EPA-600 .'2-89-065
December 1989
FURNACE SORBENT REACTIVITY TESTING FOR
CONTROL OF S02 EMISSIONS FROM ILLINOIS COALS
Final Report
Prepared by:
Brian K. Gullett and Frank E. Briden
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, North Carolina 27711
Prepared for:
Center for Research on Sulfur in Coal
Suite 200, Coal Development Park
Post Office Box 8
Carterville, Illinois 62918-0008
and
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
NOTICE
This document has been reviewed in accordance with
U.S. Environmental Protection Agency policy and
approved for publication. Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
-------
ABSTRACT
Research was undertaken to evaluate the potential of furnace sorbent injection
(FSI) for sulfur dioxide (S02) emission control on coal fired boilers
utilizing coals indigenous to the State of Illinois. Tests were run using
four coals from the Illinois Basin and six calcium hydroxide [Ca(0H)2]
sorbents, including one provided by the Illinois State Geological Survey
(ISGS).
Testing was divided into three tasks:
1. Pilot- and bench-scale sorbent reactivity testing.
2. Sorbent microstructure characterization.
3. Injection ash characterization.
Pilot-scale FSI testing gave SO2 removal percentages greater than 60, with
some tests (including those with the ISGS sorbent) exceeding 70 percent
removal for Ca/S ratios of 2:1. Bench-scale testing of injection at
economizer temperatures (538 °C) yielded comparable removals of approximately
55 percent. X-Ray diffraction (XRD) tests of the sorbents showed a strong
correlation between three measured crystallite microstructural parameters and
sorbent reactivity in the FSI tests. Extraction Procedure (EP) Toxicity Tests
with the sorbent injection ash gave values well below Resource Conservation
and Recovery Act (RCRA) limits for regulated metals.
-------
TABLE OF CONTENTS
Section Page
Abstract ...... ii
List of Figures ; . . . . iv
List of Tables v
Acknowledgements vi
1.0 Executive Summary 1-1
2.0 Introduction and Background 2-1
3.0 Experimental Procedures 3-1
3.1 Innovative Furnace Reactor Tests 3-1
3.2 Graphite Furnace Reactor Tests 3-2
3.3 X-Ray Diffraction Tests 3-5
3.4 Extraction Procedure Toxicity Tests 3-6
3.5 Short Time Differential Reactor Tests 3-6
3.6 Analytical Procedures 3-8
4.0 Results and Discussion 4-1
4.1 Innovative Furnace Reactor Tests . 4-1
4.2 Graphite Furnace Reactor Tests 4-6
4.3 Short Time Differential Reactor Tests 4-7
4.4 X-Ray Diffraction Tests 4-9
4.5 Extraction Procedure Toxicity Tests 4-11
5.0 Conclusions and Recommendations 5-1
6.0 References 6-1
Appendix A: Innovative Furnace Reactor Data A-l
Appendix B: Predicting SO? Sorbent Reactivity in the Innovative
Furnace with x-Ray Diffraction Peak Profiles for the
Illinois Center for Research on Sulfur in Coal B-l
Appendix C: Quality Control Evaluation Report C-l
v
-------
LIST OF FIGURES
F igure Page
3-1 Innovative Furnace Reactor (IFR) 3-3
3-2 Graphite Furnace Reactor (GFR) 3-4
3-3 Short Time Differential Reactor (STOR) ' 3-7
4-1 IFR Sorbent Reactivity 4-4
4-2 STDR Alcohol Sorbent Reactivity. . 4-8
4-3 Reactivity Predicted by X-ray Line Broadening Versus IFR Data. . 4-12
B-l Reactivity Predicted by XLB Versus IFR B-15
vi
-------
LIST OF TABLES
Table Page
4-1 Coal Characteristics 4-2
4-2 Sorbent Characteristics 4-3
4-3 Results from Economizer Injection Tests on GFR 4-6
4-4 Illinois Coal Reactivity and XRD Peak Shape Data 4-10
4-5 Results of EP Toxicity Tests on IFR Ash 4-13
B-l Illinois Coal Reactivity and XRD Peak Shape Data B-9
B-2 Derived Regression Functions for Individual XRD
Peak Shape Factors B-ll
B-3 Derived Regression Functions for Pairs of XRD
Peak Shape Factors . B-13
B-4 Derived Regression Functions for Triplets of XRD
Peak Shape Factors B-14
B-5 Reactivity Prediction from the Two Best XLB Regression
Functions Derived from the Observed Reactivities of
Coals #1, #6, and #9 Applied to the Individual Coals . B-17
C-l Data Quality Objectives C-3
vi i
-------
Acknowledgements
The authors wish to acknowledge: the general guidance of Mike Lin and
Ken K. Ho of CRSC (Center for Research on Sulfur in Coal), a cofunder of this
work under IDENR (Illinois Department of Energy and Natural Resources)/CRSC
subgrant agreement 88-142; the technical and editorial support of Kevin R.
Bruce and Laura 0. Beach of Acurex Corporation, supported by EPA contract 68-
02-4701; and the equipment and managerial support of George R. Gill is, U.S.
EPA.'Air and Energy Engineering Research Laboratory.
vi i i
-------
1.0 EXECUTIVE SUMMARY
Testing in the Innovative Furnace Reactor (IFR) with four Illinois coals and
six sorbents demonstrated sulfur removals in excess of 60 percent at a Ca/S
ratio of 2:1. Tests with the 1ignosulfonate modified Marblehead calcium
hydroxide and the Illinois State Geological Survey (ISGS) alcohol calcium
hydroxide gave sulfur removals above 70 percent under the same conditions.
Removal percentages were lower for the Illinois Basin Coal Sampling Program
(IBCSP) #2 coal than for the other coals. This was attributed to its higher
pyritic to organic sulfur ratio.
Bench-scale testing of reactivity at economizer temperatures (538 °C) yielded
sulfur removals of roughly 55 percent at a Ca/S of 2:1. Studies were impacted
by apparent mass transfer limitations on the rate of reaction.
X-Ray Diffraction tests of the sorbents focused on eight parameters of their
crystallite microstructure. Three of these parameters (average column length,
modal column length, and strain at maximum column length) were found to
correlate strongly with the sorbents' performance in the IFR testing. This
may allow a method of predicting sorbent reactivity without costly pilot-scale
testing.
Analyses of sorbent injection ash for potential leachability using the
Extraction Procedure (EP) Toxicity Tests gave values well below Resource
Conservation and Recovery Act (RCRA) limits for all metals tested. While
below RCRA limits, the final pH values for these leachates were high enough to
elicit concern.
1-1
-------
2.0 INTRODUCTION AND BACKGROUND
Emissions of sulfur oxides, principally sulfur dioxide (SO2), from combustion
sources have increased awareness and concern in recent years. In particular,
SOjj emissions from coal-fired boilers used by utilities and industries have
been implicated as major contributors to a growing acid precipitation problem.
While long-term ecological effects of acid precipitation are being debated, it
is clear that a reduction in S02 emissions is greatly desirable. Factors to
weigh in determining an SO2 control technology are cost, SO2 removal
efficiency, and ease of retrofitting to existing boilers. The optimum control
technology would balance removal levels with the cost to the industry or
utility (and ultimately the consumer). One technology that has received
considerable attention is Furnace Sorbent Injection (FSI), which offers a
relatively low capital cost, ease of retrofitting, and reasonable removal
efficiencies.
A large body of research on FSI is currently available. The effects of such
fundamental parameters as injection temperature, sorbent type, particle size,
and SO2 concentration have been investigated on the pilot-scale (Beittel et
al., 1985; Bortz and Flament, 1985; Snow et al., 1986). These investigations,
along with on-going full-scale demonstrations, indicate that SO2 removals of
approximately 60 percent may be expected using commercially available calcium
hydroxide [Ca(0H)2] sorbents. Noted potential impacts of FSI on the boiler
include increased slagging and fouling, increased mass loading on particulate
removal systems, and alteration of the chemical composition of boiler ash.
2-1
-------
This current investigation is designed to provide data at the pilot-scale on
SO2 removal from a combustor fired with Illinois Basin coals and injected with
a range of sorbent types. These comparative data, along with results from low
temperature testing, physical analysis of the sorbents, and chemical analysis
of the ash, will be used to evaluate FSI as a control technology for
facilities using Illinois Basin coals. Exceptionally high removal
efficiencies could expand the range of applications for Illinois high sulfur
coal at a lower cost than coal cleaning or wet flue gas desulfurization (FGD)
alternatives.
