Sie sind auf Seite 1von 9

Ind. Eng. Chem. Res.

2004, 43, 1235-1243 1235

Evaluation of Zinc Oxide Sorbents in a Pilot-Scale Transport


Reactor: Sulfidation Kinetics and Reactor Modeling
Richard Turton,*,† David A. Berry,‡ Todd H. Gardner,‡ and Angela Miltz§
Department of Chemical Engineering, West Virginia University, Morgantown, West Virginia 26506,
Department of Energy, National Energy Technology Laboratory, Morgantown, West Virginia 26505, and
Parsons, Morgantown, West Virginia 26506

Under conditions of negligible external mass-transfer resistance, kinetic studies on the sulfidation
of zinc oxide were performed using thermogravimetric analysis (TGA) for three sorbent samples
comprising approximately 60, 80, and 100 wt % zinc oxide with the balance being inert binder.
The size range of particles for each sorbent was quite broad, with average diameters in the
range of 60-80 µm. Conversion-time profiles were obtained for each sample of sorbent over
the temperature range of 482-593 °C at 101.3 kPa pressure using between 0.5 and 2 vol % of
hydrogen sulfide in nitrogen. Initial reaction rate data for the three sorbent samples were
correlated on the basis of the surface area of zinc oxide available for reaction. For the 60 wt %
zinc oxide sorbent, an additional series of experiments was performed on different size fractions.
The conversion-time profiles for all size fractions were found to be essentially identical at 593
°C, confirming that intrapellet (between-grain) diffusion was not significant. The reaction of
hydrogen sulfide with the zinc oxide sorbent and the diffusion of gas through the zinc sulfide
product layer (within-grain diffusion) were modeled successfully using a grainy-pellet model
with a bimodal size distribution of grains. Results were also obtained for repeated sulfidation
runs in a 11.54-m-long, 0.84-cm-diameter transport reactor at a temperature of 538 °C and a
pressure of 2.05 MPa, using the 60 and 80 wt % sorbents. A simple plug-flow model was developed
to describe the sulfidation of the zinc oxide sorbents, and predictions from this model were found
to be in close agreement with experimental results.

Introduction work has been conducted in modeling and predicting


the conversion-time histories of the solids. The models
The noncatalytic gas-solid reaction between zinc used by the above researchers are also summarized in
oxide containing sorbents and hydrogen sulfide contain- Table 1. Ranade and Harrison13 noted that the modeling
ing gases has been researched extensively. Lew et al.1,2 effort has basically followed one of two tracks: either a
reported the initial reaction rates of pure zinc oxide and form of the shrinking-core model (SCM) (e.g., Yagi and
zinc titanate sorbents (90-125 µm) in the temperature Kunii14) or a form of the grainy-pellet model (GPM) (e.g.,
range of 400-700 °C. Westmoreland et al.3 measured Szekely et al.15). In addition, a pore-based approach has
the initial reaction rates of zinc oxide (<170 µm) in the also been used. In general, the basic forms of the SCM
temperature range of 300-800 °C. Woods et al.4 and and GPM have not been able to predict correctly the
Jothimurugesan et al.5 measured the reaction rates of observed conversion-time data for the zinc oxide sul-
single pellets (4760-µm) of zinc oxide and zinc titanates fidation reaction. Instead, variations of these models
between 565 and 785 °C. Efthimiadis and Sotirchos6 have been suggested to account for the observed behav-
measured conversion-time profiles for zinc oxide sor- ior in zinc oxide sulfidation and other gas-solid non-
bents of different sizes (53-350 µm). Garcia et al.7
catalytic reactions. The most important variations or
determined initial reaction rates for fine zinc titanate
extensions to these models consider the effect of conver-
particles (<1 µm) in the 550-600 °C temperature range.
sion on the physical properties of the sorbent due to the
Li et al.8 measured the reaction rates of zinc oxide-
difference in molecular volume of zinc oxide and zinc
manganese oxide pellets (3300-4000 µm) at between
sulfide (Georgakis et al.16); the nonhomogeneous physi-
200 and 400 °C. Mojtahedi and Abbasian9,10 measured
cal structure of the sorbent, including overlapping
the reaction of zinc titanate (230-380 µm) between 550
grains (Sotirchos and Yu17) and size distribution of
and 650 °C in a fluidized-bed apparatus. Finally, Kont-
grains (Bartlett et al.18); or the inclusion of a reaction
tinen et al.11,12 measured the conversion-time history
of zinc titanate particles (200-308 µm) at temperatures zone or band rather than a sharp interface (Ishida and
in the range of 400-600 °C. These results are sum- Wen19 and Mantri et al.20).
marized in Table 1. The pore-based approach to modeling was originally
In addition to the measurement of initial kinetics and proposed by Ramachandran and Smith.21 More recently,
the conversion-time history of the solid sorbents, much a random-pore model (Yu and Sotirchos22) and a gen-
eralized-pore model (Efthimiadis and Sotirchos6) have
been proposed. The latter was successfully used to
* To whom correspondence should be addressed. Fax: 304- predict the conversion-time histories of sorbents of
293-4139. E-mail: riturton@mail.wvu.edu.

