Ind. Eng. Chem. Process Des. Dev.,
EY-77-S-02-4083.*000and the assistance of their Chemical and Instrumental Analysis Division in analyzing product samples is gratefully acknowledged. Also, two of the authors (S.K. and Y.T.S.) were partially supported under the above contract. Literature Cited A l b w t , L. F., Yu, Y. C., W d e r , K., "Coke Formation dving Pyrdysis Operation", 85th National Meeting of AIChE, Philadelphia, Pa., Paper 15e, 1978. Brown, S. M., Albright, L. F., ACS Symp. Ser., No. 32, 296-310 (1976). Crynes, B.L., Albright, L. F., Ind. Eng. Chem. Process Des. Dev., 8, 25 (1969). Davis, H. G., Keister, R. B., ACS Symp. Ser., No. 32,392-411 (1976). F r h , K. J., Hoppstock, F. H., H u t c h i i , D. A,, ACS Synp.Ser., No. 32,197-217 (1976). Ghaly. M. A., Crynes, E. L., ACS Symp. Ser., No. 32,218-240 (1976). Haraguchi, T., Nakashb, F., Sakai, W., ACSSymp. Ser., No. 32,99-116 (1976). Korosi, A., Woebcke, H. N., Virk, P. S., Am. Cbem. Soc., Div. FuelChem., Prepr., 21(6), 190 (1976). Kunkleman, J. J., Albright, L. F., ACS Symp. Ser., No. 32,241-260 (1976). Kunugi, T., Kunii, D., Tominaga, H., Sakai, T., Mabuchi, S., Takeshige, K., ACS Symp. Ser.,,No. 32,327-344 (1976). Kunugi, T., Sakai, T., Soma, K., Sasaki. Y., Ind. Eng. Cbem. Fundam., 8. 374 (1969). Kunugi, K., Sakai, T., Soma, K., Sasaki, Y., Ind. Eng. Chem. Fundam., 9, 314 (1970a). Kunugi, K., Soma, K., Sakai, T., Ind. Eng. Cbem. Fundam., 9, 319 (1970b).
Vol. 18, No. 2, 1979
261
Kunzru, D., Shah, Y. T., Stuart, E. B.,Ind. Eng. Chem. Process Des. Dev., 11, 605 (1972). Lettin, H. P., Newsome, D. S.,Wolff, T. J., Yarze, J. C., ACS Symp. Ser., No. 32, 373-391 (1976). Madgwick, G. G., Amberg, G. H.,Peterson, W. S., Can. J . Cbem. Eng., 37, 65 (1959). Powers, D. R., Corcoran, W. ti., ACS Symp. Ser., No. 32, 117-130 (1976). Rondeau, J. A., Come, G. M., Large, J. F.. ACS Symp. Ser., No. 32, 131-151 (1976). Sakai, T., Nohara, D., Kunugi, T., ACS Symp. Ser., No. 32, 152-198 (1976). Scotti, L. J., Merriii, R. C., McMunn, B. D., Romelczyk, S. J., Domina, D. J., Ford, L., Terzian. H. D., Jones, J. F., Char Oil Energy Development, Interim Report No. 5, U.S. ERDA Contract No. E(49-18)-1212, FMC Corporation (1975). Shah, Y. T., Stuart, E. B.,Kunzru, D., Ind. Eng. Chem. Process D e s . D e v . , 12, 344 (1973). Shah, Y. T., Stuart, E. B.,Sheth. K. D., Ind. Eng. Chem. Process D e s . Dev., 15, 518 (1976). Sundaram, K. M., Froment, G. F., Cbem. Eng, Sci., 32, 601 (1977a). Sundaram, K. M., Froment. G. F., Cbem. Eng. Sci., 32,609 (1977b). The Ralph M. Parsons Co., "Commercial Complex Conceptual Design/Economic Analysis, Oil and Power by COED Basad on Coal Conversion," R & D Report 1IbInterim Report No. 1. U S .ERDA Contract No. E(49-16).1775 (Sept 1975). Tsai, C. H., Albright, L. F., ACS Symp. Ser., No. 32, 274-295 (1976).