The primary objective of the planned research has been to evaluate FSI as a
potential SO2 emission control technology for coal fired boilers burning
Illinois Basin coals. FSI offers the benefits of being less capital intensive
than wet FGD as well as the ability to be readily retrofitted to existing
facilities with space considerations. To evaluate FSI potential the following
specific objectives have been outlined:
1. Develop a data base of sorbent SO2 removal efficiencies using six
sorbents with four coals at two Ca/S ratios in the Environmental
Protection Agency's (EPA) Innovative Furnace Reactor (IFR) at a
high injection temperature (1,200 °C) regime.
2. Obtain comparative SO2 reactivity data for the six sorbents at
mid-range temperatures (538 °C) in EPA's Graphite Furnace Reactor
(GFR).
2-2
-------
Characterize sorbent microstructure properties using x-ray
diffraction (XRD) techniques in an effort to correlate these
properties with sorbent SO2 removal efficiencies.
Determine the potential for leaching of toxic metals from FSI ash
using the EPA's Extraction Procedure (EP) Toxicity Test.
-------
3.0 EXPERIMENTAL PROCEDURES
In the following section, the experimental conditions used in satisfying the
objectives outlined in Section 2.0 are given. The four coals used in testing
were identified as Illinois Basin Coal Sample Program (IBCSP) #1, #2, #6, and
#9. Sorbents chosen for testing included three commercially available calcium
hydroxides (Marblehead, Linwood, and Snowflake), a dolomitic hydroxide
(Kemidol), a surfactant modified calcium hydroxide (1ignosulfonate modified
Marblehead), and an alcohol calcium hydroxide provided by the Illinois State
Geological Survey (ISGS). The ISGS sorbent was tested by combining equal
parts of each of the 10 batches provided. This was necessary to ensure that
adequate sorbent was on hand for FSI testing in the IFR. Limited testing of
the individual batches was done in the other reactor systems.
3.1 INNOVATIVE FURNACE REACTOR TESTS
Testing in the IFR consisted of determining S0£ concentrations in the flue gas
during sorbent injection while burning each of the coals at feed rates
sufficient to yield a firing rate of approximately 49,600 kJ/h (47,000 Btu/h).
The SO2 Continuous Emission Monitor (CEM) was calibrated using multiple
calibration gases bracketing measured concentration levels. After
establishing a stable baseline S0g concentration, sorbent was injected at a
Ca/S ratio of approximately 1:1 and the SO2 level monitored until an
equilibrium level was achieved. Sorbent injection was then stopped and the
S02 level allowed to return to the established baseline prior to repeating the
3-1
-------
test. The final S02 removal percentage was determined as the average of the
duplicate tests. The entire procedure was then reinitiated with a Ca/S ratio
of approximately 2:1. The entire test matrix consisted of testing each coal
with all six sorbents using duplicate runs (4 coals x 6 sorbents x
2 duplicates x 2 Ca/S ratios = 96.tests). Actual sorbent and coal feed rates
were determined gravimetrically. After running on consecutive test days,
slagging in the furnace required that it be run overnight with a gas lance in
order to deslag the interior prior to reinitiating testing. A diagram of the
IFR is shown in Figure 3-1. In all tests the sorbent injection location and
temperature were identical to those used in previous comparative tests on the
IFR (Snow et al., 1986).
3.2 GRAPHITE FURNACE REACTOR TESTS
Current supply to the electrically heated GFR was regulated to yield a
temperature profile with a peak of near 538 °C (1,000 °F) while declining
rapidly with residence time (or distance) in the reactor. Flow rates
sufficient to give a residence time of 0.75 s between 538 and 427 °C with an
S02 concentration of 3,000 ppm were used. Each sorbent was injected under
differential conditions with respect to S02 concentration and conversion to
calcium sulfite (or sulfate) determined on solid samples collected by a
cyclone separator. Figure 3-2 shows a cross-section of the GFR.
3-2
-------
u>
1
CO
SORBENT
FEEDER
COAL/GAS
FEEDER
SAMPLING
PORTS' —
ASH TRAP
INJECTION
PROBE
REFRACTORY
SHELL
GAS ANALYSIS PORT
~
n.
BYPASS VALVESh
ROOF LEVEL
EXHAUST STACK
FAN
,L1_,
BAG HOUSE
Figure 3-1. Innovative Furnace Reactor.
-------
CARRIER GAS AND
SORBENT PARTICLES
ALUMINA MUFFLE
TUBES
GRAPHITE
HEATING
ELEMENTS
SORBENT INJECTION
PROBE
- GLASS BEADS
HONEYCOMB FLOW
STRAIGHTENERS
PROCESS GAS
ENTRANCE
COLLECTION
PROBES-
REACTED SORBENT
PARTICLES AND GAS
Figure 3-2. Graphite Furnace Reactor.
3-4
-------
3.3 X-RAY DIFFRACTION TESTS
Each of the six sorbents was analyzed by XRD using procedures and theory
discussed more fully elsewhere (Briden and Natschke, 1988, 1989). The
Warren-Averbach method of peak analysis for separation of the crystallite size
and strain components was used to determine the following major microstructure
properties:
1. Average crystallite size.
2. Frequency of occurrence of column lengths as a function of
column length and the column length of maximum frequency of
occurrence (modal column length).
3. Half width of the frequency of occurrence of column lengths as a
function of column length.
4. Maximum determinable column length.
5. Strain at the average column length.
6. Strain as a function of column length.
7. Strain at the column length of maximum frequency of occurrence.
8. Strain at the maximum determinable column length.
The individual values of the sorbent microstructural properties were related
3-5
-------
to IFR-determined reactivities by regression functions to test the hypothesis
that various combinations of these properties could predict sorbent
reactivity.
3.4 EXTRACTION PROCEDURE TOXICITY TESTS
Toxicity tests were performed on ash taken from the baghouse during each of
the baseline coal tests excluding coal #1 for which insufficient sample was
collected. Analyses on eight Resource Conservation and Recovery Act (RCRA)
regulated metals (antimony, barium, cadmium, chromium, lead, mercury,
selenium, and silver) and pH were carried out using methods outlined in EPA
method 1310. The tests simulate leachate from a landfill or pond runoff. Ash
from sorbent injection using one coal (IBCSP #6) and all six sorbents
(Ca/S = 2:1) was also tested to determine the impact of FSI on disposal of
ash. Previous work indicated that unreacted calcium oxide (CaO) acts to
stabilize most of the metals, but an increase in leachate pH is expected
(Dahlin et al., 1986).
3.5 SHORT TIME DIFFERENTIAL REACTOR TESTS
Tests were run on the Short Time Differential Reactor (STDR) using 4 mg of
sorbent exposed to process gas consisting of 3,000 ppm S0£ in 5 percent 02 and
a Ng balance, preheated to 538 °C. The reactor, shown in Figure 3-3, is
designed to allow fixed bed sample exposure times in the range of 0.3 to 5 s,
while maintaining conditions differential with respect to SO2 concentration.
-------
CJ
I
PROCESS GAS
PREHEATER
PHOTO
SENSOR
STARTER
SAMPLE SLIDER
ASSEN/BLY
AIR PISTON
AIR IN
Figure 3-3. Short Time Differential Reactor.
-------
3.6 ANALYTICAL PROCEDURES
Products from the GFR and STDR were analyzed for calcium and sulfate/sulfite
concentration by atomic absorption spectroscopy and ion chromatography,
respectively. Samples were dissolved in deionized water, IN HC1, and ^2, and
stirred for 5 min. The sample was then filtered into a volumetric flask and
diluted for analyses. Sorbent particle size analyses were performed on a
Micromeritics Sedigraph 5100. Approximately 0.5 g of sample was dispersed in
Micromeritics dispersant All and allowed to stir for 15 min. Particle size
distributions were determined over the range of 1 to 50 /m in diameter.
3-8
-------
4.0 RESULTS AND DISCUSSION
4.1 INNOVATIVE EURNACE REACTOR TESTS
Prior to initiating tests in the IFR, baseline data were collected on the
coals and sorbents. Information on the coals, compiled in Table 4-1, was
provided by the ISGS. Table 4-2 presents data on the sorbents' physical and
chemical characteristics provided by the ISGS and our analyses. Calcium
values determined in-house were used in calculating Ca/S ratios in IFR
testing.
Results for FSI testing on the IFR are compiled in Figure 4-1. The data
presented are estimated S0£ removal percentages at Ca/S of 2:1, calculated by
extrapolating linearly from the mean removals at both Ca/S ratios run for each
coal/sorbent combination. Some repeats were necessary due to difficulties
encountered in controlling the sorbent feed rate in early tests. The compiled
raw data are given in Appendix A.
