The Florida Institute of Phosphate Research was created in ...fipr.state.fl.us/wp-content/uploads/2014/12/01-113-138Final.pdf · The Florida Institute of Phosphate Research was created

The Florida Institute of Phosphate Research was created in 1978 by the Florida Legislature(Chapter 378.101, Florida Statutes) and empowered to conduct research supportive to theresponsible development of the state’s phosphate resources. The Institute has targeted areasof research responsibility. These are: reclamation alternatives in mining and processing,including wetlands reclamation, phosphogypsum storage areas and phosphatic claycontainment areas; methods for more efficient, economical and environmentally balancedphosphate recovery and processing; disposal and utilization of phosphatic clay; andenvironmental effects involving the health and welfare of the people, including those effectsrelated to radiation and water consumption.

FIPR is located in Polk County, in the heart of the central Florida phosphate district. TheInstitute seeks to serve as an information center on phosphate-related topics and welcomesinformation requests made in person, by mail, or by telephone.

Research Staff

Executive DirectorPaul R. Clifford

Research Directors

G. Michael Lloyd Jr.Jinrong P. ZhangSteven G. RichardsonGordon D. Nifong

-Chemical Processing-Mining & Beneficiation-Reclamation-Environmental Services

Florida Institute of Phosphate Research1855 West Main StreetBartow, Florida 33830

(863) 534-7160Fax:(863) 534-7165

http://www.fipr.state.fl.us

MAGNESIUM SEPARATION FROM DOLOMITIC PHOSPHATEBY ACID LEACHING

FINAL REPORTPHASE III INVESTIGATION

Douglas H. Laird and Wendy K. Hanson

SCIENCE VENTURES3822 Tiara Street

San Diego CA 92111voice (619) 292-7354FAX (619) 571-3369

Prepared for

FLORIDA INSTITUTE OF PHOSPHATE RESEARCH1855 West Main Street

Bartow, Florida 33830 USA

Contract Manager: G. M. LloydFIPR Project Number: 93-01-113

August, 1997

DISCLAIMER

The contents of this report are reproduced herein as received from the contractor.

The opinions, findings and conclusions expressed herein are not necessarily those of the Florida Instituteof Phosphate Research, nor does mention of company names or products constitute endorsement by theFlorida Institute of Phosphate Research.

ii

PERSPECTIVE

Of all the impurities in phosphate rock magnesium can be one of the mosttroublesome. In the manufacture of phosphoric acid high magnesium increases sulfuricacid consumption and results in lower production rates and yields. When highmagnesium acid is used to make Diammonium Phosphate(DAP) solid fertilizer,supplemental nitrogen in the form of higher cost ammonium nitrate or urea must beadded.

The magnesium problem will grow increasingly critical as the Bone Valleyreserves are depleted and mining activities move into the southern extension with itshigher magnesium phosphate rock. High magnesium will mean that some phosphate rockwill not be mined and just left in the ground and some of the rock that is mined will haveto be discarded. While a number of techniques for magnesium removal have been havebeen proposed, no truly satisfactory solution to the problem has been adopted.

FIPR and others have devoted a great deal of time and money to solving thismagnesium problem. Processes for magnesium removal from both phosphate rock andphosphoric acid have been successfully demonstrated in the laboratory but have failed toreceive wide spread acceptance in the industry. The technique reported on in this papertreats the phosphate rock with a weak acid to solubilize the dolomite in the rock so that itcan be removed as a solution. The magnesium can be recovered as a readily salableproduct that contributes to the economics of the process.

Processes of this type could reduce the number of acres that have to be minedeach year, reduce the cost of producing DAP, and provide a valuable byproduct thatwould contribute to the industry’s financial stability.

iii

ABSTRACT

A chemical process was tested to leach unwanted magnesium from dolomiticphosphate ore with sulfuric acid under automated pH control. Product carbon dioxide isremoved by slurry spray in air or by boiling at reduced pressure. Magnesium is recoveredfor sale in the form of hydroxide or light oxide of good quality. By this method, up to 85%of MgO can be removed from rock with negligible loss of phosphate. Rock so purified isless costly to process into phosphoric acid by the wet process. Such technology mayextend the life of Florida phosphate reserves, as low-magnesium rock is depleted.

Three phases of batch laboratory testing have been completed using a computer tocontrol pH and log data. Neutralization of the leachate with lime was aIso tested, and therate and endpoint were measured. Correlations developed from the results provide thebasis for an economic systems model, The model predicts nearly 35% internal rate ofreturn after accepting a 40% capital cost penalty for innovative technology. The modelpredicts optimum economic return at a leaching temperature of 76°C (169°F), 40grams/100 ml slurry density, pH 3.37, 214 minutes leaching time, 40 minutesneutralization time, using a single tank each for leaching and neutralization. No particlesize reduction is indicated beyond that normally used with phosacid feeds.

V

ACKNOWLEDGEMENT

Science Ventures gratefully appreciates the guidance in planning this investigationand the financial support in its execution provided by the Florida Institute of PhosphateResearch.

International Minerals and Chemicals has kindly provided the dolomitic phosphatesamples for experimentation.

vi

NOMENCLATURE

A Drum filter area, sq meters Aim Theoretical limiting amount of acid which can be absorbed by reaction with

carbonates. Found by linear regression between acid volume and wt % MgO remaining in residue, Arh is the regression line intercept at zero MgO. Units are ml of 3 5.34 normal acid/gram.

Aeq Final acid absorption by reaction with carbonate given unlimited time at particular conditions of pH, temperature, particle size etc. From a correlation based on lab measurements, in ml of 35.34 N acid/g.

c Tank capacity in gallons, used to estimate unit 2 and 3 capital costs. [Cd21 Calcium ion concentration in leachate in moles/liter CP Purchase cost of major equipment items in dollars. d Sedimentation (thickener or clarifier) tank diameter in meters Dr, D2 Characteristic particle dimension before and after milling E f F H Ho

k Jh m n

::

Specific ball milling energy (wet) in kwh/short ton Mass fraction of particulate matter retained on sieve &c,/Alim

Q R s t T V Xi K-1 Y

Z

c

Hydronium ion concentration, mol/liter = exp(-pH) as function of time Final equilibrium hydronium ion concentration in neutralization kinetics Rate constant for acid attack on carbonates, l/mm Rate constant for neutralization with lime, l/min Equilibrium constant Sieve mesh opening in cm Mole ratio of Ca/Mg in leachable carbonates Percent washing efficiency in countercurrent decantation at units 7a and 7b. Saturation vapor pressure of water (atm) under belt filter from Clausius Clapyron formula Molar concentration of dissolved &HP04 Gas constant in cal/mole/OK = 1.987 Mean particle surface/volume ratio, l/cm Time in minutes Temperature in Kelvins Rotary kiln volume in cubic meters Value of model parameter i used in sensitivity analysis Anion (nitrate or chloride) concentration in moles/liter A measure equilibrium leaching acid consumption to limiting acid consumption defined as F/( 1 -F) = A,&Arb-A& General dependent variable in the propagation of error formula, represents IRR in

’ final report section. Prime represents a value estimated from correlation rather than measurement.

xi

EXECUTIVE SUMMARY

Although large quantities of phosphate rock remain in Florida, much of it containsunacceptable amounts of dolomite. Magnesium in the dolomite presents variousdifficulties to the wet phosacid process. Consequently, rock containing over roughly 1%MgO is not currently used.

If means can be found to reduce the magnesium content, some of the rock nowdiscarded will become a resource. Processing costs will be reduced and the life of theFlorida phosphate industry will be extended.

Science Ventures has investigated a chemical process to reduce magnesiumconcentration by leaching with sulfuric acid. Leaching pH is controlled by an electronicfeedback loop, and leaching product CO2 is purged from the solution. With theseconditions, up to 85% of the rock MgO was removed while dissolving typically 1-2% ofthe phosphate values.

Subsequent treatment of the leachate with lime returned nearly all of the dissolvedphosphate to solids for feed to the wet phosacid digester. With such a neutralization step,net phosphate loss is insignificant.

With this process, the leaching solution contains an anion such as nitrate orchloride which forms soluble salts with both Ca and Mg. Each mole of sulfuric acidconsumed in 1eaching precipitates a mole of gypsum and brings into solution up to half amole of magnesium salt.

The magnesium salt solution is treated in a subsequent step with lime toprecipitate Mg(OH)2 for sale as a byproduct, This produces the calcium salt of the chosenanion for recycle into the leaching step.

This report covers the third and final phase of an experimental investigation inwhich batch leaching was tested at laboratory scale. Results from all three phases are usedto develop various correlations relating design parameters to process performance. Theprocess has been named P-Cubed for Phosphate Purification Process.

The correlations form the basis for an economic model embodied as software in aspread sheet. The model computes internal rate of return (IRR), taking into account thevalue of raw and leached rock, byproduct value, chemicals, labor and other operatingcosts and capital. An IRR near 35 percent per annum is predicted after optimizing alldesign and operating parameters. The spread sheet has been delivered to FIPR on diskette.It is fairly simple using this software to test the effect on IRR of changes in assumptions,correlations and prices. This report includes a sensitivity study to evaluate effects ofvarying certain key parameters. The standard error of the IRR is estimated to be 11% perannum.

We have also used the spread sheet to test results of discrete design choices, suchas whether to use post-leaching neutralization, the number of leaching tanks etc. Thisreport includes the best of these combinations of choices. The model predicts optimumeconomic return at a leaching temperature of 76°C (169°F), 40 grams/100 ml slurrydensity, pH 3.37, 214 minutes leaching time, 40 minutes neutralization time, using a singletank each for leaching and neutralization. No particle size reduction is indicated beyondthat normally used with phosacid feeds.

Lab testing during all three phases concentrated on the leaching part of the processcycle. Batch tests were run in 2-liter vessels at a range of temperature, particle size, pHsetting, slurry density, anion concentration and leaching time. Reduced pressure was usedto expel CO2 by boiling at less than 100° C.

Nitrate, chloride, acetate and formate were tried in the role of carrier anion.Nitrate and chloride behaved similarly and both appear acceptable except that the chlorideneutralization product filtered more slowly. Results with the carboxylic acids were notpromising because their reversible dissociation conflicts with automated pH control at pHlow enough for rapid leaching.

Phase II testing additionally included measurements of iron and aluminum leached,which was insignificant. The effect on leaching of dissolved CO;! was measured. NOxcontamination of the CO2 off gas was also measured at 100°C and found to be just overthe Florida standard for gas turbines. We do not know of a NOx standard for phosacidplants. NOx emission is less at the optimum leaching temperature of 76°C. HF emissionwas found to be less than 2% of the limit for a phosacid plant. During Phase II, foamingbehavior was compared for raw and leached rock. Leaching dramatically reduced foamingtendency, as expected from the reduction in carbonate content. Seeding with gypsumcrystal was tested, with no observable effect.

Phase III tests included kinetics and endpoint of post-leaching neutralization.Particle sizes tested were evaluated in units of surface-mean particle diameter, and thiswas correlated to milling energy. Six additional dolomitic rock samples were tested fortheir suitability for leaching, making nine types tested in all. The principal product ofPhase III was the economic systems model and associated process design flow sheet andcorrelations.

2

INTRODUCTION

The highest quality US phosphate resources are being rapidly depleted. Yet greatreserves of less desirable phosphate ore remain in south central Florida. The magnesiumpresent in the dolomitic portion of these ores creates handling difficulties. Magnesiumincreases the viscosity of the phosphoric acid, making filtration difficult. In additionmagnesium compounds formed as a product tend to deposit in pipes. Reversibly-hydratedmagnesium compounds can interfere with handling and weighing of the final fertilizer. Aprocess is needed for the economical removal of magnesium.

Most R&D has been directed at flotation or heavy media separation of dolomitefrom the phosphatic portion of the ore, or ion exchange removal of magnesium fromphosphoric acid. Science Ventures has investigated a method to remove the magnesiumby a chemical leaching process before the phosacid digester. Leaching depends onchemical properties of mineral phases in the ore, but not on microscopic surface structure.This process should therefore be less sensitive to variation in feed characteristics than is

the case with flotation.