West Virginia University. different sizes.

National Energy Technology Laboratory. In the present work, kinetic expressions and phenom-
§ Parsons. enological reaction models that are capable of predicting
10.1021/ie030364a CCC: $27.50 © 2004 American Chemical Society
Published on Web 02/03/2004
1236 Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004

Table 1. Comparison of Particle Sizes and Models Used by Previous Researchers to Describe Conversion vs Time Data
sorbent particle BET SA, pore diameter, temperature experimental reaction
reference material sizes (µm) A (m2/g) 4V/A (nm) range (°C) setup model
Lew et al.1,2 ZnO 90-125 2.4 2000 400-700 TGA overlapping
ZnxTiyOx+2y 90-125 23.0-13.9 325-820 grain
Westmoreland et al.3 ZnO <170 4.9 - 300-800 TGA -
-
Woods et al.4,5 ZnO 4760a 3.9 130 565-785 TGA shrinking
ZnxTiyOx+2y 3.0-3.2 354-463 core
Efthimiadis ZnO 53-62, 27.7 300-600 TGA generalized
and Sotirchos6 88-105, pore
210-250,
297-350
Garcia et al.7 Zn0.8TiO2.4 <1 9.4 179 550-600 TGA -
Li et al.8 ZnO-MnO 3300-4000 36.6 50 200-400 TGA grainy pellet
and shrinking
core
Mojtahedi and Zn1.5TiO3.5 500-150 1.9-3.3 4500-5000 550-650 fluidized bed -
Abbasian9,10
Konttinen et al.11,12 ZnxTiyOx+2y 200-300 1.9-3.3 400-500 400-600 modeling only overlapping grain
and shrinking core
a Single pellets.
the change in conversion of solid zinc oxide based Table 2. Summary of Chemical and Physical Properties
sorbents with time were examined. Three sorbents with of the Three Sorbents Used in the Current Study
zinc oxide contents of approximately 60, 80, and 100 wt
ZnO content (wt %)
% were studied. The aim of this work was to predict
the performance of zinc oxide based sorbents when used physiochemical property 100 80 60
to treat hot hydrogen sulfide containing gas in a pilot- total sulfur (wt %) <0.01 <0.01 <0.01
scale transport reactor. This paper focuses on the N2 BET surface area (m2/g) 5.91 44.18 89.69
chemisorption surface area 0.543 2.892 4.992
interpretation of laboratory-scale data from thermo- (cm3 of ammonia/g)
gravimetric analysis and the comparison of conversion- ZnO content (wt %) 99.9 83.1 59.7
time predictions, for the same sorbent, in a transport mean particle size 77.3 71.3 78.5
reactor using the available models. (volume average) (µm)
mean particle size 61.1 55.4 64.1
(surface area average) (µm)
Experimental Methods bulk density (g/cm3) 0.917 1.23 1.07
particle density (g/cm3) 1.54 2.35 1.93
The conversion-time histories of the different sor- skeletal density (g/cm3) 5.42 4.81 3.99
bents were measured using a TA Instruments model Hg average pore diameter (nm) 516.9 52.6 28.3
TGA 2950 thermogravimetric analyzer with an EGA total pore area (m2/g) 7.01 46.03 97.16
porosity 73.0% 74.4% 73.2%
(evolved gas analysis) furnace. For all of the TGA tests,
a sample weight of approximately 3 mg of sorbent and The final phase of the work involved a series of
a total gas flow rate of 140 std cm3/min containing 0.5-2 experiments performed in a small pilot-scale transport
vol % H2S, with the balance nitrogen, were used. At the (or riser) reactor at the Department of Energy’s National
temperatures used in the current work (<600 °C), the Energy Technology Center, Morgantown, WV. The
thermal decomposition of hydrogen sulfide is negligible, experimental setup for this work is illustrated in Figure
and thus, it is not necessary to include hydrogen, to 1. For the studies reported here, the flow rate of gas to
suppress this decomposition reaction, in the feed gas to the riser reactor (11.54-m length and 0.84-cm inside
the TGA. In addition, at these conditions, the effect of diameter) was set at approximately 1350 std cm3/s, and
external mass transfer on the initial observed reaction its composition was 1 vol % hydrogen sulfide with the
rates was found to be negligible.
To measure the pore size distribution, samples were
analyzed using an AutoPore 9520 SEM mercury intru-
sion porosimeter (Micromeritics Instrument Company,
Norcross, GA) over a pressure range of 6.9 kPa to 413
MPa. The size distributions of the sorbents were evalu-
ated using a Coulter Counter Multisizer II particle size
analyzer (Beckman-Coulter, Hialeah, FL). BET surface
area analyses using nitrogen were carried out using a
Quantachrome Autosorb-6 automated gas sorption sys-
tem (Quantachrome Instruments, Boynton Beach, FL).
For chemisorption analysis, a Micromeritics ASAP
2010C chemisorption analyzer (Micromeritics Instru-
ment Company, Norcross, GA) was employed using both
ammonia and tert-butyl amine. Sorbent samples were
tested for metals content using a Perkin-Elmer (Shelton,
CT) Optima 3000 radial-view inductively coupled plasma
atomic emission spectrometer, and sulfur content was
analyzed using an SC-432DR sulfur analyzer (Leco
Corp., St. Joseph, MI). Basic data for the three sorbents
used in the current work are reported in Table 2. Figure 1. Schematic diagram of transport reactor.
Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004 1237

