-'
                                                           (a)
                                    o  Ni(N03)2
                                    •  Ni(NO3)2w/CI
                                    •  Ni(NO3)2 w/CI and kaolinite
                                    +  kaolinite only
ng  10s

£  10'
I lOV

O  10*.
 en
i  10'.
1  10°

    10-'
                                                           (b)
                                              Cr(N03)3(lll)
                                              Cr(NO3)3(lll) w/CI
    106
    10s.

    10V

    10'.

    10*.

    10'.

    10°.

    10-'.
                                                           (c)
       0.01
                      0.1
                                                  10
                                                                100
                                 Dp (urn)
  FIGURE 5 Particle volume distributions measured by electrical mobility
 and light scattering for. (a) Ni(NOj)2; (b) CrCNOjbqil); and (c) CrO,(Vl)
      aqueous solution feeds wilh and withoul kaolinite and chlorine.
equilibrium predictions, which aie independent of the inilial Cr speciation.
Effect nf Chlorine

Chlorine has a significant effect on the Ni number and volume PSDs. The
maximum number conccntralion now occurs al 0.03 nm diameler (data not
shown), while the maximum volume concentralion has shifted to particle
diameters less lhan 0.1 pm (Figure 5a).  These PSDs are consislenl with a
nucleation/vaporization mechanism for Ni in ihe presence of chlorine. The
Cr results, by conlrast. show no effect on the stack PSD by chlorine, as
illustrated by comparison of open and solid symbols on Figures 5b and 5c.
In the case of Cr(NOj).i(lH), the chlorine did not facilitate vaporization,
while in the  case of CrOj(VI), 'which  vaporized without chlorine, no
difference in PSD was  noted.  The Ni and  Cr results arc qualitatively
consistent with  the equilibrium predictions of Figure 4, which show the
effecl of chlorine to he that of moving the Ni dewpoint from above thai of
Cr lo below lhat of Cr. According  lo both iheory and experiment, chlorine
facilities Ni volatilization at high temperalures, hul has little effect on Cr.
Note, however, lhat  results are inconsistent with the low  lemperature
equilibrium  predictions  of Cr with chlorine, since the Cr was found to
condense at low lemperature (i.e., the predicted equilibrium yield of almost
10073(VI) underwent coagulation over  a longer period of time than
did Ni with chlorine, presumably because  it had a higher effeclive

-------
dewpoint.  Hydrated liifie also had negligible effect on  ihc PSDs from
Cr(NOj)3(lII) and CrOj(VI), and so it can he concluded that lime is an
ineffective sorben; for all forms of Cr tested, whether vaporization occurred
or not.
Cr Spec/alien

