-------
cn
i/t
0)
•o
•r-
s-
o
\ \ \\Ss
0.5
70
80 90 TOO
Average Pond Temperature (°F)
Figure 4.9, Fluoride Emission Rates for Ponds with Water Containing
0.628 g moles/liter Fluorides V,g = Wind Speed at 16
Meters in Meters Per Second.
(Source: King 1974).
117
-------
4.3.5 Field Verification of Fluoride Emission Factor
4.3.5.1 Verification/Calibration of Dispersion Models
Atmospheric dispersion modeling is an attempt to mathematically
simulate the transport, diffusion and transformation processes that
occur in the atmosphere. Models have been developed for elevated
and ground-level point sources, line sources and area sources.
In general, all dispersion models estimate ground-level concen-
trations of substances emitted into the atmosphere. The Gaussian
plume model is a theoretical treatment commonly used in dispersion
modeling. The basic formulation of the Gaussian equation assumes
that pround-level concentrations are inversely proportional to the
mean wind speed. Vertical and horizontal dispersion is simulated
by the use of standard deviations of plume concentration distribu-
tion for various stability categories, as determined from experi-
mental studies.
To be confident that dispersion model estimates are representative
of the real world, validation/calibration is necessary. Theore-
tical mathematical models have limitations which cause estimated
concentrations to be in error. The availability and accuracy of the
input data to the model and the accuracy of the mathematical algo-
rithm are the significant influences on the accuracy of the model
estimates. Similarly, calibration of the dispersion model is affect-
ed by the location, exposure and representativeness of the air
sampling sites and by the accuracy of the air quality data. In
validation or calibration, known pollutant emission rates, plume
118
-------
characteristics and meteorological data are combined to calculate
ground-level concentrations for specific time periods and locations,
The time periods and locations correspond to those for which am-
bient concentration measurements were obtained or are available.
The calculated and measured concentrations are compared and a
statistical relationship obtained (normally a linear regression
analysis). If the statistical relationship is significant the
model is calibrated and the relationship obtained is applied to all
future concentration estimates by the model,
4.3.5.2 Prediction of Fluoride Concentrations Downwind of Gypsum Ponds
Kinq developed a computer model whereby the simulation of dispersion
of fluoride emissions from gypsum ponds was attempted. Area source
emissions from the gypsum ponds were represented by a number of line
source emissions. The explicit form of the Gaussian plume equation
for a finite line source was utilized to describe dispersion of
emissions from the line sources. The Gaussian finite line source
equation is as follows:
_ 2g D Z2] Y rp2 1
X (X, 0, Z) = v/7ircrz U exp |_ 2 az J x J pl >J?Tr 6xp (-0.5 p2) dp (4.37)
where: x= concentration at (X, 0, Z)
X ^ downwind distance
Z = receptor height
q = line source emission rate per unit length
U = mean with wind speed
o, crz = horizontal and vertical dispersion parameters
119
-------
pl= Y'y
P2 " VCTy
Y!> y£ = limits of line source
The dispersion parameters, ay and az, are functions of downwind
distance as given by Turner (Turner, 1970).
Uncertainties enter into the Gaussian finite line source equation
through several parameters. These parameters include source emis-
sion rate (q), the dispersion parameters (ay and az) and the mean
wind speed (U). Accurate measurement of the mean wind direction is
also important in verification, since this establishes the source-
receptor relationship used in the model for concentration calcula-
tions. Proper verification/calibration of the dispersion model
requires an accurate knowledge of these parameters, and the use-
fulness of the model will depend upon the uncertainty associated
with these parameters.
Uncertainties in specifying the fluoride emission rate from the
pond contributes to errors in the model concentration estimates.
Since pollutant emission rate (q) enters directly into equation
18, the magnitude of the imposed error in the model due to Inac-
curacies would be equal to the error in the emission rate equation.
As described in Section 3.3.4, the emission rate equation is at
best accurate only within ± 90% at the 95% confidence level.
Another source of uncertainty in the dispersion model is in the
specification of the horizontal and vertical dispersion parameters,
ay and oz. King's use of ay and az is within the assumptions under
120
-------
which the parameters were experimentally determined, i.e,, sampling
time of about 10 minutes, height; the lowest several hundred meters
of the atmosphere, and the surface relatively open country. Even
so, for distances of travel up to a few hundred meters from the source
for all stabilities, oz may be expected to be correct only within
a factor of two (Turner, 1970).
King continuously recorded wind speed and direction with on-site
meteorological instruments in his field work. Type or manufacturer
of the instruments was not mentioned in King's work. Assuming that
the measurement instruments were properly calibrated and that they
meet Atomic Energy Commission guidelines for wind measuring devices,
they can be expected to measure wind speed within 0.5 mph and wind
direction within ± 5 degrees. With this degree of accuracy, the
wind speed measurements would be known within i 7 percent (at 7 mph
wind speed). These uncertainties in the wind speed and direction
specification in the dispersion model would add to the uncertainty
in concentration estimates by the model, and would affect the compar-
ison of those estimates with measured concentration values.
Errors are inherent in simulating an area source by a number of finite
line sources (refer to Figure 4.10). The simulation error lies in the
initial emission density or apportionment. Area source emissions are
initially emitted into a large volume of air since the emissions are
apportioned over an entire area. By dividing the area source into a
number of segments and concentrating each segment's emissions into line
source emissions in the simulation model, the emission density is altered
121
-------
Mean Wind
Direction
I
*O
X
I L
segment boundary
line source
inure 4.10. Line Source Simulation of a Rectangular Area Source.
122
-------
The magnitude of error introduced due to this technique is dependent
on the line source spacing used for the simulation, as well as the
source-receptor distance (see Section 5.3). A smaller line source
spacing will more closely simulate area source emissions, and there-
fore the model-simulated concentration pattern will closely resemble
that expected from a real area source, Receptors close to the line-
source simulated area source are affected to a greater extent than
those further away. In practice, iterations are commonly made a part
of the computer model in order to obtain an optimum spacing of line
sources, or a sensitivity analysis is performed. King does not men-
tion either of these procedures in his pond simulation model, wherein
he used a 10 meter line-source spacing. The effect of varying the
line-source spacing on concentration patterns is discussed in Section 5,3.
4.3.5.3 Ambient Fluoride Sampling Study
King used ambient samplers designed to measure HF concentrations in
the vicinity of ponds 10 and 20. Each station consisted of a sampler
with a capture element as shown in Figure 4.11, which was reproduced
from King's report. Absorbing medium was a 2 g/1 solution of Na^CO^
which flowed at a rate of 8-12 drops per minute along the inside
diameter of the capture element. Ambient air, entering at an inlet
1.67 meters above the ground, was continually drawn through 30 feet
of coiled tubing at a rate of 30 liters/minute, This allowed suf-
ficient liquid gas contact to effectively absorb 95 percent of ambient
123
-------
Air Inlet
1.67 N Above
Ground
500 ml-3 Neck
Flask Gas-Liquid
Separator
Rubber Tube
Connecter
Sampling Board -- 15
feet flat coll 6 mm
I.D. glass tubing
Sampler Post
Rotometer 30 L/m1n
Front View
Air Bleed
Valve
Rubber Tubing
Vacuum
Gauge
Vacuum
Pump
Figure 4.11. Ambient Air Sampler Used 1n
Study.
124
-------
air fluoride. The absorbing solution flowed into a liquid-gas separator
from which the sample was taken for lab analysis. The air stream was
pulled through a rotometer by a vacuum pump. The absorbing solution
was analyzed in the laboratory for its fluoride content in order to
determine ambient fluoride concentrations through the use of a fluoride
specific-ion electrode.
The use of the automatic "HF gas analyzer" is logical, since most likely
it is HF that is evolved from gypsum ponds. However, particulate fluor-
ides from the gypsum piles could cause positive interference by contributing
to the fluoride level measured by the analyzer. This possibility can
be readily appreciated since the disposal areas can be as high as 120
feet around the ponds. The original design of the fluoride sampler calls
for a vertical absorption tube. This allows the absorption of HF and,
in conditions of laminar flow, minimizes particulate-liquid contact. By
using coiled tubing, a cyclonic effect is induced which causes particu-
late matter to be scrubbed out of the gas stream, This could be averted
by the use of a teflon filter at the air inlet, but no indication that
such a filter was used was evident in King's report.
The exact errors that particulate matter would contribute are not pos-
sible to assess since particulate fluoride emission rates from the p-fles
would have to be known, as would any background concentrations in the
area. King did not address this problem in his experimental design or
analysis of results which precludes drawing any firm conclusions from
his field work regarding gaseous fluoride emission.
125
-------
4.3.5.4 Comparison of Measured and Predicted Fluoride Concentrations
Chapter 10 of King's work purports that significant correlations between
the simulation model predicted fluoride concentrations and measured
concentrations were obtained at the 95 percent confidence level for
both pond 10 and pond 20. However, these correlations were obtained
only after deletion of much of the data. Only 95 of 132 measurements
were used in the analysis. Justification for deletion of these data
in some cases does not seem to be valid, based on the information con-
tained in the report. However, the author was not contacted to discuss
the deletions.
In the analysis of pond 10, certain measured concentrations at Sampler
01 were not included in the analysis because the measured concentra-
tions were much higher than predicted concentrations calculated by the
simulation model. Figure 4.12 is a sketch of pond 10. These measured
concentrations occurred when the average wind speed was between 5 and
6.3 m/sec. Justification was based on the contention that the disper-
sion model could not accurately predict concentrations when high wind
speeds and short downwind distances existed.
A wind speed of 5 to 6.3 m/sec is well within the assumptions of the
Gaussian plume model and the conditions under which oy and oz were
derived (Turner, 1970). The model used by King should therefore be
as accurate at wind speeds of 5 to 6.3 m/sec as it is at wind speeds
of 1 to 5 m/sec. The data should not be rejected then on the basis
of model errors at wind speeds of from 5 to 6.3 m/sec.
126
-------
Study coordinate systen
Wind Speed and
Direction
corder
Active Gypsum Pile
Figure 4.12. Pond 10 Layout (After King)
127
-------
The dispersion parameters a and a used in the Gaussian plume model
are not applicable to the situation of short travel distances (less
than 100 meters) regardless of wind speed or stability (Turner, 1970;
Pasquill, 1974). Thus, if some data for Sampler 01 at pond 10 were
deleted for this reason, all data for Sampler 01 should be deleted
since the sampler was only 10 meters from the edge of the pond.
Finally, the measured concentration data rejected due to high wind
speeds at Sampler 01 show a significant correlation with calculated
concentrations at the 99 percent confidence level (r = 0.90, d.f. = 6).
However, calculated concentrations are consistently lower than measured
values. In spite of King's conclusions, the model seems to work well
for this situation, although corrections should be made in the model
so that measured concentrations .are,mot underestimated. A logical
place for this correction would be in the emission rate equation.
Several measured fluoride values collected at pond 10 were deleted
from analysis under the hypothesis that the active gypsum pile north-
east of the pond was contributing significant quantities of ambient
fluoride at the sampling locations (Samplers 01 and 02; see Figure 4.12).
The simulation model predicted concentrations well below the measured
values in most cases. The active gypsum pile could foreseeably con-
tribute to high fluoride values at the samplers from both gaseous
and particulate fluoride evolution.
128
-------
A study of the plot sketch of pond 10 and of the wind directions
presented in the Appendix of King's work for the usable measured
concentrations at pond 10 indicate that the mean wind direction must
be towards the stated direction in order for the gypsum pond to con-
tribute to fluoride levels at the samplers. The majority of these
usable measurements fell within the directions of -15 to 76 degrees
(based on the pond 10 coordinate system). The measurements not used
for pond 10 analysis were obtained with a mean wind direction of from
55 to 93 degrees, with only one measurement greater than 90 degree
wind direction. The pond 10 sketch indicates that these mean wind
directions are not directed from the active gypsum pile, but since
they are an average direction over a 60 minute period, wind flow over
the gypsum pile could have occurred a percentage of the time. King
does not indicate the percentage of the total time flow was actually
over the active gypsum pile, therefore the relative contribution or
effect of the gypsum pile on ambient levels at the samplers cannot
be estimated. It is also noted that for pond 20, the locations of
the gypsum pile, gypsum pond, and Sampler 01 (Figure 4.13) indicates
that particulate and gaseous fluoride from the pond 20 active gypsum
pile could contribute to fluoride concentrations at the samplers when-
ever the pond contributed to the measured concentration, i.e. when the
wind was directed from pond 20 towards Sampler 01.
Certain pond 20 data were also deleted in verification of the simu-
lation model. Sixteen (16) one-hour time periods at Sampler 02 were
removed from consideration because high fluoride values were measured
129
-------
Magnetic nor
ISO
-SCf
Figure 4.13. Pond 20 Layout (After King)
130
-------
during these time periods. Calculated values were much lower than
measured fluoride values. The explanation was given that some action
by the phosphate plant in the gypsum disposal area caused the high
ambient air concentrations. Sampler 01 displayed no similar increase.
However, due to the locations of the samplers, the gypsum pond and
the gypsum disposal area, it is likely that any activity in the dis-
posal area affecting Sampler 01 would also affect Sampler 02. It is
also likely that the gypsum pile would contribute to the fluoride
measured at the samplers when the wind was such that the pond would
also contribute. Since no upwind measurements were obtained, no
estimate of the background gaseous and particulate fluoride contri-
bution can be made.
