Gaseous Fluoride Emissions from Gypsum Settling and Cooling Ponds
Florida Scientist: Quarterly Journal of the Florida Academy of Sciences
Vol. 50 Spring, 1987 No. 2
Gaseous Fluoride Emissions from Gypsum Settling and Cooling Ponds
Howard E. Moore
Department of Chemistry, Florida International University, Miami, FL 33199
ABSTRACT: Previous estimates of hydrogen fluoride fluxes from cooling and gypsum settling ponds associated with the manufacture of phosphate fertilizer are reviewed and a theoretical estimate based on vapor pressures of HF-H2O solutions is presented. The latter yields fluxes from 122 to 195 kg HF/day for a 450 metric ton P2O5/day plant. Sixty percent or more of the total plant release of HF is due to the ponds. Derived atmospheric residence times for HF (1 to 5 hr.) indicate that fluoride is dispersed throughout Central Florida at ppb levels in particulate form and may be an environmental hazard to adjacent agricultural industry.
The primary purpose of this paper is to give a brief review of the literature relating to the emissions of fluorides, i.e., hydrogen fluoride and silicon tetrafluoride, from gypsum settling ponds and cooling ponds associated with phosphate fertilizer production in Central Florida (Fig. 1). A theoretical estimate of these emissions is presented for comparison to experimentally determined values. An excellent review of atmospheric fluoride emissions and the effect of fluoride on plants and animals was compiled by the National Research Council, National Academy of Sciences (NAS, 1971). A study of emission control practices in the phosphate industry was published by the Department of Health Education and Welfare (Crane et al., 1970). Based upon data obtained by Crane and coworkers (1970), the National Academy of Science study group (NAS, 1971) stated that vaporization from the ponds may constitute 90% or more of the total fluoride emissions to the atmosphere. In 1972, emission control regulations were promulgated in Florida for all phosphate production operations except the ponds (State of Florida, 1985). As a consequence of recognition by the industry that fluorides are hazardous to workers and in response to these regulations, fluorides that were formerly allowed to escape to the atmosphere in the operations are now efficiently trapped and conveyed to the settling and cooling ponds, where their ultimate fate is largely unknown (State of Florida, 1983). It now appears probable that these ponds contribute a large majority of the overall fluoride emissions from the phosphate industry to the atmosphere. The fate of fluorides in the atmosphere is poorly known. They may have injurious effects on the various agricultural industries adjacent to phosphate mining areas. This is discussed further below.
PHOSPHATE FERTILIZER PRODUCTION - Sauchelli (1960) stated that, despite the more-than-a-century period in which superphosphate has been manufactured in all parts of the world, the chemistry of the acidulation reaction is still not precisely known. This is still true to some extent today. The composition of Florida pebble phosphate is given in Table 1 (Sauchelli, 1960). The main component in the phosphate ore is fluorapatite, 3[Ca3,(PO4)2]*CaF2. The ore appears to contain excess fluoride as CaF2 or as fluorosilicates. The important reactions in the process appear to be as follows (Sauchelli, 1960):
(1) 3[Ca3(PO4)2] CaF2 + 10H2SO4 > 6H3PO4 + 10CaSO4 + 2HF
(2) H2SO4 + CaF2 > CaSO4 + 2HF
(3) 4HF + SiO2 > SiF4 + 2H2O
(4) 4H3PO4 + Ca3(PO4)2 > 3Ca(H4PO4)2
From these reactions, it appears most of the fluoride is released as SiF4. This is passed through scrubbers and converted to fluosilicic acid, H2SiF6 (Waggaman, 1969), (Eqn. 5).
(5) 3SiF4 + 4H2O > SiO2 2H20 + 2H2SiF6
A portion of the fluoride remains with the superphosphate. This amount varies with the strength of the sulfuric acid used and is from 25 to 50 percent of the fluoride originally present (Sauchelli, 1960); Waggaman, 1969; Crane et al., 1970). The remainder of the fluoride is transferred to the gypsum settling ponds.