In addition to factors influencing PSDs, chemical speciation is also of
paramount importance, especially as far as Cr is concerned.  Equilibrium
predictions of Cr speeiaiion (sec Linak ti til., 1996) suggest that, in the
absence of chlorine, the fraction of Cr(VI) is small and appears only at the
higher temperatures. When chlorine is added, two additional Cr(VI) species
are predicted to appear at lower temperatures [CrOCU(Vl) and CrCloO/I)].
EDX analysts of collected particles shows  that chlorine is found only on the
fused spherical particles, and not on the angular crystalline panicles.  One
might speculate that the fused particle may contain Cr compounds including
CrOCUCVl) and/or CrClc(Vl). However, since the equilibrium calculations
at low temperatures do not predict the existence of a solid species containing
Cr, one might conclude that kinetic limilalions prevent significant Cr(VI)
formation. Whatever the formation route to chlorinated Cr in the exhaust,
sulfur is predicted by equilibrium (Figure 4) to eliminate those species and,
thus, (potentially) eliminate one source of Cr(VI).
  Figure 6 depicts the overall partitioning between Cr(VI)  and total Cr
(Linak el al, 1996).  !n the upper panel (Figure 6a), Cr partitioning resulting
from the introduction of Cr(NO->}j(IH) is presented.  With neither chlorine
nor sulfur present, approximately 2% of the total Cr in the slack gas effluent
is hexavalcnt.  Addition of 1000 ppm (low) chlorine (stack) increased the
percent  Cr(VI) in the exhaust slightly to 2.5%. Addition of 6,700 ppm
(high) chlorine (stack) raised  the Cr(Vl) percentage in  the exhaust to
approximately 8%.  The addition of sulfur (no chlorine present) sharply
diminished the emission of Cr(Vl).  In fact, with a high concentration of
sulfur (7,900 ppm, slack), the Cr(VI) percentage was reduced to near
detection levels.
  The trends exhibited  in Figure 6a, are  consistent wiih  equilibrium
predictions, although the absolute values are not.  In the absence of both
chlorine and sulfur, some Cr(Vl) which  is stable at higher temperatures
appears to persist through to the lower temperature regime, even though
equilibrium  would  not predict its  presence there.   High chlorine
concentrations sharply enhanced Cv(Vl) emissions, possibly due to CrCI^
which was predicted to be stable at low temperatures. The effect of sulfur is
consistent with  equilibrium if it is  assumed that conversion of Cr to
Cr2(S04)j is rapid.  It is interesting that equilibrium predicts that even
relatively small quantities of sulfur can counteract the Cr(Vl) formation
tendencies of chlorine. This has been verified experimentally, where even
stoichiometric quantities of sulfur were able to prevent the formation of
measurable quantities of Cr(VI) (Linak rial.,  1997).  Both theory and
experiment suggest that sulfur is effective in eliminating Cr(VI) because it
displaces chlorine from the Cr(Vl) compounds  otherwise  formed without
sulfur. Both theory and experiment suggest lhal only a very small amount
of sulfur, determined  hy the Cr/sulfur stoichiomeiry. is sufficient to
suppress chlorinated Cr(Vl) compounds.
  In  the lower panel (6h). analogous results arc presented for CrOi(VI)
feed.  It is significant that Uiey are very similar to the results presented in the
top panel (6a) with Cr(ll() feed. The partitioning of Cr in a combustor thus
seems to tie independent of the initial valence of Cr waste feed.  Most of the
CrCVI) that entered was converted to Cr(III),  The similarity in partitioning
between the upper and lower panels of Figure 6, and the fact that the final
Cr(VI)/lotal Cr partitioning is independent of initial speciation, suggests that
prior to sampling,  some type of equilibrium controlled mechanism, with
neither  kinetic nor mixing limitations, is operable.  However, the low
conversion to Cr(Vl) in the presence of chlorine, and the finite conversion
in the absence of chlorine and sulfur, suggest that this  equilibrium is
"frozen" at a temperature higher than the sampling or exhaust temperature.
SEMI-VOLATILE METALS

While non-volatile metals  may not readily vaporize during combustion
processes and are often reported to contribute preferentially to bottom ash
and collectable portions of the PSDs, semi-volatile metals are often easily
vaporized at combustion temperatures.  These vapor-phase species  will
become supersaturated and subsequently undergo homogeneous nucleation
or heterogeneous condensation at the lower temperatures downstream to
contribute to the submicron PSD. These particles, because of their small
size, are difficult to collect in pollution control systems.  However, their
propensity to vaporize also  allows the possibility of control through
interactions with injected sorbents.  Therefore, this research investigated
sorbent injection processes in which the high temperatures of practical
combustion environments might be exploited to transform semi-volatile
metals into constituents that  are both, more easily collected, and more
environmentally benign, than metal effluents in the absence of combustion
modifications.   Semi-volatile metals of interest were Cd  and Pb,  and
kaolinile, bauxite, and hydrated lime  were chosen  as  representative
sorbents.
   Using a downflow laboratory combustor, Scolto et al. (1992) found that
Pb could be reactively scavenged, in-situ, by kaolinite powder which  was
injected into ihe post flame.  Reactive scavenging (chemisorption) of a metal
occurs at temperatures above the metal vapor dewpoint. This group  also