BACKGROUND

Similar concepts have been explored previously for dolomite leaching using weakor intermediate-strength acids such as SO2, carboxylic or hydroxysulfonic acids (Abu-Eishah et al 1991, Orlov and Treushchenko 1975, Phillips et al 1980, Soviet patent 1975).Carbonic acid is a weaker acid than phosphoric acid, Acidity of intermediate strength

between carbonic and phosphoric could dissolve carbonates and bring the magnesium intosolution without dissolving the apatite. Thus purified, the apatite ore could be used in theconventional wet phosacid process.

Weak acids, such as acetic acid, partially ionize in water. At high concentration,reversible ionization has the effect of limiting and roughly regulating pH. By choosing anappropriate weak acid, the leaching solution pH can be maintained low enough to leachdolomite, but too high to attack the apatite.

Acid concentration in a solution can now be controlled by accurately regulatingthe pH using an electronic feedback loop. In this case the acid ionization constant isunimportant, and the leaching acid can be selected for economy.

3

The leaching reaction is predicated on a low concentration of dissolved CO;?. Theuseful pH range narrows or disappears if product CO:! accumulates in solution. Forexample leaching using SO2 in a gas-tight system was found to remove Mg incompletelywhile dissolving apatite in significant amounts (Hansen, et al 1985).

Of course leaching is effective only for physically accessible magnesium associatedwith carbonates. Any non-carbonate magnesium, i.e. within the apatite crystal lattice, orphysically occluded dolomite crystals, cannot be treated. Other means of separation suchas flotation or heavy media are subject to these same limitations

Science Ventures has completed three Phases of laboratory study of an acidleaching process. Up to 85% of the magnesium has been separated while losing 1% orless of the phosphate. The Phosphate Purification Process, called P3, would enablephosphate producers to utilize their high magnesium phosphate ores.

In addition to the Phase 3 test results, this report includes an economic analysisand systems engineering model. This analysis is based on correlations developed fromphase 1 through 3 measurements, The economic analysis shows internal rate of return(discounted cash flow rate of return) with optimum values of the design and operatingvariables. The optimized IRR is near 35%.

CYCLIC PROCESS CONCEPT

Phosphate resources of interest contain calcite and dolomite which are leached byreaction 1, lumping these two carbonates together. The mole ratio of calcium tomagnesium in leachable carbonates is designated as n which is usually around 1 or more.Low cost sulfuric acid is used for leaching. For each mole of MgO removed from thepulp, sulfuric acid consumption is 1+n moles.

burning dolomite) is a well established conventional technology used to produce magnesiafrom sea water and well brines. Magnesium is thus generated as a valuable byproduct.

Mg(NO& + Ca(OH)z -+ M&OH)2 + Ca(NO& (3)

Other anions such as chloride can be used in place of nitrate in such a cyclicprocess provided that it forms soluble salts with calcium and magnesium. Phase I and IIinvestigations concentrated on nitrate as the circulating ion because of its moderate cost,and because minor filtration losses to phosacid feed could probably be tolerated and wouldcontribute to the value of product fertilizer. Most types of stainless steel resist nitric acid.

With chloride as the anion, the possibility of NOx emission is avoided, but theremay be corrosion of present-day digester materials. Hydrochloric acid is cheaper thannitric on a molar basis.

COMMERCIAL PROCESS

Figure 1 is the process flow diagram currently envisioned. Solids in the leachingtank are continuously agitated. Carbon dioxide is removed from solution by boiling or byimpeller-driven spray in air. Natural gas is indicated as the heat source to simplifyeconomic analysis, but steam from the sulfuric acid plant might be cheaper. Boiling mightbe done at reduced temperature and pressure, in equipment resembling vacuum flashcoolers. Everything in figure 1 below the belt filter comprises a conventional magnesiaseparation plant, except for waste water treatment unit 11.

Sulfuric acid addition rate is automatically controlled at a pH corresponding to theonset of acid-apatite reaction. Threshold pH may be fine tuned by prior testing of the orebeing processed.

Magnesium salts produced in the leaching tank remain in the liquid leachatefraction until reaching reactor unit 5a. Mg(OH& is then precipitated with lime as inconventional production of magnesia for refractories or in electrolytic production ofmagnesium metal. The fine magnesia solid is concentrated in thickeners, washed and thenfiltered. Wash water requirements are modest because the volume of Mg(OH)2 to bewashed is relatively small compared to phosphogypsum.

For example a ton of phosphorite containing 41% CaO and 3.8% MgO with 80%magnesia recovery will produce 1.3 ton of PG from the rock calcium plus .13 ton morefrom magnesium precipitation, compared to only .044 ton of Mg(OH)z to be filtered. Itmay be possible to use the magnesia wash water for irrigation after neutralization with PGpond water and settling to precipitate fluoride.

5

With this process, rock now discarded as tailings can become a resource. Thelifetime of the Florida phosphate industry will be extended. Byproduct magnesia willbenefit the phosphate industry and the nation.

Carbonaceous phosphorites produce foam when treated with concentrated sulfuricacid in phosacid digesters. This foam is generated by the release of CO2 when associatedcarbonates are attacked by the acid. The organic matter present in the phosphate-bearingminerals can stabilize this foam to such an extent that it interferes with processing.Mechanical defoamers add to capital costs and energy requirements. Alternative chemicaldefoamers add about equally to material costs. It is hoped that these costs will be at leastreduced by prior removal of most of the carbonates using the P3 process. Foaming has notbeen a problem in the laboratory leaching tests.

PHASE I SUMMARY

Phase I consisted of a 6-month preliminary lab testing program to evaluatetechnical feasibility of this leaching concept. Three samples of dolomitic central Floridaphosphate were provided by IMC. These are designated as types 2 through 4, since asmall alluvial sample had previously been numbered as "1". All three rock types wereanalyzed.

Type 2 rock had the highest MgO at 3.8%. Leaching tests showed proportionatelyhigher MgO removal with type 2, although the other samples did not differ greatly. Mostof the testing for all three phases was directed at the type 2 rock.

Although the amount of phosphate leached in phase I was always less than 2%,tests were done to reduce phosphate loss still further by neutralizing the leachate with asmall amount of base. Raw ore was tested using its reactive carbonate content as thebase. Although this was effective, lime was found more effective at reducing dissolvedphosphate concentration. Current thinking is to use slaked lime slurry as the base forneutralization.

Tests 9R and 9B were identical except that in 9B gypsum was added at half themass of feed ore. This was to test effect of seeding to promote crystal growth. There wasno significant effect on either the rate of acidulation or the amount of phosphate dissolved.

A final report on the phase I work was delivered to FIPR in October of 1991 undercontract 91-01-093.

PHASE II SUMMARY

A final report on the phase II work was delivered to FIPR in September of 1993 asa continuation of contract 91-01-093. The second phase extended testing of leaching andits sensitivity to variation in operating parameters. This is reported in the section on TestResults. In addition, a variety of technical issues affecting the practicality of this concept

7

were addressed, They are summarized in this section. The preliminary economic analysisdeveloped in phase II is superseded by this report.

Specialized tests were run during phase II to measure the effects on leachingperformance of dissolved CO;1 concentration, acid addition rate, final pH settings and pulpdensity. Miscellaneous tests were performed to determine foaming behavior, fluoride andnitrogen in off gases, product separation and minor element contaminants in themagnesium byproduct.

Foaming Tendency Reduced by Leaching

Except for run 28 with ultra-fine feed, no foaming was observed during theseleaching experiments. Perhaps the organic matter which stabilizes foam in digesters isassociated principally with the phosphate minerals rather than with the carbonates, Then itwould not be released during leaching. In addition, the gradual rate of acidification withleaching provides CO2 time to escape. This would explain why leaching did not normallyproduce foam.

A simple test was done to compare the foaming tendency of raw and leached rock.Type 4 had foamed more than types 2 and 3 in carbonate assays, so type 4 was selected

for study. Rapid strong acidification of the stirred slurry caused 16% volume expansion ofthe slurry due to foam. When a similar sample of leached rock was acidified, there was nomeasurable rise in the slurry surface. Apparently leaching avoids foaming problemsaltogether, in both leaching and digester steps.

NOx Emission

NOx was analyzed in off gas from leaching tests 19R and 19B, both at 1OO’C. Test19B had a .04 molar nitrate concentration and emitted 3.3 grams NOX per mt rock. Test19R with .01 molar nitrate emitted 0.80 g/mt. These figures show that NOx emission is inproportion to nitrate concentration.

A slurry density of 40.5% leads to a steady state N03- concentration of 0.746molar, which corresponds to roughly 61 grams of NO,/mt at 100°C based on theforegoing tests. A 1000 TPD P205 plant would consume just over 1 million tons ofrock/yr. Therefore NOx emission in CO2 produced by leaching will be about 61 mt/yr.This is more than the standard for gas turbines in Florida which has been quoted at 40mt/yr. We do not know of a NOx standard specifically for phosacid production. NOxmeasurements are not available at the optimized process temperature of 76°C, butsubstantial reduction in NOx emission is likely at the lower temperature.

Fluoride Emission in Off Gas

Gas produced in tests 16R, 17R, 18R, 18B 19R and 19B was contacted withdistilled water which was then analyzed for fluoride. All of these leaching tests were done

8

at 101°C with types 2, 3 and 4 rock and variations in slurry density. The highest of thesesix fluoride emission measurements for test 18R was 0.14 gram of F per mt of P2O5.

The EPA has established various limits for fluoride emission from phosphateplants, depending on the type of end product. The limit for a phosphoric plant acid is 10g/ton P205. Our highest HF emission measurement corresponds to about 1170 of thestandard for phosacid plants.

Filtration of Leached Residue

It is desirable to dewater leached solids to minimize water input to the phosaciddigester. Washing and dewatering will also reduce nitrate transfer to the fertilizer andminimize the possible release of nitrogen oxides in downstream equipment. Table 1 showsthe degree of dewatering measured with suction filtration on the Buchner funnel.

Optimum slurry density is about 40.5% (grams solid per 100 ml slurry). This issuitable for direct filtration without a preliminary sedimentation step. Leachate is easier tofilter than phosacid because its viscosity is essentially that of water, as measured bydrainage through a capillary.

Filtration time was measured at the end of phase III using optimization results, butit is reported here for convenience. Type 2 fine leached residue from test 18B was used.Eleven grams of powder were mixed with leachate at 44% slurry density to make 25 ml ofslurry. Particle agglomerates were broken up with a high-speed blender, The agitatedmixture was quickly filtered under suction on three layers of Whatman 41 fast ashlesspapers, 42.5 mm dia. Free liquid disappeared from the surface at 155 sec, and the lastdrop fell from the Buchner funnel at 205 sec leaving a filter cake 7.5 mm thick.

9

Nitrate is not Adsorbed on Residue

It is conceivable that nitrate might adsorb to the surface of leached residue. Thiswould cause a loss of nitrate from the leaching cycle, while carrying unwanted nitrate intothe digester. This question was addressed by analyzing the residue.

Leached ore was washed thoroughly with water which was combined for analysis.Duplicate tests showed that nitrate rinsed from the solid agreed closely with the expectedconcentration of NOS‘ in leachate surface moisture. Thus, there is no evidence for nitrateadsorption onto leached solids.

Effects of Dissolved Carbon Dioxide Concentration

It is expected from the law of mass action that high concentrations of productcarbon dioxide in the slurry will inhibit further leaching. In that event, it could becomenecessary to reduce pH to promote dissolution of carbonates. Reduced pH will tend tobring more phosphate into solution and interfere with the quality of separation.Experiments in phase II were designed to test effects on leaching of deliberately elevatedcarbon dioxide concentrations.

Tests 26 and 27 were run at 75°C and were alike in all conditions except fordissolved CO2 concentration. Pure CO2 was bubbled through the slurry in run 26. Tenpercent CO2 in air was bubbled through the slurry in run 27, which consequently absorbed1.95 ml more acid in the same time. With this and other data, an approximate equilibriumconstant was derived for the attack on carbonate.

CaMg(cO& + 4H ++ Ca’2 + Mg’2 + 2H20 + 2C02

KEQ = = 4.5x106

[H’14

4)

5)

Minor Elements in Byproduct Magnesia

Some of the iron and aluminum in phosphorites is dissolved in the strongly acidicconditions of digesters. If these elements dissolve during leaching, they will precipitatewith the magnesia and interfere with the value of that byproduct, However, the mildconditions of leaching should greatly reduce amounts of Fe and Al dissolved.