Certain other data were also deleted from the pond 20 verification
when model calculated concentration estimates were higher than
measured values (see Appendix 15.5.4 of King's work). Justification
is given that the model could not satisfactorily predict concentrations
during conditions of strong solar insolation and low wind speed. This
is a possible explanation, but not the only explanation. The measured
1 and estimated concentration values were in good agreement (r = 0.73,
d.f. = 9; significant at 95 percent confidence level), and certainly
agreed within the experimental error and uncertainties inherent in the
simulation model. It is questionable, then, that these data should
not be used in the model verification.
King's conclusion that background sources of fluoride (i.e., sources
other than the pond) did not contribute significantly to measured
131
-------
fluoride values at the samplers is not substantiated by the data.
Only a limited number of background measurements were obtained:
six at pond 10 and three at pond 20. This limited data is not suf-
ficient to make any definitive conclusions on background concentrations.
In addition, since King concluded that background fluoride levels were
insignificant, no background measurements were made when the model
validation data was gathered. Thus, it is impossible to assess the
exact contributions of background fluoride to the total measured con-
centrations.
Pond 10 background estimates were obtained from one sampling station.
The measured background values ranged from 0.0 to 0.92 x 10"6 gm-moles/
M
m fluoride and averaged 0.36 x 10'6 gm-noles/m3 (Table 4.3). The aver-
age measured fluoride concentration obtained at pond 10 for the study was
3.12 x 10" gm-moles/m3 and concentrations ranged from 0.64 x 10"6 to
10.48 x 10"6 gm-moles/m3. These data indicate that background fluoride
could contribute on the average 12 percent (0.36 v 3.12) to measured
fluoride levels at pond 10, which is indeed significant.
Three purported background measurements were obtained at pond 20. All
of these measurements were made when calculations showed that the pond
contributed insignificantly to the fluoride concentration at the sampler,
i.e., when the average wind direction was not directed from pond to
sampler. However, the same sampling locations were used in measuring
the fluoride concentration when calculations with the simulation model
showed the pond contributed significantly to the fluoride at the sampler.
132
-------
Table 4.3. Measured background ambient fluoride
concentrations and simulation model
estimates.
Measured Fluoride
Concentration
Pond 10 o.O
0.0
0.0
0.92
1.06
0.68
0.20
Pond 20 0.53
0.37
Calculated Fluoride
Concentration Due
To Pond Emissions
(x 10-6g"moles/m3)
0.0
0.0
0.0
0.0
0.30
0.0
0.03
0.0
0.0
Background Fluoride
Concentration From
Unaccounted Sources
0.0
0.0
0.0
0.92
0.76
0.68
0.17
0.53
0.37
Thus, no upwind background measurements were made when the wind was
blowing from the pond toward the samplers. This is very disturbing
since the gypsum disposal area would be upwind of the pond and samplers
under these conditions, and could contribute to gaseous and particulate
fluoride levels at the sampling locations. It is also disturbing
since the author concluded for pond 10 that the active gypsum pile
there contributed significantly to fluoride levels near the pond. From
133
-------
these observations it must be concluded that background fluoride con-
centration could not be specified with any degree of confidence for
pond 20.
To obtain a better degree of confidence as to the performance of the
simulation model, all of the pond 10 and pond 20 measurements were
plotted on a scatter diagram of calculated and measured fluoride con-
centrations (Figure 4.14 and Figure 4.15). The pond 10 data, when
including all of the deleted observations, results in a significant
correlation (r = 0.85, d.f. = 69) at the 1 percent level of signifi-
cance. Thus, it can be stated with 99 percent confidence that a linear
relation does exist between the measured and calculated values. The
variation in the data due to uncertainties and errors in the simulation
model is equal to 1-r2. For pond 10, this unexplained variation accounts
for 28 percent of the total variation in the data. Linear regression
analysis of all of the pond 20 data resulted in a poor correlation
(r = 0.15, d.f. = 60). Thus, a linear relationship does not exist
between the measured and calculated concentrations for pond 20.
Ninety-eight (98) percent of the variation in the data cannot be explained.
134
-------
• Measurements used in data analysis
® Measurements not used in data analysis
0 Wind flow over gypsum pile data
© High wind speed data
data not used in verification
x
X
•J? »'
1 s^'
• S
1 1 1 1
1234
m
a
1 1 1 1
S * 7 I
*\
MEASURED x 10-6 g moles/m3
Figure 4.14. Pond 10 - Measured Vs. Calculated Ambient A1r Concen-
trations at Samplers.
135
-------
CO
CO
E
QJ
1 2
01
VO
I
o
o
UJ
Measurements used in data analysis
Strong solar radiation, low wind speed
High fluorine measurements ©I
MEASURED x 10'6 g moles/m3
>f Data not used in verification
Figure 4.15. Pond 20 - Measured Vs. Calculated Ambient Air Concentrations at Samplers.
-------
4.3.5.5 Verification of King's Simulation Model
The preceding discussion has alluded to many sources of uncertainty
and error associated with the application of King's simulation model
of gypsum pond emissions and dispersion. These uncertainties are
not uncommon in verification and calibration of other dispersion
models, and it is often difficult to obtain a good correlation between
observed and model calculated values. Very rarely is a one-to-one
correspondence between measured and predicted values obtained. As
sampling time decreases, however, it is expected that calculated
concentrations will more closely approach measured concentrations in
magnitude. This is true since the variability of the parameters de-
scribing the atmospheric phenomena is decreased for shorter sampling
times. In this respect, King's model is superior in that sampling
times were very short (10 minutes).
Contrary to King's statement that a simulation model is accurate only
if a one-to-one correspondence between measured and calculated values
is satisfied (calculated = 1.0 x measured), any linear relationship
between the two is accurate 1f the correlation is statistically signi-
ficant, or if the model predicts concentrations within some defined
confidence limits. Dispersion models typically overcalculate ob-
served concentrations. Overcalculatlon results from model assumptions
of steady-state conditions for emissions and meteorological parameters.
137
-------
Elevated point source models have been found to overcalculate concen-
trations by a factor of 2 to 10. This in itself, however, does not
invalidate the model.
Dispersion models cannot be used to validate parameters used as inputs
to the model. This is because our imprecise knowledge of atmospheric
phenomena does not insure that the dispersion of emissions into the
atmosphere is being correctly or completely simulated. In this respect,
verification of a dispersion model does not verify an emission factor,
for the emissions could be in error by a proportionate amount, and
verification of the model still obtained. (For instance, the slope of
the linear regression line would change). It is imperative that
before a dispersion model can be properly verified, the pollutant
emissions must be known.
In conclusion, the analysis of King's modeling work points up the
following major shortcomings:
1. Verification or calibration of a diffusion model for
fluorides does not substantiate an emission factor
for fluorides.
2. Several measured concentrations were deleted from
analysis but seemingly lacked justification for such
removal.
138
-------
3. Background concentrations of fluoride could not be esti-
mated based on the available data obtained during the study.
4. Since background concentrations were not available, the
contribution of fluoride from the gypsum ponds and that
from other sources of fluoride could not be delineated.
5. Based on 1-4 above, a reliable emission factor for
fluorides from gypsum ponds has not been substantiated.
4.4 COMPARISON OF EMISSION FACTORS FROM THE LTTFMTllpF
Since Cross1 and Ross1 study provided no useful data regarding an
emission factor, it will not be considered in this section. As stated
in Section 4.2.3, Tatera measured wind speed in a wind tunnel at a
point 0.1 meters above the water. It is important to note, first of
all, that the velocity distribution in the wind tunnel is such that
the velocity will approach zero at all four walls due to surface fric-
tion. As Tatera stated, there are no accepted hydrodynamic scaling
parameters which allow wind tunnel results to be applied directly in
large scale systems. The velocity profile above a pond generally
follows a logarithmic distribution. It is zero at the surface and
increases with height.
Nevertheless, in order to make a comparison of the two studies, a
relationship, empirically developed by King, was used to scale Tatera's
velocities to a height of 16 meters.
Vl zl
e*P [0-2909+ 0.06154 (L) - 0.01164 (V16)] (4.38)
139
-------
where:
V1 = wind speed at Z1 (0.1 meter)
V2 = wind speed at Z2 (16 meters)
L = measured lapse rate (between 2 and 16 meters)
V^ = wind speed at 16 meters
Comparisons of emission factors as determined by King and Tatera (incor-
porating the correction above) are presented in Table 4.4 for three tem-
peratures. Comparisons at 75°F and 85°F are not readily applicable
since most gypsum ponds operate at temperatures above 90°F. Typically,
process waters enter the pond at 115°F and are returned to the process
at 95°F. Unfortunately, Tatera obtained all of his data between 75°F
and 95°F, while King conducted vapor pressure studies between 70°F
and 100°F. It would have been far more useful to have obtained data
between 85°F and 115°F.
Tatera's emission estimates at 95°F are approximately a factor of 2
greater than those of King. In accordance with our previous discussion,
a judgement cannot be made as to the validity of either method of
estimation when applied to a particular gypsum pond. It can be argued
from purely physical chemical grounds that the emission rate should be
about one-fifth the values determined by King*. Thus, the best statement
* Special communication with Dr. Alonzo Coots of International Industrial
Consultants; based upon assumption of saturated calcium sulfate with
calcium fluoride controlling free fluoride levels found in gypsum ponds.
Attempts to verify these assumptions in the laboratory were confounded
by the complexity of the solution.
140
-------
Table 4.4. Comparison of Emission Factors Predicted by King
and Tatera at Various Temperatures.
Vlg M/Sec
V1 M/Sec
Ft/Mi n
75°F
King (Pond 10)
Tatera (Process Water)
85°F
King
Tatera
95°F
King
Tatera
EMISSION
1
0.25
50
0.96
0.41
0.75
0.52
0.92
1.5
FACTOR IN LB/ DAY-ACRE
2
0.54
106
1.6
0.86
1.3
1.1
1.6
3.2
4
1.2
238
2.9
1.9
2.2
2.4
2.8
7.3
141
-------
that can be made at this point is that the characteristic emission
factor appears to lie in the range of 0.1 to 10 Ib/acre-day and that
a careful field study would seem appropriate to determine the contri-
bution of volatile fluorides from the ponds to the atmospheric fluoride
levels around the ponds.
142
-------
5.0 DETERMINATION OF GASEOUS FLUORIDE CONCENTRATIONS IN THE VICINITY
OF A GYPSUM POND
In order to estimate concentrations it was necessary to assume
a typical gypsum pond, servicing a phosphoric acid plant, producing
1,000 TPD of PgOs- Secondly, since no accurate or reliable emission
rate is known, downwind x/Q isopleths were calculated. Concentration
•
isopleths can then be calculated as a function of an assumed emission
factor.
5.1 DESCRIPTION OF TYPICAL GYPSUM POND
The gypsum pond is an integral part of a typical phosphoric acid
plant. It serves as a receptacle for wastewater which during the
manufacturing process comes into direct contact with raw materials,
intermediate products, by-products, waste products or finished
products.
Gypsum pond water is recirculated to the phosphate processes and
used mainly in scrubbers, barometric condensers, and for slurrying
waste gypsum. After settling of gypsum solids in the ponds, the water
is allowed to cool by evaporation and'reused.
Figure 5.1 shows the configuration of the "typical" gypsum pond.
Process waters enter at about *15°F. The area around the entry
point would, therefore, be the area for maximum evaporation and
fluoride emission. The water cools by evaporation down to about
95°F which is the temperature at which it reenters the process.
A recirculation rate of 30,000 GPM would be expected.
143
-------
GYPSUM PILE
FERTILIZER
PROCESSES
FRESH
MAKE-UP
WATER
figure 5.1- Tj^fcal typsun fond Servicing A 1000 TPD-PzOs Pfttotf
-------
The average concentrations of species in gypsum pond water are
summarized below. These figures are based on the mean measurements
from thirteen ponds provided by The Fertilizer Institute.
Concentration (mg/1)
Al 170
Ca 1600
Cl 3500
F 5100
Fe 210
K 110
Mg 165
Na 930
NH3 - N 520
p 4500
S04 4400
Si 1560
PH 1.6
The important factors to be considered and the concentrations of
free fluoride ions' (F-), pH and temperature. The value for fluorides
given above are not free fluoride measurements, but are total soluble
fluoride. These three factors- should be sufficient to specify an
equilibrium vapor pressure for HF in the. pond water, which is directly
related to the emission factor.
A typical gypsum pond, handling both slurry and process water from
a 1.000TPD-P205 plant would have about 350 acres of wet area.
Water depth would be about 10 feet. Most likely it would be located
adjacent to the plant and surrounded by mined out land of sparse
vegetation or swamp. Assuming that the pond is used for both
gypsum settling and cooling, there would be a region where the
stream from the sluicing operation would join the pond. This area,
known as the gypsum flats,, is where the gypsum settles. It would
145
-------
constantly be worked by draglines which would remove settled wet
gypsum and transfer it onto an active gypsum pile to dry. The
gypsum pile would be about 80 feet high on about 150 acres adjacent to
the wet pond.