Figure 2. Conversion histories of different size cuts of 60% ZnO


sorbent.
Figure 3. Surface area and volume of NH3 absorbed as functions
balance nitrogen. Solids flow rates were set using a loss- of binder content.
in-weight feeder (K-Tron, Pitman, NJ) located at the
bottom of the bed that fed solids to a mixing chamber product layer formed at the surface of the grains
prior to their flowing upward through the riser reactor. (within-grain diffusion) are important for the sorbents
The reaction between the solids and gas leaving the used in this work. Therefore, the initial reaction rate
riser was quenched, and the phases were separated data from the TGA tests can be used to establish the
using a cyclone and filter. Gas samples were taken intrinsic kinetics for the reaction of hydrogen sulfide and
during the experimental runs, but H2S concentrations zinc oxide
in the gas leaving the reactor were found to vary
considerably and were unreliable for estimating overall
ZnO(s) + H2S(g) f ZnS(s) + H2O(g)
conversion. Instead, the conversion of the gas-phase
hydrogen sulfide was calculated from the sulfur content -rH2S ) k1CH2SmCZnOn (1)
of the solids retrieved from the spent sorbent collector
at the end of each run. Previous researchers1,2,11 have assumed or shown that
the dependence of the reaction rate on the concentration
Results and Discussion of zinc oxide is zeroth-order, whereas others3,7 have
assumed a first-order dependence on the solid oxide
Intrinsic Reaction Kinetics. To determine intrinsic concentration. For this work, the value of n is assumed
reaction kinetics from initial reaction rate data, it is to be zero, consistent with the work for pure zinc oxide
important to establish the roles played by the different and zinc titanates by Lew et al.1 Therefore, for TGA
diffusion-related resistances. As mentioned above, care studies, the slope of the conversion-time plot for a given
was taken to avoid external mass-transfer resistances sorbent can be related to the rate constant k1 by
by using a high gas flow that eliminated any gas flow
dependence of the rate of reaction. However, when
considering intraparticle diffusion, the size and internal
pore structure of the particles determine whether this k1 )
(1 dx
MWZnO dt ) | t)0
(2)
resistance is important. For some particles undergoing (SA)(CH2S)
gas-solid reaction, intraparticle diffusion might always
be present, in which case the interpretation of initial where
rate data becomes more difficult (Sofekun and Do-
raiswamy23) For the case when intraparticle diffusion Wt - W0

( )
plays a role, the initial rate of reaction is influenced by x ) conversion of sorbent )
MWZnS
particle size. - 1 (1 - R)W0
Previous workers1,2 have manufactured sorbents con- MWZnO
taining large fractions of macropores and noted that,
for their particles, below 90-125 µm, the effect of the To use eq 2 to determine the rate constant, k1, the
particle size on the initial reaction rate was negligible. specific surface area of the sorbent (SA) must be known.
However, other researchers6 have noted a significant For pure sorbents, the active surface area can be taken
influence of particle size on initial reaction rates with as the BET surface area or the total pore surface area
particles as small as 65 µm. Conversion-time histories from mercury porosimetry measurements. However, for
for different particle size cuts were compared for the sorbent-binder pairs, the surface area reflects both the
60 wt % ZnO sorbent at a gas concentration of 1 vol % active (sorbent) and inactive (binder) surface area. This
H2S in nitrogen at 593 °C, the highest temperature level fact is clearly shown in Table 2, where the BET surface
examined in this work. The results from the TGA tests areas for the 100, 80, and 60 wt % ZnO sorbents are
are shown in Figure 2. For these tests, the conversion given as 5.91, 44.18, and 89.69 m2/g, respectively. The
profiles for all particle sizes overlap, and it is clear that large increase in surface area is clearly a function of
particle size does not have an appreciable effect on the the increasing content of alumina binder that overshad-
reaction rate. Because the intraparticle (between-grain) ows the decreasing active surface area of the zinc oxide.
diffusion path should clearly be dependent on the Figure 3 shows the relationships among the BET
particle size, the intraparticle diffusion resistance can surface area, volume of ammonia absorbed during
be assumed to be negligible and eliminated from the chemisorption studies, and pore area from mercury
analysis. Hence, only the reaction of H2S with ZnO in porosimetry as functions of the weight percent of sorbent
the grains and the diffusion of gas through the ZnS in the three samples shown in Table 2. From Figure 3,
1238 Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004