Table 1: Composition of Florida Pebble Phosphate
CaO -- 46-50%
P2O5 -- 30-36%
CO2 -- 1.5-4.4%
F -- 3.3-4.4%
SiO2 -- 7.3-9.8%
Other Oxides -- 1.7-5.7%
H2O -- 0.3-2.6%
State of Florida (1985) emission standards, which apply to the various steps in the manufacture of phosphate fertilizers, allow for a total emission of 0.21 kg of fluoride per metric ton of P2O5 produced. While there is probably no such thing as a typical phosphate complex, a model plant is assumed (Cross and Ross, 1969) that produces 450 metric tons of P2O5 per day and thus is allowed to discharge to the atmosphere 95 kg of fluoride (State of Florida, 1985). As mentioned above, these regulations do not cover emissions from the settling and cooling ponds. The model plant above has associated with it 65 ha of settling ponds and 40 ha of cooling ponds. Measurements by Schiff and coworkers (1981) show both types of ponds have similar fluoride contents.
Sauchelli (1960) states that under certain conditions the fluosilicic acid may decompose. Thus, the ponds may appear to be in a quasi-equilibrium of the type:
2HF(aq) + SiF4(aq) = H2SiF6(aq)
H2SiF6(aq) = H+(aq) + HSiF6-(aq)
HF(aq) = HF(g)
SiF4(aq) = SiF4(g)
This is in accord with the fact that HF and SiF4 are found in the vapor state over solutions of fluosilicic acid (Jacobsen, 1923) and that fluosilicic acid cannot be isolated in the pure state (Colton, 1958). The ratio of HF to SiF4 in the gaseous state increases with a decrease in fluosilicic acid concentration (Baur, 1903). The concentrations of soluble fluoride in the ponds range from 4,000 to 14,000 ppm with value over 10,000 ppm being quite common (Cross and Ross, 1969; Crane, 1970; Tatera, 1971. Schiff et al., 1981).
Most of the measured fluoride must exist as HF. Pond water analyses are usually performed with a fluoride specific electrode on samples that are either made basic or are highly diluted with water thus raising the pH. Fluoride electrodes are not sensitive to combined fluoride such as HF or SiF4, but are sensitive to OH- at pH greater than 8 (Skoog, 1985). Also, the ponds generally have pH values between 1.2 and 1.8 (Schiff et al., 1981). For a pond with 10,000 ppm soluble fluoride and pH of 1.5, 98% of the fluoride exists as HF rather than F (K for HF = 7.2 * 10-4 at 25 C; Weast, 1979). This analytical technique and calculation gives no indication as to the concentration or volatility of SiF4. "Surveillance is advisable (Crane et al., 1970)" as, "the fate of the material is largely unknown (State of Florida, 1983)."
FLUORIDE FLUX ESTIMATES -- Three previous estimates of fluoride emissions from ponds, which are based on experimental measurements of atmospheric HF over ponds, have been made. Crane and Ross (1969) installed a "greenhouse" with a stack and exhaust fan directly on a settling pond. They then pulled air through the assembly at a rate which gave concentration values in the enclosure equal to those on the bank adjacent to the pond. They then estimated the emission rate to be 11.8 kg/day for 65 ha pond. The 40 ha cooling pond would be expected to add 7.3 kg/day for a total of 19.1 kg/day. Crane and Ross stated that they thought these to be minimum emission values.
Crane and coworkers (1970) reported on a similar experiment in which air was drawn at a known rate through an enclosure over a twenty foot long air, gypsum-pond water interface. The values obtained were shown to be temperature dependent and ranged from 0.56 kg/ha day at 50 C to approximately 2.8 kg/ha day at 80 C. For the example above these would lead to 59 to 294 kg/day. Crane and coworkers (1970) state that they do not believe the air reached its ultimate saturation value and that the values are low. The water as it enters the pond may be at somewhat higher temperatures initially, thus fluxes may be much higher than over the pond as a whole.
Tatera (1970) made measurements on pond water that was transported to a laboratory where studies were conducted in a wind tunnel. He also could vary wind speed and temperature. He determined emission rates could vary from 0.36 to 20.5 kg/ha day, the actual emission rate being very dependent on temperature and wind speed and less dependent on fluoride content of the pond water. He also stated that other mechanisms such as perculation and photochemical reactions must account for additional emissions of fluorides to the atmosphere. Measurements by Tatera (1970) yielded 12.1 kg/ha day for a typical pond. This would result in 1270 kg/day emission from the ponds in the example above.