When rainfall exceeds evaporation, a certain amount of the water
would be discharged from the pond. Prior to entering a water
stream, it would be treated by double stage liming which would
raise the pH to 9.0. This would effectively remove fluoride as
fluorspar through a series of reactions. It would be necessary
to use this system on an intermittent basis for 3 months per year.
5.2 GROUND LEVEL GASEOUS FLUORIDE CONCENTRATIONS IN THE VICINITY OF A
GYPSUM POND — _
The Gaussian finite line source equation was used to estimate
gaseous fluoride concentrations downwind of a hypothetical gypsum
pond. A computer program was developed to calculate concentrations
due to a large number of spatially distributed line sources at a
number of downwind receptor points. The line source model was
chosen for the area source simulation because it is considered the
best available calculation technique.
A 350 acre rectangular pond, similar to the configuration shown in
Figure 5.1, was chosen for simulation. The pond's length was
taken to be twice its width. The pond was divided into a number
of equal sections, 10 meters wide. A line source was placed at the
center of each segment oriented cross-wing, and the segments emissions
concentrated into the line source contained therein.
146
-------
Estimations of ground-level fluoride concentrations were obtained for
two different meteorological situations (typical average day conditions
and worst conditions). Isopleths of concentration/emission rate
(x/Q) ratios were obtained for each situation (Figure 5.2 and Figure 5.3)
Average conditions consisted of stability class land a wind speed of
3 m/sec. Typical worst conditions consisted of stability class 6
and a wind speed of 2 m/sec. The x/Q value gives the normalized
ground-level concentration in relation to the emission rate. For
example, if the x/Q ratio is 1.0, the ground level concentration
will be equal to 1 ug/m3 if the emission rate is 1 Ib/acre/day. If
the emission rate is 10 Ib/acre/day and x/Q is equal to 1.0, the
3
ground-level concentration would be 10 yg/m .
The concentrations predicted by the simulation line source model
should not be construed to represent actual concentrations which
would exist downwind from a real gypsum pond. Verification and
calibration of the model is required before it can be confidently
used to predict concentrations near gypsum ponds. The greatest
utility in the present simulation is in the ground-level concen-
trations patterns generated by the mode. For meteorological conditions
similar to those used in obtaining Figures 5.2 and 5.3, the patterns
shown are expected to be similar to those actually occurring near
the pond. However, the magnitude of the concentrations may not be
those displayed.
The simulation assumes a typical gypsum pond as described previously,
and assumes a constant wind velocity, unaffected by topographic
147
-------
O.5
1.0
KILOMETERS
cx>
GYPSUM POND
I
Figure 5.2 Isopleths of Calculated Ground-Level Fluoride x/Q Ratios Downwind of a Hypothetical
Gypsum Pond. (Stability Class 6, Wind Speed = 2 m/sec.)
-------
0.5
1.O
KILOMETERS
GYPSUM POND
Figure 5 3 Isopleths of Calculated Ground-Level Fluoride x/Q Ratios Downwind of a Hypothetical
Gypsum Pond. (Stability Class 4, Wind Speed = 3 m/sec.)
-------
features such as the gypsum pile. It further assumes contributions
from only the gypsum ponds in gaseous form. Background levels would
have to be known and added, as would particulate or gaseous contri-
butions from the piles and plant. If it is true, however, that
gypsum ponds are the main sources of fluorides, then Figures 5.2 and
5.3 should be accurate in showing the relative dispersion of fluorides
around a gypsum pond.
5.3 COMPUTER MODEL SENSITIVITY ANALYSIS
In order to test the sensitivity of the model to changes in line source
spacing, the computer model was executed for a typical gypsum pond
configuration utilizing various meteorological conditions and downwind
receptor distances. Each of the ponds King studied were approximately
500m x 900m. Fluoride samples were located within about 100m of the
edge of the ponds. In his simulation model, King employed a 10m line
source spacing to calculate downwind fluoride concentrations.
For comparison purposes, a gypsum pond 500m x 840m was employed for
the sensitivity study. The mean wind direction was situated crosswind
(90°) of the 840m sides. Concentrations were calculated for receptor
distances of 10m, 25m, 50m and 150m, and for a range of meteorological
conditions. Calculations were made for both a 10m line source spacing
and a 5m spacing. The results are tabulated in Table 5.1. Examination
of this table indicates that the effects of line source spacing are
most critical at closer receptor distances, but are also dependent on
meteorological conditions. For the parameters listed in the table,
150
-------
Table 5.1 Sensitivity Analysis for the Finite Line Source Model
Fluoride concentration (pg/m3)
Line Source
Meteorological
Stabi-
lity Wind
1
3
4
4
5
Conditions
Speed (m/s)
3
5
1
10
3
Receptor
10
0.9
1.1
8.5
0.8
3.7
25
0.7
0.9
7.0
0.7
3.1
Spacing
Distance
0
0
5
0
2
50
.5
.7
.8
.6
.6
= 10m
(m)
150
0.4
0.5
4.5
0.5
2.1
Line Source
Receptor
10
0.8
1.0
8.1
0.8
3.6
25
0.7
0.8
6.8
0.7
3.0
Spacing
Distance
50
0.5
0.7
5.7
0.6
2.6
= 5m
(m)
150
0.4
0.5
4.5
0.5
2.1
-------
the largest difference between conditions calculated using 10m and
5m line source spacings is 0.4 yg/m3 (stability class 4 and wind speed
of 1 m/sec). This value represents only a 5 percent difference in the cal-
culated concentration. Based on the sensitivity analysis, it is
concluded that a 10m line source spacing, such as that employed by
King, will introduce a certain small error in the calculated concen-
trations, but that a 10m spacing results in sufficient accuracy for
model validation/calibration purposes.
152
-------
6.0 IDENTIFICATION OF CONTROL TECHNIQUES
Several control processes for reducing fluoride emissions from gypsum
ponds are evaluated in this section. All of the candidates examined are
unusual with respect to air pollution control equipment in that they are
indirectly applied to the air emissions. In general, air pollution con-
trol equipment includes mechanical collectors, bag filters, electrostatic
precipitators, and scrubbers. It is not surprising to find special pro-
blems associated with the processes considered herein due to the complex
chemistry involved in the pond waters, the large water volumes involved,
and the acidic nature of the fluorides themselves.
A search of the literature and discussions with members of the phosphate
industry and the EPA revealed six potential candidates. They are:
1. Kidde Process for treating barometric condenser and scrubber water.
2. Swift Process for treating barometric condenser and scrubber water.
3. Liming the gypsum pond water to raise the pH and produce insoluble
calcium fluoride.
4. Dry conveyance of the gypsum to waste stacks instead of sluicing with
water.
5. Calcining the rock prior to treatment in the process.
6. Changing the basic process to the Hemi/Dihydrate process for phosphate
production.
In addition to these six processes, consideration is given to complete
segregation of the gypsum and cooling ponds. This segregation is con-
sidered tantamount in applying any of the first four control options;
153
-------
therefore a brief discussion will follow concerning the benefits to be
derived from a two pond system.
Two Pond System in a Phosphoric Acid Plant
Phosphoric acid plants utilize a wide range of gypsum/cooling pond
arrangements. In most cases process and gypsum sluicing waters are
transported to a common pond allowing these waters, which are vastly
different in properties to mix, with the ultimate result that both
process and gypsum pond waters will be highly contaminated //ith acidic
P2°5' H2S04* as wel1 as H2SiF6.
In some cases, separate cooling and gypsum ponds are utilized. All
process waters except gypsum sluicing water are sent to evaporative
cooling ponds. Gypsum slurry is pumped from the filtration operation
to a gypsum pile where the gypsum settles. The supernatant water is
subsequently recycled through the cooling pond, thus contaminating it
with P205, H2S04, and fluorides from the filtered gypsum.
A third possible method involves employing a "Two Pond System" in which
the cooling pond can not be contaminated with waters from filtration. This
"=thod was discussed by Parish and Enriquez in a recent paper (1973). It
-: recommended as part of a two-step solution for reducing fluoride
ssions from gypsum and cooling ponds to be used simultaneously with the
S.ift Absorption Process".
154
-------
The reasons for separate ponds were given by Enriquez and Parish were:
1. The required size of the gypsum slurry pond is small (about 5 acres)
since no area is required for cooling. This water would be the
most contaminated and acidic water in the plant due to the presence
of P205, H2S04> iron and aluminum complexes, and fluorides from
the filtration operation.
2. The size of the pond required for the barometric condensers is
determined by the cooling duty requirements. This area is estimated
to be 0.1 acre/TPD P205 .
Since the cooling pond will receive condensed vapors from the flash
cooler and evaporators, it is possible that entrained phosphoric acid
could be present as a contaminant. This, however, may be minimized by
the addition of entrainment separators, the result being that the main
contaminant entering the cooling pond will be limited to fluorides.
Figure 6.1 schematically shows the proposed system. One pond would service
the filtration operation by receiving gypsum slurry. The othar pond would
receive waters from the fume scrubbers, flash coolers, and the evaporation
trains. The cooling pond will require about 200 acres for a 1,000 TPD P205
plant, while the gypsum pona will require approximately 5 acres.
The costs required to segregate the two ponds at an existing plant are
site specific, but are considered negligible compared to the other control
costs presented herein. Since each plant will have different problems
155
-------
Filter Wash Water
en
tn
Gypsum
Pond
(3 to 4 Acres)
Flash
Coolers
Fume
Scrubber
Figure 6.1. Two Pond System for Phosphoric Acid Plant.
-------
switching to a two pond system, costs have not been calculated. These
costs, however, are expected to be well below the cost of the fluoride
control equipment.
Judging the cost-effectiveness of each process was facilitated by using
the following criteria:
1. Ability of the process to lower gypsum and cooling pond emissions
(efficiency)
2. Total capital investment requirements
3. Cost per ton of P20s produced
4. Commercial availability
5. Demonstrated capability
Our findings regarding cost and efficiency are summarized in Table 6.1.
Examination of this table reveals a large range of control costs ($1.25
to $7.46 per ton P205) and efficiencies that warrant some discussion.
The Swift Process is, in this study's estimate, the most cost-effective
method for control of emissions from gypsum and cooling ponds when
employed in conjunction with a segregated two pond system. A saleable
product (H2SiF6) is recovered. The product is sold to large municipalities
for fluoridation of water at approximately $200/ton H2SiF6 (delivered)
and to aluminum companies for production of fluorosilicates and other
fluorine compounds used in processing aluminum at $60/ton H2SiF6
(FOB Plant). The economics of this process are considerably improved
if all of the fluosilicic acid produced is sold at the latter price.
157
-------
en
Co
Table 6.1. Capital Investment and Operating Costs For Fluoride Control of 1,000 TPD PoOn Plant.
2U5
Process
Kidde
Swift Absorption
Liming
Conveyor
Pre-Calcination
of Rock
Total Capital Annual ized Operating
Where Fluorine Removal Investment Costs
Applied Efficiency $MM Total , MM $/Ton P205 By-Product
Barometric
Condensers 95-98%a 2.57 2.31 $ 7.46 (NH,)9 SiF
42 6
Barometric
Condensers 90%b 1.30 0.17 1.25° H SiF
2 6
All Cooling b
Pond Water 90% 2 10 0 97 3J3 None
Gypsum Filter After
Acidulation a i.Q7 0.43 1 40 None
Crushed Phosphate
Rock b 29.81 7.57 24.42 None
a) Not calculated due to uncertainties in fluorides evolved from filter cake.
b) 90% removal of fluoride in rock is achieved. However, the fluorides evolved are scrubbed
and transferred to ponds.
c) A credit of $2.26/ton P205 is realized if all fluosilicic acid produced is sold at $60/ton
(100% HSiF).
-------
In this case a credit of $2.26/ton P205 may be realized. Although this
study did not conduct a market survey, discussions with several indi-
viduals involved in the production and marketing of phosphates and
aluminum lead the authors to believe the market for H2SiF6 can absorb
some expansion, primarily in the aluminum industry.
The second most cost-effective method for reducing emissions involves
liming the pond waters to a pH of about 4.0. This reduces emissions by
lowering the fluoride vapor pressure. Unfortunately, this technique does
not produce a saleable byproduct and is subject to serious limitations.
Perhaps the most severe of these is the increased potential for equipment
scaling caused by precipitation of silica. Current practice tends to
eliminate this problem which has plagued the industry in the past.
Additional problems arise from secondary impacts associated with the
mining of limestone necessary for this control. While it is outside
the scope of work for this contract to assess secondary environmental
impacts outside of the phosphate complex, we feel it significant that
approximately 37,200 tons per year of limestone will be required to line
the pond waters associated with a 1,000 TPD P205 complex.
The Kidde Process has the highest potential recovery of all candidate;
for fluorine in the weak phosphoric acid. Since it is believed that th^
fluorine from the barometric condensers is the primary source of emissions
from the cooling pond, this process receives the highest technical merit.
Unfortunately, the high annualized costs put this process at a severe
disadvantage when compared with its nearest competitor, the Swift vapor
absorbers. The high costs of the Kidde Process reflect the study's
159
-------
assessment that the process be limited to that required to control fluor-
ide emissions. Thus, the byproduct in this instance is ammonia silico-
fluoride and not aluminum fluoride, as Kidde suggests. The limited market
for the ammonia salt (ca. 2,000 TRY) renders this byproduct a waste
stream rather than a saleable product. Regardless, it is the opinion
of the authors that the increased process complexity and additional
capital investment required to produce aluminum fluoride is unwarranted
as a control scheme. Perhaps these are the very reasons why Kidde has
been unsuccessful in marketing this process.