"O
0)
3
CO

-------
0.8.
0.6.
0.4.
0.2'.
ed species
p ^ p
OS O O
c. 0.6.
m
|0.4.
« 0.2.
£ 0.0.
•D 1.0
•5 °-8-
H 0.6
o
§ 0.4
| 0.2
I 1-°'
0.8
0.6
0.4
0.2.
0.0
\\ (a) Cl = 0 ppm, S = 0 ppm
\ \
Pb\ \Cd
i *
\ \
\ \
,
"V (b) Cl = 0 ppm, S = 2500 ppm
V"

\

\ \ (c) Cl = 2500 ppm, S = 0 ppm
\ \
\ Pb ', Cd
\ \
\ \

\ \(d) Cl = 2500 ppm, S = 2500 ppm
\ \
\ \
Pbl 'tCd
\
\ \
	 , 	 _, 	 , 	 . 	 , 	 . 	 , 	 J 	 , 	 > 	 , 	 . 	 . 	 , 	 , 	 , 	 , 	 T 	
       300      600     900      1200    1500
                           Temperature (K)
                                                  1800
                                                          2100
  FIGURE 7 Cd and Pb equilibrium predictions for four conditions: (a) 0
ppm chlorine, 0 ppm sulfur; (b) 0 ppm chlorine, 2500 ppm sulfur; (c) 2500
ppm chlorine, 0 ppm sulfur; and (d) 2500 ppm chlorine, 2500 ppm sulfur;

dewpoinL For Cd, CdO(s} is formed at 1300 K and CdSO«(s) is preferred
over CdCl2(s) forming at 1000 K.
Pb/Kaolinite/Chlorine System

In contrast to Ni (Figure 5a), the Pb data (Figure 8a), without chlorine,
indicate the presence of a distinct submicron mode with a mean particle
diameter between O.I and 0.2 urn.  This behavior is consistent with Pb
vaporization followed by subsequent aerosol formation and growth and is
consistent with the known volatilities of elemental Pb and Pb-oxide.  With
chlorine added, this mode is shifted towards even smaller panicle sizes
(between 0.03 and 0.1 nm) possibly indicating delayed  nucleation and a
less mature aerosol  at the sampling location.  There is  also evidence of
bimodal behavior in the presence of chlorine which may  indicate the
formation of at least two Pb species with different nucleation characteristics.
The impactor data (Figure 8b) indicate that between 80 and 82% of the
measured Pb is associated with particles smaller than 1.1 (im.
   With the addition of kaolinite, boih the DMPS and impactor data indicate
substantial reductions in the submicron aerosol volume and Pb mass
fraction (72 and 98%, respectively) compared to the  corresponding PSDs
without chlorine.  Similar reductions are  also  evident comparing the
distributions with chlorine (49 and 86%).  Morphological observations
indicated that  much of the kaolinite melted, both with and without Pb
present.  These results are consistent with those  of Scotto et a!.  (.1992),
where high uptakes of Pb on  sorbent particles were  associated  with
formation of melts on sorbent surfaces.
Cd Interactions with Sorbenfs