Table 3. Experimental Matrix of TGA Experiments


variable (units) values
nominal zinc oxide content of 60, 80, 100
sorbent (wt %)
temperature (K) 700, 755, 811, 867
concentration of H2S (vol %) 0.5, 1.0, 2.0

it is clear that the trends for all three of these surface


area measures, as functions of the percentage of active
sorbent, are linear. This indicates that the ammonia is
not preferentially absorbed on the zinc oxide sites, which
is because of the near equally acidic nature of the
alumina binder. The total pore area measured by
mercury intrusion porosimetry shows a close correlation
to the BET surface area results.
In this work, initial reaction rates should be compared
using the active (zinc oxide) surface area. However,
although Lew et al.1,2 were successful in correlating
initial reaction rates for a variety of zinc titanates by
normalizing the rates using the surface area from BET
analysis, such a normalization is not possible here, as
attempts to estimate the surface area of the zinc oxide
only, using chemisorption, were unsuccessful.
On the basis of the results in Figure 3, it is postulated
that, for unreacted sorbent, the surface area can be
linearly partitioned into the amount due to the binder
and that due to the zinc oxide

SATotal ) (1 - R)SAZnO + RSABinder (3)


Figure 4. Comparison of reaction rates for the 60% and 80% ZnO
Using the BET surface area in eq 3, the best-fit sorbents with that for the 100% ZnO sorbent at the same
regression gives a value of SABinder ) 211 m2/g, which conditions: (a) using the same grain size as for the 100% sorbent,
is consistent with the range of specific surface areas for (b) using the best-fit grain sizes.
alumina supports, namely, 80-350 m2/g.24 If it is further
assumed that the zinc oxide is distributed within the Table 4. Comparison of Constants for Initial Reaction
Rate with Literature Values
alumina support in the form of small spherical grains,
then the initial active surface area of the sorbent can activation
be expressed as energy, frequency 95% CI
Ea factor, for Ea
reference, sorbent (kJ/gmol) k0 (cm/s) (kJ/gmol)
(1 - R) 6
SAZnO|t)0 ) (4) current work, ZnO 31.4 0.333 38.2-24.5
FZnO dgrain Lew et al.,1 ZnO 43.1 1.31 -
Lew et al.,1 ZnxTiyOx+2y 38.9 0.40 -
where FZnO is the density of pure zinc oxide (5600 kg/ Westmoreland et al.,3 ZnO 30.3 1.22 -
m3). The grain diameter is obtained from the total pore
area of 100% sorbent (R ) 0, SA ) 7.01 m2/g) as dgrain ) Figure 4b, along with the best-fit values for the grain
0.153 µm. This grain dimension is similar to that dimensions.
reported by Lew et al.2 Scanning electron photomicro- As noted previously, initial reaction rate data were
graphs of the pure sorbent show grain sizes in this recorded for the three sorbents in Table 2 under the
range, although the distribution of sizes and shapes is experimental conditions given in Table 3. The activation
quite broad. The active surface areas for each of the energy and frequency factor were found using a stan-
other sorbents were estimated from eq 4, using the grain dard Arrhenius plot of ln(k1) as a function 1/T. The
radius given previously. Experiments were run to results are given in Table 4 along with the 95%
determine the initial rate of reaction for each of the confidence interval (CI). Also included are the results
sorbents at a variety of temperatures and H2S concen- of Lew et al.1 and Westmoreland et al.3 Although the
trations. The test matrix for these runs is summarized data from these researchers showed significant scatter,
in Table 3. Initial reaction rates were found to be a confidence intervals were not reported for the rate
linear function of the concentration of H2S, which is constants. From Table 4, it is clear that the activation
consistent with a first-order mechanism, m ) 1 in eq 1, energy measured in the current work is consistent with
as noted by other researchers.1-3 The surface areas from the work of Westmoreland et al. However, it appears
eq 4 were used with the initial reaction rate data from to be significantly lower (at a 95% significance level)
these tests, and a comparison of all three sorbents is than the results of Lew et al.,1 although the scatter in
shown Figure 4a. It can be seen from this figure that the work of these researchers is not taken into account
the rates for the 100 and 80 wt % ZnO sorbent agree in this comparison.
well with each other but that the results for the 60 wt Gas-Solids Reaction Modeling. A comparison of
% ZnO sorbent are significantly higher. A better fit was the results from the present study with phenomenologi-
obtained by using the effective average grain dimension, cal models used previously to describe the sulfidation
dgrain, as a fitting parameter. The results are shown in of zinc oxide was made. Data for the three sorbents
Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004 1239

volume than the reactant, the governing equation is


3
4π R0 FZnO dx
-rH2S ) - ) 4πR02CH2S ×
3 MWZnO dt

( R0{[Zv + (1 - Zv)(1 - x)]1/3 - (1 - x)1/3}


[Zv + (1 - Zv)(1 - x)]1/3(1 - x)1/3Deff
+

)
-1
1
(5)
(1 - x)2/3k1

where Zv is the ratio of the molar volume of the product


to that of the reactant, R0 is the initial radius of the
grains, Deff is the effective diffusivity through the
product layer, CH2S is the concentration of hydrogen
sulfide at the surface of the grains, and x is the
conversion of the grains. Under conditions of constant
hydrogen sulfide concentration, eq 5 can be integrated
to yield the conversion-time history, which is given by