The other three processes (dry conveyance of gypsum, pre-calcination
of the phosphate rock, and the hemi/dihydrate process) all suffer from
the disadvantage that a major process change or change in industry
practice is required. None of these three is currently used on a
large scale in the U.S.
The remainder of this section is devoted to a detailed assessment of
these six candidates regarding cost and effectiveness. In reviewing
this section, the reader is reminded that these estimates have been
derived from literature and other sources and not from detailed engi-
neering estimates.
160
-------
6.1 KIDDE PROCESS
6.1.1 Process Description
The Kidde Process, used to defluorinate wet-process phosphoric acid and
convert the extracted fluorine into various fluorine compounds, is a
patented process developed by Kidde Process Company, Pasadena, California.
Fluorine compounds that may be produced include aluminum fluoride, syn-
thetic cryolite (Na3AlF3), hydrogen fluoride, and synthetic fluorspar.
Figure 6.2 is a detailed process flow diagram of the Kidde Process for
a 1,000 TPD P205 facility. The reaction sequence is as follows. React-
ive silica is mixed with phosphoric acid as it is fed to the acid
concentrators. The silica has been dried to about 10 percent moisture
and is premixed with the raw 30 percent P205 acid in an agitated tank
with about 15 minutes retention time. The amount of silica required is
determined by the amount of fluorine in the raw acid. The resulting
slurry is fed to the concentrators through the usual control system. The
principal reaction at this point is:
Si°2(<0 + *HF(aq) SiF4(g) + 2H2° (6.1)
The overhead vapors from the acid concentrators are now condensed with an
aqueous ammonium bifluoride solution; the pH of this condensing solution
is approximately 4.0. As currently practiced, these vapors are condensed
with cooling water or weak fluosilicic acid with a pH of 1.5 to 2.0. Herein
lies a major advantage of the Kidde Process. Scrubbing with the higher pH
bifluoride solution produces condensed vapors with a lower fluoride vapor
161
-------
u-? 6.2. Kidde Closed-Loop Fluorine Recovery for a 1,000 TPD P205 Plant.
-------
pressure. At this point, the condensed SiF4 reacts with the bifluoride
as follows:
SiF4(g) + 2(NH4)F-HF(aq) (NH4)2SiF6(a + H2S1Fg(aq) (6.2)
In the next step of the Kidde Process, the condensed vapors from the
various stages of evaporation are combined. This combined stream is then
neutralized with ammonia which neutralizes the acid produced in the baro-
metric condensers (eq. 6.2). This proceeds according to the following
equation:
H2S1F6(aq) * 2NH3(aq)
Concentration of the resulting solution is then achieved in multiple
effect evaporators. The product, ammonium silicofluoride, is then stored
for shipment or further reaction, as follows..
•
The product ammonia silicofluoride can be reacted further to produce
additional products. An example of such a scheme is depicted in the
following reaction scheme:
+ 4NH3(aq) + 2H2° 6NH4F(aq)
NH4F'HF(aq) * NV (6-5)
6NH4F.HF(aq) + Al203-3H20(s) 2(NH4)3A1F6* + 6H20 (6.6)
900° F
(NH4)3AlF6(s) A1F3(S) + 3NH4Ft (6.7)
163
-------
This scheme is that recommended by Kidde to produce aluminum fluoride
which can be marketed to the aluminum industry. It has added advantages
in that the silica and ammonia are recovered; these can then be recycled
to the P205 plant for further use.
For the purposes of this study, reactions 6.4 through 6.7 are not incor-
porated in the fluoride recovery scheme. This is because the emission
reduction estimated in the following section is effected by reaction
6.1 through 6.3; i.e., removal of fluoride from recyclable process waters
is achieved at that point.
6-1-2 Process Applicability for Fluoride Emission Reduction
The applicability of this process as a candidate for reducing fluoride
emissions from gypsum ponds is due to the lowering of the fluoride vapor
pressure through neutralization with ammonia. An estimate of the emission
reduction potential is arrived at as follows. As currently practiced
with no fluorine recovery, the overhead vapors are condensed with
recycled cooling pond water; the pH of this solution is approximately
1.5 to 2.0. At this pH, the fluoride vapor pressure is approximately
13.8 x 1CT6 rrniHg (25 C) (HEW, 1970). In the Kidde Process, condensed
vapors are not sent to the cooling pond as in the above scheme. Instead,
the solution is concentrated and neutralized to a pH of 5.0. At this
pH, the fluoride vapor pressure is only 0.65 x 10'6 mmgHg (25 C). There-
fore, in the final condenser of the ammonium silicofluoride evaporation
train, the vapor pressure is approximately reduced by the following amount:
Approximate fluoride reduction = (13.8 - 0.65) x 100%/13.8 = 95.3%
164
-------
This is approximately the emissions reduction to be expected since the
ammonium silicofluoride evaporation train operates at this pH which is why
the fluorine content of the product streari is less than 0.2% F. (Figure 6.2).
6.1.3 Process Evaluation
In April, 1974, the Kidde Process Company submitted a technical and
economic study to Shahpur Chemical Co. Ltd. of a proposed fluorine-
recovery and aluminum fluoride-manufacturing installation at the company's
phosphate plant at Bandar Shahpur, Iran. The study predicts costs of
less than $10.00 per ton for extracted fluorine and less than $100.00
per ton for aluminum fluoride. The selling price for aluminum fluoride
at that time (1974) was over $300.00 per ton. However, no full-scale
plant actually exists today, or is currently under construction.
Adaption of a fluorine-recovery system to an existing wet-process phos-
phoric acid plant should present few problems. The major items of
equipment that must be added are the triple-effect evaporators for
ammonium silicofluoride concentration, the shell and tube heat exchangers
used for cooling the condensed overhead products, and a small cooling
tower for removing heat associated with the ammonium silicofluoride evap-
orators (see Figure 6.2).
The existing barometric condensers used in concentrating the P20g can
he used with the Kidde Process sincn the condonsing medium (ammonium
bifluoride) is less corrosive than the currently USPC! pnnri viaters.
165
-------
Additional benefits that should result from installation of the Kidde
Process include:
1. Reduction or elimination of liming for pond water discharged from
the plant's cooling ponds.
2. Improvement in the physical properties of the phosphoric acid due to
a reduction in its fluorine content.
3. Increased recovery of fluoride otherwise lost in the evaporation
stage.
Of these, the third was estimated in Section 6.1.2. The first two are
not easily quantifiable; i.e., it is not easy to assess cost benefits
associated with then. However, botli of these are benefits that sliould be
attainable with this system.
6•1•4 Process Economics
Estimated incremental capital investment requirements and annualized
operating costs for a 1,000 TPD phosphoric acid facility are summarized
in Tables 6.2 and 6.3, respectively.
The data were determined in the following manner. Equipment sizes rec-
commended by Kidde for a 560 TPD P205 plant were examined for reasonableness.
These sizes were then adjusted to process the flows corresponding to a
1,000 TPO plant. Costs for major equipment items were obtained from pub-
lished sources; standard engineering estinates were then used to arrive
at installed plant costs. As Table 6.2 shows, approximately 2.G million dol-
lars is required to install the equipment associated with this control process,
166
-------
Table 6.Z. Capital Investment Requirements for Kidde Process Producing
29,000 TRY (NH4)2SiF6 (1,000 TPD P205).
DIRECT COSTS
EQUIPMENT, E
PIPING
CONCRETE
STEEL
INSTRUMENTS
ELECTRICAL
INSULATION
PAINT
FIELD MATERIALS, M
DIRECT MATERIAL, E + M = M
DIRECT FIELD LABOR
SUB-TOTAL DIRECT COSTS
$ 652,000
297,300
33,250
20,200
66,500
13,000
32,000
3,000
465,250
1,114,250
410,000
1,524,250
INDIRECT COSTS
FREIGHT, INSURANCE AND TAXES
OTHER INDIRECTS (CONTRACTORS FEES, INTEREST DURING
CONSTRUCTION, ENGINEERING)
52,160
565,300
SUB-TOTAL INDIRECT COSTS
CONTINGENCY (20% DIRECT AND INDIRECT COSTS)
617,460
428.342
TOTAL CAPITAL INVESTMENT (TCI)
$ 2,570,052
167
-------
Table 6.3. Annualized Operating Costs for Kidde Process.
Quantity
Unit Cost Annual Cost
RAW MATERIALS
SILICA
AMMONIA
14,700 TPY
4,200 TPY
$ 35/TON
$190/TON
$
$
514,500
798,000
1,312,500
UTILITIES
ELECTRICITY
STEAM
COOLING TOWER MAKE-UP
OPERATING LABOR
OPERATING
SUPERVISION
MAINTENANCE
LABOR AND MATERIALS (L&M)
SUPPLIES
350 HP
37,500 LBS/HR
100 GPM
2 MEN/SHIFT
OF OPERATING LABOR
4% OF TCI
15% OF L & M
20 MILLS/KWH
$0.50/M LBS
0.20/M GAL
$ 5.50/HR
$ 35,500
139,500
8,930
$ 183,930
$ 102,300
15,345
$ 117,645
$ 102,950
15.450
$ 118,400
OVERHEAD
PLANT
PAYROLL
50% OPERATION & MAINTENANCE
20% OPERATING LABOR
$ 118,000
23.530
$ 141,530
FIXED COSTS
DEPRECIATION (5% OF TCI)
TAXES AND INSURANCE (2% OF TCI)
CAPITAL CHARGES (10% OF TCI)
TOTAL ANNUAL COSTS
COSTS PER TON OF P205
$ 128,685
51,475
257.370
$ 437,530
$ 2,311,535
$ 7.46
168
-------
Annualized operating costs for the Kidde Process were obtained using
estimates from Kidde's report and standard engineering estimates. A
310 day year was assumed to allow for equipment repairs and maintenance.
Examination of Table 6.3 shows a net annual operating cost of $7.46 per
ton P205. One will notice that no credit is given to the 29,000 TRY
of (NH4)2SiF6. This is explained by the fact that only one firm
(Agrico Chemicals) in the U.S. produces this chemical. This firm indi-
cated that they only market 2,000 TRY to commercial laundries where it
is used to neutralize residual alkalinity between the wash and rinse
cycles. Thus, even though this chemical has a current market value of
$340 per ton, credit cannot be assigned since a market does not exist.
The question now arises concerning the feasibility of extending this
process to include production of AlFj. Although a detailed assessment
was not made to determine the additional capital investment to do this
(i.e., produce A1F3), an estimate was made by pro-rating Kidde1s data.
This estimate indicates an additional investment of approximately $6.7
million to install a plant to process the (NH4)2 SiF6 to A1F3. This
plant would produce roughly 13,000 TRY of A1F3, recycle NH3 and Si02
to the fluorine recovery stages, and improve the economics of this con-
trol process significantly. A simplified flow diagram of this plant
is included in Figure 6.3.
The above figures indicate that approximately 9.3 million dollars would
be invested in a plant capable of servicing a 1,000 TPD P20S facility
169
-------
.._*•!
— _ — (*.{,,'
ci/^7 3.f ^_ _
Otrft -
; ££*STt"c'
-------
(310,000 TRY). The present capital investment requirements for a 1,000
TPD P205 complex is approximately 15 million dollars. This is mid-1975
battery limits costs for an existing plant. Thus, the total investment
required for the complete Kidde Process is of the same order of magnitude
as that required for the P205 complex itself.
Although the current market for A1F3 was not evaluated in this study,
one aluminum manufacturer that was contacted (Kaiser) indicated that
some market does exist for these chemicals used in producing aluminum.
In fact, this company operates a small facility in the central Florida
area where the sodium salt of fluosilicic acid is produced, shipped,
and then used to make cryolite, a flux used in producing aluminum.
6.2 SWIFT PROCESS
The Swift Process, was developed and patented by W. R. Parrish (patent
No. 3273713), assignor to Swift and Co., Chicago, 111. It involves the
removal of fluorine compounds from wet phosphoric acid manufacturing ir
a manner similar to the Kidde Process.
6.2.1 Process Description
Figure 6.4 presents a process flow diagram as well as a fluorine and
partial overall material balance for a wet process phosphoric acid plant
producing 1,000 tons/day• P20g. Evaporation from dilute (28.5% P 0 ) to
concentrated (52.5% P205) is effected by three evaporation units in
series. Vapors from each unit are scrubbed by an absorber which removes
90 percent of the fluorides evolved during each evaporation stage.