Grant (1984) stated that about 100 kg of fluoride are produced per metric ton of P2O5. Thus for a 450 metric ton/day plant, 45,000 kg of fluoride would be processed and 1270 kg would be only a 2.8% release to the atmosphere. These authors believe their fluxes to be the minimums and because they vary so widely, an independent method of estimation seems appropriate.
Very approximate estimates of the fluoride flux from gypsum ponds can be made on the basis of HF vapor pressures over HF-H20 solutions and H2O evaporation rates along with a consideration of atmospheric conditions. The use of HF vapor pressures from HF solutions rather than fluorisilicic acid solutions appears as the concentration of HF in the pond waters is measured by the fluoride specific electrode as mentioned above.
VAPOR PRESSURE OF HYDROGEN FLUORIDE SOLUTIONS -- The vapor pressure of water and hydrogen fluoride over pure hydrogen fluoride solutions has been measured by Fredenhagen and Wellman (1932); Khaidukou and coworkers (1936); Brosheer and coworkers (1947) and Munter and coworkers (1949). The best data at lower concentrations appear to be those given by Brosheer and coworkers (1947). These data are given in Fig. 2 and 3 for temperatures of 25 C. Tatera (1971) also has given data, but his values appear low.
Waters in the settling and cooling ponds are saturated with several calcium salts and hence are not pure HF-H2O solutions. However considering them as pure solutions the mole fraction of HF corresponding to 10,000 ppm fluoride is 0.010. Addition of other solutes will tend to reduce the mole fraction of HF, but to what extent is not certain. More important may be the nature of the solutes and the pH of the solutions. As mentioned above, at a pH of 2 or below most of the fluoride will be in the form of HF and not F-. Molecular HF is volatile while F- is not. Other solutes in the pond may cause a salting-out effect and increase the vapor pressure of HF over the ponds. Field measurements of the vapor pressure of HF over the actual cooling and settling ponds need to be made.
The vapor pressure of water above the cooling and settling ponds will be effected by the same factors as for hydrogen fluoride. The biggest variable in determining both vapor pressures may be temperature as shown (Fig. 2 and 3).
WATER EVAPORATION RATES -- The flux of hydrogen fluoride to the atmosphere can be estimated from the evaporation rate of water and the vapor pressures of hydrogen fluoride and water (MacKay and Walkoff, 1973). The estimation of evaporation rates is not an easy task (Chow, 1964). The evaporation rate of water depends on the temperature of the water and air, wind velocity, atmospheric pressure, chemical purity of the water, and the depth of the water body.
Evaporation rates can be estimated in several ways. The two most common methods are from averages of evaporation as determined by evaporation pans and from evaporation equations based on modifications of Dalton's law.
The most commonly used pan is the U.S. Weather Bureau Class A Land Pan (Chow, 1964). Evaporation rates using this pan have been published by the Weather Bureau (Kohler et al., 1955). True evaporation rates for open ponds and shallow lakes are generally about 0.70 of the pan value. Evaporation rates depend upon the depth of the pond or lake. Deeper lakes have lower evaporation rates. Also the evaporation rate decreases as the salinity increases. About 1 percent decrease in the evaporation occurs with 1 percent increase in specific gravity. The evaporation rate occurs with 1 percent increase in specific gravity. The evaporation rate of ocean water is thus 2-3 percent lower than fresh water. This may have a decided effect on estimates of evaporation from settling ponds, but may be offset by the shallow depth of these ponds.
Evaporation rates derived from pan measurements are 152-165 cm/yr for Central Florida. Actual measurements on lakes are given as 122-132 cm/yr. The 0.70 factor applied to the pan values yields 114 cm/yr or 0.31 cm/day. No correction is made for the salinity of the ponds.