Cd baseline and Cd/chlorine data (without kaolinite) are similar to
corresponding Pb data presented above. Elemental Cd, CdO, and CdCb
vapor pressures are similar to those for elemental Pb, PbO, and PbClj, all
of which are notably high at the peak temperatures seen in the combustor.
As with the Pb system, the Cd behavior is indicative of particle formation
via a vaporization mechanism. The impactor data (Figure 9a) show that 88
and 85% of the Cd mass are associated with particles less than I.I u.m for
the Cd baseline and Cd/chlorine experiments, respectively. Also consistent
with the Pb data, the Cd data show that the addition of kaolinite causes
substantial decreases in both the DMPS submicron volume concentration
(61%-data not shown) and the <1.1 ujn impactor Cd mass fraction (97%-
Figure 9a).
                                                                          These results differ from those of Uberoi and Shadman (1991) in two
                                                                        important respects: (1) the amount of Cd removed here (97%), in a time-
                                                                        scale of seconds, is far higher than the 5% removed by kaolinite in  their
                                                                        moderate temperature bench-scale studies, and (2), the sorbent particles that
                                                                        removed Cd here were melted, with no observable (by XRD) Cd-rclated
                                                                        crystalline structure, while in  the  bench-scale studies  they remained
                                                                        crystalline.  The melt appeared  to avoid limitations of pore blockage by
                                                                        reaction products, as identified by Uberoi and Shadman (1991).
                                                                          With the addition of bauxite, the DMPS distributions (Cd baseline and
                                                                        Cd/bauxite) illustrate significant removal of particles <0.2 |lm (distributions
                                                                        not shown). The impactor data (Figure 9b) indicate that  97% of the Cd
                                                                        originally associated with particles < 1.1 (im in diameter was removed  from
                                                                        that particle size range through the addition of bauxite. These results are in
                                                                        agreement with those of Uberoi and Shadman (1991) which  suggest bauxite
                                                                        to be an exceptional sorbent for use  with Cd.  Furthermore, in both this
                                                                        combustor study and in the  previous bench-scale studies, the  sorbent
                                                                        panicles remained unmelted and crystalline. Therefore, sorbents that do not
                                                                        melt can also be effective in reactively scavenging vapor-phase metals, if
                                                                        pore blockage is not a factor.
                                                                          As with the two other sorbents, hydrated lime acts as an effective agent
                                                                        to scavenge Cd which would otherwise contribute to the submicron aerosol
                                                                        fraction (Figure 9c). However, these results are in contrast to those of
                                                                        Uberoi and Shadman (1991), and one would not expect reactive scavenging
                                                                        to occur. It is interesting to note that hydrated lime seems to be particularly
                                                                        effective even in the presence of chlorine.  The 99% reductions in  both
                                                                        submicron volume (DMPS data not shown)  and Cd submicron mass
                                                                        fraction (impactor)  with chlorine present (Figure 9c), represent the greatest
                                                                        measured removals seen for any  chlorinated system examined here. In the
                                                                        absence of Cd, the lime sorbent particles were crystalline, angular, and had
                                                                        not melted. With  the addition of Cd, the calcium-rich sorbent particles
                                                                        melted. Calcium oxide, which is basic, is known to enhance formation of
                                                                        eutectic melts with acidic metal oxides.
                                                                       Mechanisms and Conclusions

                                                                       There appear to be  two high temperature mechanisms  that allow the
                                                                       scavenging of metals at combustion temperatures above their dewpoints.
                                                                       The first mechanism  involves reaction between metal vapor and a sorbent
                                                                             100000
                                   0 w/o Cl or kaolinite
                                   D w/ Cl only
                                   » w/ kaolinite only
                                   " w/ Cl and kaolinite
   £    0.0
                            1              10
                               Dp (pirn)
                                                         100
FIGURE 8  DMPS submicron volume distribution and impactor Pb mass
       fraction distribution for the Pb/kaoiinUe/chlorinc system.