Product leachate from type 2 ore in test 17R was precipitated with lime to producemagnesia. The dried magnesia was then analyzed for iron and aluminum by atomic

10

absorption. Iron was undetected at the 20 ppm level, and the aluminum content was.04%. This corresponds to leaching of 0.27% of the aluminum content of the rock, andless than 0.006% of the iron.

Some of the leachate from run 17R was equilibrated with fresh feed to raise its pH.This is a mild form of post-neutralization. Equilibrated leachate contained 0.017%

fluoride, compared to 0.04% in unequilibrated leachate. Use of lime in place of fresh feedwill have a stronger effect. However, fluoride is probably not significant as a contaminantin the magnesia at this level.

11

METHODOLOGY

FEED CHARACTERIZATION

Dolomitic phosphate ores from International Minerals and Chemicals were kindlyprovided by G. Michael Lloyd of FIPR. The nine phosphorites are designated as Types 2,through 10. Type 1 was a dark sandy material used only in preliminary tests prior tophase I.

Types 2-10 were damp mixtures containing coarse pebbles up to roughly 1 cmdiameter along with finer sandy mineral. Microscopically, each type appeared highlycrystalline. Colorless quartz crystals were mixed with amber phosphate and carbonate.Black inclusions appeared to be organic. The black material appeared primarily onsurfaces when these ores were pulverized, suggesting that large particles break most easilyat the internal black surfaces. Type 2 ore contained microscopic marine fossils. Type 4was darker than the others, with greater foaming tendency.

Oneida Research Service (ORS) and Science Ventures (SV) analyzed the ores forvarious elements. Only magnesium content was analyzed for types 5-10, by Huffman Labsusing atomic absorption. Results are listed in table 2. A charge balance was computed bySV to estimate the CaO in types 2-4 ore, assuming .018 moles alkali/100 g ore and 1/9mole of OH- or F- per mole Pod.

FEED PREPARATION

The ores were all dried at 105°C and disk milled to pass 50 mesh. At this step, thepowders were mixed with care and sampled with a rotating riffler. Previous work showedthat coarse phosphate powders spontaneously separate by particle size and then givenonuniform samples for analysis. Riffled powder samples were then tumbled dry with 1/2inch Burundum rods in a rubber barrel ball mill of 12 pound capacity to reduce particlesize.

Type 2 “coarse” was produced by cyclically disk milling and sieving until all of itpassed 50 mesh. Type 2 “medium I” was prepared from the Coarse grade by tumbling in aball mill dry for 15 hours. “Fine” type 2 ore was prepared from medium I by ball millingan additional 19 hours.

13

Type 2 “medium II” was prepared during phase II of this project by ball milling dryfor 12 hours, Most of the testing used this 2 medium II material. This product was usedto prepare “very fine” by tumbling wet with 1/4 inch Burrundum rods for about 48 hours.Types 3 through 10 were prepared similarly to 2 medium II. The dried mineral was diskmilled until all passed 50 mesh, and then a representative portion was tumbled dry with 1/2inch burrundum for 12 hours.

PARTICLE SIZE MEASUREMENT

It is necessary for process optimization calculations to express particle finenesswith a single parameter. We have used the inverse of the area-mean diameter, designatedS. This is equivalent to dividing the total particle surface area of an ore sample by the sumof its particle volumes.

Because of the fineness of some of these feed materials, particle size was measuredby the Blaine method of air permeation. By this method a cylindrical sample of powder iscompacted by a standard procedure. A known volume of air is then passed through thesample at known pressure and the time is measured. The Blaine apparatus was calibratedversus a standard powder of known particle size.

Sieve analyses were used to verify and supplement the Blaine measurements. Ballmilling times and all the particle size data are shown in Table 3.

14

LEACHING EQUIPMENT

A schematic of the laboratory leaching equipment is shown in figure 2. Two suchleaching systems were built to operate simultaneously. The slurry was sealed in one gallonglass beakers heated with silicone heater tapes. Because of breakage, later experiments(runs 9-28) were performed using heated oil baths, and then with 2-liter glass pressurereactors in mantles. A shaft driven glass stirrer kept the solids in suspension with vigorousmixing. The slurry contacted no metal,

As noted in the introductory section of this report, Hanson 1985 showed that CO;?.produced by carbonate leaching suppresses further leaching by the law of mass action, Toavoid this limitation, these experiments were designed to continuously expel CO2 byboiling, except for runs 24-28 which used controlled-CO2 gas bubbling. Water vaporizedby the gas stream was returned to the slurry by refrigerated condensers. Leaching tests 30and 46 which were below 45°C also did not use boiling to expel CO2. Instead, a stream ofair was bubbled through the slurry. This removed some water by evaporation, which wasreplaced by a stream of distilled water. That water was used to dilute the leaching acid.

When boiling at less than 100°C, the slurry temperature was controlled indirectlyby the pressure setting. The heaters were then set for maximum reflux short of floodingthe condenser/mixer. Programmable temperature controllers were used with gas/airbubbling experiments 24-28, 30 and 46.

Water jacketed reflux condensers were optionally supplemented by refrigeratedrefIux condensers mounted above. Concentrated sulfuric acid (17.93 M) was introducedat the tops of the condensers. The lower halves of the condensers were filled with 6 mmglass beads to promote mixing reflux water with acid, so that the acid had been diluted at

15

least 20-fold before contact with slurry. A commercial leaching process would userecirculating filtered leachate rather than distillate for dilution of leaching acids.

The pH was monitored using an industrial probe, Cole Parmer model 27003-00.Temperature compensation was automatic via 100 ohm Resistive Temperature Detector(RTD). The pH signals were recorded by a computer which relayed control signals to theacid-dispensing syringe pumps. These syringe pumps are made from 50 ml burettes, sothat acid consumption can be read directly from their scales.

The two leaching systems are designated R and B for the red and blue lines ontheir strip charts. Sample 10R2 for example is the second intermediate sample taken fromthe red leaching system during run 10. After run 28, the red and blue leaching systemswere no longer run simultaneously, and the R and B designations were dropped.

In preparation for a run, the reaction vessels were heated long enough to establishsteady state reflux. Then experimental periods with computer controlled addition of acidbegan.

Periodic grab samples of the slurry were removed via a sampling valve and diptube. Vacuum (if any) was momentarily released and 5-15 ml of slurry were drawnquickly into a syringe. An effort was made to collect a representative sample of the slurrysolids, including large particles. During Phase 3, the sampling syringe was actuated byvacuum designed so that sample size and withdrawal rate were uniform.

ANALYSIS OF PRODUCTS

“Intermediate” samples taken at periodic intervals during each experiment and thefinal solutions were analyzed for phosphate content, After letting solids settle out formore than an hour, the liquid to be analyzed was passed through a 0.45 micron syringefilter. It was then diluted as necessary and treated with Hach molybdovanadate reagent at1 drop/ml diluted sample.

Measurements for phosphorus were performed using a Milton Roy Spectronic 20Dspectrophotometer at 430 nm At least one calibration standard was measured with eachuse of the spectrophotometer. A background measurement was taken for each leachatesample. The analyses were precise to plus or minus .005 millimoles/liter or .4 PPM or.001% of total feed phosphorus, based on the standard deviation of our calibrationstandards.

The samples were diluted, and calibration standards were prepared with acalcium/magnesium nitrate solution rather than plain water. Thus the nitrate concentrationwas always the same as for undiluted leaching solution, and any slight backgroundabsorbance due to nitrate is controlled.

17

Some of the solid residues were analyzed for magnesium by contractorsEnvironmental Engineering Laboratory (EES) and ORS to correlate magnesium removalwith acid consumption. Solids from each of the grab samples from run 5 were washed anddried for analysis of the magnesium content, Magnesium was measured at EES usinginductively coupled plasma emission spectroscopy on aqua regia extracts, and by ORSusing X-ray fluorescence directly on dried solids, The corresponding liquids from run 5were also assayed by EES for magnesium. The solid residues from experiments 12through 28 and the intermediate solid samples from experiments 18R and 18B wereassayed for magnesium by ORS.

TEST CONDITIONS

Phase I Experiments 5 - 10

Experiments l-10 used 200 grams of Type 2 ore mixed with 910 ml of startersolution to make about 1000 ml of slurry at 20% pulp density. The 910 ml of startersolution was saturated with gypsum and contained .185 moles of calcium nitrate. This isenough nitrate to combine with all of the magnesium in Type 2 ore if it is leached intosolution. The starter solution composition is designed to simulate recirculating solution ina cyclical commercial process with a magnesia byproduct, as described in the introductorysection.

Temperatures were set during different runs at 50°, 75° or 101°C. Thosetemperatures were maintained throughout individual experiments to within 2°C usingpressure control. The 101°C experiments were run at atmospheric pressure while otherswere operated in a partial vacuum. Carbon dioxide was removed by vigorous boiling.

Phase II Experiments 12-28

Most of the phase II experiments had an initial ore weight of 200 grams whichyields a 20% slurry density. The 910 ml starter solution contains the same ingredientsdescribed in experiments 1 - 10. However, calcium nitrate concentration was 0.01016 M,which is nearly 20 times less than in the phase I experiments. That was a mistake whichwas corrected in phase III.

Run temperatures for all tests except 26 through 28 were set to about 101 °C atatmospheric pressure (boiling). Tests 12-18 varied feed type and reaction rates.

Slurry density of experiments 19 through 22, except for experiment 19B, is 10%.That is, about 100 grams of fresh ore were used with liquid sufficient to make 1 liter ofslurry. Run 19B used 400 grams of ore which corresponds to 40% slurry density. Thepurpose of varying slurry density was to study concentration effects on leachingperformance. Initial Ca(NO& concentration was varied in proportion to slurry density inphase II because a commercial process would have to operate that way. In phase IIIslurry density and nitrate concentration were varied independently.

18

Phase III experiments 31-51

To improve reliability and control, phase III leaching used 2-liter cylindrical pyrexpressure reactors heated by electric mantles. Each test used 500 ml slurry to allow headspace for stirring and boiling. Most phase III tests were done at 50% slurry densitycontaining 250 grams of pulverized rock, but some 20% tests were done. The saturationamount of gypsum was added. Except as noted, calcium nitrate was added in an amountcorresponding to leaching 80% of ore MgO in commercial conditions. For 50% slurrydensity this amounts to .189 moles or 44.5 grams of Ca(N0&.4HzO in the 500 ml slurry.

Tests 31-34 were done to survey ore types 5-10 for their suitability for leaching.Those tests all were done at 100°C and 20% slurry density.

Tests 36 and 37 were “lump dump” experiments using an amount of acidconsumed in previous tests under pH control This acid was added rapidly to the slurrywithout pH control and then the stirred mixture was allowed to equilibrate. Results werenot promising due to excess attack on phosphate. In other words, this chemistry is notentirely reversible.

Post-Leaching Neutralization

The pH of some phase I leachates was raised to precipitate dissolved phosphate anreturn it to the pulp. The motivation was to improve purity and value of the magnesiaproduct, since phosphate losses from the leached rock were not economically important.

Carbonate minerals in fresh phosphate feeds were successful in precipitatingphosphate, but lime was more effective and much less volume was required. Moreextensive testing was done during phase III using only lime for neutralization, Phase Iproduct samples with the suffix “E” in their names were equilibrated with fresh feeds orneutralized with lime.

During phase III, product leachates were titrated with saturated calcium hydroxidesolution using a few milligrams of calcium orthophosphate as seed crystal. Sincemagnesium hydroxide precipitates above pH 9, that was the limiting pH for thesetitrations, Leachates neutralized to various pH levels were analyzed for equilibriumphosphate by molybdovanadate spectrophotometry. Kinetics of reactions producingCa@O& and CaHPO4 were determined from time-varying measurements of hydroniumion concentration.

19

TEST RESULTS

SUMMARY OF LEACHING RFSULTS

Table 4 below summarizes results of all Phase I through Phase III leachingexperiments 3 through 51. (Missing test numbers had to be discarded or aborted forvarious reasons.) Feed type designations including particle size were described in Table 3.

Final % P loss is the mole percent of phosphorus in feed solids that has beenleached into solution, prior to neutralization, To compute PPM of P205 in solution at20% slurry density for types 2, 3 and 4, multiply % P loss by 533, 598 and 670respectively.