FZnOR02
t) ×
2MWZnODeffCH2S

{Zv - [Zv + (1 - Zv)(1 - x)]2/3


Zv - 1
- (1 - x)2/3 + }
FZnOR0
[1 - (1 - x)1/3] (6)
MWZnOk1CH2S

In eq 6, the only unknown parameter is the effective


diffusion coefficient of the product layer, Deff. However,
if the time for complete reaction of the sorbent is known,
then the effective diffusion coefficient can be determined
by setting the conversion (x) equal to unity in eq 6.
Using average diameters of 0.105 and 0.153 µm, respec-
tively, for the 60 and 80 wt % zinc oxide containing
sorbents, the conversion histories obtained from the
TGA are compared in Figure 6 with predictions from
eq 6 at a temperature of 593 °C and a hydrogen sulfide
concentration of 1 vol %. The corresponding values of
Deff for the 60 and 80 wt % sorbents at 593 °C were
found to be 4.3 × 10-14 and 9.2 × 10-14 m2/s, respec-
tively. These values compare with 1.9 × 10-12 m2/s,
calculated using an overlapping grain model developed
by Lew et al.1, and 1.4 × 10-13 m2/s, calculated using a
variable property grain model proposed by Ranade and
Harrison.13 From Figure 6, it is clear that, compared to
the experimental data, the model underpredicts the
conversion at any given time. The lack of fit of the basic
Figure 5. Comparison of reaction rates for an H2S mole fraction
of 1% at 482 and 593 °C for (a) 60%, (b) 80%, and (c) 100% sorbent. grain model for the sulfidation of zinc oxide is consistent
with other work that has shown that the size distribu-
listed in Table 2, at two different temperatures, are tion of the grains, the shape of the grains, and the
shown in Figure 5. It is clear that, although the trends overlapping of the grains all significantly influence the
of conversion with time are similar for all three sor- predicted shape of the conversion-time history.
bents, some differences do exist. The differences in To obtain a better fit of the data, several modifications
conversion histories between the different sorbents to the basic grain model were investigated. These
might be due to structural differences between the modifications included using a grain size distribution
sorbents, such as different grain size distributions and obtained through image analysis of SEM images, using
different grain shapes. In accordance with the previous a distribution of shape factors obtained through the
discussion, it was observed that the resistances to both image analysis to describe the grains, and using a
external mass transfer and to intraparticle (between- swelling grain model. However, none of these methods
grain) mass transfer can be ignored. Therefore, the basic produced a satisfactory fit. The best fit was obtained
starting point for comparison is the grain model in using a bimodal size distribution of grain sizes (equal
which only the product-layer (within-grain) diffusion weight fractions) that satisfied the criterion that the
and kinetic resistances are important. For noninteract- total surface area of the grains be conserved. The grain
ing grains in which the product has a different molar sizes obtained for the 60 wt % sorbent were 0.072 and
1240 Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004

of the conversion history, the fit is good for both the 80


and 60 wt % sorbents.
The values for the effective diffusion coefficient in the
product layer, Deff, calculated using the bimodal size
distribution of grains were in the range of 1.6-5.1 ×
10-13 m2/s at 583 °C and 2.8-8.4 × 10-14 m2/s at 482
°C; these results compare well with the predictions of
Ranade and Harrison13 of 1.4 × 10-13 and 2.1 × 10-14
m2/s at 583 and 482 °C, respectively.
The bimodal grain size distribution was used for
comparing the results of the TGA experiments with
those of the pilot-plant study. Because the maximum
conversion obtained in the pilot-scale experiments was
less than 40%, the use of this modification to the grain
model did not cause significant error for the conditions
used in the pilot-scale experiments.
Transport Reactor Modeling. The objective of this
section of the work was to compare the findings of the
TGA studies with the conversions obtained for similar
sorbents in the transport reactor shown in Figure 1. Two
series of 10 runs each were performed for the two zinc
oxide containing sorbents (60 and 80 wt % sorbent)
studied previously at essentially identical conditions
(temperature of 538 °C and pressure of 2.05 MPa), and
these results are summarized in Table 5. It should be
noted that no regeneration of the sorbent was carried
Figure 6. Comparison of predictions of the grain model with data
out during the series of 10 runs, and therefore, the
at 593 °C for (a) 60% and (b) 80% sorbent. weight percent of sulfur measured was the cumulative
amount of sulfur in the sorbent up to that number of
reactor passes. The conversions, as a function of the
number of passes, for the two sorbents are listed in
Table 5 and plotted in Figure 8 for both sorbents. The
results show very similar conversion histories for the
two sorbents. However, the conversion at six riser
passes for the 60 wt % sorbent appears to be an
anomalous result.
To compare the results of the TGA studies with those
from the pilot-plant unit, a simple reactor model for the
riser was developed. Starting with eq 5 and assuming
that the flow pattern for both the solid and gas in the
riser are close to plug flow, the relationship between
the gas-phase concentration and the solid-phase conver-
sion for isothermal operation is given as