-------
                                ° w/o Cl or sorbent
                                   w/ Cl only
                                   w/sorben| only
                                   w/Clandsorbent
           0.1
                                                          100
 FIGURES Interactions hclwccn Cd and three sorbenls.  Impactor mass
 fraction distributions lor; (a) Cd/kaolinite; (h) Cd/bauxile; and (c) Cd/lttne.
                      with and without chlorine.
crystalline surface (Uheroi and Shadman, 1991). SEMs provided evidence
of this sorption mechanism for the Cd/hauxile system, which, by exhibiting
a I/Dp dependence of Cd mass fraction on the sorbent, also suggested pore
diffusion or external reaction controlled processes (Linak and Wendt,
1993). The second mechanism allows scavenging of meiul vapor by a
liquid melt on the sorbenl.  Melting appears to improve capture. Cd was
scavenged by kaolinile in the combustor used here, because the sorbenl
melted (as depicted in SEMs), bul it was only poorly scavenged at lower
temperatures (Uberoi  and Shadinan, 1991) where no melting was observed.
In the Cd/lime  system,  the melt was created by  the very interaction  of
dissolution. Melts were also observed alter the scavenging of labile Ni (in
the presence of chlorine).  Thus, metal capture by sorbenls  may be more
practical in high temperature combustion environments, where melting is
more  likely, than was initially  suggested by ihe moderate temperature
bench-scale thermogravimeirie reactor studies (Uheroi and  Shadinan. 199!),
1991).
  The effect of chlorine is to significantly increase the submicrort volume
concentrations and suhmicron metal mass fractions, in  the absence  of
sorbents, and, for the  Pb/kaolinite, Cd/hauxite, and Cd/kaolinite systems, to
diminish sorbent effectiveness when they are present.  An explanation for
this behavior is as follows: for Pb and Cd interaction with alumino-silicate
sorbents, the true metal  reactant with the substrate is  probably the metal
hydroxide or oxide. As chlorine is introduced, equilibrium is shifted away
from these reactive metal species towards unreactive metal chlorides. This
has been shown for the Na/kaolinite system (Mwahe, 1993). and suggested
for Pb (Scotto  etui.. 1992).  Ni, however, can be  vaporized only by
interaction with chlorine, and then (possibly through equilibrium with other
reactive Ni species) scavenged by kaolinite. The capture of Cd by hydrated
lime is through a different (physical) mechanism (suggested  by the  SEMs)
where the pertinent mechanisms involving chlorine are currently unknown.
VOLATILE METALS

Volatile metals including Hg 'jnd Se can be distinguished from semi-volatile
metals in combustion systems by the fact that they exhibit significant vapor
pressures even at low to moderate temperatures typical of flue gas cleaning
equipment. This typicaliy results in poor emission control of these species.
!n fact, of all the trace metals emitted during fossil fuel combustion and
waste incineration processes, Hg is likely considered the most problematic.
This is not because it is the most toxic  or  typically present in highest
concentrations, but rather, because current control  processes designed for
paniculate, nitrogen oxide (NOX), and acid gas emissions are minimally
effective in controlling vapor-phase Hg species.
   Hg exists in two valent states; elemental Hg (Hg°), and oxidized (ionic)
Hg (Hg*2). Effective Hg control using methods designed for paniculate
and acid gas emissions will depend largely on Uw form (vapor or condensed
phase) and specialion (elemental or oxidized) of the Hg in post-combustion
regions prior to air pollution control devices.  It has been noted by Senior et
al. (1997)  that oxidized Hg*2 js more likely to be captured by residual
carbon or  removed by existing  flue gas desulfurization  units,  while
elemental Hg° is more likely to escape the air pollution control devices and
be emitted (o  the atmosphere.  Several  investigations are in progress
examining  the potential of carbon and inorganic-based sorbenls for Hg
control at moderate to low temperatures typical of Hue gas cleaning systems
(Miller etal.. 1995; Krishnan «til, 1995, 1997).