Final % MgO is from analysis rather than correlation. Actual MgO in 9B solidproduct was 0.60% of dry solid. This has been adjusted to compensate for the 200g ofgypsum that were added to the phosphate feed as seed crystal. The adjusted value in table4 is on a gypsum-free basis for comparison to other experiments.

Table 4 summarizes the final state of each test. Intermediate time-dependent acidconsumption, leachate analyses and pH are detailed for each test in Appendix A.Appendix B gives magnesium analyses of residues along with acid consumption andelapsed time.

ACID ADDITION CORRELATION TO MAGNESlUM REMOVAL

The volume of acid dispensed is continuously recorded in each leachingexperiment. Since magnesium removal depends on acid addition, it should be possible todevelop correlations between acid consumption and magnesium removal that are accurateat least for fixed ore types and leaching conditions. Such correlations are used to estimateore MgO content remaining, which is slow and costly to measure directly. Figures 3-11show the relationship between acid consumption and magnesium removal for all the testedfeeds.

Figure 3 includes results of all magnesium analyses done on raw and leached Type2 ore. The points include many experimental variations in temperature, particle size, acidaddition rate, etc. Those variations apparently do not have much effect on the relationship

21

MgO Leaching vs Added Acid type 2 ore - all samples

0.02 0.04 0.06 0.08 0.1 0.12 I Acid ml/gram of ore

Figure 3. Magnesium Removal vs Acid Consumption for Type 2 Ore

MgO Leaching vs Added Acid type 3 ore


0.02 0.04 0.06 0.08 Acid ml/gram of ore

23



0.025 0.03


I

3

p2.5 .- c .-

ff 2 2 0 31.5

0.5

0 0.02 0.03 0.04 0.05 0.06 0.07 Acid ml/gram of ore


25


0.01 0.015 Acid ml/gram of ore

0.02 0.025 0.005

Figure 7. J

Magnesium Removal vs Acid Consumption for Type 6 Ore

r MgO Leaching vs Added Acid

type 7 ore

3 ’ 0.8

0.6

Figure 8. Magnesium Removal vs Acid Consumption for Type

0.01 0.02 0.03 Acid ml/gram of ore

0.04 0.05

7 Ore

26


3

.E2.5 c .-

5

2 2

r”

1



MgO Leaching vs Added Acid

2.6

type 9 ore

2.4

g2.2 .- .- g 2

2 1.8

0 al.6 5 s 1.4

3 1.2

1

0.8 1

0

I I

0.02 0.03 Acid ml/gram of ore

I I Ea

0.04 0.05


27


0.8

0.6

0 0.01 0.02 0.03 0.04 Acid ml/gram of ore

Figure II. Magnesium Removal vs Acid Consumption for Type 10 Ore

28

between acid volume and MgO removal. The X axis is scaled as acid ml/gram of ore sothat tests with variations in pulp density can be presented on an equivalent basis. The acidwas always concentrated reagent grade sulfuric acid at 35.34 normal concentration.

The linear regression line in figure 3 shows correlation between acid consumptionand MgO remaining in the solid. The intercept of that line was forced to pass through theraw rock analysis at 3.8% MgO. The slope of the regression line can be used to calculatethe leachable Ca/Mg mole ratio, n in the dolomite formula of equation 1. For type 2 ore,the slope is equivalent to a value of n=0.92. Other rock types have generally larger values.Type 6 has a change in slope consistent with near-equal moles of calcite and magnesite.

Also shown on figure 3 are acid volumes corresponding to total magnesium, andtotal carbonate content. They are .0526 ml/g and .123 ml/g respectively. If Mg+2 werethe only cation to react with acid, reaction would be complete at .0526 ml/g. Additionalacid consumption reflects reaction with calcium.

The carbonate endpoint corresponds to the sulfuric acid molar equivalent of thecarbonate analysis. This would be the acid consumption if all of the carbonate wereleachable under these pH-controlled conditions. Carbonate analysis is not needed forsystems analysis or control under this process concept.

LIMITS OF MgO REMOVAL

Inspection of Table 5 values for final MgO removal show that the fraction ofmagnesium which can be removed depends principally on ore type. Type 4 leaches rapidlyuntil 2/3 of its magnesium is gone. Reaction then stops at pH 4.2 with negligible loss ofphosphate. If pH setting is reduced, phosphate dissolution increases but little moremagnesium dissolves. This is consistent with presence of roughly 1/3 of the magnesium ina phosphate mineral such as francolite, with Mg substituting for some of the Ca in thephosphate crystal lattice.

Ore types 2 and 3 leach rapidly while about 60% of their magnesium dissolves. Leachingrate then becomes slower and slower without usually reaching a distinct end point.

Leaching is promoted by fine milling, long reaction time and high temperature.Table 5 summarizes the maximum amounts of MgO removal at pH 4.2, based on favorabletest results for each ore type tested. Percent phosphate loss for types 5-10 was computedassuming 26% P205 in raw rock. The figures suggest that pH below 4.2 might promoteleaching in types 5-10 without undue phosphate loss.

Replicate tests on type 2 rock are shown in table 5 with 83 and 87% magnesiumremoval. The difference is attributed to random variation in feed quality, and a mean valueof 0.85 is used to represent maximum Mg removal from type 2 rock in these conditions.

29

LIMITS OF ACID ADDITION

Test results show that the amount of acid which can be absorbed withoutexceeding pH limits is principally dependent on rock type if ample time is allowed, Thefinal equilibrium acid proportion depends rather weakly on other independent variableswhich are temperature, particle size, pH and slurry density. The rate of acid addition onthe other hand is strongly accelerated by particle size reduction, as expected.

In order to make use of laboratory test results in the systems analysis it isnecessary to develop engineeking correlations from the data. The form of thesecorrelations is based on curve fitting numerical experiments trying various combinations ofvariables and various algebraic forms. Physically implausible forms were rejected and thebest fit was selected without introducing excessive numbers of terms. Temperature inthese correlations is always in Kelvins.

This section describes the correlation which predicts equilibrium acid addition fortype 2 rock as a function of the independent variables. The following section makes use ofthe equilibrium acid correlation to develop a correlation for leaching kinetics. Thepreceding section described linking MgO leaching to acid addition using the slope of theregression line. Another correlation will be developed to estimate phosphate leachingfrom acid addition along with time and the other independent variables. Finally, acorrelation is developed to model the rate of phosphate precipitation by neutralizationwith lime. These correlations link laboratory tests to the economic systems analysis.

30

Figure 12 is a strip chart record showing total acid volume dispensed as a functionof time. Under pH control, acid addition is rapid at first and then slows and approaches anasymptote which is always short of dissolving all the magnesium. Strip chart records fromeach test were extrapolated by hand on log paper to estimate the equilibrium acid additionendpoint, In a few cases, extrapolation was not necessary because pH control had beencontinued for long periods, The equilibrium amount of acid actually absorbed in particularconditions of temperature etc. given ample time is defined as L9. The equilibrium shiftstoward more acid addition if for example the pH setting is reduced.

A few “lump dump” tests were run with little or no pH control to explore thatprocessing alternative. Those experiments were not included in the data for theequilibrium acid correlation. The number of type 2 experiments deemed sufficientlycomplete and valid for use in this correlation was 28.

There is a theoretical limiting value of leaching acid addition Ai;n (ml/g) defined byextrapolating regression lines such as those in acid/MgO plots figures 3-10 to zeroresidual magnesium. A value of Arb=O.O668 ml/g is indicated on figure 3 for type 2 ore.Aside from random measurement error, Aiim is never reached in practice. The equilibriumacid addition L9 is always a fraction F < 1 of Ah, where F is correlated to leachingconditions. To make sure that F extrapolates properly and never exceeds 1.0, it is definedin terms of another dimensionless variable, Y which is greater than 0.

F=Y/(1 +Y) or Y=F/(1 - F) 6)

Aq = FAiim = Arin;y/( 1 + Y)

Measured values of Y defined by equation 6 were related to the independentvariables by multiple regression to give a correlation Y' used in equation 7. T istemperature in Kelvins. [X] is anion concentration in moles/liter. S is mean particlesurface/volume ratio in 1/cm. Y' is limited in the model so that F cannot exceed 0.85, thegreatest mean fraction of magnesium removal found experimentally for type 2 rock. Thisprevents extrapolation of correlations in the systems model beyond their valid range andinto a regime where incremental acid attacks phosphate rather than magnesium. However,this inequality constraint proved to be unnecessary and was never invoked.

Y' = -25.18 + 63.2/pH + .0383T - 3.88[X-] + .000245S

Figure 13 is a scatter plot of measured versus correlation values of equilibrium acid&,g, illustrating the quality of fit.

31

33

LEACHING KINETICS

Time-dependent acid addition volume, A(t), approaches the final equilibrium value& given ample time. First-order decay kinetics have been assumed.

A(t)/&4 = 1 - exp(-kt) 9)

where t is leaching time (minutes), and rate parameter k is a correlation depending on theindependent process variables. A total of 125 samples were selected from among thetime-dependent type 2 data. These were used to correlate a regression k' for use inequation 9. The quality of fit is seen on scatter plot figure 14.

k'= exp(-3.9276 + 6.287~10~~s + 3.877x10m5T - .002890t)

PHOSPHATE LOSSES TO LEACHING

The amount of phosphate leached was usually under 1% of the resource in thesetests. This is far less than beneficiation losses for example, and is not important tophosphoric acid yield. However dissolved phosphate precipitates as magnesium orcalcium phosphate in the magnesia byproduct, Phosphate impurities degrade byproductquality, reducing its price.

Phosphate dissolution can be minimized with fine grinding and slow leaching.Another alternative is accept a less careful separation where some phosphate dissolves,and then precipitate it in a subsequent step. The amount of additional acid and baseconsumed thereby is minor.

Very tine type 2 rock in test 28 also gave rapid leaching with a distinct endpointafter 73% magnesium removal. Other tests featured slower reaction and more gradual,asymptotic approach to maximum acid addition. We attribute this to physicalmicrostructure of the minerals, with occlusion of dolomite in large particles possiblyaggravated by surface deposition of gypsum during leaching.

Measurements of phosphate leaching are complicated by slow reversibility in whichsome of the dissolved phosphate reprecipitates. This leads to scatter in the leachateanalyses of Appendix A. From a practical processing point of view, the significantquestion is how much phosphate is in solution at the time slurry leaves the leaching tank.The pH measurements were taken at the time of sample withdrawal, but dissolvedphosphate was measured often hours or days later, In hindsight it is not surprising that thepH data do not correlate closely with dissolved phosphate.

The amount of phosphate in these leachates is governed by equilibrium betweeninsoluble calcium orthophosphate and slightly soluble monohydrogen calcium phosphate.

CaHp04 + .5Ca+2 t) .5Ca@O4)2 + IFf+

34

11)

EQUILIBRIUM CONDITION IN NEUTRALIZATION WITH LIME

The minor amounts of phosphates leached into solution can be precipitated andreturned to the pulp by neutralizing the leachate with a small amount of base. Carbonatesin fresh rock gave adequate results in phase I testing, but lime proved more effective as abase. Phase III results with lime are reported below.

Tests 44 and 48 were both run at 50% slurry density at 65°C with type 2 rock.The tests were similar except that nitrate was the carrier anion in test 48, whereas chloridewas used in test 44. Leachates produced in these two tests were selected forneutralization experiments.

36

Precipitation was done also at 65°C with saturated calcium hydroxide solution,Results with chloride were closely parallel to those for nitrate for neutralization as well asleaching. Those two anions seem to be interchangeable with regard to their effects ondissolution of magnesium and phosphates, and subsequent precipitation of phosphate.However, phosphate precipitated by neutralization of nitrate solution was easier to filterthan was the case with the chloride solution.

A volume of 84 ml of final leachate from test 48 was stirred with lime water addedin eleven increments. After each aliquot of Ca(OPQ2, a 1 ml sample was withdrawn andallowed to equilibrate several hours, Then the samples were filtered through .45 micronmedia, and the filtrates were analyzed for phosphate and pH. Figure 16 relates pH to bothequilibrium phosphate concentration and the volume of lime water consumed.

Phosphate precipitation is nearly complete at pH 6 and above. Yet pH can bepermitted to range at least as high as 9 before magnesia begins to form. As a processingstep, the leachate can be neutralized near the high end of this range for rapid reaction.Near pH 9, equilibrium phosphate concentration is insignificant and can be treated as zeroin the model. Only the kinetics of precipitation and the rate of filtration need to berepresented in the systems model.