171
-------
TO SCRUBBERS AND CONDENSERS
FROM SCRUBBERS AND CONDE
COOLING
POND
43OOLB-F/HR
ATMOSPHERE
4SOLB-F/HR
REACTION/
FROM
cooLwe
POND
FROM
COOUN6
POND
5090LB-F/HR
2S24OO LB/HR
m
PHOSPHATE
MOCK
BAROMETRIC
CONDENSER
CONDENSER
CONDENSER
SCRUBBCR
I8»LB-F/>W
TOCOOUWPOND
I29LB-F/MR
TO COOLING
56LB-F/HR
TO COOLING POND
58OLB-F/HR
20SOOLB/HR
rrseoo LB/HR
196700 U/HR
62.5%
225200 LB/HR
4SOLB-F/HR
tSfOOLB/Mft
NOOOL»/H«
T.79%
I7MOLV/HR
15.5% HgSl^
HBdOLB/HR
Figure 6.4 _Sw1 ft Process for FT uori neRecovery at a 1000 TPD
Plant
-------
A scrubber with a 96 percent fluoride removal efficiency is utilized
for control of vapors generated in the acidulation reactor. Cooling
pond water is used as the absorbing medium and a stream of dilute (3.85%)
HgSIFg is continuously removed. This stream is introduced into a recir-
culation tank from which a solution of 7.75% H2SiF6 is drawn and pumped
into the Swift vapor absorber servicing the third evaporator. The
scrubbing liquor is introduced into the scrubbing unit through nozzles
in the form of a spray. After absorbing the volatile fluorine material,
the scrubbing liquor falls into a barometric leg, the end of which is
immersed in the recirculation tank. The temperature of the liquor being
recycled is maintained at a level which will ensure the absorption of
volatile fluorine compounds while minimizing water vapor condensation at
the operating pressure.
A solution of 7.75% HgSiFg is drawn from the recirculation tank servic-
ing the third unit and pumped into a recirculation tank servicing the
second unit. The operational descriptions of the Swift components
(absorber, recycle tank, and pump) servicing evaporation units 1 and 2
are identical to that described for evaporation unit 3. However, each
unit is operated at a different temperature, pressure, and recycle
rate, and a more concentrated stream of fluosilicic acid is drawn from
each recycle tank. The product (25% HgSiFJ is continuously removed
and transferred to storage tanks for shipment.
173
-------
6.2.2 Process Applicability for Fluoride Emission Reduction
The original purpose of the Swift Process was for the economic recovery
of byproduct fluorides and the realization of a return on investment.
In the original report on gypsum pond emissions (November, 1974), the
potential of the Swift Process as a means of controlling volatile
fluorides emanating from these ponds was discussed. Since that date
the original inventor, Parrish and his associate, Enriquez, have in
fact recommended this process as a pollution control strategy when
coupled with a two pond system. At least one phosphoric acid plant
presented this process as its solution to potential fluoride emissions
resulting from once-through river water for its process water require-
ments.
Table 6.4 indicates the level of fluoride recovery possible by the
installation of the Swift Process. Of particular interest is the fact
that fluorides entering the cooling pond are reduced by approximately
90 percent. Bearing in mind the fact that the gypsum pond is not
effected and that it comprises only about 5 wet acres, it is the author's
opinion that fluoride emissions should be reduced by 90 percent. The
basis for this conclusion comes from the work of Tatera, who correlated
fluoride concentrations in pond water with gaseous emissions.
6.2.3 Process Economics
A summary of capital costs is presented in Table 6.5. The primary pieces
of equipment are three sets of recirculation tanks, pumps, and absorbers.
174
-------
Table 6.4. Fluorine Material Balance in a 1,000 TPD P205 Plant
Utilizing the Swift Vapor Absorption Process.
With Swift Without Swift
In Phosphate Rock 10,000 Ib/hr 10,000 Ib/hr
Recovered
As 25% H2SiF6 3,855 Ib/hr 0
Not Recovered
Stays in gypsum 4,500 Ib/hr 4,500 Ib/hr
Lost at reactor 20 Ib/hr 20 Ib/hr
Stays in 52.5% P205 . 1,250 Ib/hr 1,250 Ib/hr
Lost to cooling pond 375 Ib/hr 4,230 Ib/hr
175
-------
Table 6.5. Capital Investment Requirements for the Swift Vapor
Absorption System.
Capital Investment (Installed Costs)
1. 2 Scrubbers handling 140,000 ft3/min each
0 $282,000 each $ 564,o0o
2. 1 Scrubber handling 180,000 ft3/min 362,000
3. 3 Rubber-lined recirculation tanks handling
1,050 gal each
(a $29,100 each 87,300
4. 3 Stainless steel recirculation pumps
(900 gpm @ 100 psi) $40,900 each 122,700
5. 4 Rubber-lined storage tanks 100,000
gal capacity @ $42,000 each i68tQQO
TOTAL CAPITAL INVESTMENT j 1,304,000
176
-------
The rest of the equipment shown in Figure 6.4 is required even without
the Swift Process. Storage facilities consisting of four rubber-lined
tanks are required to hold one week's supply of 25% H SiF produced.
2 6
Costs will vary greatly depending upon whether the plant is existing
or new. These costs were derived for retrofit to an existing plant.
The figures presented compare well with those quoted by a manufacturer
of $500,000 for the equipment servicing each evaporator of $1,500,000
total capital investment for three modules.
Annualized operating costs are presented in Table 6.6. Assuming no
income from the sale of fluosilicic acid, the increase in price of one
ton of P00_ would be $1.25. At the present market value of $60/ton
L 5
H SiF *, the Swift Vapor Absorption Process will operate at a profit
2 6
of $2.26/ton of PJ)_ if all fluosilicic acid produced is sold. It was
2 5
this potential profit which prompted several firms to utilize the Swift
Process on the basis of economics. As previously stated, a marketing
survey was not conducted. However, it is doubtful that all plants in the
U.S. could utilize the process and realize a profit on their venture.
Its use would have to be accompanied by a vigorous marketing effort and
possible price reduction in order to make by-product fluosilicic acid
a more attractive source of fluorides than Mexican fluorspar,
If gypsum and cooling pond are high emitters of fluorides and control
equipment is required, the Swift Process is an attractive low cost or
low profit abatement system.
*FOB Plant
177
-------
Table 6.6. Annualized Operating Costs for Swift Process.
Quantity
Unit Cost Annual Cost
UTILITIES
ELECTRICITY
OPERATING LABOR
OPERATING
SUPERVISION
MAINTENANCE
LABOR AND MATERIAL
SUPPLIES
OVERHEAD
PLANT
PAYROLL
FIXED COSTS
250 hp
1/2 MAN/SHIFT
15% OF OPERATING
4% OF CAPITAL INVESTMENT
15% OF LABOR AND MATERIAL
50% OPERATION & MAINTENANCE
20% OPERATING LABOR
20 mills/KWH $ 27,730
$ 5.50/HR
DEPRECIATION (5% OF CAPITAL INVESTMENT)
TAXES AND INSURANCE (2% OF CAPITAL)
CAPITAL CHARGES (10% OF CAPITAL)
TOTAL ANNUAL COST
COST PER TON OF P205
CREDIT
iai!4 TONS 100% H.SiFc
c o
ADJUSTED ANNUAL CREDIT
CREDIT PER TON P0
$ 60/TON
$ 24,000
3.600
$ 27,600
$ 52,160
7,840
$ 60,000
$ 43,800
5,520
$ 49,320
$ 65,200
26,080
130,400
$ 221,680
$ 386,330
$ 1.25
$1,086,870
$ 700,530
$2.26
178
-------
6.3 LIMING OF COOLING PONDS
Treatment of effluents by "double-liming", or two-stage lime neutraliza-
tion, has been practiced by wet-process phosphoric acid plants for
many years. The process simultaneously raises the pH while lowering the
concentration of soluble fluorides.
6.3.1 Process Description
Under the assumption that a two pond system will be used to segregate
gypsum sluicing water from other process water, cooling pond water will
contain fluorides and small amounts of P00C. Thus it is necessary only
f. 0
to add sufficient lime to precipitate out fluoride compounds as fluor-
spar (CaF2). Most fluorides from the scrubbers and barometric condensers
will be in the form of fluosilicic acid. Upon addition of sufficient
lime, fluorides will be precipitated according to:
H2S1F6(aq) + 3Ca°(s) + H2° 3CaF2(s) + 2H2° + S102(s) <6'8>
Holding the mixture in a quiescent area allows the particulate CaF2 to
settle.
Table 6.7 shows the reported effect that liming has on actual cooling
pond water. As this Table shows, a pH of 3.9, soluble fluorides are
only one percent of the value at a pH of 1.4. Figure 6.5 shows a graph
of fluoride vapor pressure versus pH from the data in Table 6.7. Use
of this figure allows one to calculate the potential emissions reductions
to be expected by liming the pond waters. Thus, at a pH of 3.9, the
179
-------
Table 6.7. Effect of Liming on Fluoride Evolution From Gypsum-Pond Water
PH
1.4
2.6
3.0
3.3
3.9
4.5
6.1
6.25
7.72
9.7
12.1
12.3
12.5
Soluble
fluoride,
ppm
8125
4000
NR
450
106
100
106
NR
NR
NR
NR
NR
16
Ca(OH)o,
Ib/gallon
0.116
0.145
0.156
0.157
0.160
0.192
0.193
0.207
0.213
0.222
0.246
0.346
Vapor pressure
of fluoride
@25°C, mmHg
13.8 x 10"6
6.22 x 10"6
NR
NR
NR
0.86 x 10"6
0.45 x 10'6
NR
NR
NR
NR
NR
NR
SOURCE: (HEW, 1970)
NR: Not Reported
180
-------
£3
3 8
V-i VO
5) O
U) •—I
2.0
1.0
• 9
.8
.7
.6
.5
w
.
PH
Figure 6."). Fluoride Vapor Pressure vs. pH (Source: HEW, 1971)
181
-------
fluoride vapor pressure obtained through interpolation of this figure
is 1.3 x 1(T6 mmHg. The expected emission reduction obtained in liming
from pH 1.4 to pH 3.9 is:
Reduction = (13.8 - 1.3)100%/13.8 = 90.6%
A pH of 3.9 can be achieved by single stage liming. Assuming that a
two pond system will be employed, it will be necessary to lime the
entire cooling pond contents from pH 1.4 to 3.9 only once and thereafter
to add sufficient lime to handle the theoretical amount of fluorides
entering the cooling pond. Thereafter the major acidic components, P 0
2 5
and H2$04, are confined to the gypsum sluicing water.
Figure 6.6 is a schematic diagram of a modern well-controlled single
stage liming system. Although not shown in this figure, a rotary kiln
for calcining limestone and the auxiliary equipment required for its
operation are included in the proposed scheme and cost estimates.
6-3.2 Fluoride Emission Reduction by Single Stage Liming
The reduction of fluoride emissions by liming is a consequence of the
reduction of fluoride vapor pressure upon the addition of lime as indi-
cated in Table 6.7. It is estimated that for a cooling pond having an
initial pH of 1.4, emissions can be reduced by 90 percent by the addition
of 0.157 pounds Ca(OH)2/gallon to the contents of the pond (Table 6.7).
182
-------
OUST COLLECTOR
LIME
FEEDER
FROM
BAROMETRIC
CONDENSERS
COOLING POND
Figure 6.6. Proposed Single Liming System
-------
Table 6.8 presents an overall fluoride material balance. The quantity
of fluorides entering the cooling pond is reduced by 90 percent. Since
the gypsum pond occupies only about 5 acres, and since fluorides present
in the gypsum are mostly solid CaF2> it is this study's estimate that
overall gaseous fluoride emissions can be reduced by 90 percent by liming
to a pH of 3.9 to 4.0.
Table 6.8
Fluorine Material Balance in a
1,000 TPD P20s Plant Utilizing Liming
as Fluoride Control Strategy
Basis: 10,000 Ibs/hr F
Item
Gypsum cake
Precipitated by liming
Vapors from reactor
c? c«y p^n
«/£. • \J lo r O **C
Cooling pond
With Liming
4,500
3,807
20
1,250
423
Without Liming
4,500
20
1,250
4,230
6.3.3 Process Considerations
Installation of single-liming systems at existing phosphoric acid
plants should pose few, if any, problems. Most plants already have
double-liming treatment systems for pond water overflow discharge.
184
-------
This study proposes use of a single-liming facility as shown in Figure
6.6. Lime requirements for a 1,000 IPS P205 plant are arrived at as
follows.
First, an initial, one-time dose of lime is required to brinq the
existing cooling pond water up to the required pH of 3.9 to affect
a 90 percent emissions reduction. Acidic constituents of the existing
pond requiring neutralization at this point are P205, l^SO^, >and ^SiFg.
The initial lime requirements are calculated based on the following
assumptions, applicable to a 1,000 TPD plant:
1. 300 acres of cooling pond waters at an average depth of 0.5 meters
(160 MM-Gal).
2. 0.157 Ibs Ca(OH)2 is required to treat each gallon (see Table 6.7)
Based on these assumptions, 9,530 tons of lime (as CaO) are initially
required to raise the pH to 3.9. Before this lime is added, the gypsum
and cooling ponds are segregated to ensure the integrity of the proposed
system.
The second phase of the liming program occurs once steady-state has been
achieved after the initial dosage has been applied. The quantities of
lime required during steady-state are determined from equation 6.8. The
lime requirements according to this equation are 1.47 Ibs CaO per Ib
fluorine (3 moles CaO/6 moles F). Based on data presented in Section 2.2
of this report, we estimate that approximately 91 Ibs F/ton PoOr will end
185
-------
up in the cooling pond waters. This includes fluorine scrubbed from the
reactor vapors and that removed in the barometric condensers. Based on
these data and assumptions, 135 Ibs CaO per ton P205 will be required
during steady-state operation (67.3 tons per day).