Equilibrium Predictions

Figure 10 presents equilibrium  predictions of the behavior of Hg within a
combustion environment.  Sixteen Hg species and physical states were
considered.  These calculations were made in the  same manner as those
presented for the non-volatile and semi-volatile metals (Figures 4 and 7),
with the exception that the HE concentration (stack) used was 100 ppb
instead of KK3 ppm.  Meihane/air combustion at a stoichiomelric ratio of 1.2
was used. Chlorine  and sulfur concentrations (stack) were both 100 ppm.
The top two panels (Figures Kb and  Itth) plot condensed tnass fraction of
Hg and oxidized Hg*3 mass  fraction  as a function of temperature. Four
conditions (with and without chlorine and sulfur)  are presented on each
panel.  Figure  lOa indicates  a  Hg dewpoint (without chlorine or sulfur)
helween 4(KS and S!X) K through ihe condensation of HgO(s). The effcci of
sulfur is to increase this threshold temperature approximately 100 K, while
the effect of chlorine (even  with sulfur present)  is 10  decrease (he Hg
dewpoint through the formation of HgCljCs).  Note that,  in contrast to Cr,
HgSO4(s) is not preferred to HgCI;i(s).  Figure lOb indicates thai at high
temperatures (>9(X) K) almost all the Hg exists as Hg".  Without chlorine
or sulfur, HgQ  and then HgO(s) are predicted with decreasing temperature.
With sulfur (but without chlorine), HgO and then HgS04(s) are predicted.
However, HgO, HgClj, and then HgClj(s)  are predicted to be formed
whenever chlorine  is available  (<9()() K), and vapor-phase HgClj  is
predicted lo be present ax the dominant Hg species at temperatures between
900 and 400 K.  This may offer the possibility of Hg control through
interactions of oxidized HgClj with sorbents.
  It should be re-emphasized that equilibrium calculations can be used only
to determine which species are thermodynamically  possible.  They do not
include kinetic or mixing considerations which may'severely limit die
attainment of equilibrium in the short times available  in practical combustion
systems.  Senior et til. (1997) report that, in a survey of 14 coal combustion
systems, oxidized Hg*5 concentrations upstream of the air pollution control
devices (750-900 K) ranged between 3(1 and 95% of the total Hg measured
(averaging 75%).  They went on to conclude that equilibrium could not be
used to quantitatively predict Hg specialion in ihe flue gas. However, even
with this liability, equilibrium calculations are  useful lo test hypotheses and
suggest experiments.
  Figures  lOe and  Hid present the  results  of equilibrium calculations
designed to examine possible interactions between Hg and calcium that
might be present in  the ash or introduced as  a sorhent.  While searching
thermodynamic data bases for data on  Hg species, we discovered dala for a
set of ealctum-Hg amalgams (TAPP, 1995), and these "species" were
included in the data  base for the calculations presented in Figures lOc and
lOd.  Figure  lOc presents  predicted Hg speciation as a function  of
temperature for a fuel-lean (SR=1.2) scenario with calcium.  Sulfur was
also included because of its possible interactions with calcium. As expected
calcium oxides, hydroxides,  carbonates, and CaSO,t(s)  are the preferred
species over all  temperatures examined.   However, under fuel-rich
conditions,  such as those used 10 stage combustion  for NO, control, these
oxidized species  are not predicted,  and Figure lOd indicates that two
calcium-Hg amalgams are ihermodynamically stable within a temperature
window between 900 and 1300 K. This result may be significant because it
represents conditions that may be achieved in practical combustion systems.
Experiments are currently underway to examine these predictions.  Other
experiments are examining  Hg speciation  as  functions of combustion
conditions and chlorine and sulfur content, and factors which promote the
formation of Cl? over HCI which may influence concentrations of HgCI 3
(Gullett eiat.. 1990;  Senior « al.. 1997).
  Preliminary experiments (without chlorine or sulfur addition) indicate
that Hg speciates primarily as elemental Hg° (>95%) and that hydrated lime
(introduced as a sorbenl) is minimally effective in reducing the vapor-phase
oxidized Hg*2 under fuel-lean (SR=1.2)  conditions. However, similar