dCH2S Ms,0(1 - R) dx
v0 ) (7)
dt MWZnO dt

where v0 is the superficial volumetric gas flow rate and


Ms,0 is the mass flow rate of unreacted sorbent into the
riser. To predict conversion levels per pass, eqs 5 and 7
must be integrated simultaneously along the length of
the riser. In addition, because a bimodal size distribu-
Figure 7. Comparison of predictions of the bimodal grain model tion of grain sizes has been assumed, the conversion
with data at 482 and 593 °C for (a) 60% sorbent (50% with dgrain,1 level for each grain size must be followed through the
) 0.072 µm and 50% with dgrain,2 ) 0.194 µm) and (b) 80% sorbent integration. To integrate these equations, the relative
(50% with dgrain,1 ) 0.100 µm and 50% with dgrain,2 ) 0.326 µm). slip between the sorbent and gas phase must be as-
sessed. For very dilute flows, a first approximation
0.194 µm, which yielded an initial reaction surface area might be to assume that the solids are traveling relative
to the gas at their terminal velocity, which was esti-
of 6.02 m2/g. For the 80 wt % sorbent, the grain sizes
mated to be approximately 15 cm/s. Alternatively, the
obtained were 0.100 and 0.326 µm, which yielded an
slip velocity between the solids and gas can be calcu-
initial reaction surface area of 5.67 m2/g. Using this lated in terms of the void fraction in the riser, according
approach, the predicted conversion-time histories are to the relation
compared with the experimental data in Figure 7. This
approach is essentially empirical, given that the actual ug LMs
size distributions of the grains were quite broad for both uslip ) - (8)
sorbents. However, it can be seen that, over the majority  Vriser(1 - )Fpart
Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004 1241

Table 5. Experimental Results from Small Pilot-Scale Reactor


nominal gas flow nominal temp sorbent recovered sorbent feed rate run time sulfur pickup conversion,
run (std cm3/s) (°C) (wt %) (g/s) (s) (wt %) x
80% Sorbent
1 1333 538 94 0.63 2880 3.19 0.0993
2 1333 538 91 0.63 2700 4.81 0.1509
3 1333 538 100 0.63 2460 6.74 0.2135
4 1333 538 100 0.63 2460 7.41 0.2356
5 1333 538 100 0.63 2460 8.21 0.2621
6 1333 538 100 0.63 2400 8.79 0.2814
7 1333 538 100 0.63 2400 9.25 0.2969
8 1333 538 100 0.63 2400 9.83 0.3165
9 1333 538 100 0.63 2400 10.20 0.3290
10 1333 538 100 0.63 2400 10.86 0.3515
60% Sorbent
1 1361 538 98 0.63 2280 2.11 0.0909
2 1361 538 101 0.63 2220 3.76 0.1632
3 1361 538 100 0.63 2220 4.60 0.2006
4 1361 538 88 0.63 2220 5.30 0.2319
5 1361 538 99 0.63 1980 5.87 0.2577
6 1361 538 100 0.63 1920 8.66 0.3857
7 1361 538 100 0.63 1920 7.34 0.3246
8 1361 538 99 0.63 1920 7.33 0.3242
9 1361 538 101 0.63 1860 8.18 0.3634
10 1361 538 88 0.63 1860 8.14 0.3616

Table 6. Summary of Changes to Chemical and Physical Properties of the Sorbents Due to Sulfidation
sorbent description
80 wt % ZnO 80 wt % 60 wt % ZnO 60 wt %
property unreacted sulfideda unreacted sulfideda
total sulfur (wt %) <0.01 10.86 <0.01 8.14
N2 BET surface area (m2/g) 44.2 23.5 89.7 68.2
chemisorption surface area (cm3 of ammonia/g) 2.892 1.935 4.992 3.477
mean particle size by volume (µm) 71.3 78.7 78.5 68.2
mean particle size by surface area (µm) 55.4 62.5 64.1 56.9
bulk density (g/cm3) 1.23 1.4815 1.0667 1.2362
skeletal density (g/cm3) 4.81 3.99 3.99 3.64
Hg average pore diameter (mm) 0.0526 0.0454 0.0283 0.0261
porosity (%) 74.4 62.8 73.2 66.0
a End of run 10.