Final leachate from run 44 was selected to test phosphate precipitation ratebecause of its high phosphate content. That leaching experiment was run at 65 °C, andneutralization was done at the same temperature.

After equilibrating the solution between pH 5 and 6, saturated lime water wasquickly run in, raising the pH to 6.2. This reduced the hydronium ion concentration, H,below its final equilibrium value, HO. Solution pH was recorded periodically as a measureof H and the curve was extrapolated to determine Ho. The displacement from equilibrium,Ho - H is a measure of precipitation reaction progress

The natural log of & - H has been plotted in Figure 17 against time in minutes.The measurements cluster closely around a straight line with a correlation coefficient of 8= .996, supporting a hypothesis of first-order kinetics. The slope of that line gives a ratecoefficient at 65°C of .135/minute and a half life of 5.1 minutes. The systems modelcalculates the fraction of leached phosphate remaining unprecipitated as a function ofneutralization time, t, as exp(-. 135tJ.

The foregoing rate constant is valid for the reaction in equation 11, at pH aboveroughly 4. At lower pH, monohydrogen phosphate is produced from dissolveddihydrogen phosphate with faster kinetics.

38

ACETATE AND FORMATE AS CARRIER ANIONS

The carrier anion circulates between leaching and magnesia precipitation tanks. Itscalcium and magnesium salts must be soluble in water. Most of the experiments weredone with nitrate, but chloride also turned out to be satisfactory. Most other anions areinfeasible due to cost, environmental or corrosion considerations. However, simplecarboxylic acids might be feasible as carrier anion.

Experiments were done with both acetate and for-mate anions and type 2 ore at 101°C. However, these acids proved to be too weak. Their reversible ionization near pH 3made it impossible to control the pH by feedback without addition of excess acid. Anypossible use of carboxylic acids with this process concept must be limited to relatively highpH and slow leaching, probably without post-neutralization.

41

SYSTEMS MODEL ELEMENTS

Correlations derived from laboratory measurements link test results to theeconomic systems model. They were described above in the preceding section, Otherelements of the systems model include economic assumptions, price estimates andeconomic analysis which are described in this section. The following section presentsoptimized model results which provide the basis for process design.

Much of the model consists of straightforward algebraic programming detailswhich are beyond the scope of this report. Reference is made to the model itself for thosedetails. The model is embodied in a spreadsheet in QUATTRO PRO 6.0 format (.WB2file) in FIPR possession and also available from Science Ventures (619) 292-7354, 3822Tiara Street San Diego CA 92111. Formulas, parameters and assumptions in the modelcan be studied in detail, modified and tested using this spreadsheet on diskette.

PROCESSING COST COMPONENTS OF MAGNESIUM CONTAMINATION

A principal goal of leaching is to improve the value of rock initially containingundesirable amounts of magnesium. In order to evaluate and optimize this process, it isnecessary to estimate the relationship between magnesium removal and value of theleached residue. The following analysis assumes that dolomitic rock must be used whichtoday would be discarded. Incremental processing costs due to its increased magnesiumcontent are estimated. This treatment of the data may underestimate benefits of leaching.As noted in the second report section, leaching virtually eliminates foaming in the digester.Also, digester acid requirements are reduced by prior removal of most of the carbonate.

The value of discarded tailings or overburden is zero. However, if rock ofmarginal quality is used it must be priced at least according to the cost of tax, depletionand royalties plus transportation to the plant. Those minimum costs are avoided if thesame rock is instead discarded.

The following analysis assumes a linear price relationship to magnesium content.One point on this line is defined by rock which is barely worth processing, considering theminimum costs of transportation etc. Another point on this line corresponds to rockwhich might currently be considered as typical or standard. Rock of standard quality issold at higher prices which have been established by market supply and demand.

Most of the data used in this section are derived from Baumann 1995, “Definingthe MgO Problem and its Economic Impact on Phosphoric Acid Production”. Page

43

numbers in this section refer to that publication unless otherwise noted. Baumann’s datahave been re-analyzed here using raw measurements as much as possible.

Baumann’s standard or control rock contains 0.65% MgO and 31.78%P2O5 and isvalued at $21.50 in 1993 dollars. Table 6 shows the effect on various cost components ofan increase in MgO from 0.65 to 1.65%. In other words, rock containing the baseline0.65% MgO has value considered as standard, but hypothetical rock with 1% additionalMgO (all else equal) suffers the decline in value shown in the right column of table 6. Thetotal at bottom right corner of the table is the slope of the line relating estimated rockprice to percent MgO, based on incremental processing costs.

Increased Phosphate Loss

Based on Baumann page 2-4 at 26% acid 1.8% MgO and 3.7 hrs retention, a 1%increase in MgO results in a product loss of 1.346% of feed P205. According toBaumann p 5-12 a ton of standard (control) rock produces 1/3.25=.308 tons of PZOSvalued at $190/ton, or .308*190=$58.5 worth of P2O5/ton rock. A loss of 1.346% ofthat is valued at .01346*58.5 =$0.787/ton rock. However, Baumann p5-16 notes that labrecoveries are substantially better than can be expected at the plant. Let us assume afactor of 1.2 plant/lab loss so that the penalty becomes 0.787* 1.2=$.95/ton rock per 1%increase in MgO.

Slower Filtration

According to Baumann p5-12 labor, supplies, plant overheads, maintenance, ins,taxes and depreciation total $44.49/ton PzOs or 13.70/ton rock processed. This totalscales with processing time which is filtration rate-limited. Page 2-5 indicates a factor of1.190 increase in filtration time per 1% increase in MgO based on measurements for 1.8 vs0.65% MgO with 26% acid and 3.70 hrs retention. Thus, the filtration time costincrement for 1% increase in MgO is .190* 13.7=$2.60/ton of rock processed.

Excess Steam Requirement

Becker page 474 quotes total operating costs for concentration of phosacid at$28.46 per ton P2O5 in 1989 dollars, Let us take that at $34/ton in 1993 dollars.Baumann 5-12 notes that an increase to 1.23% MgO will increase acid concentration costsby 10%. This amounts to .1*34=$3.4/tonP2O5 or .308*3.4=$1.05/ton rock with MgOincreasing from 0.65 to 1.23%. A full 1% increase in MgO will give 1.05/.62=$1.69increase in steam costs.

Additional Maintenance due to Sludge

Magnesium compounds in acid react slowly to deposit MgSiF6.6H2O, magnesiumpyrophosphate and possibly other solids (Becker p130 and 532). The additionalmaintenance cost is unknown, Table 6 estimates $.20/%MgO per ton of rock to deal withincremental sludge formation.

Additional Gypsum Disposal

According to the above, a 1% increment in rock MgO leads to about .0254 moleadditional CaSO4 precipitated from associated dolomite per 100g rock. At a formulaweight of 172, this amounts to 172*.0254*.01=.0437 additional ton of gypsum /tonrock/l% MgO.

45

Phosphate bound to Insoluble MgNHdPOd

When phosacid containing magnesium is ammoniated, a nearly insoluble salt isformed which binds an equimolar amount of phosphate. Although this solid may beslowly available as a plant nutrient, standard fertilizer assay methods do not measure it.This can be considered a loss to the available phosphate and nitrogen content of thefertilizer. It is evaluated here at the value of the PsOs only, although it represents a loss ofnitrogen values as well.

Using the total at bottom right of table 6 as the slope, the estimated relationshipbetween rock transfer price and MgO content is derived. Note that this estimate is basedon incremental processing costs only.

Price ($/ton) = 28.26 - 10.4*[%MgO] + .6765*(%P205 - 31.78) 16)

Baumann p 5-25 cost estimates can be used to compute $1.70/ton fortransportation and taxation, etc. This is used as the minimum transfer price at which it is abreak-even choice to use the rock or discard it. Equation 16 gives a corresponding MgOconcentration of 2.55% if P2O5 is at 31.78%. Standard phosphorite according toBaumann contains 31.78% P2O5 compared to 30.69% in the type 2 rock used in most ofour tests. The valuation formula above has been extended to represent rock value asdirectly proportional to its phosphate content, assuming equal levels of magnesium andother properties.

VALUE OF PRODUCT MAGNESIA

Magnesium oxide of high quality has been quoted as high as $475/ton, and thebyproduct credit for magnesia is important to viability of this process concept. ApparentUS consumption of magnesium and its compounds is approximately 750,000 tons/yr on ametal basis. Of this, 17% is actually converted into metal while 62% is used in basicrefractories, in steel mills for example. Magnesia of particularly high purity and price isused in hot-face firebrick, backed up with cheaper magnesia of lower purity. High purityis also required in magnesium metal production. Less pure magnesia is used in sulfite-process pulping of wood, in flue gas desulfurization, and to precipitate metals from wastewater.

The magnesia produced as a byproduct of a 1000 ton/day P205 plant wouldprovide for about 1.5% of US consumption.

It is expected that the P-cubed process will produce magnesia comparable to thatmade conventionally from sea water and oyster shell lime, of 97 to 99% purity. Thepurity is sensitive to our process conditions, such as retention time in the neutralizationtank where dissolved phosphate recrystallizes. It is also sensitive to the amount ofsediment which passes the belt filter in figure 1.

46

To evaluate and optimize this process, it is necessary to express the value ofmagnesia as an algebraic function of the percentage of impurities it contains. This sectiondescribes such a correlation based on market prices of magnesia ranging from low to highquality.

The P-cubed process would produce directly a low-temperature calcined, reactive"light" magnesia MgO powder. Light magnesia permanently loses its reactivity at hottertemperatures, becoming the "heavy" magnesia used in refractories. We have found littleprice difference between light and heavy grades of similar purity, so the two types arelumped together here. The cheapest magnesia is calcined directly from magnesite rockand is used principally in its reactive form as a chemical. Better purity is obtained byprecipitation of magnesium hydroxide from brine or sea water with lime or dolime. Oystershell lime can produce superior purity.

Recent price quotes for various grades are plotted on figure 18. A linearregression of these prices against the inverse square root of percent impurity has acorrelation coefficient of 0.65. This correlation is represented by the solid curve on thefigure. The straight dashed line extrapolated from the curve is based on the concept thatlow grade magnesia is used as a chemical which is valued in proportion to its MgOcontent, Zero market value is assumed arbitrarily at 25% MgO. The systems modelimposes an upper limit value of $475/ton corresponding to the highest price quote so thatthis correlation will not be extrapolated improperly, That upper limit was not invoked bythe model in optimization calculations.

PRICES OF OTHER FEED MATERIALS AND PRODUCTS

Raw phosphate rock price is evaluated from its MgO content as described twosections above, using the same formula as for leached rock. The producer’s sulfuric acidprice of $28/ton is that used in Baumann 1995. Leaching loss of phosphate is evaluated at$190/ton of P2O5 also according to Baumann. However, that loss is insignificant in theoptimized model. The Chemical Marketing Reporter provided prices for hydrochloric acidand slaked oyster shell lime. HCl of 22°Be concentration (specific gravity 1.17 at 35wt%) ex works in the Gulf coast is quoted at $85/ton. Slaked lime was quoted at $60/ton.The true (inflation adjusted) cost of credit was taken at 4% per annum.

Mr. John Carrol of Nitram estimated a price of $167/ton (100% basis) for 54%nitric acid FOB Tampa. A fuel price of $2.20/million BTU for gulf coast natural gas wasgiven by Chemical Week. Electric power value was estimated at $.07/kwh by G. M.Lloyd of FIPR.

Value of Carbon Dioxide

The possibility of separating product CO2 for sale was investigated. Mr. JasonBramhall of Carbonic Industries in Florida kindly provided much detail regarding themarket and quality specifications for carbon dioxide. It is clear from that discussion that

47

Value of Nitrate in Fertilizer

A small credit was taken for nitrogen in the final product fertilizer derived fromnitrate carried out of the leaching cycle as filtration loss. This nitrate is equivalent to anequimolar amount of ammonia in the fertilizer assay, and reduces the amount of ammonianeeded to reach specified nitrogen content. The nitrate credit is evaluated at $1.7/lb molebased on anhydrous ammonia at $200/ton from Chemical Marketing Reporter.