Figure 4-1 shows several trends in the data. SO2 capture levels for the
IBCSP #2 coal are substantially lower for all sorbents tested (with the
possible exception of the Marblehead hydroxide) than for the other coals. It
is interesting to note that, while the sulfur content of the IBCSP #2 coal
(3.23 percent) is bracketed by the other coals, it differs from them in one
important aspect. Unlike the other coals tested, pyritic sulfur accounts for
the majority of the sulfur present in IBCSP #2, giving a pyritic/organic
4-1
-------
sulfur ratio of 2.53:1 compared to values less than 1:1 for the other coals.
No explanation for the apparent adverse effect of a high pyritic/organic
sulfur ratio on FSI is currently.available. Future work to verify and
determine the cause of this phenomenon is needed.
TABLE 4-1. COAL CHARACTERISTICS
IBCSP #1 IBCSP #2 IBCSP #6 IBCSP #9
Moisture
14.14*
13.62
10.42
6.60
Volatile Matter
44.12
43.34
39.56
35.90
Fixed Carbon
45.62
49.92
51.44
56.10
Ash
10.28
6.66
9.00
8.00
Carbon
67.66
73.31
71.64
74.72
Hydrogen
4.86
5.21
4.73
5.06
Nitrogen
1.18
1.47
1.78
1.77
Oxygen
11.63
10.09
9.05
9.32
Pyritic Sulfur
1.20
2.34
1.83
+
Organic Sulfur
3.00
0.92
1.94
+
Pyritic/Organics
Ratio
0.40
2.53
0.92
+
(no units)
Total Sulfur
4.26
3.23
3.77
1.17
Calorific Value
(kJ/kg)
29,797
31,971
31,314
31,425
Calorific Value
(Btu/lb)
12,606
13,526
13,248
13,295
* All values in wt % except where noted.
+ Measurements not taken.
4-2
-------
TABLE 4-2. SORBENT CHARACTERISTICS
Sorbent Ca(wt %) Median Particle Diameter (/an)
Kemidol
34.3
3.93
Linwood
49.9
2.88
Marblehead
50.1
4.33
Modified Marblehead
50.0
3.47
Snowflake
47.5
3.46
ISGS Mix
47.4
2.06
ISGS BH-16
*
1.90
ISGS BH-20
*
2.67
ISGS BH-22
*
2.24
ISGS BH-24
*
2.01
ISGS BH-29
*
1.68
ISGS BH-30
*
2.09
ISGS BH-31
*
2.01
ISGS BH-32
*
2.35
ISGS BH-33
*
2.08
ISGS BH-34
*
2.41
* Measurements not taken.
Furthermore, when the data from the three other coals (IBCSP #1, #6, and #9)
are viewed collectively, the SOg removal by individual sorbents does not
differ radically from coal to coal. In each case the relative standard
deviation of the mean S02 removal percentage (standard deviation of mean
removal divided by the mean) is less than 10 percent. This could be an
indication that the pyritic/organic sulfur ratio of each coal is the largest
coal-specific factor in FSI performance using the same sorbent.
The commercially available Ca(0H)2 sorbents (Linwood, Marblehead, and
Snowflake) all yield approximately the same values for SO£ removal percentages
when excluding the data from IBCSP #2. The sorbents hydrated under special
4-3
-------
80
KEMIDOL LINWOOD MARBLEHEAD MODIFIED SNOWFLAKE ISGS
MARBLEHEAD MIX
SORBENT NAME
COALS: ¦¦ BCSP #1 ^ BCSP #2 YZZ1 BCSP #6 BCSP
Figure 4-1. IFR Sorbent Reactivity.
-------
conditions (the 1ignosulfonate modified Marblehead and the ISGS alcohol
hydroxide) clearly exhibit superior performance. Past tests (Kirchgessner and
Jozewicz, 1988) attribute the enhanced performance of the modified Marblehead
to its ability to resist sintering at the high temperatures seen in FSI. The
performance of the ISGS sorbent may be related to its very small particle
size. Recent EPA in-house tests (Gullett et al., 1988) have demonstrated the
importance of sorbent particle size to sulfur capture. Mixing studies have
shown that, in many instances, sorbent injection takes place under conditions
likely to result in limitations on mass transfer rates of SO2 to the reacting
particle (Newton et al.,1988). In such a regime, ultimate sorbent reactivity
will be inversely related to the size of the reacting particle. More work on
such mixing phenomena is needed to find ways of injecting sorbent in such a
manner as to maximize reaction.
It is interesting to note that for these tests the dolomitic hydroxide,
Kemidol, did not outperform the purely calcium-based sorbents. Past tests
noted in the literature (Snow et al., 1986) indicate that Kemidol would be
expected to yield higher SO^ removal percentages than the calcitic hydroxides
at the same Ca/S ratios. In the tests shown in this work, removal percentages
for the dolomitic hydroxide were comparable to those seen for the calcitic
hydroxides and were not high enough to offset the increased solids loading on
particulate removal systems that necessarily accompanies FSI with dolomitic
sorbents due to the unreactive magnesium oxide portion.
The overall impression of the applicability of FSI as an S02 control
4-5
-------
technology for Illinois coals is positive. With the exception of IBCSP #2,
which gave lower results for unknown reasons discussed earlier, S02 removals
for each of the coal/sorbent tests approached or exceeded 60 percent at a Ca/S
ratio of 2:1. Indeed, tests with the specially modified sorbents routinely
exceeded 70 percent. These test results strongly recommend FSI as a cost
effective means of controlling SO2 emissions from coal-fired combustors.
4.2 GRAPHITE FURNACE REACTOR TESTS
Results from economizer temperature (538 °C) sorbent injection testing on the
GFR are shown in Table 4-3. The data show a clear inverse relationship to
sorbent particle size as measured using the sedigraph; as particle size
decreases, the conversion of the sorbent to the calcium sulfite product in the
GFR increases. Again, this is indicative of mass transfer resistances acting
to control the rate of reaction, rather than other potentially faster
TABLE 4-3. RESULTS FROM ECONOMIZER INJECTION TESTS ON GFR*
Sorbent
Mean Conversion (%)
Marblehead
Modified Marblehead
9.8 ± 0.7
11.0 + 1.1
11.7 ± 1.0
15.2 + 2.9
17.7 + 1.0
17.6 ± 1.1
19.9 + 2.2
15.3 + 2.5
Snowflake
Linwood
ISGS BH-20
ISGS BH-24
ISGS BH-29
Kemidol
* Data obtained from minimum of 10 runs at 538 °C, residence time = 0.75 s,
3,000 ppm S02, 5 percent 02, N2 balance.
-------
mechanisms such as inherent chemical kinetics. Removing these resistances
may show a faster true rate of reaction.
4.3 SHORT TIME DIFFERENTIAL REACTOR TESTS
The STDR is designed to operate under differential conditions with respect to
S02 concentration. Using small particles, such as in this work, along with a
high process gas throughput (resulting in a high gas velocity with respect to
the sorbent particles), serves to remove film layer and pore diffusion mass
transfer resistances as potential rate limiting steps for the sorbent/S02
reaction. Results from testing in the STDR with an SO2 concentration of
3,000 ppm using ISGS BH-29 sorbent are shown in Figure 4-2. Similar
conversions were obtained with Linwood hydroxide over the same time range.
These results predict an SO2 removal of roughly 55 percent for a 1 s residence
time and Ca/S ratio of 2:1 when injecting sorbent at or near 538 °C. This
removal percentage is slightly lower, than those reported in previous works
(Bortz et al., 1986). More work is needed to accurately quantify the
fundamental rate of the sorbent/S02 reaction under economizer injection
conditions using reactors like the STDR prior to predicting potential SO2
removal levels. The effects of parameters such as SO2 concentration, sorbent
surface area, and sorbent porosity on reaction rate have not been thoroughly
investigated.
4-7
-------
60 j 1 1 1 1 1
S?
Z 50 - . -
O
w .n
C 40 ¦ ¦
£ ¦
O 30 " I
o ¦
2 20 ¦ |i
LU
CQ ¦
o 10 "
O)
0 ' 1 1 1 1
0 2 4 6
TIME, s
Figure 4-2. STDR Alcohol Sorbent Reactivity
4-8
-------
4.4 X-RAY DIFFRACTION TESTS
It has been proposed that crystallite size can affect the gas/solid reactions
by modifying the interface between the two phases. It is further proposed
that crystal lattice strain could contribute to reactivity by decreasing the
stability of the solid and producing a source of activation energy from the
strain energy stored in the lattice (Briden and Natschke, 1988). Some of the
crystallite size and strain data for the samples analyzed, along with the XRO
peak half widths, are shown in Table 4-4. This table gives the percent
conversion data from the sorbents tested in the IFR. The percent conversion
data are reported in units of moles of Ca reacted divided by moles of Ca
available multiplied by 1,00. The complete XRD results and discussions are
contained in Appendix B.