Evaporation rates for shallow ponds and lakes can also be determined from a Dalton's law type equation given by Rohwer (1933):
(10) E = 1.958 {(1.465-0.0186B) (0.44 + 0.190W) [exp Pw/Pa)]}
where B is the barometric pressure, W is the wind velocity, and Pw an Pa are the saturation and actual water vapor pressures. B, Pw, and Pa are in Torr, W is in km/hr, and E in cm/da.
Schiff and coworkers (1981) give a wind speed of 19.6 km/hr, an air temperature of 85 C and a relative humidity of 95 percent during the times when they were sampling. Using these values (Eqn. 10) gives an evaporation rate of 0.174 cm/da or about half of that estimated from pan values.
The flux of gaseous HF to the atmosphere from the pond may not follow the evaporation rate of water exactly. If equilibrium for both HF and H2O is reached and the air mass over the pond is exchanged with surrounding air, the replacement air may have a high humidity, thus suppressing the evaporation rate of water, but be void of HF. Under these conditions the HF would continue to escape to the atmosphere and the HF flux would be greater than estimated from the evaporation rate of water. Since no information is available, the calculations which follow do not consider salinity or exchange of air over the pond with high humidity air which is void of HF.
ATMOSPHERIC FLUXES OF HYDROGEN FLUORIDE -- The flux of HF from the surface of the pond, F, is given by:
(11) F = E[(PM)HF/(PM)W]
where E is the evaporation rate of water. P and M are the saturation vapor pressures and molecular weights of HF and water. The vapor pressures are both temperature and concentration dependent as shown above in Figures 2 and 3. However, the ratio of vapor pressures is less dependent on temperatures, so the effect of temperature on the flux is largely governed by the change in evaporation rate with temperature.
Using a value of 10,000 ppm HF for a gypsum settling pond (Schiff et al., 1981), the vapor pressures of water and hydrogen fluoride are 23.48 and 0.024 Torr at 25 C. Introducing the evaporation rates calculated above leads to fluxes at 2.1 x 10-4g/m2 min (pan evaporation) and 1.3 x 10-4 g/m2 min (Dalton's law equation). Using vapor pressure values at 40 C gives similar values. A summary of flux values is given in Table 2.
TABLE 2: Summary of HF Flux Data for a 105 Hectare Pond
Cross and Ross (1969) - 19 HF, kg/day
Crane et al. (1970) - 59-294 HF, kg/day
Tatera (1970) - 38-2140 HF, kg/day
Present Estimate - 123-195 HF, kg/day
ATMOSPHERIC CONCENTRATIONS OF HF -- Surface air concentrations may be related to the flux by (Crank, 1956):
(12) C = F/(D•A)1/2
where D is the turbulent diffusion coefficient and A is the removal rate of HF from the atmosphere. The mean residence time or the time to remove 1/e of a species from the atmosphere is given by T = 1/A. Eqn 12 assumes a uniform flux over an infinite horizontal plane. This may not be applicable in the strict sense as the ponds have finite areas and replacement air may be void of HF as mentioned above. However, Eqn 12 should provide some insight into the residence time of HF as related to air concentrations near the settling ponds.
Values of D vary with altitude and generally range from zero at the ground to 1200 m2/min in the open troposphere (Jacobi and Andre, 1963; Ikebe and Shimo, 1972). With a wind speed of 20 km/hr over open land at 1 m altitude, D has an approximate value of 12 m2/min.
The value of the atmospheric residence time for HF has not been determined. It must be very short as HF is a very reactive gas and will undergo deposition on plant and other surfaces including aerosols in the atmosphere. An estimate of the value of T can be obtained from measured values of HF near the ponds through the use of Eqn 12. Schiff et al. (1981) found atmospheric concentrations ranging from 20 to 60 ppb. Earlier values ranged from 0 to 170 ppb (Hendrickson, 1961; Huffstutler and Starnes, 1970). If a value of 60 ppb (50 ug/m3 is assumed, calculated values for T, range from 0.5 min to 2.2 hr.