Figure 8. Comparison of conversion histories for 60% and 80%


sorbents as a function of number of passes through the riser
reactor.

where ug is the superficial gas velocity,  is the void


fraction in the riser, Ms is the mass flow rate of sorbent
entering the reactor, L is the length of the reactor, Vriser
is the volume of the empty riser, and Fpart is the particle
density.
The predictions of this model (eqs 5, 7, and 8) are
compared with the experimental data from the riser
reactor for different void fractions in Figure 9. From
these results, it is clear that the conversion histories
are well represented by riser voidages in the range of
0.997-0.998, which are only slightly greater than the Figure 9. Comparison of experimental data with predictions of
value obtained by assuming that the solids move at their the riser-reactor model for (a) 60% and (b) 80% sorbent.
terminal velocity. Attempts were made to estimate
independently the solids hold-up in the riser using with the high absolute pressures, made reliable mea-
pressure drop data. However, the low voidages, coupled surements of differential pressure very difficult.
1242 Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004

Finally, the physical and chemical properties of the Subscripts


sorbents were compared before and after the sulfidation 1, 2 ) grain sizes used in the bimodal grain size model
experiments in the transport reactor and are shown in Binder ) binder
Table 6. The changes in surface area, skeletal density, g ) gas
and porosity of the sorbents that occurred during the 0 ) initial or superficial
course of the experiments were all significant. part ) particle
Conclusions riser ) in the riser
In conclusion, for the sorbent-binder combinations s ) sorbent
examined in this work, the effect of conversion history slip ) relative velocity of the solids with respect to the gas
was independent of pellet size. A simple grainy-pellet t ) time
model was unsatisfactory in describing the conversion- ZnO ) zinc oxide
time history obtained in TGA experiments, which is ZnS ) zinc sulfide
consistent with other researchers findings. A bimodal
size distribution of grains was used to predict success- Literature Cited
fully the conversion-time profile of the 60 and 80 wt % (1) Lew, S.; Sarofim, A. F.; Flytzani-Stephanopoulos, M. Sul-
sorbents at temperatures of 482 and 593 °C. Effective fidation of Zinc Titanate and Zinc Oxide Solids. Ind. Eng. Chem.
diffusion coefficients derived from this model were found Res. 1992, 31, 1890-1899.
to agree well with previously determined values. Fi- (2) Lew, S.; Sarofim, A. F.; Flytzani-Stephanopoulos, M. Model-
nally, the bimodal grain size model was used with a ing of the Sulfidation of Zinc-Titanium Oxide Sorbents with
simple plug-flow reactor model to predict the conversion Hydrogen Sulfide. AIChE J. 1992, 38, 1161-1169.
per pass in the riser section of a small pilot-scale reactor (3) Westmoreland, P. R.; Gibson, J. B.; Harrison, D. P. Com-
parative Kinetics of High-Temperature Reaction Between H2S and
operating at 533 °C and 2.05 MPa. The conversion Selected Metal Oxides. Environ. Sci. Technol. 1977, 11, 488-491.
histories were found to be strongly dependent on the (4) Woods, M. C.; Gangwal, S. K.; Jothimurugesan, K.; Harri-
relative slip between the solids and the gas. The best son, D. P. Reaction between H2S and Zinc Oxide-Titanium Oxide
fits between the model and experimental results were Sorbents. 1. Single-Pellet Kinetic Studies. Ind. Eng. Chem. Res.
found for void fractions in the range of 0.997-0.998. 1990, 29, 1160-1167.
These values are slightly lower than the values obtained (5) Jothimurugesan, K.; Harrison, D. P. Reaction between H2S
by assuming that the solids move at their terminal and Zinc Oxide-Titanium Oxide Sorbents. 2. Single-Pellet Sulfi-
dation Modeling. Ind. Eng. Chem. Res. 1990, 29, 1167-1172.
velocity relative to the gas. For the first time, data from
(6) Efthimiadis, E. A.; Sotirchos, S. V. Reactivity Evolution
a laboratory TGA apparatus have been used to predict During Sulfidation of Porous Zinc Oxide. Chem. Eng. Sci. 1993,
successfully the performance of a pilot-scale transport 48, 829-843.
reactor. (7) Garcia, E.; Cilleruelo, C.; Ibarra, J. V.; Pineda, M.; Palacios,
J. M. Thermogravimetric Study of Regenerable Sulphur Sorbents
Acknowledgment for H2S Retention at High Temperatures. Thermochimica Acta
The authors thank William Grimes, Lanny Golden, 1997, 306, 23-30.
Don Floyd, and K. David Lyons for their assistance in (8) Li, Y.; Guo, H.; Li, C.; Zhang, S. A Study on the Apparent
conducting the TGA and transport desulfurization tests. Kinetics of H2S Removal using a ZnO-MnO Desulfurizer. Ind.
Eng. Chem. Res. 1997, 36, 3982-3987.
Funding for R.T. was provided through the Oak Ridge (9) Mojtahedi, W.; Abbasian, J. H2S Removal from Coal Gas
Institute for Science and Education’s (ORISE) Educa- at Elevated Temperatures and Pressure in Fluidized Bed with Zinc
tional and Research Experiences for Faculty program, Titanate Sorbents. 1. Cyclic Tests. Energy Fuels 1995, 9, 429-
Oak Ridge Institute for Science and Education, Oakridge, 434.
TN. (10) Mojtahedi, W.; Abbasian, J. H2S Removal from Coal Gas
at Elevated Temperatures and Pressure in Fluidized Bed with Zinc
Titanate Sorbents. 2. Sorbent Durability. Energy Fuels 1995, 9,
Notation 782-787.
C ) concentration, mol/cm3 (11) Konttinen, J. T.; Zevenhoven, C. A. P.; Hupta, M. M. Hot
Gas Desulfurization with Zinc Titanate in a Fluidized Bed. 1.
dgrain ) diameter of the grain, µm Determination of Sorbent Particle Conversion Rate Model Param-
Deff ) effective diffusion coefficient through the product eters. Ind. Eng. Chem. Res. 1997, 36, 2332-2339.
layer, cm2/s (12) Konttinen, J. T.; Zevenhoven, C. A. P.; Hupta, M. M. Hot
k1 ) rate constant in eq 1, cm/s Gas Desulfurization with Zinc Titanate in a Fluidized Bed. 2.
L ) length of the riser, cm Reactor Model. Ind. Eng. Chem. Res. 1997, 36, 2340-2345.
M ) mass flow rate of the sorbent, g/s (13) Ranade, P. V.; Harrison, D. P. The Variable Property Grain
MW ) molecular weight, g/mol Model Applied to the Zinc Oxide-Hydrogen Sulfide Reaction.
R0 ) initial radius of the grain, µm Chem. Eng. Sci. 1981, 36, 1079-1089.
(14) Yagi, S.; Kunii, D. Studies on Combustion of carbon
SA ) active surface area of the sorbent, cm2/g
particles in flames and fluidized beds. In Proceedings of the 5th
t ) time, s Symposium (International) on Combustion; Hottel, H. C., Ed.;
u ) velocity, cm/s Reinhold: New York, 1955; pp 231-244.
v ) volumetric flow rate, cm3/s (15) Szekely, J.; Evans, J. W.; Sohn, H. Y. Gas-Solid Reactions;
V ) volume, cm3 Academic Press: New York, 1976.
W ) mass of the sorbent, g (16) Georgakis, C.; Chang, C. W.; Szekely, J. A Changing Grain
x ) conversion of the sorbent Size Model for Gas-Solid Reactions. Chem. Eng. Sci. 1979, 34,
Zv ) ratio of the molar volume of the product to that of 1072-1075.
the reactant (17) Sotirchos, S. V.; Yu, H.-C. Overlapping Grain Models for
Gas-Solid Reactions with Solid Product. Ind. Eng. Chem. Res.
Greek Letters 1988, 27, 836-845.
(18) Bartlett, R. W.; Krishnan, N. G.; Van Hecke, M. C.
R ) mass fraction of binder in the sorbent Micrograin Models of Reacting Porous Solids with Approximations
 ) void fraction in the riser to Logarithmic Solid Conversion. Chem. Eng. Sci. 1973, 28, 2179-
F ) density, g/cm3 2186.
Ind. Eng. Chem. Res., Vol. 43, No. 5, 2004 1243