The reactivities for IBCSP coals #1, #6, and #9 were quite similar. The
analysis of variance showed a significance level of 0.9 for the sorbents and
0.06 for the coals. Consequently, it was considered reasonable to average the
coal reactivities to increase the reliability of the sorbent characterization.
To test the hypothesis that the individual x-ray line broadening (XLB) factors
were related to reactivities, regression functions were derived using the
observed experimental reactivities for coals # 1, #6, and # 9. For the first
stage of this study, it was found that the best single estimator of IFR
reactivity was maximum column length, a value representative of the maximum
dimension within the distribution of dimensions measured in a crystallite.
4-9
-------
Table 4-4. ILLINOIS COAL REACTIVITY AND XRD PEAK SHAPE DATA
Kenidol
Linvood
Karblehead
Modified
Karblehead
Snowflake
ISGS Mix
Reactivity IBCSP il*
58.7
57.9
57.6
59.8
66.6
76.6
Reactivity IBCSP 12
42.6
38.1
57.3 ' "
49.5
47.7
58.8
Reactivity IBCSP #6
67.8
60.1
53.8
67. ' .
57.6
74.2
Reactivity IBCSP i9
52.7
52.3
63.6
75.5
57 .
69.4
Reactivity Avg #1, #6, #9
59.733
56.766'
58.333 ¦
67.433
'• 60.4
73.4
Avg Col Lnth**
9.8
11.5
9.9
8.7 ¦
15
11.5
Max Col Lnth
45
38 ;
41
32 ,
50
33
Max Freq Col Lnth
3.37
5.33
3.57
3.93 f
' 7.4V
6.53
Strain (Avg Col Lnth)
2.18 E-3
2.21 E-3
2.2 E-3--
2.34 E-3
1.71 E-3
2.22 E-3
Strain (Max Col Lnth)
1.16 E-3
1.34 E-3
1.25 E-3
1.49 E-3
9.1 E-4
1.38 E-3
Strain (Max Frq Col Lnth)
3.27 E-3
2.99 .E-3
3.26 E-3
2.84 E-3
.' 2.55 E-3
2.86 E-3
* Reactivities in units of SO, cecovai percentage at Ca/S = 2:1
** Lengths in units of nanometers, ni
In the next stage, regression equations were derived for relating the XLB
factors two at a time to the observed IFR reactivities. For the 15 pairs of
factors, the correlation coefficients varied from 0.13 to 0.79. The best pair
of estimators was maximum column length and the strain at maximum column
length with a correlation coefficient of 0.79. This value is considered quite
significant, considering its derivation was subject to coal and furnace
variability.
Since the increase in correlation was vastly improved by using two factors,
the third stage was to use three factors for the analysis. For triplets, the
correlation coefficient varied from 0.40 to 0.99. It would appear that it is
possible to almost completely characterize the microstructural relation to
reactivity with three XLB factors. The best correlation coefficient of 0.99
4-10
-------
was derived from the average column length, modal column length, and strain at
maximum column length.
Results and predictions obtained from the XLB regression data on the sorbents
evaluated can be seen in Figure 4-3. The three XLB factors mentioned above
appear to be the best estimators of reactivity from the number of samples
analyzed to date. The linear correlation coefficient of 0.99 for the relation
is conclusive and is considered reliable for ranking sorbent materials.
Future studies of other sorbents could further establish the reliability of
this method and its application to ranking sorbent reactivity without
undergoing large-scale testing.
4.5 EXTRACTION PROCEDURE TOXICITY TESTS
Results from the EP Toxicity tests are given in Table 4-5. As anticipated,
values for all of the regulated metals are below the RCRA limits. Sorbent
injection would appear to stabilize many of the metal species, particularly
arsenic and cadmium. While the final pH values are below RCRA limits, they
are high enough to elicit some concern. Methods for stabilizing the ash or
neutralizing leachate from the ash may bear investigation.
4-11
-------
[ ISGS MIX
/
/
/
/
/
— /
/
/
/
/
1 „ ' MODIFED MARBLEHEAD
~
/
/
/
/
1 KEMIDOLji SNOWFLAKE
/
¦-MARBLEHEAD
. /* UNWOOD
i i ' i i i i i'ii '''i ¦ ¦ ¦ i
55 60 65 70 75
PREDICTED CAPTURE, %
Figure 4-3. Reactivity Predicted by X-ray
Line Broadening Versus FR Data •
4-12
-------
TABLE 4-5. RESULTS OF EP TOXICITY TESTS ON IFR ASH
Parameter
EPA Method
RCRA Limit
(mg/L)
Coal #2
(mg/L)
Coal #6
(mg/L)
Coal #9
(mg/L)
Kemidol
(mg/L)
Linwood
(mg/L)
w
Maitolehead
(mg/L)
Modified
Marblehead
(mg/L)
Snowflake
(mg/L)
ISGS
Mix
(mg/L)
Spike
Percent
Recovery
Arsenic
206 2
5
0.049
0.034
0.069
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
109
Barium
200.7
100
026
0.21
0.26
029
0.15
0.12
0.12
0.12
0.093
87
Cadmium
2007
1
0.11
0.088
0.082
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
97
Chromium
200.7
5
0.043
0.19
0.36
0044
0.063
0.072
0.054
0.075
0 071
69
Lead
239.1
5
0.23
0.025
0.13
0.14
0.029
<0.001
<0.001
0.035
<0.001
104
Mercury
245.1
0.2
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
95
Selenium
270.2
1
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
71
Silver
200.7
5
0.032
<0.01
0.022
<0.01
<0.01
<0.01
<0.01
<001
<0.01
45
Final pH
12.5
5.1
5.0
4.9
12.4
11.9
122
11.8
11.9
11.8
NS*
* Not Sampled.
-------
5.0 CONCLUSIONS AND RECOMMENDATIONS
Pilot-scale testing of the S02 removal potential of FSI with Illinois Basin
coals showed that removals in excess of 60 percent can be readily achieved
using commercially available sorbents and a Ca/S ratio of 2:1. The ISGS
alcohol sorbent and the marblehead 1ignosulfonate modified sorbent gave
removals in excess of 70 percent. Lower removals were noted for the coal high
in pyritic sulfur (as opposed to organic sulfur). Further investigation is
necessary to verify and explain this phenomenon. The greatest removals were
seen using the ISGS alcohol hydroxide. It is believed that its performance is
enhanced by its small particle size and the resultant mixing benefits. More
investigation of both bench- and pilot-scale mixing phenomena is necessary to
understand how these parameters affect sorbent reactivity. Better injection
techniques could minimize particle size effects through greater mixing,
yielding increased removals for larger particles.
Testing of sorbent injection at economizer temperatures showed that removals
of roughly 55 percent at a Ca/S ratio of 2:1 can be expected. However, not
much is currently known about fundamental reaction kinetics for this
mid-temperature sorbent/S02 reaction. In order to more accurately predict the
full-scale performance of injection of sorbent in this temperature region,
more work is needed to clarify the effects of temperature, SO2 concentration,
and sorbent characteristics on reactivity.
XRD tests indicated that the sorbent microstructural characteristics of
5-1
-------
average column length, modal column length, and strain at maximum column
length can provide a basis for prediction of sorbent performance in FSI
applications. Verification of the predictive power of these parameters is
warranted through testing with additional sorbents.
Analyses of the FSI ash showed that it could be considered nonhazardous in
terms of RCRA limits for leaching of heavy metals. The pH of the leachate is
a concern, however, because of its alkaline nature. Further investigation
into methods of disposal or utilization of FSI waste is necessary.
5-2
-------
6.0 REFERENCES
Beittel, R., J.P. Gooch, E.8. Dismukes, and L.J. Muzio. 1985. "Studies of
Sorbent Calcination and SC^-Sorbent Reactions in a Pilot Scale Furnace."
Proceedings: First Joint Symposium on Drv SO^ and Simultaneous S0o/NQx
Control Technologies. Vol. 1, U.S. Environmental Protection Agency,
AEERL, Research Triangle Park, NC, EPA-600/9-85- 020a (NTIS PB85-
232353), p.16-1.