ENERGY REQUIREMENT FOR SUPPLEMENTAL FEED MILLING

Tests showed that magnesium leaching was faster with finely milled feeds. Acorrelation was developed to relate ball milling energy to product particle size so that thisparameter can be evaluated and optimized. The correlation is concerned with costs of

48

milling to greater fineness than is conventional for phosacid feeds. Costs of conventionalmilling are attributed to the phosacid process and are not included in this leaching systemsmodel.

With baseline design choices, the optimized model indicates no advantage tomilling beyond normal, conventional preparation of phosacid feeds. However, there issubstantial advantage to use of some acidic pond water in the ball mill. This improves theleaching water balance, reduces acid requirement, reduces retention time in the leachingtank (not modeled) and recovers phosphate dissolved in the pond water.

As noted in the second section of this report, a single parameter is used here tocharacterize mean particle fineness, This is the ratio of total surface area in a sample tothe aggregate volume of all the particles (cm-1) designated as S. In reality, particle sizedistribution is roughly logarithmic and covers a considerable range. In order to relate S tocommonly-available sieve data we have used the particle size curves in figure 4.17 ofBecker 1989. Those curves show particle size distributions for wet open-loop ball millingof Florida pebble as measured by Agrico. Four different specific energies were testedranging from 14 to 24 kwh/ton.

Each of those four curves was integrated numerically to estimate S for each energyinput.

S = j 6/m df 17)

where f is the mass fraction retained and m is the corresponding sieve mesh size incentimeters. The factor of 6 is based on an approximation of cubic symmetry. Table 7shows the results. Figure 19 plots measured S against specific energy, E and showscorrelation lines.

49

Bond’s Law predicts that particle surface increases as the square of milling energy.Although this is a good approximation at low and moderate energy, it fails at high energy.In fact, the amount of surface which can be exposed by mechanical size reductionapproaches an asypmtote, even with extreme energy input. This phenomenon is illustratedby the upper data set in figure 19, which represents ball milling of slate. Those pointsfrom Perry 1973 page 8-56 have been fit empirically to a function S = cl + Q./,/-E. The

50

correlation used for phosphate makes use of cz from the slate correlation to extrapolatebeyond available measurements toward higher energy input, as shown by the dashed line infigure 19.

This approximate, qualitative extrapolation of the data is necessary in absence offull-scale ball milling tests to fulfill mathematical requirements for economic modeling andoptimization. As noted above, the computed optimum indicates that no additional millingis needed with base-case design assumptions, The extrapolated part of the milling energycurve is unused, and the “new” ball mill in figure 1 is not needed.

A particle size S for conventional phosacid feeds is taken to be 1388/cm, based onspecific milling energy of 16.7 kwh/ton.

LEACHING PROCESS WATER BALANCE

The water balance calculation in the systems model represents Hz0 circulating atsteady state in a loop from leaching tank to neutralization tank unit 3 to magnesiaprecipitation and separation at units 5a and 6, and back to the leaching tank unit 2. Seefigure 1 process flow diagram in the first section of this report. The water balance sectionof the model expresses all Hz0 inputs and outputs in units of ml of water per 10 liters ofleaching slurry.

Hz0 Inputs

Nitric and sulfuric acid feeds contain water diluting the acids to less than 100%strength. This dilution water is accounted for according to the strength of each acid asspecified as inputs to the model. In addition, a small amount of water is produced by thechemical reaction between nitric acid and carbonates in the rock.

2HN03 + MC03 -+ M(No3)2 + coz + Hz0 20)

Phosphorite from the mine contains some surface moisture. According to Becker1989 page 261, this averages 9.5% of central Florida rock. It constitutes the secondlargest Hz0 input and accounts for about 19% of the total.

The final source of water to the leaching cycle is rinse water to the belt filter. Thisstream is computed from the water balance as total water outputs minus the foregoinginputs, Rinse water recovers nitrate and magnesium to the leaching cycle and reducesnitrate loss to the digester. Rinse water is the largest input at about 78% of the total.

Lloyd and Leyshon 1993 indicate that up to 20% of the ball milling liquid can bemade up of pond water from the phosacid process. This systems model would imposethat 20% limit if it were necessary and would complete the leaching water balance withfresh water, However, the limit is not needed with current values and formulas in themodel. Currently, all of the rinse water is pond water, and pond water constitutes about17% of total water to the ball mill.

51

The model currently uses 1% as the input value of P205 concentration in the pondwater, Besides reducing leaching acid requirement and leaching retention time, the pondwater contributes a small amount of phosphorus which reports to the leached residue.This contribution exceeds leaching losses of phosphate after neutralization, so that theoverall percent phosphate loss is slightly negative. Our lab results indicate that the acidityin this amount of pond water can be neutralized by the highly reactive fraction of leachablecarbonate in feeds, without adversely affecting pH control in leaching.

Hz0 Outputs

The largest output in the leaching water balance is surface moisture on leachedresidue passing to the digester at 48% of the total. Likewise, 19% of total water leavesthe loop in the underflow from thickener unit 6 as surface moisture associated withmagnesia. According to Perry 1984 page 19-64, the slurry from a magnesia thickenercontains about 25 grams of Mg(OH)2 per 100 ml.

About 1.6% of the water leaves the loop in chemically bound form as magnesiumhydroxide. Similarly, 3.2% is removed as chemically bound hydration water in gypsum.For each mole of sulfuric acid consumed in leaching carbonates, one net mole of water isremoved from the liquid phase and is bound to solid gypsum.

Substantial amounts of water leave the loop through evaporation. Carbon dioxideconcentration in the leaching tank is kept low to promote leaching reactions. This can bedone with impeller-driven spray resembling foam breakers in digesters. Or it can be donewith barometric legs similar to vacuum flash coolers, We have assumed that 5 moleculesof water are to be evaporated for every molecule of carbon dioxide removed this way.Such evaporation accounts for roughly 17% of water outputs.

The belt filter in process flow diagram figure 1 operates by suction, similarly toconventional tilting-pan or rotary drum filters. In order to produce adequate suctionwithout boiling the filtrate, slurry fed to the filter must not be too hot. We have assumed amaximum temperature of 60 °C (140°F) for that slurry, so that vacuum up to 0.8 atm (608mm Hg) can be used. Since the optimized model calls for a temperature hotter than 60°,some evaporative cooling is needed at the neutralization tank. This is indicated on figure1 by a fan directed parallel to the slurry surface. If the model indicates a leachingtemperature below 60°, neutralization tank evaporative cooling is skipped. Currently, themodel indicates that 10% of water outputs is evaporation due to neutralization tankcooling. The model includes effects of water evaporation in heat and fuel requirements.

The belt filter also removes some water by evaporation. According to Becker1989 page 432, 30 Kg of air can be expected to penetrate each square meter of filter in anhour for a filter operating on a l-minute cycle as is typical of a belt filter. Based on Lloydand Leyshon 1993, the belt filter may have a capacity of 9.75 ton/square foot/hr, or 3940Kg/m2/hr. The ratio of these two numbers is .0076 Kg air passing through each Kg ofleached residue during the filtration cycle. The vapor pressure of water from the Clausius

52

Clapyron formula at the belt filter temperature is used with the air/solid ratio to computebelt filter evaporation.

MAGNESIA WASH WATER NEUTRALIZATION WITH POND WATER

Purified fresh water which was used to wash product magnesia leaves the systemat wash water separation unit 7a and drum filter, unit 8. This water has a pH near 12 andcontains a small amount of dissolved and suspended magnesium and calcium hydroxides.We envision a process step in which this alkaline water is contacted with additionalphosphatic pond water in proportion measured for pH neutralization. Water soneutralized will precipitate most of its fluoride content as MgF2 and CaF$ sludge. Aftersedimentation, it may be suitable for irrigation or other use. The net effect would be torecover in useful form all of the fresh water used to cleanse the magnesia, plus the volumeof pond water used to neutralize it.

According to Becker 1989 page 553, a similar method is used in Louisiana. Pondwater must be discharged there because local rainfall exceeds evaporation. The effluent isneutralized to pH 8.8 by a two-stage calcium carbonate/lime treatment, This bringsdissolved fluoride and phosphate down to only a few ppm.

The proposed mixture brings together all of the conditions to precipitatechukhrovite CaS0&4lFr3. 10H20; namely calcium, fluoride, fluosilicate and sulfate ionsand dissolved aluminum from the pond water, plus elevated pH and calcium from themagnesia wash. Chukhrovite separation could help reduce concentration of dissolvedfluoride and meet environmental requirements.

Regardless of neutralization, the magnesia wash water will contain nitrate left overfrom two-stage countercurrent washing in units 7a and 7b. This amounts to about 0.1molar concentration, or 1200 ppm nitrogen. With current figures, alkali metals from pondwater equal just over half of the nitrate moles. Therefore the nitrate which is in excessover alkali metal can be expected to form a magnesium salt. Principal solutes expected inthe treated water are nitrate, sodium and magnesium in declining order.

Data from page 547 of Becker 1989 are used to characterize the pond water, usingthe “typical” composition in table 9.4. Summing the cations, total hydronium equivalent is.165 moles/liter. Anions total 1.133 mol/l and the difference equal to ,968 mol/l representshydronium ion acidity. This is balanced against calcium and magnesium hydroxides in themagnesia wash water to compute the volume of pond water needed to neutralize it. Toestimate resulting precipitation of sludge, it is assumed that all the cations in Becker table9.4 except for sodium and potassium form their crystalline fluorides. Remaining fluorideprecipitates as MgF2, and all the phosphate precipitates as Mg3(PO&.8H20, Sulfate isassumed to remain in solution as the magnesium salt. This approximation is a book-keeping artifice to estimate the amount of solids to be processed in unit 11, and thereforeits size and cost. Actual solid product species may differ.

53

Based on Forbath data, about 5% of the magnesia entering washer 7a will be lostin suspension in the overflow. This is the principal source of alkalinity to unit 11 wastewater treatment along with the smaller saturation amount of calcium hydroxide.

A design goal of w=90% washing efficiency was selected arbitrarily. This meansthat 90% of solutes in unit 6 underflow are to be separated from product solid bycountercurrent decantation. Assuming perfect mixing in each of two washing stages, theproportion of wash water needed relative to unit 6 underflow surface moisture is:

From the sludge formation rate and total waste water volume the dimensions ofclarifier unit 11 are computed. A midrange overflow rate of 1.0 cubic meter per squaremeter per day is used, based on Perry 1984 table 19-8.

PRODUCTION OF MAGNESIA

For purposes of this analysis, the base used to produce magnesia is taken to beslaked oyster shell lime rather than ordinary mineral lime or dolime. By purchasing aproduct which has already been slaked, the process flowsheet and economic analysis aresimplified. Moreover, the high purity of oyster shells makes possible production ofpremium quality magnesia.

In process flow diagram figure 1, everything below the belt filter (except wastewater treatment unit 11 previously described) represents conventional production ofmagnesia. It was adapted from an article in Chemical Engineering by Peter Forbath March24, 1958. The subject was a plant in Moss Landing California operated by KaiserChemical Co. using sea water as the brine source. In addition to the abundant technicaland economic data in that article, much information could be extracted from the severalphotographs.

Lime is fed to reactor unit 5a in slight excess to form a suspension of finely-particulate magnesia. The excess lime and insoluble matter underflow 5a and then 5b andis disposed to the gypstack. The overflow from 5a is saturated in calcium hydroxide andcarries the magnesia in suspension.

The larger thickener unit 6 allows sedimentation of the magnesia to separate itfrom liquid which recycles to the ball mill. Unit 6 underflow is washed in twocountercurrent stages with purified water. The washed magnesia is dewatered on a drumfilter to about 50% water and then is calcined in a rotary kiln. We assume that a reactivelight magnesia is produced, so that calcining is done at about 700°C to remove hydrationwater and produce dry MgO powder for sale.

54

Heat Requirements of Magnesia Calciner

According to Forbath 1958, the drum filter unit 8 produces a filter cake containing50 grams of dry solid per 100 ml. Given the literature value of 2.36 for specific gravity ofMg(OH)2, for every gram of dry solid there is 100/50 - 1/2.36 = 1.58 ml of surface waterto be evaporated. It is assumed that resulting water vapor will be swept out of the rotarykiln at a temperature moderately hotter than its boiling point, such as 150°C. It takes 573calories to evaporate a gram of water, and another 48 cal to heat the vapor from 25°C to150, for a total of 621 calories required to evaporate each gram of surface water.