During the initial dosage two quantities of lime will be required—that
necessary to raise the pH and the daily steady-state requirements.
If we now assume a 40-day transient period to achieve steady-state,
the initial daily requirement will be 305 tons CaO/day (9530/40 + 67.3).
6.3.4 Process Economics
A summary of capital costs associated with liming is presented in Table
6.9. For the large quantities of lime required, it will be more econom-
ical for the plant to produce its own lime from limestone. Thus, included
in the capital costs are a kiln, limestone bins, and a scrubber for the
control of particulate emissions from the kiln. The other major expenses
are the chemical feed system as well as the initial lime requirement
which have been included into the total capital investment.
Annualized operating costs are presented in Table 6.10. The greatest
costs are those associated with fuel to calcine the limestone. Depre-
ciation and capital charges comprise approximately 33 percent of annualized
costs. Total annualized costs using this strategy will result in an
estimated cost increase of $3.13/ton
186
-------
Table 6.9. Capital Investment Requirements for Single Liming System.
Capital Investment (Installed Costs)
1. Lime kiln (81 x 140" with motor) $ 1,100,000
2. Lime kiln scrubber (includes fan, motor, ducting) 61,000
3. Limestone bins and conveyors 120,000
4. Chemical feed system (including slaker, pump,
storage, reaction vessel) 283,000
5. Miscellaneous (20% of equipment costs) 95,400
6. Initial lime requirements (9,530 tons
@ $47/ton) 447,910
TOTAL CAPITAL INVESTMENT $ 2,107,310
187
-------
Table 6.10. Annualized Operating Costs for Single Liming.
Quantity
Unit Cost Annual Cost
RAW MATERIALS
LIMESTONE
120 TPD
$ 4/TON $ 149,000
UTILITIES
ELECTRICTY
FUEL
OPERATING LABOR
OPERATING
SUPERVISION
50 hp
3.1 mm-BTU/Ton CaCO,
1 MAN/SHIFT
15% OPERATING
20 mills/KWH
$2/mm-BTU
$ 5.50/HR
$ 5,070
230,600
$ 235,670
$ 48,200
7.200
$ 55,400
MAINTENANCE
LABOR AND MATERIAL (L&M)
SUPPLIES
4% OF TCI
15% OF L & M
$ 81,624
11.884
$ 93,508
OVERHEAD
PLANT
PAYROLL
50% LABOR & MAINTENANCE
20% OPERATING LABOR
$ 74,454
11.080
85,534
FIXED COSTS
DEPRECIATION (5% OF TCI)
TAXES AND INSURANCE (2% OF TCI)
CAPITAL CHARGES (10% OF TCI)
TOTAL ANNUAL COST
$ 102,030
40,812
204.060
$ 346,902
$ 971.084
ANNUAL LIMING COST
$/TON CaO
$/TON P0
$ 46.62
$ 3.13
188
-------
6.4 GYPSUM TRANSPORT BY CONVEYOR BELT
Transporting the gypsum by-product to storage piles using dry conveyor
belt rather than slurry pipeline could possibly reduce the fluoride
emission from this segment of P205 production. The benefits (i.e.,
emissions reductions) are not as apparent as the annual costs are cal-
culable. This is because the gypsum filter cake contains some phosphoric
acid and between 18-35 percent free moisture. Thus the pH of this water will
most likely be low and the fluoride vapor pressure high. If it were pos-
sible to keep these solids dry once the free water evaporated, the emissions
would be negligible due to the very small vapor pressure of most solids
at ambient temperatures. However, operation of a gypsum stack does not
readily accomodate covering the stack to reduce emissions either with
dirt or some other impervious barrier.
Although several plants in Belgium are reported to be using a belt conveyor,
the authors have knowledge of only one dry-belt system in the U.S., in
California. Apparently it is economically justified in those cases where
the gypsum can be marketed for agricultural and other uses.
Capital investment requirements for a one-mile by 24-inch covered con-
veyor are presented in Table 6.11. These costs were estimated assuming
4.5 tons gypsum per ton of P205 with 25 percent free moisture included
(Slack, 1968). A six year life was assumed based on vendor estimates
of a useful life of 8 to 10 million tons transported.
189
-------
Table 6.11. Capital Investment Requirements for Dry Gypsum Conveyor.
Capital Costs
Belt conveyor—one mile @ $250/ft $ 1,056,000
Feeders including motors (2 each) 10,000
TOTAL CAPITAL INVESTMENT $ 1,066,000
Annualized Operating Costs
Depreciation (1°P_ = 16.7%) $ 178,000
b
Capital charges (10%) 106,600
Operating costs @ $ 0.08/ton-mile (vendor estimate) 148,800
TOTAL ANNUALIZED COSTS $ 433,400
$/TON P0 PRODUCED $ 1-40
190
-------
Annualized operating costs, also presented in Table 6.11, of approximately
$1.40 per ton of P205 were estimated. Although other measures would most
likely need incorporation for this system to be effective, the additional
costs were not estimated. These measures might include some form of
covering material on the dry gypsum stack and water treatment facilities
for surface runoff and seepage.
Although approximately 45 percent of the original fluorine in the rock
is retained in the gypsum, the potential emissions from this source are
most likely the smallest when the gypsum arid cooling ponds are separated.
This should be true even with commonly used slurry transport techniques
due to the small area requirements for the gypsum piles.
6.5 PRETREATMENT OF ORE BY CALCINING
6.5.1 Process Description
Precalcining the phosphate rock prior to accidulation with sulfuric acid
was considered as one candidate control process. Calcination involves
heating the rock in a rotary kiln or fluid bed reactor. Although direct
and indirect firing may be used, the latter method has some disadvantages
as discussed below. As currently practiced in producing animal feeds,
the rock is calcined by direct firing methods.
Defluorination of phosphate rock occurs in two regimes. In the first,
low-temperature regime, up to 66 percent of the fluorine can be removed
by heating the rock to approximately 2,000°F. Heating the rock to its
191
-------
fusion temperature, which ranges from 2,500°F to 2,700°F volatizes another
21 to 24 percent which increases the fluorine removal to approximately
90 percent. The prevalent gaseous species involved is HF, although SiF^
is present in smaller amounts, with the ratio HF:SiF4 increasing with
increasing temperature.
The off-gases from this reaction must be cooled and scrubbed to remove
particulates and fluorides before the gases are vented to the atmosphere.
Herein lies one of the chief disadvantages to this approach. First, the
gas volumes handled with direct firing are large (approximately 500 ACFM/
ton P205 at 2,000°F) and second, the fluorides scrubbed from the gases
must be disposed of, generally in the scrubber water cooling pond. As
currently practiced, cooling pond water is used to scrub the gases and is
neutralized before returning to the cooling pond, Thus, indirect-firing
only removes the fluorine from the rock to have it placed in the pond.
Indirect-firing in a rotary kiln reduces the quantity of off-gases that
must be handled. However, the temperatures that are obtained by this
method are limited to 1,200°F for stainless steels; use of special metals
will increase this to 2,000°F, but at a great additional expense.
As proposed herein, two direct-fired rotary kilns will treat 3,000 TPD
of phosphate rock. Prior to scrubbing, the hot (ca. 2,000°F) off-gases
will be cooled to 500°F by waste heat boilers which will produce approx-
imately 300,000 Ibs/hr low pressure (150 psi) steam. This last provision
will improve the economics somewhat by providing a steam credit.
192
-------
6.5.2 Process Economics
Capital investment and annualized operating costs are presented in Table
6.12. As this Table shows, the capital required to install kilns to treat
3,000 TPD of rock is roughly $30 million, nearly twice the investment of
the P20g plant itself. This is without any fluorine byproduct recovery.
Since scrubbing the gases only produces a one to two percent solution
of H-SiFg, additional equipment would be needed to concentrate this to
the 25 percent, which is the normal saleable concentration.
6.6 HEMI/PIHYDRATE PROCESS
The hemihydrate-dihydrate process for making wet-process phosphoric acid
has been in use in Japan for several years. Its primary advantages are
the higher overall yield of ?2®5 ancl ^e production of high quality
gypsum, suitable for gypsum plaster and gypsum boards. About 40 percent
of fluorides present in the rock are volatilized in the acidiculation
stages and may be recovered as fluosilicic acid.
6.6.1 Process Description
Although there are several variations of this process, they basically
differ from the common dihydrate process in that hemihydrate is formed
during decomposition of the phosphate rock and is subsequently recrystallized
as dihvdrate before filtration. The decomposition and recrystallization
steps are carried out in separate reaction vessels to allow optimum process
control. The resulting dihydrate crystals, as compared with those made in
the common dihydrate process, are larger and retain less P20e and other
impurities in the lattice.
193
-------
Table 6.12. Capital Investment Requirements for Pre-Calcinin§ of
Phosphate Ores.
Capital Costs
Rotary kilns: 2 - 16.5' x 600' direct fired
(installed costs, mid-1975) $ 28,400,000
Waste-heat boiler (2) 700,000
Gas-scrubbers (2) 710,000
TOTAL CAPITAL INVESTMENT $ 29,810,000
Annualized Operating Costs
Fuel $ 1,060,000
Electricity (1,400 hp @ $ 0.20/KWH) 175,000
Depreciation (10% TCI) 2,981,000
Maintenance (5% TCI) 1,490,500
Capital Charges (10% TCI) 2,981,000
TOTAL ANNUALIZED COSTS $ 8,687,500
Steam credit (300,000 Ibs/hr @ $ 0.50/M-lb) 1,116,000
NET ANNUAL OPERATING COSTS $ 7,571,500
$/TON P205 PRODUCED $ 24.42
194
-------
The key to the process is growing large, easily filterable crystals at
a rate fast enough to give economic operation, reducing the phosphate
substitution in the process. The diyhydrate is thus crystallized under
quiescent, carefully controlled conditions, thereby making it possible
to maintain a uniform excess of sulfuric acid throughout the slurry.
6.6.2 Effect of Hemi/Dihydrate Process on Fluoride Emissions
There is virtually no data available to allow a judgment to be made on
any fluoride emissions reduction which would result from employing the
Hemi/Diiiydr'ate process.
In order to make an estimate of potential fluoride emissions reductions,
the following will be assumed:
1. 45 percent of the fluorides initially present in the rock will remain
in the by-product gypsum.
2. 12 1/2 percent of the fluorides will be present in the product acid.
3. 0.2 percent of the fluorides will be released to the atmosphere.
4. 42.3 percent (or the remainder) of the fluorides will either be
recovered or end up in the cooling pond.
5. 30 to 40 percent of the fluorides are recoverable as fluosilicic-
acid (as claimed by the manufacturer).
Based on the above, it is estimated that 70 to 95 percent of the fluorides
which could contribute to cooling pond emissions can be recovered as by-
product H2SiF6.
195
-------
The common dihydrate process is based on a rather violent reaction which
produces local deficiencies of sulfuric acid, and thus leaves p205 in the
dihydrate lattice. As a result of the better control in the hemihydrate-
dihydrate process, loss of phosphate by lattice substitution is very low.
Recovery is thus improved and the by-product gypsum is more suitable for
use in building products.
Figure 6.7 presents a flow diagram of the Fisons Ltd. scheme for wet
process phosphoric acid manufacture. The notable differences with this
scheme from those employing dihydrate routes are that:
1. No evaporators are utilized since a 50 percent P205 product is obtained
in the reactors.
2. A much higher amount, which is estimated to be 30 to 40 percent of
all fluorides initially present in the rock, is evolved in the reac-
tion/filtration step. These vapors are sufficiently rich in fluorides
to allow the production of fluosilicic acid as a saleable by-product.
6.6.3 Process Economics
Since the hemihydrate-dihydrate process is an alternative to the exist-
ing U.S. wet phosphoric acid processes, the economics cannot be evaluated
in the same manner as the previous "add-on" control systems. In order to
use the hemihydrate-dihydrate process, the existing plant must be exten-
sively changed and markets found for the gypsum and fluosilicic acid
produced. The success of the process in Japan rests largely on the facts
that Japan must import all phosphate rock and most of its gypsum. Thus
196
-------
Cooling
Water
Process water
Process water
Reactor «l
Reactor«2
Filter Feed Tank
i Effluent
--' pumping Tcnk
50% PgOgproducTocid
Figure 6.7. Hani/Dihydrate Wet Process for Phosphoric Acid Manufacture.
Fisons Ltd. Scheme.
-------
the small added yield of PoOc in combination with the production of a
useful gypsum makes it most attractive to the Japanese.
In the United States, these factors are not present, and the relatively
low cost phosphate rock and gypsum do not create the favorable economics
here as in Japan. By way of comparison, however, a cost comparison
between the dihydrate and the hemihydrate-dihydrate process was develop-
ed by Fisons Ltd. for a European plant using Morocco phosphate rock and
built in 1971. This analysis, which is summarized in Table 6.13, indicates
the process is competitive for a new plant. Capital costs are also included
for a 1,000 TPD plant built in 1975.
The Hemi/Dihydrate routes are presently not practiced in this country.
Reasons given include:
1. Florida rock is not of sufficiently high quality to allow its use.
2. Problems are encountered in the filtration operation when using Florida
rock.
3. The relatively large amount of clay fines and other impurities make
this process difficult to control.