Bortz, S., and P. Flament. 1985. "Recent IFRF Fundamental and Pilot Scale
Studies on the Direct Sorbent Injection Process." Proceedings: First
Joint Symposium on Dry SQg and Simultaneous SCWNO^ Control
Technologies. Vol. 1, U.S. Environmental Protection Agency, AEERL,
Research Triangle Park, NC, EPA-600/9-85-020a (NTIS PB85-232353),
p. 17-1.
Bortz, S., V.P. Roman, R.J. Vang, and G.R. Often. 1986. "Dry Hydroxide
Injection at Economizer Temperatures for Improved SO2 Control."
Proceedings: 1986 Joint Symposium on Dry S02 and Simultaneous S02/N0..
Control Technologies. Vol. 2, U.S. Environmental Protection Agency,
AEERL, Research Triangle Park, NC, EPA-600/9-86-029b (NTIS PB87-
120457), p. 31-1.
Briden, F.E., and D.F. Natschke. 1988. "Calcium Hydroxide Sorbent Reactivity
with Sulfur Dioxide from X-Ray Diffraction Peak Profiles." Proceedings:
6-1
-------
First Combined FGD and Drv SOo Control Symposium. Vol. 3, U.S.
Environmental Protection Agency, AEERL, Research Triangle Park, NC, EPA-
600/9-89-036c (NTIS PB89-172175), p. 7-83.
Briden, F.E., and D.F. Natschke. 1989. "The Characterization of a Solid
Sorbent with Crystallite Size and Strain Data from X-ray Diffraction
Line Broadening." Accepted for publication in Adv. in X-rav Analvsis.
32.
Dahlin, R.S., C.L. Lishawa, and N. Kaplan. 1986. "Analysis of LIMB Waste
Management Options." Proceedings: 1986 Joint Symposium on Drv SOo and
Simultaneous SQ2/N0X Control Technology, Vol. 2, U.S. Environmental
Protection Agency, AEERL, Research Triangle Park, NC, EPA-600/9-86-029b
(NTIS PB87-120457), p.39-1.
Gullett, B.K., J.A. Blom, and R.T. Cunningham. 1988. "Porosity, Surface Area,
and Particle Size Effects of CaO Reacting with SO2 at 1100 °C." React.
Solids 6:263.
Kirchgessner, D.A., and W. Jozewicz. 1988. "Laboratory- and Commercial-Scale
Results of Testing Surfactant-Modified Sorbent." Proceedings: First
Combined FGD and Drv S02 Control Symposium. Vol. 2, U.S. Environmental
Protection Agency, AEERL, Research Triangle Park, NC, EPA-600/9-89-036b
(NTIS PB89-172167), p. 5-121.
6-2
-------
Newton, G.H., D.K. Moyeda, G. Kindt, J.M. McCarthy, S.L. Chen, J.A. Cole, and
J.C. Kramlich. 1988. Fundamental Studies of Dry Injection of Calcium-
Based Sorbents for SO2 Control in Utility Boilers. U.S. Environmental
Protection Agency, AEERL, Research Triangle Park, NC, EPA-600/2-88-069
(NTIS PB89-134142).
Snow, G.C., J.M. Lorrain, and S.L. Rakes. 1986. "Pilot Scale Furnace
Evaluation of Hydrated Sorbents for SO2 Capture." Proceedings: 1986
Joint Symposium on Drv SO^ and Simultaneous SOo/NO^ Control
Technologies. Vol. 1, U.S. Environmental Protection Agency, AEERL,
Research Triangle Park, NC, EPA-600/9-86-029a (NTIS PB87-120465), p.6-1.
6-3
-------
APPENDIX A
IFR DATA
A-1
-------
IFR DATA
Baseline Injected
Coal Sorbent SO2 (ppm) SO2 (ppm) Removal (%) Ca/S
IBSCP #1 Kemidol 3167
Kemidol 3167
Linwood 3468
Linwood 3468
Marblehead 3180
Marblehead 3180
Mod. Marblehead 3394
Mod. Marblehead 3394
Snowflake 3161
Snowflake 3161
ISGS Mix 3120
ISGS Mix 3120
IBSCP #2 Kemidol 2340
Kemidol 2340
Linwood 2469
Linwood 2469
Marblehead 2433
Marblehead 2433
Mod. Marblehead 2332
Mod. Marblehead 2332
Snowflake 2541
Snowflake 2541
ISGS Mix 2288
ISGS Mix 2288
IBSCP #6 Kemidol 2817
Kemidol 2817
Linwood 2746
Linwood 2746
Marblehead 2722
Marblehead 2722
Mod. Marblehead 2788
Mod. Marblehead 2788
Snowflake 2918
Snowflake 2918
ISGS Mix 2862
ISGS Mix 2862
IBSCP #9 Kemidol 882
Kemidol 882
Linwood 1075
Linwood 1075
Marblehead 908
Marblehead 908
Mod. Marblehead 869
Mod. Marblehead 869
Snowllake 1032
Snowflake 1032
ISGS Mix 960
ISGS Mix 960
1866 41.1 1.12
1066 66.3 2.25
2403 30.7 0.87
1703 50.9 1.75
2302 27.6 0.91
1504 52.7 1.83
2186 35.6 0.93
1508 55.6 1.86
2249 28.8 0.85
1371 56.6 1.70
1979 36.6 0.79
1204 61.4 1.58
1632 30.2 1.12
1170 50.0 2.24
1852 25.0 0.94
1536 37.8 1.91
1560 35.9 0.92
1139 53.2 1.85
1765 24.3 0.91
1271 45.1 1.82
1890 25.6 0.88
1472 42.1 1.75
1538 32.8 0.85
1136 50.3 1.70
1488 47.2 1.12
659 76.6 2.25
1546 43.7 1.15
842 69.3 2.30
1517 44.3 1.03
1187 56.4 2.07
1436 48.5 1.11
702 74.8 2.22
1852 36.5 1.10
1058 63.7 2.21
1151 59.8 1.07
610 78.7 2.14
498 43.5 1.33
260 70.5 2.66
766 28.7 1.06
464 56.8 2.12
624 31.2 0.87
404 55.5 1.74
533 38.7 1.08
144 83.4 2.16
611 40.8 0.92
488 52.7 1.84
508 47.1 1.16
186 80.6 2.32
A-2
-------
APPENDIX B
PREDICTING S02 SORBENT REACTIVITY IN THE
INNOVATIVE FURNACE WITH X-RAY DIFFRACTION PEAK PROFILES
FOR THE
ILLINOIS CENTER FOR RESEARCH ON SULFUR IN COAL
B-l
-------
PREDICTING S0? SORBENT REACTIVITY IN THE
INNOVATIVE FURNACE WITH X-RAY DIFFRACTION PEAK PROFILES
FOR THE
ILLINOIS CENTER FOR RESEARCH ON SULFUR IN COAL
INTRODUCTION
It was observed early in the history of x-ray diffraction (XRD) analysis
that the spectral peaks were not discrete lines as the Bragg equation would
predict for a perfect crystal lattice. Instead, the intensities decrease
gradually from the maximum at the Bragg angle. XRD line broadening (XLB)
comes about from decreasing crystallite (grain size), increasing crystal
lattice distortion, and instrument effects.
It is apparent that crystallite size can affect gas/solid reactions by
modifying the interface between the two phases. It has been further proposed
that crystal lattice strain could contribute to reactivity by decreasing the
stability of the solid and providing a source of activation energy from the
strain energy stored in the lattice. The first attempt to relate sorbent
conversion to crystallite microstructure factors was reported by Briden and
Natschke (1).
THEORY
X-ray powder diffraction (XRD) is today, the most powerful technique
available for the identification of chemical compounds or phases present in
B-2
-------
solid materials. XRD is possible because of the geometric order of solids.
The degree of order can vary from almost nothing, as in glasses, to very high
order in materials such as quartz. Even such amorphous materials as glass,
rubber, and petroleum jelly do exhibit some order which is manifested as XRD
spectra with a few very broad peaks. Most solids show sufficient XRD spectra
to enable their identification by comparison to standard recorded spectral
files. Unknowns consisting of pure phases or mixtures can be identified by
executing systematic searches against a standard data base by manual methods
or with computer assistance. As a general rule, the limits of detection for
individual components of a mixture are about 5 percent.
The XRD spectra, as mentioned earlier, arise from the order within a
solid material. From the angle of reflection, 8, of an x-ray beam, with
wavelength X (1.54056 A for copper Kal radiation), one can determine the
spacing between the planes, d, from the Bragg equation, which states that the
sine of the angle of diffraction is equal to the reflection order times the
wavelength divided by two times the d spacing.