Data on gaseous HNO3, a similar compound, is available for comparison. Parrish and coworkers (1936) measured particulate NO3 along with gaseous HNO3, NO, and NO2 in a non-urban site west of Boulder, CO. Based on their measurements, they concluded that most of the gaseous HNO3 is removed by deposition rather than by attachment to aerosols and subsequent rainout or washout. A T for gaseous HNO3 was estimated to be between 17.5 and 8.8 hr for the winter and summer months, respectively. The shorter residence time corresponds to more humid conditions in the Rocky Mountains.
Jacob and coworkers (1986) studied the H2SO4 - HNO3 - NH3 system at high humidities and in fogs. They concluded that the residence time of gaseous HNO3 was less than 12 hr. They developed a stirred-tank model to explain their experimental values and adopted a deposition velocity for gaseous HNO3 of 3 cm/s. Measurements of the deposition velocity in clear air over a grass field in Central Illinois yielded a mean value of 2.5% + 0.9 cm/s (Huebert and Robert, 1985). Deposition velocity is related to residence time by
(13) I/T = (v/h) + Aa + Kc
where v is the deposition velocity, h is the atmospheric mixing height, Aa is the removal rate by aerosol attachment and kc is the removal rate by chemical processes. The latter is important for HNO3, but not for HF as gaseous HNO3 undergoes photolysis while HF does not.
Residence times calculated from deposition velocities are very dependent on the atmospheric mixing height. Parrish and coworkers (1986) assumed a mixing height of 1100 m, while Jacob and coworkers (1986) assumed 400 m. These appear to be typical values except under extremely stable conditions when the mixing height may be on the order of 100 m (Wallace and Hobbs, 1977). Values of 300 and 1000 m for the mixing height combined with the data of Huebert and Robert (1985) yield 3.3 and 11 hr, respectively, for T, for gaseous HNO3. Aerosol attachment should result in even shorter residence times as indicated by Eqn 13.
Attachment reactions for ions to particles in the atmosphere with diameters less than 0.1 um is quite rapid. For Rn-222 daughters the attachment rate has a T, corresponding to 0.33 t5o 1.85 min. (Porstendorfer and Mercer, 1980). Once formed, aerosol particles undergo coagulation until they reach a radius of approximately an hour (Nakatani, 1980). The attachment of HF to aerosols may be somewhat slower as attachment in the case of HF depends upon having sufficient NH3 to partially neutralize the acid. HF would not be expected to attach to a highly acidic sulfate particle, however, if attachment rates are similar to deposition rates then T for HF, assuming it is similar to HNO3, would vary from 1.2 to 5.5 hours. Best estimates of T for gaseous HF thus appear to be on the order of 1 to 5 hours.
Once attached to aerosols, fluoride will have a residence time corresponding to the residence time of the aerosols. Based on Rn-222 daughter measurements, Moore and coworkers (1973; 1980) calculated mean residence time for atmospheric aerosols of about 5 days. Based on a long range transport model for sulfate over the Eastern United States, Kleinman (1983) estimated a residence time of 4.4 days. In the same model calculation, Kleinman (1983) found the average distance from source to receptor for sulfate to be 583 km. Thus it appears most of the gaseous HF will be confined to the region of emission, but particulate fluoride may be spread throughout Central Florida. In Sarasota County, FL a mean of 0.14 ppb gaseous fluoride has been determined in ambient air samples for the years 1980-1985; however, no data on particulate fluoride are available for comparison (Sarasota County, 1985).
The above discussion gives no insight into the flux or fate of SiF, in the atmosphere. It is a less reactive gas which should have a longer residence time and be transported over greater distances before reaction. This should be investigated further.
BIOLOGICAL EFFECTS OF FLUORIDE EXPOSURE -- In plants, exposure to elevated fluoride concentrations is most evidenced by the development of leaf margin and tip necrosis. Injury to plants at lower fluoride levels may result in decreased plant dry weight, vigor, leaf size and fruit quality and quantity. The threshold levels at which these symptoms become evident is difficult to ascertain accurately. Possible effects of citrus become of interest in Florida. While citrus, like other species, exhibits a range of tolerances, threshold levels for growth effects appears to be in the range of 50-100 ppm (dry weight) on the leaves.