It is important to point out that, in fact, several Japanese firms use
Florida rock successfully. Sometimes calcining is practiced to improve
the quality of the rock. However, it is foreign firms which have the
experience in Hemi/Dihydrate acid production and no doubt hold many
198
-------
Table £.13. Comparative Process Economics*
(Basis: European plant (built In 1971) using Morocco rock to make 50,000 metric tons of P205/year as SOX phosphoric acid)
<£>
V£>
Extraction efficiency. X
(based on Measured cake losses)
Operating costs/ton
Phosphate rock, tons
Sulfuric Ic1d (100X), tons
Low-pressure stream, tons
Fresh Mater, tons
Cooling water, tons
Electricity, kwh
CAPITAL COSTS, $
Depreciation, Interest, taxes
and Insurance (201)
Maintenance (7X)
Operating .labor, men/shift
OVERALL PRODUCTION COST. I/METRIC TON
Typical D
-------
patents and rights associated with these schemes, It is therefore very
difficult at the present time to calculate cost comparisons for such plants
when built by American firms in' the United States.
200
-------
7.0 DEMONSTRATION COSTS
In this section demonstration costs for two items are considered.
First, a methodology for determining fluoride emissions from gypsum
ponds is discussed; costs for the proposed method are then pre-
sented. In the second portion of this section demonstration costs
are presented for determining the effectiveness of the control
method selected by this study as the most cost-effective candidate.
In summary, the costs for these items are:
Field verification of emission rate (2 ponds) $ 70,925
Demonstration of Swift system (1 pond)
a) pilot plant $295,760
b) modification of existing plant $274,475
As pointed out in section 7.2.2, the last item listed could be in
error by approximately $25,000. This is due primarily to the uncer-
tainties in estimating costs for pond segregation without site-
specific data.
7.1 VERIFICATION OF PREDICTED FLUORIDE EMISSIONS
Based on our findings concerning the emissions of fluoride from gypsum
ponds, it was concluded that no investigator had as yet established
experimentally the fluoride emission rate from gypsum ponds. While
King has conducted experiments in which ambient fluoride concentrations
were measured, he did not directly measure the emission rate. Even
201
-------
though his predicted concentrations correlated quite well with his
measured values, the mathematical dispersion model used in his predic-
tions is not exact and, as pointed out in Section 5.0 of this report,
might tend to overestimate.
For the above reasons, the following experimental program is proposed
which would 1) directly measure fluoride emissions as a function of
velocity and 2) verify the ambient ground-level fluoride concentrations
predicted in section 4.3 of this report.
Fluoride emissions would be measured as follows. All fluorides emit-
ted from the pond must pass through an imaginary vertical plane pass-
ing through the downwind dike of a gypsum cooling pond and extending the
length of the pond. If, at any height z^ above the dike, the fluoride
concentration and wind velocity are known, the mass flux at that point
can be determined. Mathematically stated:
E = VWAZC
where: W = width of pond
AZ = difference in height
C = fluoride concentration
By determining the flux at several heights above the dike, the total,
mass of fluorides passing through the plane is:
l (7.1,
where: C = 0
202
-------
One possible method for measuring concentrations at various heights
would be to construct towers at each end of the gypsum pond. A remote
long path infrared sensor could then establish the average concentra-
tion over the total length of the pond at different heights. The
concentration profile obtained from this operation could then be used,
together with the velocity profile, to calculate the fluoride emission
rate.
The above approach presents some practical problems that might be hard
to overcome. The biggest problem would be alignment of the remote IR
source with the detector. As the system got progressively higher up
the towers, the wind would most likely make it difficult, if not im-
possible, to keep the units aligned.
One advantage to the above approach is that by using a dispersive infra-
red detector, one could determine which fluoride compounds were being
emitted.
i
Another approach that could be taken to determine the emissions
is to install ambient fluoride samplers at different heights on
the towers. However, it is suggested that the towers now be located
at points equidistant from the ends of the pond (1/3L and 2/3L).
Figure 7.1 shows a theoretical concentration profile predicted by the
finite line source dispersion model discussed in Section 5.3 of this
report. As the curves in this figure indicate, the ambient concen-
tration 30 meters above ground level should be about ten and twenty-
five percent of ground-level concentrations for stability class four
203
-------
Height (m)
•o
fD
O
(D
O
o
fD
Stability Class 4
Stability Class 2
Figure 7.1 Theoretical Ambient Concentration Profile as a
Vunction of Height
-------
and two respectively. Thus, a 100 foot tower with five ambient samp-
lers equally spaced should be sufficient to establish the concentra-
tion profile.
Velocity profiles can be established with three cup-anemometers placed
at different heights, e.g., 2, 8 and 16 meters. Recorded wind speeds
can be used to establish the value of the power low exponent in equa-
tion 4.33 for each experiment conducted.
Once the concentration and velocity profiles are determined, the emis-
sion rate is calculated through equation 7.2.
To further verify King's and Tatera's equations for fluoride emission
rates, the following data should be collected for each experiment.
Ten centimeter wind velocity (Tatera), average temperature for the
pond water in each line source (Tatera and King), average liquid
fluoride concentration in each line source (vapor pressure) and am-
bient temperature and cloud cover (for stability).
To establish the accuracy of the ambient fluoride concentrations pre-
dicted in Section 5.0 of this report, a network of ambient samplers
should be placed downwind of the pond, e.g,at 100, 200 and 300 meters.
One sampler should also be placed upwind to establish the background
fluoride concentration.
7.1.1 Sampling Methods
It is important that an appropriate sampling method be used that will
ensure results that are as accurate as the state-of-the-art permits.
" 205
-------
For several years the State of Florida Department of Pollution Control
has been monitoring ambient fluoride levels in the vicinity of the
phosphate mining and processing area near Lakeland, Florida. The
sampling method used by this agency consists of a series of midget
impingers containing either sodium hydroxide or distilled water. This
method is under question, however, as one investigator (Sholtes, 1973)
found that in the concentration range of 10-100 ppm gaseous fluorides,
the collection efficiency decreased in direct proportion to the gas
concentration. Collection efficiencies as low as 30 percent were
measured in this range.
The ambient sampler used and evaluated by King was also tested by the
EPA (Baumgardner, CPL). This sampler consisted of a 15 foot length
of 6MM I.D. coiled glass tube coated with Na2C03. Ambient air is
drawn through this tube at approximately 30 LPM for one hour at which
time the glass tube is flushed with a buffer solution. The fluoride
content of the buffered Na2C03 solution can then be determined with
a specific ion electrode.
Evaluation of a commercial model of the above sampler by the EPA
indicated unusually large response times when the unit was calibrated
using an 8 foot length of small bore teflon tubing placed upstream
of the glass tubing. King reported a collection efficiency of 85 per-
cent for a 15 foot length of glass tubing and 95 percent for two 15
foot lengths connected in series which should be sufficient for the
purposes of the experiments proposed herein.
206
-------
7.1.2 Experimental Design
In order to gather sufficient data to verify the fluoride emission
estimates of King and of Tatera, it is proposed that two ponds be
tested, preferably at two different plants. Ponds of roughly rectangu-
lar shape should be utilized in order to minimize the number of as-
sumptions made in determining the fluoride flux.
Two telescoping meteorological towers, installed as shown in Figure 7.2
will house the ambient samplers and meteorological equipment. Prior
to performing each experiment, it will be necessary to measure the
pond water temperature at a minimum of two points in each line source,
as shown in Figure 7.2. One pond water sample from each line source
should be obtained for fluoride determination.
Each experiment should be approximately one hour long as longer time
periods might allow for larger variations in wind speed and direction.
This is important since the ground level wind direction should be approx-
imately normal to the finite line sources.
In order to verify ground level concentrations arid to determine the
fluoride species emitted from the ponds, it is proposed that a long
path remote sensing infrared detector be placed on top of the dike.
As the total path length of this instrument will be approximately
300-400 meters (pond width ), accurate determination will be possible.
7.1.3 Program Costs
In order to evaluate the fluoride emissions as a function of wind
velocity, it will be necessary to measure the emission rate at several
207
-------
•— o
0) •—
OJ
<_)
O)
A
A
A
A
A
C
* ' !* M4 m
D
A
A
A
A
A
E
C -
Ml /O *J
L/3 M
D
D
KEY: A - Liquid Temperature Measurement
B
C
D
E
- Upwind Ambient F, Sampler
Meteorological Tower
Downwind F^ Sampler
Remote Infrared Sensor
Figure 7.2 Ambient and Pond Emission Sampling Network - Aerial View.
208
-------
different velocities. As there will be no control over this variable,
it is anticipated that approximately one month will be needed for each
pond tested. Utilizing four men in the field, this will amount to
about 640 man-hours field testing for each pond.
Table 7.1 gives a breakdown of the estimated costs and man-hours for
a two-month field sampling program on the above basis (two ponds
tested).
7-2 VERIFICATION OF THE SWIFT VAPOR-ABSORPTION SYSTEM
Several experimental approaches were considered that would allow an actual
demonstration of fluoride emissions reduction upon the cooling pond. In
selecting an approach, the primary criterion considered useful in demon-
strating emissions reductions is the soluble fluoride level in the cooling
pond waters. As shown in Figure 6.5 of this report, the correlation
between pH, soluble fluorides and fluoride vapor pressure is such that
one can use this parameter in judging emissions reductions.
Two approaches at demonstration are considered. The first approach
involves construction of a pilot plant including acid attack system,
scrubbers, evaporators, gypsum and cooling ponds. This would conven-
iently be done near an existing plant where small quantities of steam,
electricity, and water can be obtained. The second approach would
involve segregation of the gypsum and cooling ponds at an existing
facility equipped with Swift vapor absorbers. After an equilibration
period the soluble fluoride content of the cooling water will be measured
209
-------
ro
o
Table 7.1 Cost Breakdown for Fluoride Emissions Determination from Two Cooling Ponds
Man-Hours
Engineer Technician
(=> S30/hr
-------
over a time period. When the fluoride content has reached a steady state
value, the system will be operated in that mode (i.e., Swift scrubbers
either on or off) until sufficient data has been gathered to allow con-
clusions to be drawn concerning operation in that mode. In the following
pages each approach will be discussed at length and program costs pre-
sented.
7.2.1. Pilot Plant Approach
This study estimates that a small pilot plant, approximately one ton
P2°5 Per day, should be sufficient to obtain results for demonstration
purposes. All unit-processes contained in a full-scale (500-1000 TPD)
facility should be incorporated with the exception of a sulfuric acid
plant. Sulfuric acid will be purchased in bulk for the experimental
program. A 10,000 ft cooling pond will be used to provide cooling for
the barometric condenser waters. This pond will be monitored daily for
fluorides. After approximately three months operation using the Swift
absorbers (and producing 20 percent H2SiF6), the absorbers will be turned
off and acid production stopped. The system will then be operated for
another three month period as before. Now the pond waters should be-
come more acidic and the fluoride levels increase. Comparison of the
data from each mode will then allow conclusions to be drawn concerning
the effectiveness of this system in reducing fluoride emissions.
Program costs for this approach are presented in Table 7.2 These costs
were arrived at by the following methods:
- Pilot plant construction costs, sans the sulfuric acid plant, were
estimated using a size exponent of 0.6 as suggested by Guthrie (1970)
211
-------
Table 7.2 Pilot Plant Program Costs for the Swift Vapor Absorption System
I. Plane Investment (installed 1975)
$150,000
II. Six Month Operating Costs
Quantity
Raw Materials
Phosphate rock
SuIfuric acid
Utilities
Electricity
Steam
550 tons
440 tons
720KWH/ton P205
4,0001b/ton
Pilot Plant Operation
Technician 3,120
Engineers 520
Unit Cost
$35/ton
$50/ton
$0.02/KWH
$0.50/M-lb
$20/hour
$30/hour
Laboratory Analyses
Fluoride
PH
Rock Samples 400 @ $7/sample
Water Samples 800 @ $5/sample
800 @ $3/sample
Data Analyses and Interpretation
Chemist4~0
Engineer/Scientist 80
Reports and Project Management
Engineer 200
$25/hour
$30/hour
$30/hour
Total Cost
$ 19,250
$ 22.000
$ 41,250
$ 2,640
$ 370
$ 3,020
$ 62,400
$ 15.600
$ 78,000
$ 2,800
$ 4,000
$ 2.400
I9,200
$ 1,000
$ 2,400
$ 3,400
$ 6,000
III. Miscellaneous Expenditures
A. Travel (2 man-trips P $250 each)
B. Telephone
C. Reports (60 copies @ 300 pages each)
D. Computer
E. Housing subsidence for on-site engineers
$ 500
$ 100
$ 900
$ 500
$ 2,000
IV, TOTAL Estimated Program Costs
$295,760
212
-------
Table 7.3 Estimated Costs for Demonstration of Swift Process at a
500 TPD - P205 Facility
I. Phase 1 Costs - Plant Selection
A. Labor (80 engineer hours @ $30/hour) $ 2 400
B. Non-labor (travel for site selection) $ 'SQQ
$ 2,900
II. Phase 2 Costs - Plant Modification
A. Engineering and supervision (200 hours) $ 6,000
D. Non-labor (travel and communication) $ 1,000
C. Subcontractor costs (dredging, dozing, piping, etc.) $25,000
D. H2SiF6 storage tanks (for additional $ 20,000
product storage one 50,000 gallon $ 52,000
rubber-lined tank)
III. Phase 3 Costs - Operation of Swift System
A. Labor
1) Engineering and supervision (260 hours) $ 7,800
2) Chemist (200 hours) (F", P?0c, etc.) $ 5 QOO
B. Non-labor ° ,
1) Travel (4,000 miles @ $0.20/mile) $ 800
2) Per diem (15 man-days @ $30/day) $ 450
3) Communications $
$ 12,350
IV. Phase 4 Costs - Operation without Swift System
A. Labor
1) Engineering (260 hours) $ 7 800
2) Chemist (200 hours for F~, P90R, etc.) $ 5*000
B. Non-labor c °
1) Subsidy for loss of H2SiFc to cooling
pond (25 TPD of 100% H2$iF6): 75% @ $60/ton $174,375
2) Travel (4,000 miles @ $OT20/mile) $ 800
3 Per diem (15 man-days @ $30/day) $ 450
4) Communications $ 100
$186,725
V. Phase 5 costs - Data Interpretation and Report Preparation
A. Labor
1) Engineer (160 hours @ $30/hour) $ 4 800
2) Chemist (80 hours 0 $25/hour) | 2'ooo
B. Non-labor * ^'UUU
1) Reproduction (60 copies - 300 pages @ $0.05/copy) $ QQQ
$ 7,700
Estimated Project Cost $274 4?5
213
-------
- Raw materials and utilities requirements were estimated using
guidelines presented by Slack (1968).