If a crystalline material were perfect and every respective distance, d,
for each family of planes in the material were precisely repeated (and if the
x-ray measuring device were perfect), all of the x-ray intensity would, for
each given d space, be reflected only at the respective Bragg angle 8.
However, no crystalline material is perfect (true also for x-ray measuring
devices); consequently, the reflected x-ray intensity for the d value does not
all occur as an infinitesimally narrow line but is distributed in a pattern
B-3
-------
with maximum intensity at the Bragg angle and decreasing in both directions
going away from the Bragg angle.
Crystalline materials can be imperfect in a number of ways. Most solid
materials are not one single contiguous crystal lattice, but are composed of
agglomerations of many small crystallites. The crystallite is the largest
uninterrupted crystal unit. When crystallites are smaller than 1000 A, the
width of diffraction peaks starts to increase. This enables a measurement of
crystallite dimensions down to a limit where they approach the dimensions of
the x-ray radiation.
Another type of imperfection which can occur in a crystal is distortion.
Distortion is caused by significant variations in the d spacing. This can
happen when an alien species is introduced into the crystal lattice; e.g.,
potassium impurity in sodium chloride. It can also come about from other than
ideal crystallization conditions such as rapid temperature change or the
presence of surface active agents. Other types of distortion can occur from
slippage of crystal planes which takes place during cold working of ductile
materials, such as metals.
It has been described how the profile of the XRD peak is affected by
three factors: the instrument effects, the crystallite size, and the
distortion within the crystal lattice. The first attempt to relate peak shape
to crystallite size was by Scherrer (2), who derived the following expression
for the peak half width B, in radians, as a function of the radiation
B-4
-------
wavelength, the crystallite dimension L, in A, and the Bragg angle:
B (28) = 0.94/L cos 0
It has been postulated that peak broadening due to small crystallite
size is of the Cauchy or Lorentzian form while peak broadening due to lattice
distortion is of the Gaussian form. Several techniques for separation of size
and strain effects attempt to utilize this property (Schoening, 3). These
methods have an advantage in needing XRD data for only one peak but are
limited in the amount of information given and theoretical rigor.
The most powerful method for separation of size and strain effects
available, at present, is that originally proposed by Warren and Averbach (4),
Warren (5), and Klug and Alexander (6). According to Warren, the distribution
of diffracted power can be expressed as a Fourier series. If only symmetrical
contributions to peak broadening are considered, then only two cosine terms of
the peak profile function need to be considered. One cosine coefficient is
independent of XRD peak order and is related to crystallite column length
while the other coefficient is dependent on XRD peak order and is related to
crystallite strain. Before the size and strain effects can be separated the
instrument effects must be removed. To do this it is first necessary to
remove the asymmetry of the experimental XRD peak which is caused by the Ka2
peak contribution to the larger Kal peak. Fourier coefficients are then
derived to fit the corrected experimental data. The instrument effects on
line broadening can be removed by dividing the Fourier coefficients of the
B-5
-------
sample XRD peak by those of a standard material which has relatively minimal
line broadening from size or strain. Finally, a set of Fourier coefficients
are available which contain only size and strain effects.
The Warren-Averbach method for separating the size and strain components
entails the determination of Fourier coefficients to fit the diffraction
peaks. This calculation was highly complicated before the development of
modern computers. Today the necessary calculations take relatively little
time. The Ka2 interference is then corrected and the instrument effects are
removed. Finally the size and strain effects are separated through
development of equations where the strain is derived from the slope of the
expression while the size comes from the Y intercept. From these data come
the frequency of occurrence of column lengths and mean squared strains as a
function of column length.
EXPERIMENTAL
Six samples of sorbent were evaluated with four different coals in the
innovative furnace reactor operating as described earlier.
The experimental data were taken on a Siemens D-500 diffractometer. A
copper x-ray tube was used running at 50 kV and 40 mA. A 1 degree entrance
aperture and 0.05 degree receiving si it were used. A scintillation detector
equipped with a graphite monochrometer was used. The sample powders were
packed in a side loading cell. The diffractometer was run and the data were
B-6
-------
processed with the Siemens DIF 500, version 1.1 operating system.
At the beginning of the study, a complete ,XRD scan was taken on a sample
of the sorbent material. The phase was identified as JCPDS No. 4-733.
Because of the relative high intensities and their isolation, the first (101)
and second (202) order peaks seen at the two-8 angles of 34.1 and 76.8
degrees were chosen for the first Warren-Averbach analysis. Zinc oxide (JCPDS
No. 36-1451) was chosen as the standard, to account for instrument line
broadening, because it had peaks close by (36.26 and 76.77 degrees) which were
very narrow. Before running the Warren-Averbach analysis, the experimental
XRD peak data were fitted to one of several available functions. The adoption
of this step has been an extremely important advance in peak analysis because
it smooths out random noise and removes smaller interfering peaks. Without
very smooth data it is not possible to get meaningful results, according to
Zorn (7) and Langford and Delhez (8). The split Pearson VII function was used
for our data, allowing flexibility in fitting complex peaks ranging from pure
Gaussian to pure Lorentzian.
RESULTS AND DISCUSSION
The Warren-Averbach analysis provides much information about the
microstructure. The major properties delineated are:
1. Average crystallite size.
B-7
-------
2. Frequency of occurrence of column lengths as a function of column
length, and the column length of maximum frequency of occurrence
(or modal column length).
3. Half width of the frequency of occurrence of column lengths as a
function of column length.
4. Maximum determinable column length.
5. Strain at the average column length.
6. Strain as a function of column length.
7. Strain at the column length.of maximum frequency of occurrence.
8. Strain at the column length of maximum determinable column length.
Some of the crystallite size and strain data, along with the XRD peak
half widths, are shown in Table B-l. This table gives the percent conversion
data from the reactions of six Ca(0H)2 sorbent samples with four coals plus
the average of three coals. It was seen that the reactivities for coals 1, 6,
and 9 were quite similar. The analysis of variance showed a significance
level of 0.9 for the sorbents and 0.06 for the coals. Consequently, it was
considered reasonable to average the coal reactivities to increase the
reliability of the sorbent characterization to test the hypothesis that the
B-8
-------
TABLE B-l. ILLINOIS COAL REACTIVITY AID XRD PEAK SEAPE DATA
Modified
Keaidol
Linvood
Harblehead
Harblehead
Snovflal?
:scs
Reactivity IBCSP 1
58.7
57.9
57.6
59.8
66.6
76.6
Reactivity IBCSP 2 +
42.6
38.1
57.3
49.5
47.7
58.8
Reactivity IBCSP 6 +
67.8
60.1
53.8
67
57.6
74.2
Reactivity IBCSP 9 +
52.7
52.3
63.6
75.5
57
69.4
Reactivity Avg 1,6,9+
59.733
56.766
58.333
67.433
60.4
73.4
Avg Col Lnth*
9.8
11.5
9.9
8.7
15
11.5
Max Col Lnth*
45
38
41
32
50
33
Kax Pre
-------
individual XLB factors were related to the reactivities. Regression functions
were then derived to convert each of the factors to reactivity, using the
observed experimental reactivities for coals 1, 6, and 9. The percent
conversion data are reported in units of moles of Ca reacted divided by moles
of calcium available times 100.
In a previous study, it had been found that each of the factors listed
above was related to reactivity (Briden and Natschke, 1). In that study, the
same starting materials were used, and they varied only by the concentration
of an additive. Conditions were more precisely controlled in a flow reactor
with SOj supplied in a gaseous stream to sorbent material suspended on quartz
wool. In the current study, involving a number of quite dissimilar sorbents,
the same simple factors were not sufficient to predict reactivity. The linear
correlation coefficients (LCC) for each factor are shown in Table B-2. The
regression function used was Y = AX + B where Y was observed reactivities,
A was the regression coefficient, X was the XLB factor, and B was the
regression equation constant. From the table it is seen that the best
estimator of reactivity was maximum column length with an LCC of only 0.41.
Under some conditions, with some types of data, some investigators might
consider an LCC of 0.41 as having some significance, but a higher LCC was
desired, in this case, to accept the possibility of a relationship.
For the data reported on in the previous study (Briden and Natschke, 1)
the LCCs were computed after publication. The LCCs varied from about 0.62 to
0.94 for single factors. However, when pairs of factors were used in the
B-10
-------
iable b-2. derived regression functions for individual xrd peak sbape factors
Linear'
Regression
Regression
Correlation
Function
Function
Coefficient
Coefficient
Constant
Average Column Length*
¦ 0,02
-0.38
66.9
Maxima CoIumi Length*
0.41
-0.589
86.2
Modal Column Length*
• 0.07
1.03
57.5
Strain at Avg. Coluin Length
a. 07
7830.