It takes 3 1.39 Kcal/mole to dehydrate magnesium hydroxide to oxide or 538cal/gram of Mg(OH)2. Assuming dehydration at 900K (627°C) it takes 6.652 Kcal to heatthe product MgO from ambient. Vapor produced from the bound water requires another5.24 Kcal/mole to reach 627°C.

To summarize, the calciner requires the following amount of heat to produce lightmagnesia, in cal/g neglecting losses.

Magnesia Precipitation Reactors 5a and 5b

Forbath 1958 process flow diagram shows a small third reactor which extracts thelast available alkalinity in dolime to form magnesia. Because our design calls for more-reactive oyster shell lime and is smaller in scale, we have omitted this third reactor. FromForbath data it can be computed that retention time in reactor 5a is 73 minutes. This ismultiplied in the model by leachate flow rate in cubic meters per minute to arrive at avolume for this reactor. Its diameter is then determined assuming a height of 2. S meters.

It is assumed that the ratio in surface area of secondary reactor 5b to 5a will besimilar to that reported by Forbath, which was 0.32. This leads to a diameter for 5b(rounding upwards) which is 0.6 times that of reactor 5a.

Magnesia Thickener Unit 6

Forbath data on thickener unit 6 indicates a production rate of 0.117 metric ton ofmagnesium hydroxide per square meter per day. This is consistent with data in Perry 1984page 19-64 which gives values of 0.1 to 0.2 for brine and 0.1 to 0.33 for sea water. P-cubed process production of magnesia in mt/day divided by 0.117 gives the requiredthickener area in square meters.

Magnesia Washers Units 7a and 7b

Unit 6 underflow is slurried with rinse water and is allowed to separate bysedimentation in thickener unit 7a. That underflow in turn is slurried with purified freshwater and then is separated in unit 7b. Forbath reports that two similar sized thickeners

55

are used as 7a and 7b, and that the surface area of each is l/3 of total unit 6 area. Thus,the diameter of each is a factor of 1/35 = .577 times that of unit 6.

Other Magnesia Production Equipment

The Forbath plant produced 337 mt/day of magnesia. A ratio of our MgO outputto this figure is useful to estimate required areas of sedimentation units and otherequipment, This was the approach used to estimate the dimensions of sedimentation unit10, which is used to purify fresh water for washing magnesia using a small amount of lime.Our diameter estimate for that unit is 7.0 meters.

Analysis of Perry 1984 table 20-36 for lime sludge shows that capacity of a rotarykiln is directly proportional to its volume, and is almost independent of its aspect ratio orscale. A kiln reported by Forbath calcined 135 mt/day of magnesia in a kiln of 523 cubicmeters volume with a length 23.5 times its diameter. This corresponds to capacity of 0.26mt of MgO per cubic meter per day. It is used by the model in conjunction with ourmagnesia production to estimate kiln dimensions.

Maintenance for the calciner is relatively high due to elevated temperatures andwear on the refractories. From Perry page 20-37 annual maintenance is estimated at S to10% of total installed cost, We have adopted a midrange maintenance rate of 8%/yr.

CAPITAL COSTS OF MAJOR PLANT EQUIPMENT

Sedimentation Units Cost

The magnesia production half of the flowsheet contains seven sedimentation unitsto thicken solids or clarify liquids. The section directly before this one describedcalculation of their dimensions. The capital costs of these units are estimated from acommon formula based on their diameter. The formula is derived from a log plot on page311 of Ulrich 1984. It gives the purchased equipment cost, Cp, in 1982 dollars as afunction of diameter in meters, d.

23)

Mixing Tanks Cost

The volume of leaching tank unit 2 is determined by the volumetric rate of slurryto be leached (m”/min) times the batch retention time in minutes derived from correlationson lab data. The result in cubic meters is multiplied by a dimensionless factor whichrelates batch leaching in the lab to continuous stirred reactor design. This factor accountsfor effects of backmixing in the CSTR, which requires longer retention and larger vesselsfor conversion equivalent to batch or plug flow reaction, The retention time penalties forCSTRs relative to batch process are tabulated at model spreadsheet address B175 as afunction of the number of uniform-sized stirred tanks. This factor rapidly approaches 1 asthe number of series tanks increases.

56

Peters and Timmerhaus 1991 page 539 gives a log plot for several types of mixingtanks. We have used the correlation line for cylindrical 304 stainless steel tanks withimpeller for both leaching and neutralization tanks. Purchase cost in 1990 dollars, Cp, isgiven as a function of tank capacity in gallons, C.

24)

Ball Mill Cost

As noted above, the model does not call for supplemental particle size reductionbeyond that used conventionally to prepare digester feeds. Nevertheless, it is necessary toinclude in the model a realistic estimate of ball mill cost and energy requirements. If thosecosts are not included, the model will indicate falsely that supplemental milling should beused, since it costs nothing.

Energy requirements for supplemental milling were discussed in figure 19 andassociated text. Peters and Timmerhaus 1991 give a log plot on page 563 showingpurchase cost of various types of wet milling equipment. This cost includes installation,classifier, motors, drives, and an average allowance for foundations and erection. The linewhich represents finest milling (98% minus 325 mesh) was fit to give the followingformula, where M is mill capacity in short tons per hour,

Belt Filter Cost

Order of magnitude costs of suitable belt filters were estimated by Lloyd andLeyshon 1993. These estimates might be appropriate to both Eimco and Delkor filters,both of whom have significant experience in phosphoric acid. Lloyd and Leyshonestimated costs of complete installation at four levels of PZOS production ranging from 900to 2700 tons per day. Our design basis is 1000 metric tons or 1102 short tons.Interpolation on Lloyd and Leyshon data leads to a cost estimate of 4.12 million for a beltfilter of 61.2 square meters completely installed.

Drum Filter Cost

Forbath 1958 reports that a drum filter 14 ft diameter by 18 ft long operating at 22inches Hg vacuum processes 375 tons per day of magnesia on a dry MgO basis. This filterproductivity is equivalent to 4.62 mt/m2/day. Our magnesia production of roughly 60mt/day MgO is divided by this productivity in the model to estimate filter area required.

Ulrich 1984 presents a log plot on page 314 showing purchased costs of varioustypes of liquid filtration equipment as a function of filter area, A in square meters. Thecorrelation line representing vacuum drum filters has been fit to the following formula.

57

Cp = 23988 A.5885 1982 dollars

This basic cost is multiplied by a factor of 3.0 to represent the premium cost ofpolypropylene filter media.

Calciner Cost

As noted above, capacity of a rotary kiln correlates well with volume and is almostindependent of aspect ratio. Perry 1984 page 20-38 table 20-16 gives costs of rotary kilnsas a function of size. The purchase cost includes drive, burner and controls but not bricklining. Kiln volume V (m”) in that table is highly correlated to purchase cost according tothe following formula. The correlation coefficient r2 is 0.9927.

Cp = 222000 + 2177V 27)

Based on figure 20-42 of Perry 1984, about half of the length of this rotary kilnshould be lined with MgO firebrick refractory. Again from Perry page 20-37 the cost ofthis refractory lining is $7000 per axial meter for a kiln 9 ft or 2.7 meters diameter. Themodel computes kiln volume as described above, and kiln length calculation is based on2.7 meters diameter.

Perry page 20-37 notes that present-day rotary kilns are typically 65 to 75%efficient in terms of fuel heating values transferred to the load irrespective of any externalheat recovery equipment. The systems model conservatively uses the lower 65% figure toestimate fuel costs.

TOTAL CAPITAL INVESTMENT

Purchase costs of major plant equipment items are tabulated in column C, lines 245through 278 of the model. As noted above, individual purchase cost data have been takenfrom a variety of sources which vary by year and may be on an installed basis or not.Column D of the model applies cost escalation multipliers from the Marshall and SwiftEquipment Index, adjusting all column C costs to 1995 dollars. In column E, factors areapplied as appropriate to convert all column D costs to a fixed capital investment basis.This includes piping, electrical utilities, site preparation, foundations, instrumentation,auxiliaries, field expense, engineering, contractor’s fees, and contingency. The factors arethose from Perry 1984 page 25-70, table 25-50 for grass roots solids-fluid processingplants.

Fixed capital investment in major equipment items is totaled at B287 of the model.Factors are then applied to arrive at total capital investment. Ulrich 1984 page 279 figure5-2 illustrates contingency margins for innovative process steps or operations. We haveadopted a 40% contingency factor for technical uncertainty, based on two unconventionalsteps in processing of solids (leaching and neutralization). Also from Ulrich p326working capital is estimated at 15% of total fixed capital investment. These are summedfor total capital investment at model address B290.

58

The systems model does not include cost of land. Land costs may be minorcompared to other components of capital, and phosphate producers may already ownsuitable land adjacent to their phosacid plants.

LABOR AND OTHER DIRECT AND INDIRECT EXPENSES

Ulrich 1984 page 329, table 6-2 gives typical labor requirements for operation ofvarious generic types of process equipment in operators per unit per shift. The systemsmodel adopts those values except for the gas-solid contacting rotary kiln unit. The modelconservatively uses twice Ulrich’s labor for the latter unit because it operates at fairly hightemperature. Total labor costs are then corrected from 1982 to 1995 dollars,

Supervision and clerical costs are taken by Ulrich page 330 to be 15% of directlabor cost. Laboratory expenses are also taken to be 15% of direct operator labor.Maintenance is estimated as 6%/yr of fixed capital investment. Operating supplies areestimated at 15% of maintenance. The burden applied to labor is 60% of the totaloperators plus supervision/clerical plus maintenance. 2% of fixed capital is attributed toproperty tax and insurance. Administrative costs are estimated to be 25% of overheadcost, where overhead is defined as labor burden.

INTERNAL RATE OF RETURN

The internal rate of return (IRR) is also known as the discounted cash flow rate ofreturn, It measures economic merits of this process concept, and is the objective functionwhich is maximized by adjusting all the independent variables, The independent variablesare located in model C9..C15 and are:

1. % MgO in raw feeds2. Neutralization retention time in minutes,3. Leaching retention time, min4. Slurry density in grams of solids per 100 ml slurry, or percent5. Mean particle size, S as ratio of surface/volume, l/cm6. Leaching pH7. Leaching Temperature, Kelvins

The model has also been optimized by experimenting with certain discrete design choicessuch as how many leaching and neutralization tanks to use, and whether or not to includethe neutralization step.

IRR is computed from a series of annual cash flows in millions of 1995 dollars atmodel lines 305..331. It is assumed that the plant will take three years to build and thenwill operate for twenty years. It is further assumed that solid wastes can be disposed of ata nearby gypstack.

59

The volume of phosphate processed is based on an operating factor of 85%,together with plant capacity of 1000 TPD P205. The result is slightly over one milliontons per year of rock processed. Annual process earnings are computed as the value ofmagnesia produced plus the improvement in phosphate value minus feed costs, labor andsupervision, maintenance, fuel, power and other regular production costs, This includes aroyalty at 2% of total costs.

The bottom-line IRR is computed at model address C298 and is shown also ataddress B4.

60

SYSTEMS MODEL RESULTS AND PROCESS DESIGN

Model results presented below result from complex interplay between laboratory-basedcorrelations, economic parameters, and design choices. Substantial changes in any of theseelements can produce substantial changes in optimized results.

OPTIMUM VALUES OF THE INDEPENDENT VARIABLES

All of the independent design and operating variables have been adjustedsimultaneously by the model to find the global maximum internal rate of return as defined in thepreceding section. Table 8 shows optimized values of the independent variables, based on theprocess flow diagram in figure 1. Some variations from that design are also considered in thissection, along with their impact on IRR. The table includes optimized variables for a processvariation omitting neutralization in unit 3, along with resulting IRR.

Table 8. Optimum Values of the Independent Variables

61

The model includes constraints on some of these independent variables to assure thatthey do not violate physical or practical limitations. Leaching temperature for example, isexplicitly limited by spreadsheet settings to the range from 25 to 100 degrees Celcius. Itwould be impractical to operate at cooler than ambient or above the boiling point of water. Amaximum slurry density of 70% is based on Becker 1989 page 266 which notes that wetmilling is done at 62-70% slurry density, limited by pumping requirements, A lower pH limit of2.9 has been established to avoid extrapolating correlations beyond the range of ourmeasurements. As it happens, none of these explicit inequality constraints has actually beeninvoked by the model in the base case calculation. The optimum solution lies within all explicitconstraint boundaries.