Received f o r review D e c e m b e r 27, 1977 Accepted S e p t e m b e r 28, 1978
Plasma Jet Process for Making Elemental Phosphorus John D. Chase, Joseph F. Skrivan,' Daniel Hyman, and James E. Longfield American Cyanamid Company, Chemical Research Division, Stamford, Connecticut 06904
A process for making elemental phosphorus which is based on an experimental investigation is described where phosphate rock is reduced by heating with a hydrocarbon gas in a plasma jet. The effect of plasma power level on extent of reduction is shown. Up to 80% of the phosphorus in the rock is reduced for feed rates of 2-3 Ib/h. The extent of reduction is measured by analysis of both product gas and slag. A tubular graphite reactor was more durable than an alumina reactor. Heat transfer calculations and similar results for -200 mesh ground rock and smaller (25 km) jet-milled rock suggest that heat transfer to the rock does not limit the conversion in the system used. The minimum electrical power requirement obtained in these experiments, where the plasma gas composition varied from 0 to 30% H2 in argon, was 24 kWh/lb of phosphorus. A process concept envisages the use of a recycled 2:l CO-H2 product gas mixture as the plasma gas with an enthalpy which is double the 6000 Btu/lb of the 30% H2-Ar mixture used in these experiments, which would presumably lower the power requirement. The thermochemistry for several reducing agents with and without silica flux is considered and the implications are discussed with respect to the proposed process.
Introduction Since the turn of the century elemental phosphorus has been made by the electrothermal furnace process. This process involves heating phosphate rock, coke, and silica in a carbon-lined electric furnace where electricity provides the energy to heat the furnace charge to 1250-1500 "C whereby carbothermal reduction of the rock occurs to CaSi03 slag, elemental phosphorus, and CO. During the past decade there have been a number of schemes described in the scientific and patent literature. These are similar to the furnace process in that phosphorus is obtained by high temperature reduction of the ore, but differ in that reaction does not take place in a massive molten phase. Three of these schemes utilize plasma jets to provide the high reaction temperature. One proposed process (Goldberger and Baroch, 1966) uses a plasma jet to fluidize a bed of coke and silica. Another, described by Mosse et al. (1968), involves passing ground phosphate rock with or without silica through an N2 plasma, then rapidly
* Address correspondence to this author a t American C y a n a m i d Co., B e r d a n Avenue, W a y n e , N.J. 07470. 0019-7882/79/1118-0261$01 .OO/O
quenching with water to dissociate the ore such that the treated ore can be leached to recover Pz05. This scheme involves physical treatment rather than chemical since no reducing gas is used and elemental phosphorus is not made. Another plasma process by Foex et al. (1971) incorporates a sophisticated reactor design wherein an inclined rotating furnace is heated by an arc between two plasma jets at each end of the furnace. Ore, silica, and coke are fed through the top, and molten slag and phosphorus vapor are removed from the bottom of the furnace. According to thermodynamics, methane will reduce phosphate rock more readily (at lower temperature) than carbon. Since this reaction will occur below the melting point of phosphate rock (1950 K), a gas-solid reaction is indicated which suggests feeding ground phosphate rock suspended in methane to a plasma jet reactor without the addition of silica, conventionally added to lower the melting point of the molten slag in the furnace process. This approach is followed in the present work. Heat transfer between plasma gas and the solid ore particles typically is the limiting process in ore reduction in arcs and plasmas. Since this heat transfer is greatly enhanced by increasing the surface area by grinding the solids, finely ground ore (--a00 mesh) and further jet milled ore were
e 1979 American Chemical Society
262
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979 11501
,
,
,
,
3
,
I rl=kr
1
COOLED
TEMPERATURE,
‘K
Figure 1. T h e r m o d y n a m i c s of p h o s p h a t e rock r e d u c t i o n .
used in the present investigation.
Thermodynamics of Phosphate Rock Reduction Free energy calculations were made for reduction of phosphate rock by hydrogen, silica and hydrogen, carbon, methane, acetylene, and carbon and silicon in reactions 1 through 6, respectively. Thermodynamic data for the calculations were obtained from National Bureau of Standards Circular 500 (1952), Kelley et al. (19501,and the “JANAF Thermochemical Tables” (1970). In the following equations, thermodynamic properties of actual rock, fluorapatite, were used. However, in the chemical equations, for simplicity, tricalcium phosphate is written in place of fluorapatite although the latter contains about 10% silica and 2-3% fluorine as calcium fluoride.