-------
       0.0
         300    600
                         900    1200    1500
                           Temperature (K)
                                                 1800    2100
 FIGURE 10 Hg equilibrium predictions: (a) condensed fraction with and
 without chlorine and sulfur; (h) oxidized fraction with and without chlorine
and sulfur; (c) Hg species with calcium and sulfur - fuel lean (SR=1.2); and
       (d) Hg species with calcium and sulfur - fuel rich (SR=0.6).
experiments performed under staged conditions (SRprim;uy=0,8) suggest
that hydrated lime might he an effective sorhent. indicating that vapor-phase
oxidized Hg*J was reduced to less than detection limits.  However, these
same experiments also indicated that the elemcnfa! Hg° remained >95% of
the total Hg measured and was unaffected hy the presence of hydrated lime.
Therefore, the locus of ongoing efforts includes strategies to promote the
formation of oxidized Hgc/mo/., submitted.
Markowski. O.R., Ensor, D.S., Hooper, R.G., and Carr. R.C.  (1980). A
   Submicron Aerosol  Mode in Flue Gas from a Pulverized Coal Utility
   Boiler. Environ. Sci. Technol, 14(11). 1400-1402.
Maisoukas, T.  and Friedlander,  S.K. (1991).  Dynamics of Aerosol
   Agglomerate Formation. /.  Colloid Interface. Set.. 14C(2), 495-506.
Miller, SJ., Laudal. D.L.. Dunham. G.E., Chang, R., and Bergman, P.O.
   "Pilot-Scale Investigation of Mercury Control in Baghouscs," EPRI/DOE
   Third International Conference on Managing Hazardous and Paniculate
   Air Pollutants. Toronto, Canada, August (1995).
Mulholland, J.A. and Sarofim, A.F, (1991). Mechanisms of Inorganic
   Particle Formation Dunne Suspension Heating of Simulated Aqueous
   Wastes. Environ. Sci. Techno!.. 25(2). 268-274.
Mulholland, J.A., Sarofim, A.F., and Yue, G. (1991), The Formation of
   Inorganic Particles During Suspension Healing of Simulated  Wastes.
   Environ. Prog.. 10(2), 83-88.
Mwahe, P.O. (1993). Mechanisms Governing Alkali Metal Capture by
   Kaolinite in a Down flow Combusior, Ph.D dissertation. University of
   Arizona, Tucson. AZ.
Nettleton, M.A. (1979). Paniculate Formation in Power Stations Boiler
   Furnaces. Prog. Energy Combust, Sci., 5, 223-243.
Neville, M. and Sarafim, A.F. (1982). The Stratified Composition of
   Inorganic Submicron Panicles Produced During Coal Combustion, 19th
   Comb. (Int.) Symp..  1441-1449. Comb.  Inst.. Pittsburgh, PA.
Quann, R.j. and Sarofim, A.F. (1982). Vaporization of refractory oxides
   during pulverised coal combustion.  19tli Comb, (Int.) Symp., 1429-
   1440, Comb. Inst., Pittsburgh, PA.

-------
Queneau P.ff..  May,  L.D., and  Cregar. D.E., "Applicaiioii of Slag
   Technology  to  Recycling  of Solid  Wastes."  199!  Incineration
   Conference. 69-85. Knoxville.TN. May 1991.
RCRA - Resource Conservation and Recovery Act (1986).  Subtitle C,
   Sections 3001-3013, 42 N.S.C.. Sections 6921-6934 (1976)  and
   Supplement IV (1980) amended (1986).
Rizeq, E.G.,  Hansel!. D.W., and Seeker, W.R. (1994).  Prediction of
   Metals Emissions and Partitioning in Coal-fired Combustion Systems.
   Fuel Processing  Techno!., 39, 219-236.
Scotto,  M.A., Peterson, T.W., and Wendt, J.O.L. (1992). Hazardous
   Waste Incineration:  The In-stiu  Capiure of Lead by Sorbents  in a
   Laboratory Down-How Combustor.  24ih Comb.  (Ins.) 5ym/;., 1109-
   1118, Comb.  Insl., Pittsburgh, PA.
Seigneur, C. and Constaminou, E. (1995). Chemical Kinetic Mechanism
   for Atmospheric  Chromium. Environ, Sci. Ttchnol., 29,222-231.
Senior, C.L.,  Bool, L.E., 111, Huffman, O.P., Huggins. F.E., Shah, R.
   Sarofirn, A., Olmez, I., and Zeng, T. "A Fundamenlal Study of Mercury
   Partitioning in Coal Fired Power Plant Flue Gas," AWMA 90th Annual
   Meeting, Toronto, Canada, June 1997.
Smith, R.D. (1980).  The Trace Element Chemistry of Coal During
   Combustion and  the Emissions from Coal Fired Plants. Pros- Energy
   Combust, Sci., 6, 53-119.
TAPP - Thermodynamie Properties Software, V2.2, ES Mieroware  Inc.,
   Hamilton, OH (1995).
Uberoi, M. and  Shadrnan, F. (1990). Sorbents for the Removal of Lead
   Compounds from Hot Flue Gases. AtChE Journal. 36(2), 307-309.
Uberoi, M. and Shadman, F.  (1991). High-temperature  Removal of
   Cadmium Compounds Using Sorbents. Environ. Sci. Trchnol., 25(7),
   1285-1289.
Wendt. J.O.L. (1994). Combustion Science for Incineration Technology.
   25A Comb, (Int.) Symp.. 277-289, Comb. Inst., Pittsburgh, PA.
                                                                10