45.9
Strain at Max. Coluin Length
0.22
14800.
44.2
Strain at Mod. Coluin Length
0.12
-8130.
86.9
* Lengths in units of nanoneters, na
-------
regression, LCCs ranging from 0.721 to 0.993 were observed. When triplets of
factors were used, the LCCs ranged from 0.94 to 0.993. Since the use of more
factors offered better correlation, presumably because of the consideration of
more microstructure factors, it was obvious that the same might help for the
current study.
In the next step, regression equations were derived for relating the XLB
factors, two at a time, to the observed reactivities. For the 15 pairs of
factors, the LCCs varied from 0.13 to 0.79. The regression equation used was
Y = AXj + BX2 + C where again Y was observed reactivities, A and B were the
respective coefficients for the factors, Xj and were the XLB factors, and C
was the equation constant. The results are shown in Table B-3. The next best
pair of estimators was average column length and modal column length, with an
LCC of 0.78. These values were considered quite significant, considering
their derivation was subject to variation in the coal and furnace variability.
Since the increase in correlation was vastly improved by using two factors
instead of one, it was natural to go on to three factors for the analysis.
The results are given in Table B-4. For triplets of factors, the LCCs varied
from 0.40 to 0.99. It would appear that it is possible to almost completely
characterize the microstructure contribution to reactivity with three XLB
factors. The best LCC of 0.99 came from the parameters average column length,
modal column length, and strain at maximum column length. A plot showing the
regression of these data and the observed reactivity data is given in Figure
B-l.
B-12
-------
TABLE B-3. .DERIVED REGRESSION FUNCTIONS FOR PAIRS OF
XRD PEAK SEAPE FACTORS
Function Variable Coefficients
Average
Max i mii
Modal
Strain
Strain
Strain
Linear
Column
Colon
Coluui
at Average
at HaxituE
at Hodal
Function
Correlation
Lenath*
Length*
Lenqth*
Colutn Lenath
Colura Lenath
Column Lenath
Constant
Coefficient
1.19
-0.89
—
—
—
—
82.1
0.52
-5.49
—
7.43
—
--
—
86.1
0.78
2.18
—
—
27900
--
—
-21.3
0.16
1.50
—
—
—
27100
—
12.0
0.33
-2.08
—
—
—
-19700
144.0
0.39
—
0.67
1.6
—
—
—
81.3
0.58
—
:.i8
—
-22700
—
—
158.0
0.61
—
-2.82
—
~
-78900
274.0
0.79
—
-0.614
—
—
—
- 9220
114.0
0.57
__
3.41
26200
--
—
-10.6
0.49
—
—
2.14
—
22270
—
23.9
0.48
—
—
-0.303
—
—
- 9740
93.1
0.13
—
—
-32000
46900
—
72.5
0.40
--
--
--
21800
—
-18700
71.2
0.49
—
--
--
18900
- 1180
73.8
0.46
* Lengths in units of nanoneters, nc
B-13
-------
IABLE B-4. derived regression fuhciiohs for triplets of
XRD PEAK SEAPE FACTORS
Function Variable Coefficients
Average
Haxiiui
Modal
Strain
Strain
Strain
Linear
Column
Cohan
Column
at Average
at Maxitur
at Hodal
Function
Correlation
Lenath*
Lenath*
Lenath*
Column Lenath
Column Lenath
Coluin Lenath
Constant
Coefficient
-16.3
1.6
18.6
—
--
--
85.1
0.96
- 2.2?
-1.5
—
-51900
—
—
259.
0.67
- 2.56
-4.45
—
—
-154000
—
462.
0.93
0.011?
-0.616
—
—
—
- 9164
114.
0.57
-10.7
—
10.3
-35160
—
«
205.
0.90
-13.8
—
14.8
—
- 47200
—
201.
0.99
- 5.47
—
7.32
—
—
610
88.2
0.78
- 0.67
—
—
-42050
515000
—
95.7
0.40
0.12
—
—
22770
—
-18480
67.2
0.49
- 0.504
—
—
—
15500
-13920
90.
0.46
--
-1.27
-0.32
-26650
~
..
172.
0.61
—
-4.99
-2J1
—
-161000
--
477
0.90
—
-0.65
1.08
—
—
- 3700
94.3
0.59
—
-4.12
—
33880
-156400
—
350.4
0.87
—
-1.62
--
-40490
--
8556
189
0.62
—
-3.45
—
~
-102000
5330
312
0.81
—
—
2.91
15570
9320
—
2.9
0.49
—
—
1.98
26490
—
-10670
27.4
0.55
—
—
1.46
„
21540
- 4720
42.3
0.50
--
—
—
88690
- 60530
-39340
64.8
0.52
* Lengths in units of nanoieters, ni
-------
75
70
ISGS MIX Q, <
/
/
/
/
/
/
/
/
/
65 -
„ ' MODIFIED MARBLEHEAD
60 _ KEMIDOLji SNOWFLAKE
/
-MARBLEHEAD
55
UNWOOD
I L—l I 1— I l_
J 1—1 l_
55
60
65
70
75
PREDICTED CAPTURE, %
Figure B-1. Reactivity Predicted by XL£ versus IFR
B-15
-------
All of the foregoing regressions were derived using the average
reactivities for coals 1, 6, and 9. Since the average reactivities were used,
the question arises as to what results would be obtained if the same
regression function were used to estimate the reactivities of individual
coals. The correlation coefficients for the individual coals are given in
Table B-5. The LCCs for coals 1, 6, and 9 for the two best triplets of XLB
factors were very good: while those of coal 2 did not indicate any significant
relation at 0.19 and 0.33. Of course, coal 2 was a high pyrite coal and
therefore had a different range of reactivities, making it impossible to fit.
CONCLUSIONS
The testing of sorbent materials with pilot scale and even bench scale
facilities is quite resource intensive. A reliable laboratory method is
needed to evaluate the numerous sources and treatments for calcium hydroxide
sorbents so they can be optimized. Even after the best sorbents have been
characterized, it will still be necessary to utilize the method for quality
control. An effective testing method should be capable of ranking candidate
sorbents in a reliable reproducible manner in any properly equipped
laboratory. The ranking should be independent of gas stream makeup, reactor
configuration, system temperature, and other variables. The ranking should be
based on the microstructural properties of the material alone. The described
technique involves the determination of reactivity from the regression of the
samples' three XLB factors over a range of experimentally observed
reactivities. The XLB factors of average column length, maximum column
¦ J-U
-------
TABLE B-5. REACTIVITY PREDICTION FROM THE TWO BEST XLB REGRESSION
FUNCTIONS DERIVED FROM THE OBSERVED REACTIVITIES
OF COALS #1, #6, and #9 APPLIED TO THE INDIVIDUALS COALS
IBCSP Coal
Correlation Coefficient
Average Column Length
Modal Column Length
Maximum Column Length
Average Column Length
Modal Column Length
Strain Modal Column Length
1
2
6
9
Average #1, #6, #9
0.86
0.19
0.82
0.63
0.96
0.89
0.33
0.72
0.68
0.99
B-17
-------
length, and strain at the modal column length, appear to be the best
estimators of reactivity from the number of samples analyzed so far. The
linear correlation coefficient of 0.99 for the relation is very strong.
Even though the relation between reactivity and the three factors has
been established, further work is needed. First, it is difficult to gain
access to samples which have well documented reactivity. Every opportunity
should be taken to run further XLB studies as more sample sorbent sets are
tested for reactivity. The availability of more samples could show another
triplet of XLB factors to be better. The processing of more samples, in any
case, can further establish the reliability of the method. More data could
also lead to the development of theory to explain the true physical meaning of
the column length and mean squared strain. A new method for the
interpretation of the Warren-Averbach mean squared strains has been proposed
by Turunen et al. (9). The application of this theory to our data could lead
to a better understanding of reactivity enhancement.
This is a new technique for the evaluation of the reactivity of solids
for gases. It should be applicable wherever it is needed to optimize the
reaction between solids and gases.
-------
REFERENCES
1. Briden, F.E., and D.F. Natschke. 1989. "The Characterization of a
Solid Sorbent with Crystallite Size arid Strain Data from X-ray
Diffraction Line Broadening." Accepted for publication in Advances in
X-rav Analvsis. 32.
2. Scherrer, P. 1918. "Bestimmung der Grosse und der Inneren Struktur von
Kolloidteilchen Mittels Rontgenstrahlen." Nachr. Gottinger Gesell. 2,
98.