--
+ 5H20 + P2 (1) Ca3(P04)2+ 3Si02 + 5H2 3CaSi03 + 5H20 + P2 (2) Ca3(POJ2 + 5C 3Ca0 + 5CO + P2 (3) Ca3(P0,)2 + 5CH4 3Ca0 + P2 + 5CO + 10H2 (4) Ca3(PO4I2+ 2.5C2H2 3Ca0 + P2 + 5CO + 2.5H2 (5) 3CaSi03 + P2 + 5CO (6) Ca3(P04)2+ 3sio2 + 5C Ca3(P04)2+ 5H2
--
3Ca0
-
-
The free energy for reducing phosphate rock at various temperatures is plotted in Figure 1. On the basis of the reactions as written, the best reductant is methane followed by acetylene, followed by the furnace reaction (6) where the reactants are wetted by CaSiO, slag at relatively low temperatures (above about 1600 K), and finally carbon alone in reaction 3. Hydrogen with or without silica is not effective. A large amount of heat must be supplied both to heat the reactants to reaction temperature and to supply the endotherm for the reduction reaction. The thermochemistry for the commercial furnace process, eq 6, will be compared to the present process, eq 4. Reaction 4 is favorable at a lower temperature (1400 K) than reaction 6 (1650 K). The minimum theoretical energy required to provide the sensible heat to the reactants and the endotherm is 690 and 600 kcal/g-mol of P2for reactions 4 and 6 at 1400 and 1650 K, respectively. In this calculation only CaSiO, was considered to be in the liquid phase. Thus, if reduction for the present process proceeds between solid rock and gaseous methane (eq 4), then according to the thermochemistry the reaction will take place at lower temperature although it will require slightly (16%) more energy. Kushnir and Klimovich (1968) have also shown from thermodynamic calculations that CHI is more effective in reducing phosphate rock than carbon. Experiments by
MAJOR PORTION OF PRODUCT
\HEAT EXCHANGER
IN PHOSPHORUS COLLECTING DRUM
Figure 2. E q u i p m e n t for p l a s m a j e t p h o s p h a t e rock r e d u c t i o n .
Fastorskii (1947) have shown that in the absence of an active pyrolysis catalyst it is difficult to achieve equilibrium decomposition of methane below 1300 K and that purely methane reduction occurs at this temperature (Klimovich, 1958). Kushnir (1968) pointed out that high-temperature pyrolysis of methane is minimized when hydrogen is added to the system and that the kinetics of phosphate rock reduction are improved when the gas phase is dilute in methane. Thus in the present case in the plasma jet reactor where the system was dilute (less than 10%) in methane and a substantial amount of hydrogen was usually present, it is reasonable to assume that undecomposed methane was present and this reacts with the phosphate rock as illustrated in eq 4. Also, even if methane loses hydrogen and decomposes to acetylene, from eq 5 acetylene is shown to be nearly as effective a reducing agent as methane.
Experimental Section A. Equipment. A diagram of the equipment used is shown in Figure 2. A dc plasma torch rated at 50 kW (made by Thermal Dynamics Corp., Model H50A) equipped with thoriated tungsten cathode and a copper anode was used. Four 12-kVA Miller welding rectifiers connected in parallel (each with voltage taps at 40, 80, and 160 open circuit V) served as the dc power supply. Torch power was metered and controlled with a Thermal Dynamics Corp. power console. The nozzle of the plasma torch was connected to the top of a 15 in. diameter, 20 in. long stainless steel water-cooled reactor housing. Into the effluent plasma, ground ore suspended in a stream of methane reducing gas was fed through a water-cooled feed probe at an angle of 45” to the axis of a graphite reactor. 0-Ring construction was used to maintain a gas-tight reactor enclosure for the reducing atmosphere. The ground ore was metered at feed rates between 7 and 40 g/min using a vibrating screw feeder with variable speed drive and a gas-tight hopper. This feeder (Vibra Screw Corp., 0.5 in.) was calibrated before each run and had an accuracy of better than *2%. In order to suspend the ore with methane a glass cone, which had a 4 mm outlet at the bottom, was attached to the feeder spout. Metered methane gas and suspended ore passed through a transport line to a water-cooled probe into the top of the reactor. Most experiments were done with a 19.5 in. long 3 in. i.d. 4 in. 0.d. graphite tube (Ultra Carbon Co.). Since the graphite reactor is operated hot to prevent solidification of the molten ore on the walls, high-temperature insulation was required between the reactor and the cooled outer housing. A fibrous low-density alumina-chromia mixture
Ind. Eng. Chem. Process DES. Dev., Vol. 18, No. 2. 1979 263
was used which is suitable for service to 2700 "F. The reactor housing was supported on a 6 in. i.d., 15 in. long Inconel-lined mild steel tee. A flange was attached to one end and a 5 ft long 6 in. i.d. mild steel, Inconel-lined heat exchanger was attached to the other end. Both the tee and the heat exchanger were wound on the outside with copper coils through which steam flows to cool the solids and gases leaving the reactor and to prevent condensation of phosphorus. Fitted to wire supports attached to the exit flange of the heat exchanger were two 1.5 ft long 2 in. diameter Teflon filter bags which served to collect any fine solids suspended in the gas stream. The significantvolume of the heat exchanger also provided a settling chamber for removal of some suspended solids. Some slag at the base of the reactor was removed through the tee flange after the run. In order to prevent high solids loadings on the bag filters leading to high pressure drop, a periodic reverse pulse of nitrogen was