(19) Ishida, M.; Wen, C. Y. Comparison of Kinetic and Diffu- (23) Sofekun, O. A.; Doraiswamy, L. K. High-Temperature
sional Models for Solid-Gas Reactions. AIChE J. 1968, 14, 311- Oxidation of Zinc Sulfide: Kinetic Modeling under Conditions of
317. Strict Kinetic Control. Ind. Eng. Chem. Res. 1996, 35, 3163-3170.
(20) Mantri, V. B.; Gokarn, A. N.; Doraiswamy, L. K. Analysis (24) Satterfield, C. N. Heterogeneous Catalysis in Industrial
of Gas-Solid Reactions: Formulation of a General Model. Chem. Practice, 2nd ed.; Krieger Publishing Co.: Malabar, FL, 1996.
Eng. Sci. 1976, 31, 779-785.
(21) Ramachandran, P. A.; Smith, J. M. A Single-Pore Model
for Gas-Solid Noncatalytic Reactions. AIChE J. 1977, 23, 353- Received for review April 28, 2003
361. Revised manuscript received December 22, 2003
(22) Yu, H.-C; Sotirchos, S. V. A Generalized Pore Model for Accepted December 23, 2003
Gas-Solid Reactions Exhibiting Pore Closure. AIChE J. 1987, 33,
382-393. IE030364A

Das könnte Ihnen auch gefallen