Concentrations on the leaves of plants can be related to atmospheric concentration by:
ÆL=KCT
where ÆL is the increase in fluoride concentration on leaves above normal background levels due to an atmospheric concentration, C, maintained for time, T. K is the apparent accumulation rate factor. Accumulation is not constant as implied by the equation but appears to decrease with time. Thus exposure to high concentrations for short periods of time may not have the same effect on exposures to lower concentrations over longer periods of time. This is illustrated by the work of McCune (1969) on citrus which indicates that exposure to 1-10 ppb fluoride for 10-12 months will produce foliar markings as readily as exposure to 5,000-10,000 ppb fluoride for 1 hour. McCune suggests an ambient exposure limit of 1 ppb to prevent damage to citrus. Data by Brewer and coworkers (1969) support these findings.
Also of interest is data on plants used as forage for cattle. Data on orchard grass appears typical of other forage crops tested. The work of Benedict and coworkers (1965) indicates a value of K equal to 1.1 for orchard grass at fumigation levels of 0.6 to 0.9 ppb and times ranging from 21 to 120 days. Thus orchard grass would accumulate 50 ppm fluoride in a month at ambient fluoride levels of 1.5 ppb.
Crippling skeletal fluorosis has been observed in cattle exposed to high levels of fluoride (McClure, 1970). A concentration of 40 ppm has been cited as an official allowable maximum of fluoride on plants which are used as forage for cattle (Lewis, 1970). State of Florida (1978) analyses on grass samples from Polk, Hillsborough, and Manatee counties in Florida during the years 1976-78 show mean levels of 9 to 117 ppm for monthly samples. These levels must accumulate airborne fluoride. The Peace River, which flows through Polk, Hardee and Desoto Counties, FL was found to contain 46 ppm fluoride in 1961 (Waldbott, 1978). More recent samples run in the author's laboratory showed 4.2 ppm at Bowling Green, FL and from 30.5 to 48.5 ppm (11 samples) in the remainder of the river down to Arcadia, FL. While these levels are equal to or above those considered injurious to cattle, only two cases of crippling fluorosis in cattle have been reported in Florida in recent years (State of Florida, 1986).
Ambient concentrations of atmospheric fluorides from the phosphate production operations do not appear to be at levels which would be injurious to humans at the present time. The 1 ppb level proposed by McCune (1969) to protect citrus would be adequate for both livestock and humans. The State of Florida does not presently have a standard regulation on ambient fluoride levels in the atmosphere. Some insight into what the level should be for the protection of humans can be gained by comparison to levels for other atmospheric pollutants. Limits for human exposure to several air contaminants are given in Table 4 (Weast, 1979) along with the State of Florida ambient standards. The exposure limits are concentrations above which "exposures...shall be avoided, or protective equipment shall be provided and used." By comparison proposed ambient standards for protection of humans would be in the range of (1) 0.3 ppm for 3 hours; (2) 0.06 ppm for 24 hours; and (3) 0.03 ppm annual mean. These regulations should apply to a combined total of gaseous and particulate fluoride as they appear to be equally injurious.
CONCLUSIONS -- Based on the evidence above, settling and cooling ponds associated with phosphate fertilizer production account for 60% or more of the total fluorides released to the atmosphere in the process. The HF is rapidly attached to aerosol particles and may be dispersed throughout Central Florida in the particulate form. These low levels of fluoride can result in accumulations which may be injurious to plants and animals. Several awards were given to farmers in suits brought against aluminum refining plants for damage due to fluoride emission during the period 1095-1962 (Waldbott et al., 1978). Waldbott and coworkers also suggest that the decline in the cattle industry in Florida from 1954 to 1965 was due to fluoride emissions from phosphate plants.
More information is needed on the airborne concentrations of gaseous and particulate fluorides in Central Florida. More information should be gained on the chemistry of the ponds with the objective of seeking a suitable way to lower fluoride fluxes. Liming is one obvious way to lower fluxes but is too costly. Understanding the chemistry may aid in developing less costly control methodology.
To learn more about the phosphate industry, see www.fluoridealert.org/phosphate/overview.htm
You can help support our efforts to initiate legislative changes
and spread our message.