The magnesia content of feed rock is considered to be an independent variable in thesense that the mine may choose whether to discard rock of a particular quality or to send it tothe plant, and can blend different grades of rock to meet specifications. The model finds thatthe most economic rock for leaching has a dollar value which equates to the minimum cost ofbringing it to the plant, as estimated from table 6. Rock with more MgO than that should bediscarded. Rock with less MgO would produce less magnesia for sale to support leachingcosts. This is in effect, an implicit, indirect inequality constraint which fixes the optimum valueof type 2 rock MgO at 2.48%.

Similarly, the model finds that the optimum particle size is that which requires nosupplemental millng. A coarser grind leaches more slowly with no compensating process costadvantage. A finer grind does not reduce leaching costs enough to compensate for incrementalmilling costs. This implicit constraint fixes the mean particle size, S in the base caseoptimization, When the neutralization step is omitted, the model invokes the explicittemperature maximum at 100°C, and relaxes the implicit constraint of no supplemental milling.In the latter case, the model calls for substantial particle size reduction.

All other independent variables are optimized at unconstrained values at the peak of asmooth continuous convex surface in hyperspace. The first derivatives of IRR with respect tothese variables are all zero, so they are not included in the sensitivity analysis later in thissection.

NUMBER OF LEACHING AND NEUTRALIZATION TANKS

Among design choices is the number of stirred reactors in series which is mosteconomical. Multiple tanks reduce backmixing, so that a specified degree of conversion can beachieved with less total retention time and reactor volume. This is offset by the capital costs ofmore numerous but smaller reactors, The optimization problem has been solved with 1, 2 or 3tanks to measure effects of that choice on IRR.

IRR is maximized with a single tank for leaching and a single tank for neutralization.However, results are nearly as good with two or even three series tanks of uniform size. Table9 gives IRR as a function of tank numbers,

62

Another process variation concerns the choice of carrier anion. Lab results indicatethat either chloride or nitrate are promising candidates, although most of this project hasfocused on nitrate. If chloride is used instead, there is a small saving in feed costs which ispartially offset be the loss of nitrate contribution to nitrogen values in the fertilizer. Theoptimized independent variables are unchanged to two significant figures, and the IRRincreases by 0.074 percentage point to 35.023%/yr. The economic difference between thesetwo anions is insignificant, and the choice will depend on environmental and corrosionconsiderations.

NO POND WATER TO BELT FILTER RINSE

It has been assumed in most of this analysis that phosacid pond water will be used inthe belt filter rinse to balance the quantity of water circulating in the leachingmagnesiaproduction loop. What is the effect on process design and economics if fresh water is usedinstead? To study this question, model cell L27 pond water % P205 was temporarily changedfrom 1 to zero, and sulfuric acid consumption was increased to provide for magnesia washwater treatment in unit 11. Results are shown in the third column of table 8. The IRR declinesby 0.70 % not counting the cost of the fresh water used to replace pond water.

SENSlTIVTY ANALYSIS

This section analyzes the sensitivity of the IRR to variations in several inputs andcorrelations in the model. Results are used to measure components of uncertainty and to guidefuture investigation so that it will most efficiently reduce economic risk. The partial derivativesof IRR with respect to these various parameters have been computed numerically byincrementing the parameter and then repeating optimization. Results are shown in table 10.

The contribution of each parameter to the total variance of IRR is computed accordingto the standard formula for propagation of error. The internal rate of return is designated as Z,

63

and the value of parameter i is designated as Xi. If there is no covariance between the X’s, thenthe variance of Z is:

28)

The variance of each parameter, $i, is known from statistical analysis in the case of thefour designated correlations and the MgO price, which is also based on a correlation developedin this project. Variance of parameter 5, the Ca/Mg ratio designated as n, was also determinedstatistically. For other parameters the variance was estimated from judgment of the quality ofthe measurements or other inputs. Although there is a subjective element in some of thestandard errors, this analysis is useful to estimate total uncertainty in IRR, and to identify theprincipal contributions to that uncertainty.

The total at lower right corner of table 10 represents the variance of IRR. Its squareroot estimates the standard deviation of the IRR which is 11% per annum. Of course the tableis not exhaustive, and there may be additional line items which could be included. We havetried to identify and include the most significant contributions to total variance. With thesefigures, the 1-sigma range of the IRR is from 24 to 46 percent per annum inflation-adjustedreturn on investment. The two-sigma range is 13 to 57%.

CONCLUSIONS AND RECOMMENDATIONS

Now that the economic systems model is complete, the best combination of designchoices is known and the optimum values of all process variables have been estimated, futurelab testing can be more productive. Tests can be designed and run at conditions which are nowknown to practical and relevant. Test results can be used promptly in the model to moreaccurately define optimum conditions. Feedback from the model will guide efficient testingand accelerate commercial development.

From table 10 it is obvious that uncertainty in economic performance is dominated byvariance in the transfer price of the leached rock. This should be evaluated in more detail inconsultation with phosacid engineers and managers. Half as much variance is contributed bythe correlation for the carbonate leaching rate constant. That term can be improved withfurther laboratory measurements, aimed specifically at conditions which are now known to benear-optimum. These two terms account for 65% of the total variance in IRR.

Each of the other sources of IRR variance contributes 1/13 or less of the total. Indeclining order of variance contribution, they are raw rock MgO content (negative incrementassumed), the Y factor which represents the final equilibrium fraction of MgO which isleachable, the contingency for unconventional process, the price of magnesia, the plant capacityfactor, and the mass fraction of solids suspended in filtrate passing the belt filter.

64

Table 10. Contributions to Uncertainty in IRR

65

Phase II tests on NO, emission indicated that emission might exceed small sourcestandards with leaching at 100°C. Now that we know optimum temperature is near 76°, NOxtests should be repeated at that temperature and using a more appropriate anion concentrationand other relevant conditions.

If the NOx tests reveal potential environmental problems, the use of chloride as thecarrier anion should be tested in more detail, with regard to both leaching and neutralization.The technical and economic feasibility of chloride can then be evaluated and maximized.

This research has concentrated principally on a single type of dolomitic phosphatedesignated as type 2. Any commercial process based on P-cubed technology must acceptconsiderable variation in MgO and other feed composition variables. Other ore samples shouldbe tested and modeled to evaluate their economic merits and estimate optimum conditions.From this process should emerge an average combination of design and operating variableswhich works well for a wide variety of rock, and which offers good economic returns with allof them.

In order to represent a new type of rock, the model requires four correlations based onlab measurements:

1. Acid consumption should be plotted against residual rock MgO content for samples leachedat a relevant range of the independent process variables. A regression line is then computedforcing the intercept at the MgO analysis of raw rock. The regression slope is used to estimatemolar Ca/Mg ratio leached, n. The intercept on the abcissa at zero MgO defines the limitingacid addition, &ti.

2. Long term leaching tests should be run at a range of pH, T, S, etc. Results are to be used inmultiple regression to fit parameter Y which relates equilibrium acid consumption 4 to &.

3. Samples should be taken at intervals during tests proposed above. Their magnesiumanalyses are correlated to time and the other process variables to develop an expression forleaching rate constant, k.

4. Also using samples withdrawn at intervals, dissolved phosphate should be promptlymeasured to avoid time-dependent changes. Results are to be used to correlate phosphatedissolution to leaching conditions.

Current results suggest that the P-cubed process merits further investigation. Even atthe worst-case (minus two sigma) IRR condition, it promises to return 13% on investmentafter inflation. The process may assure survival and prosperity of phosacid producers inFlorida and wherever dolomitic phosphates are found.

66

REFERENCES

Abu-Eishah, S. I., et al. 1991. Beneficiation of calcareous phosphate rocks using diluteacetic acid solutions: optimization(sic) of operating conditions for Ruseifa (Jordan)phosphate. Int. J. of Miner. Proc. 31: 115-126.

Anon 1991. Strong growth for hydroxide. Chemical Week May 15.

Anon 1975. Phosphorous pentoxide enrichment in phosphorite-by treating with maleicacid and regenerating side products. Soviet Union patent 469664.

Association of Florida Phosphate Chemists. 1980. Methods Used and Adopted by theAssociation of Florida Phosphate Chemists. 6th ed., 9-37 to 9-39.

Baumann, A N and Jacobs Engineering, 1995. Defining the MgO Problem and itsEconomic Impact on Phosphoric acid Production. FlPR 92-10-102, Florida Institute ofPhosphate Research, Bartow FL.

Becker, Pierre. 1989. Phosphates and Phosphoric Acid. 2nd Ed. New York: MarcelDekker Inc. 628-717.

Bureau of Mines. 1987. Mineral Commodity Summaries. US Dept of Interior.

Chilton, T. H. 1968. Strong Water: Nitric Acid Sources, Methods of Manufacture, andUses, Cambridge: MIT Press.

Forbath T P, March 24, 1958. Magnesia from sea water. Chemical Engineering.McGraw Hill NY.

Hansen, J. P., B. E. Davis and T. O. Llewellyn. 1985. Removal of magnesia fromdolomitic southern Florida phosphate concentrates by aqueous SO2 leaching. BuMines RI8953.

Kirk-Othmer. 1985. Concise Encyclopedia of Chemical Technology. New York: JohnWiley & Sons.

67

Lloyd, G M and D W Leyshon 1993. Repulping and rewashing phosphogypsum. AIMERegional Conference, Lakeland FL.

Orlov, E.A. and N. N. Treushchenko. 1977. Possibility of combining the process ofmagnesium removal from Karatu phosphorites with the purification of waste containingsulfur oxides gases. Chemical Abstracts. 86: 31547p.

Perry, R. H. and D. W. Green. Eds. 1984. Perry’s Chemical Engineers’ Handbook 6th ed.New York: McGraw-Hill Co.

Peters, M S and K D Timmerhaus 1991. Plant Design and Economics for ChemicalEngineers, 4th Ed. McGraw Hill, NY.

Phillips, J. F., G. H. McClellan and J. F. McCullough. 1980. Chemical beneficiation ofphosphatic limestone and phosphate rock with a-hydroxysulfonic acids. U.S.Patent4,238,459.

Stull, D R and H Prophet 1971. JANAF Thermochemical Tables, 2nd Ed. NSRDS-NBS37.

Ulrich, G D 1984. A Guide to Chemical Engineering Process Design and Economics.John Wiley and Sons, NY.

68

APPENDIX ALeaching Conditions, Acid addition and Phosphate Dissolution

All samples in this table are liquids from sulfuric acid leaching which were assayed for phosphate.Suffix “a” or “b” following the sample number signifies a replicate measurement.A dash in the Sample Number separates sample numbers from run period numbers. When two leaching

tests ran simultaneously, the dash is replaced by “R” for red apparatus and “B” for blue apparatus.“E” in the sample number means that the leachate was stirred with a small amount of fresh feed toprecipitate phosphate and raise the pH.

“F” represents the final leachate sample for the run period. It is taken from bulk liquid after stirringhas stopped and most solids settled, rather than syringe sample of stirred slurry.

“mmols acid/g ore” means millimoles of H2SO4. Multiply by 2 to get milliequivalents of hydronium.Acid used for tests 1-28 was 17.93 molar, or 35.86NAcid used in tests 30-51 was 17.67 molar

A-2

A-3

A-4

A-5

A-6

A-9

APPENDIX BMgO Content of Raw and Leached Ores

All MgO analyses are on solid residues except for samples designated“L” under run 5. The latter leachate assays were converted to asolid-equivalent basis. Sample numbering corresponds to nomenclatureand leaching times in Appendix A.

B-2

B-3

B-4

I APPENDIX CEconomic Model of the Leaching System

PHOSPHATE PURIFICATION PROCESS

C - l

c-3

COST ESCALATION TABLES

c-4

CAPITAltRELATED EQUIPMENT DESIGN DATA

C-5

CAPITAL COSTS

C-6

C-7

C-8

c-9

LABOR AND MISCELLANEOUS EXPENSES

c-10

The Florida Institute of Phosphate Research was created in ...fipr.state.fl.us/wp-content/uploads/2014/12/01-113-138Final.pdf · The Florida Institute of Phosphate Research was created

Documents