As shown, total program costs are estimated to be approximately $295.760
for a six month program.
7.2.2 Conversion of Existing Facility
The other approach considered in the report for demonstrating control
feasibility is as follows. An existing P205 plant, equipped with Swift
vapor absorbers and currently producing fluosilicic acid will be modi-
fied by completely segregating its pond systems. This will insure that
fluorine contained in the gypsum does not find its way to the cooling
pond waters. The plant will then operate in two modes. In the first,
fluosilicic acid will be produced 100 percent of the time the plant is
in operation. At present, our best estimates indicate the Swift absorbers
at most plants are not in operation at all times. This will probably
require purchase of additional storage facilities to handle the additional
fluosilicic acid produced during this phase of the experimental program.
Assuming this additional recovered fluosilicic acid requires five day
storage, one 50,000 gallon tank will be required. Once the facilities
are prepared, the plant will be operated in this mode for approximately
six months.
During the second operational phase of this program, the Swift absorbers
will not be operated. Instead, all of the unrecovered fluorine will be
condensed in the barometric condensers and go to the cooling pond. This
214
-------
should cause the fluorine content of the cooling waters to increase.
This increase will be measured until the steady-state concentration
Plateau is reached. At that time the plant will be run for an additional
period (approximately one month) to allow data collection and analyses.
During this phase it will be necessary to subsidize the P205 company
for loss of valuable HgSIFg. The current price of which is $60/ton on a
100 percent basis. For a 500 TPD - P205 plant, this will amount to approxi-
mately 25 TPD - H2SiF6 (100 percent).
Program costs for this approach are summarized in Table 7.3. The
most uncertain figure in this table is that required to separate the
two ponds (i.e. gypsum and cooling). For our purposes it is assumed
that approximately $25,000 will be required to affect separation.
However, the uncertainty in this figure is unknown but possibly large.
215
-------
BIBLIOGRAPHY
Atkin, Sydney, et .aU, Ind. Eng. Chem.. 53, 705 (1961).
Bjerrum, J.G. Schwarzenbach, and Sillen, L.G., Stability Constants of
Metal-Ion Complexes, with Solubility Products of Inorganic
bubstances7~Cnem.Soc.(London), Spc.Publ.No. 7 (1958).
Brosheer, J.C.; Lenfesty, F.A.; Elmore, Kelly L., Ind. and Eng. Chem.,
39, 423 (1947). a
Buslaeu, V.A and Gustyakoya, M.P. Zh. Neorqa. Khlm. 10, 1524 (-1965) In
John S. Judge, J. Electrochem. Soc., 118. 177? (1971).
Chilton, T.H. and Col burn, A.P., Ind. Eng. Chem., 26:1183 (1934).
Considine, D.M., Chemical and Process Technology Encyclopedia,
McGraw-Hill Co., New York, 1974.
Crosby, N.T., J. Appl. Chem.. 19, 100 (1969).
Cross, F.L. and Ross, R.W., J. Air Pollution Control Assoc.. JJJ, No. 1.
Dahlgren, Sven-Eric, "Chemistry of Wet Process Phosphoric Acid
Manufacture", Phosphoric Acid, edited by A.V. Slack, Vol I, Part I.
pp. 91-156,
Development Document for Effluent Limitations Guidelines and New Source
Performance Standards for the Phosphorus Derived Chemicals Segments
of the Phosphate Manufacturing Point Source Category.
Development Document for Interim Final Effluent Limitations Guidelines
and Proposed New Source Performance Standards for the Other Non-
Fertilizer Phosphate Chemicals. Segment of the Phosphate
Manufacturing Point Source Category, EPA sp. EPA 440/1-75/043,
Jan. 1975.
Development Document for Proposed Effluent Limitations Guidelines and New
Source Performance Standards for the Basic Fertilizer Chemicals
Segment of the Fertilizer Manufacturing Point Source Category, EPA,
sp. EPA 440/1-73-011, Nov. 1973.
Economic Analysis of Proposed Effluent Guidelines, EPA sp.
EPA 230-1-74-043, Sept. 1374.
EPA, EPA-440/l-74-006-a, Jan. 1974.
ESE, Inc., Personal Communication from J.C. Kutt to Jack Sosebee (1975).
216
-------
Rossotti' F'J'C-> J- Inorq- Nuclear Chem. 26, 1959
Forster, J.H. "Direct Contact Evaporation", Phosphoric Acid, edited by
A.V. Slack. Vol. I, Part II. pp. 579-607. -
Fox, E.J., Stinson, J.M., and Tarbutton, G., Superphosphate. U.S. Dept.
of Agriculture and Tenn. Valley Authority. 1964. Chapter 10.
P • uO / •
Getsinger, J.G., "Hemihydrate by the Foam Process", Phosphoric Acid
edited by A.V. Slack. Vol. I, Part I. p. 369. - -
Guthrie, K.M., "Capital and Operating Costs for 54 Chemical Programs "
Chem Eng.. June 14, 1979. 140-156.
Harbe<*» E-G'« Jr*' USGS Prof. Paper 272-E, U.S. Government Printing
Office, Washington D.C., 1962.
Hein, L.B., "Removal of Impurities (from Phosphoric Acid)," Phosphoric
Ac_[d, edited by A.V. Slack. Vol. I, Part II, pp. 687-708. -
Huffstutler, K.K., "Pollution Problems in Phosphoric Acid Production "
Phosphoric Acid, edited by A.V. Slack. Vol. I, Part II
pp. 727-737. '
Inorganic Fertilizer and Phosphate Mining Industries Water Pollution and
Control, EPA, 12020 FPD 09/71. Sept. 1971.
Judge, John S., J. Electrochem. Soc.. 118, 1772 (1971).
Kern Edward F. and Jones, T.R., Trans. Am. Electrochem. Soc. 49.
t / «3
King, W.R., and Farrell , J.K., Fluoride Emissions from Phosphoric Acid
Gypsum Ponds. EPA sp. EPA bb(J-^-M-(J9b. October,
Legal, C.C., and Myrick, O.D., "History and Status of Phosphoric Acid."
Phosphoric Acid, edited by A.V. Slack. Vol. I, Part I. pp. 14,
«it , 4U«
Long, Harold, Personal communication, Feb. 1975.
Lutz, W.A. and Pratt, D.J., "Principles of Design and Operation"
Phosphoric Add, edited by A.V. Slack. Vol. I Part I '
pp. 159-212.
Mesmer, R.E., and Baes, C.F., Jr., Inorg. Chem, 8, 6(1969).
Munter^Paul A.^Aepll. Otto T., Kossatz, Ruth A.. Ind. and Eng. Chem..
217
-------
Munter, Paul A., Aepli, Otto T., Kossatz, Ruth A., Ind. and Eng.
Chem.. 41, 1504 (1949). a~
Murakarrri, K., Hari, S., "Hemihydrate-Dihydrate Processes in Japan,"
Phosphoric Acid, edited by A.V. Slack. Vol. I, Part I. p. 287.
Pasquill, F. Atmospheric Diffusion. John Wiley & Sons, New York, 1974.
Sanders M.D., "Recovery of Fluorides as By-Products," Phosphoric Acid.
edited by A.V. Slack. Vol. I, Part II. pp. 765-778.
Schlichting. H. Boundary Layer Theory. McGraw-Hill, New York. (1955).
Lnapter 21. '
Shindo Kazuo S. Harada Y., "NKK Process," Phosphoric Acid. Edited by
A.V. Slack, Vol. I, Part I. pp. 328, 329^—^
Shireye. R.N. Chemical Process Industries. (16). McGraw-Hill Co., New
iorK, iyb/.
Sho1t"' £-s:> — T^' "Evaluation and Modification of Fluoride Sampling
and AnalyticaTllethods." EPA-650/2-73-007 (October, 1973).
Slm0nYork*U954)" -1uor1ne Chemi'strv- Vols. I-III, Academic Press, New
Acid, Vol. I, Part I,
Stumn, Werner, and Morgan, James J., Aquatic Chemistry.
Wiley-Interscience, New York (1970J:: - -
Takeuchi H. Tayama, I., "Mitsubishi Process," Phosphoric Acid. Edited
by A.V. Slack, Vol. I, Part I. p. 301. - -
* v£a.ramf.ters Which Inf1"e"ce Fluoride Emissions from Gypsum
Pn.u. Dissertation, University of Honda (1970). - -
Teller, A.J. and Reeve, David, "Scrubbing of Gaseous Effluent."
Phosphoric Acid. Edited by A.V. Slack, Vol. I, Part II, pp. 741-778.
Thiesenhusen, H., Gesundh-Ing. 53, 113-19, 1930.
' : Workbogk, ?f ...r^.on..smes- '-
h Service Publication, No. y99-AP-26, May 1970.
U.S. Department of Health, Education and Welfare, Atmospheric Emissions
from Wet-Process Phosphoric Acid Manufacture. National AIV -
Pollution control Administration, Raleigh, N.C. (1970).
Yasuda, T., Miyamoto, M. , "Nissan Process." Phosphoric Acid. Edited bv
A.V. Slack, Vol. I, Part I. pp. 307, 30?: — ^ - y
218
-------
TECHNICAL REPORT DATA .
(Ptvasc read Instructions on the reverse before completing)_
RtPORT NO.
EPA-600/2-78-124
2,
4. TITLE AND SUBTITLE Evaluation of Emissions and Control
Techniques for Reducing Fluoride Emissions from
Gypsum Ponds in the Phosphoric Acid Industry
'IEN
iCCES
REPORT DATE
June 1978
. PERFORMING ORGANIZATION CODE
7. AUTHOH(S)
A. A. Linero and R. A. Baker
. PERFORMING ORGAN
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Science and Engineering, Inc.
P.O. Box 13454
Gainesville, Florida 32604
10. PROGRAM ELEMENT NO.
1AB604
11. CONTRACT/GRANT NO.
68-02-1330, Task 3
12. SPONSORING AGENCY NAME AND ADDRF.SS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND
Task; 12/73-1/75
PERIOD COVERED
14. SPONSORING AGENCY CODE
EPA/600/13
,9. SUPPLEMENTARY NOTES IERL-RTP project officer Edward L. Wooldridge is no longer witn
EPA. For details contact R.A. Venezia, Mail Drop 62, 919/541-2547.
16. ABSTRACT
repor|. g^es results of a study of gaseous emissions from gypsum
disposal and cooling water ponds to determine their potential as sources of airborne
fluorides from the manufacture of phosphoric acid. A model of the chemistry within
the pond environment was developed. Previous emission rate studies were evaluated
with respect to assumptions, methodologies, and conclusions: none provided a suf-
ficient basis for determining a fluoride emission factor. The ponds, found in major
phosphate production areas (e.g. , Florida, North Carolina, and the Western U.S.),
usually cover several hundred acres. Process water enters a pond at over 100 F,
and leaves at lower temperatures after evaporative cooling. Ambient concentrations
near a typical pond were calculated by predict! ve modeling methods , assuming
emission rates of 0.1, 1, and 10 Ib per acre per day. At the higher rates, control
of fluoride emissions appears necessary, based on TLV criteria adjusted for the
general population. An analysis of possible control methods indicated that liming
is too expensive an alternative. Capital costs are too high for the complete Kidde
process. The most promising method appears to be the Swift process for fluoro-
silicic acid recovery, coupled with segregation of the cooling and gypsum pond
waters. The hemi/dihydrate process also appears to be promising. _ __
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
Pollution
Gypsum
Ponds
Phosphoric Acids
Industrial Processes
Fluorides
Mathematical Models
Calcium Oxides
c. COSATI Held/Group
Pollution Control
Stationary Sources
Gypsum Ponds
Liming
Fluorosilicic Acid
Kidde Process
Swift Process
13B
08G
08H
07B
13H
12A
13. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This:Report)
Unclassified
21. NO. OF P,
228
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73J
219
-------