-------
  NRMRL-RTP-P-250
           TECHNICAL REPORT DATA
     (Please read Instructions on the reverse before camp
I. REPORT NO.
PA/600/A-97/085
4, TITLE AND SUBTITLE
 Metal Partitioning in Combustion Processes
                                                       5. REPORT DATE
                                                       6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S>

 William P.  Linak
                                  8. PERFORMING ORGANIZATION BEPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
                                                       10, PROGRAM ELEMENT NO.
 See Block 12
                                  11. CONTRACT/GRANT NO.
                                  NA (Inhouse)
12. SPONSORING AGENCY NAME AND ADDRESS
 EPA, Office of Research and Development
 Air Pollution Prevention and Control Division
 Research Triangle Park, NC  27711
                                  13. TYPE OF REPORT AND PERIOD COVERED
                                  Published paper; 10/89-6/97
                                  14. SPONSORING AGENCY CODE
                                   EPA/600/13
15. SUPPLEMENTARY NOTES ^PPCD project officer is William P. Linak.  Mail Drop 65, 919/
 541-5792. Presented at 4th Int. Conf. on Technologies and Combustion for a Clean
 Environment. Lisbon.  Portugal.. 7/7-10/97.	
16. ABSTRACT ,
          The paper summarizes ongoing research efforts at the National Risk Man-
 agement Research Laboratory of the U. S.  Environmental Protection Agency examin-
 ing (high temperature) metal behavior within combustion environments. The parti-
 tioning of non-volatile (Cr and Ni), semi-volatile (Cd and Pb), and volatile (Hg) me-
 tals in combustion systems was investigated theoretically and experimentally.  Theo-
 retical predictions were based on chamical equilibrium and suggested that such cal-
 culations can be useful in predicting relative volatility and speciation trends,  and to
 direct experimental efforts. Equilibrium studies employing a 59 kW laboratory scale
 eombustor examined the behavior (volatility, particle size, and speciation)  of metal
 vapors and particles produced by aqueous metal solutions sprayed through a swirling
 natural gas diffusion flame. These experiments were designed to study metal trans-
 formation mechanisms in a relatively simple combustion environment without the
 complex effects of additional species. Further  experiments examined the potential
 use of common inorganic sorbents  (kaolinite, bauxite,  and hydrated lime) to adsorb
 metal vapor, offering a potential means of metal emissions control.
17.
                              KEY WORDS AND DOCUMENT ANALYSIS
                 DESCRIPTORS
                                           b.lDENTIFIERS/OPEN ENDED TERMS
                                               c. COSATI Field/Group
 Pollution
 Metals
 Emission
 Separation
 Combustion
 Volatility
Sorbents
Pollution Control
Stationary Sources
Trace Metals
Partitioning
13B    11G
11F, 07B
14G

21B
20M
18. DISTRIBUTION STATEMENT
 Release to Public
                                           19. SECURITY CLASS (This Report}
                                           Unclassified
                                                                     21. NO. OF PAGES
                      20, SECURITY CLASS (This page)
                      Unclassified
                                               22. PRICE
EPA Form 2220-1 (9-73)

-------