3. Schoening, F.R.L. 1965. "Strain and Particle Size Values from X-ray
Line Breadths." Acta Crvstanoaraphica. 18 (5), 975.
4. Warren, B.E., and B.L. Averbach, 1950. "The Effect of Cold Work
Distortion on X-ray Patterns." J. Appl. Phvs. 21, 595.
5. Warren, B.E. 1969. X-rav Diffraction. Addison Wesley, Reading, MA.
6. Klug, H.P., and L.R. Alexander. 1974. X-rav Diffraction Procedures
for Polvcrvstalline and Amorphous Materials, 2nd ed. Wiley, New York.
7. Zorn, G. "Pitfalls in the Evaluation of Diffraction Line Shape."
Private communication, submitted for publication. Siemens, A.G.,
Corporate Research and Development, Munich, FRG, October 1987.
B-19
-------
8. Langford, J.I., and R. Delhez. "Profile Analysis for Microcrystalline
Properties by the Fourier and Other Methods." Private communication,
submitted for publication. Department of Physics, University of
Birmingham, Birmingham, UK, October 1987.
9. Turunen, M.J., T.H. Oekeijsen, R. Delhez, and N.M. Van der Pers. 1983.
"A Method for the Interpretation of the Warren-Averbach Mean-Squared
Strains and Its Application to Recovery in Aluminum." J. Add!. Crvst.
16, 172.
B-20
-------
APPENDIX C
QUALITY CONTROL EVALUATION REPORT
C-l
-------
QUALITY CONTROL EVALUATION REPORT
This research was sponsored by the Illinois Center for Research on Sulfur
in Coal (CRSC) and did not formally require adherence to Quality Assurance
guidelines as established by the Air and Energy Engineering Research Laboratory.
Nonetheless, measurement activities were conducted as if the research fell under
Category III project requirements. These procedures for instrument operation
and quality control checks are detailed in the Quality Assurance Project Plan:
LIMB Support Laboratory, prepared under EPA Contract No. 68-02-4701 by Acurex
Corporation, dated May 2, 1988, and the Qual ity Assurance Project Plan: Limestone
Injection Multistage Burner Furnace, prepared under EPA Contract No. 68-02-3988,
dated May 5, 1985.
Discussion of Data Quality
The data quality will be discussed below for each measurement activity in
terms of precision, accuracy, completeness, representativeness, comparative
value, and corrective action. The data quality objectives for the major
laboratory measurements are included in Table C-l. Four criteria were used to
validate data integrity: 1) acceptable performance during calibration and
accuracy tests; 2) internal consistency among observations when plotted
graphically against time for a given set of conditions; 3) consistency with prior
data; and 4) agreement among results obtained through independent analytical
methods.
Values included in Table C-l were calculated in the following manner:
Accuracy = mean measured value - known value x jqq
known value
with the exception of the Innovative Furnace Reactor (IFR) coal and sorbent feed
rates where the desired value is substituted for the known value:
Precision = standard deviation of the mean x
mean measurement
Completeness ¦ amount of valid frt* x 10„
amount of total data
Since the SO? concentration is a continuous measurement, completeness has
no meaning as a data quality indicator for this parameter.
The IFR coal and sorbent feeders were cal ibrated over the operational range
prior to starting testing, then checked at the desired setting before and after
each run. Checks were performed by measuring the weight of sample fed into a
tared beaker as a function of time.
The SO2 monitor was calibrated over the range of operation using three
certified gases of known SO2 concentration and a zero gas prior to the start of
testing. The instrument was zeroed and spanned daily before and after testing,
with the drift of the span indicated in the table.
C-2
-------
Table C-l Data Quality Objectives
Coal feed rate
(screw rate)
Sorberit feed rate
(screw feeder)
^ SO? concentration
(ultraviolet)
Sulfur (ion chromato-
graphy)
Calcium (atomic absorp
tion spectroscopy)
Particle size (x-ray
sedimentation)
Surface area (N2
adsorption)
* NA = Not applicable
Accuracy
Goal Achieved
1%JL __UU
±10 8.2
±10 1.4
±10 5.0
± 5 2.7
± 5 3^5
±10 9.5
±10 2.3
Precision
Goal Achieved
i%j_ -m
±10 1.0
±10 6.5
±10 5.0
± 5 1.5
± 5 2.5
±8 1.4
+10 5.4
Completeness
Goal Achieved
i%l_ _1%J
90 100
90 86
NA* NA
90 100
90 100
90 100
95 100
C-3
-------
Sulfur was measured as sulfate on the ion chromatograph calibrated with
three standard solutions made from a commercially available stock (Fisher
Scientific). A spiked sample made independently from analytical grade NapSO^
was run daily to monitor data quality along with daily laboratory blanks. Data
quality for calcium determinations by atomic absorption spectroscopy was
calculated in the same manner using a calcium stock solution (Fisher Scientific)
and spikes made from analytical grade calcium carbonate.
Quality control checks on the x-ray sedigraph were done by running a
reference garnet sample with an independently determined mean particle diameter
of 2.4 iim supplied by Micromeritics. Surface area quality control checks were
performed using a sample from Duke Scientific with a known surface area of 24.3
m/g.
All goals for data quality were met with the exception of completeness for
the IFR sorbent feed. Problems with clogging of the injection probe necessitated
the repeat of four tests.
Limitations on the Use of the Data
Due to inherent variability in reactor systems' time/temperature regimes,
removal percentages reported in this work should not be considered as predictive
of actual performance in full scale applications. The data should be considered
as indicative of probable performance trends on the full scale. Likewise, values
stated for coal or sorbent specific parameters (surface area, sulfur, or calcium
weight percent) are not to be considered as representative of all materials
produced by the respective manufacturers, but of the individual batches used in
this work.
Significant QA/QC Problems
As detailed in an earlier section, plugging of the IFR sorbent injection
probe led to rejection of data from four tests and necessitated their being
repeated. The sorbent injection probe is currently being redesigned to enlarge
the inner diameter, making clogging less likely. Procurement of a new sorbent
feeder to improve data precision is also underway. As suggested by earlier
researchers, the furnace was deslagged on alternating days by operation of a mid-
furnace combustion lance. However, no effect on data was seen as a result of
such action.
C-4
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. 2.
EP.A-600/2-89-065
3. RECIPIENT'S ACCESSIOf* NO.
4. TITLE AND SUBTITLE
Furnace Sorbent Reactivity Testing for Control of SC2
Emissions from Illinois Coals
5. REPORT DATE
December 1989
6. PERFORMING ORGANIZATION CODE
7. AUTHOR®
Brian K. Gullett and Frank E. Briden
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING OROANIZATION NAME AND AODRESS
See Block 12
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
NA (Inhouse)
12. SPONSORING AGENCY NAME ANO ADDRESS
EPA, Office of Research and Development
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final: 10/88 - 8/89
14. SPONSORING AGENCY CODE
EPA/600/13
15. supplementary notes project officer is Brian K. Gullett, Mail Drop 4, 919/541"
1534.
i6. abstract xhe report gives results of an evaluation of the potential of furnace sorbent
injection (FS1) for sulfur dioxide (SC2) emission control on coal-fired boilers burning
coals indigenous to Illinois. Tests were run using four coals from the Illinois Basin
and six calcium hydroxide--Ca(CH)2--sorbents, including one provided by the Illinois
State Geological Survey (ISGS). The evaluation included pilot- and bench-scale sor-
bent reactivity testing, sorbent microstructure characterization, and injection ash :
characterization. Pilot-scale FSI testing gave SG2 removal greater than 60%, with
some tests (including those with the ISGS sorbent) exceeding 70% removal for Ca/S'
ratios of 2:1. Bench-scale testing of injection at economizer temperatures (538 C)
yielded comparable removals of about 55%. X-Ray diffraction (XRD) tests of the sor-
bents showed a strong correlation between three measured crystallite microstruc-
tural parameters and sorbent reactivity in the FSI tests. Extraction . procedure (EP)
toxicity tests with the sorbent injection ash gave values well below Resource Conser-
vation and Recovery Act (RCRA) limits for regulated metals. ./•
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. 1DENTIF1 ERS/OPEN ENDED TERMS
c. COSATl Field/Group
Pollution Calcium Hydroxides
Coal
Combustion
Sulfur Dioxide
Emission
Sorbents
Pollution Control
Stationary Sources
13B
21D
21B
07B
14 G
11G
19. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
r 64
20 SECURITY CLASS (Thispage)
Unclassified
22. Pffl'"- - -
EPA Form 2220-1 (9-73)
i
------- |