provided with a Pnlsair (Pulverizing Machinery Co.) collection unit. After leaving the filter, the gas passes through a heated line to a scrubber where the phosphorus gas is condensed with a water spray. A t the end of a run the condensed phosphorus was melted with a heating coil a t the base of the scrubber and the phosphorus was collected. The gas from the scrubber was diluted with nitrogen to bring the mixture below the flammability limit and then sparged into a 10-gal water tank to remove any uncondensed phosphorus and to provide a water seal against leakage of air into the system. B. Procedure. When H,-Ar mixtures were used as plasma gas the plasma torch was started with argon only and then the required hydrogen was added. The argon plasma gas flow was about 2.3 g-mol/min in all runs. Prior to feeding phosphate rock, the plasma was operated (at somewhat reduced power level) for 1.5 h to bring the reactor to thermal equilihrium. The entire system was operated under a slight positive pressure to prevent air from leaking in and mixing with the potentially explosive mixture which contains up to 30% hydrogen. During each run, where the reactants were fed for at least 20 min, product gas samples were taken downstream of the filter for phosphorus analysis. Also, solid products were collected from below the reactor, inside the settling heat exchanger, and on the filter bags. C. Analyses. The fraction of phosphorus in the rock which is reduced was determined by measuring the amount of phosphorus condensed from a sample of product gas. The phosphorus concentration was related to inert argon gas in the sample, and the yield was then calculated from known flow rates of argon in the plasma gas and the feed rate of the phosphate rock. The gas sample bulbs used contained about one-fourth of their (250 cm3) volume of Cu(NO& solution which reacts with phosphorus in the gas to form solid copper phosphide. The weight of phosphorus in the bulb was then determined by a calorimetric technique, which was found to be reproducible to *3%. A l-cm3 syringe was used to remove some gas from the sample bulb for gas chromatographic analysis of HZ,CO, CH,, AI, and Nz.Prior to each run the chromatograph was calibrated with a known gas mixture. The phosphate rock and in some runs the product solids were analyzed by conventional wet analytical techniques to determine percent PzOs, CaO, SiOz, CaCO,, C, and F. Where the product solids were analyzed, the yield of phosphorus was determined from the ratio (P20s/CaO)in the product solids and ore feed. An analysis of feed rock (Florida Land Pebble Phosphate rock BPL grade 68/66), supplied as ground to 85%
Table I. Typical Analysis of Feed Rock and Product Solids %
%
%
P,O, SiO, CaO rock productsolids
30.9 10.2 45.9 10.1 5.6 63.8
%
%FCaCO, 1.01 2.2 8.4
%C
1.89
4.5
2.8
Figure 3. Micrographs of jet-milled ore feed and product solids, run 6 9 a, jet-milled ore feed; b, product solids with jet-milled ore feed, 81% reduction; Pn= 3.5 kWh/lb.
through 200 mesh, and product solids for run 69 is given in Table I. This analysis corresponds to a 77% conversion. Results a n d Discussion A. General. The main series of experiments consisted of 15 initial runs with a pure argon plasma and 15 runs where a mixture of argon and hydrogen (up to 30% Hz) were used. In these experiments a graphite reactor tube and the larger, well-insulated reactor housing were used which have been previously descrihed. The major variables investigated were power level or reactor temperature, ore feed rate, ore particle size, and reactor diameter. In all runs 13% excess methane according to eq 4 was used. Silica flux was not used. B. Mode of O r e Reduction. Most of the product solids were a dark powder deposited just below the base of the reactor and the remainder was either slag which solidified after dropping off the reactor wall or very fine particles deposited in the heat exchanger or filter. From the appearance and location where the solids were collected it appears that most of the reduction takes place with gas-borne particles. In all experiments, 13% excess methane was used and the carhon formed on cracking probably contributed to the dark color of the product solids. When ore, but no methane, was fed to the plasma no phosphorus was found in the product gas. This experiment indicates that no significant reduction occurs on the surface of the graphite reactor with any ore particles which came in contact with it. X-ray diffraction measurements made on the product solids indicated that CaCz was not present. Thus the CaO formed carhide in the reactor. Diffraction lines for Ca,PzOg were observed, however, indicating that the following reaction between reactant and product of reduction reaction (eq 4) occurs Ca3(P04)2+ CaO Ca,PzOg (7)
-
Micrographs of ore feed particles and corresponding product solids were taken using transmitted light at 150X and these are shown in Figures 3 and 4. The product solids were heated in a furnace in air to oxidize carbon present before the micrographs were taken. In Figure 3
264
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979
Table 11. Data for Some Typical Phosphate Rock Reduction Experiments run no.
Figure 4. Micrographs of -200 mesh ore feed and product solids, run 70: a, -200 mesh ore feed; h, prcduct solids with -200 mesh ore feed, P, = 1.9 kWh/lb; 55% reduction.
the jet-milled ore feed is seen to have an average particle size (by volume) of about 25 wm and the product solids are of similar size or slightly larger since there are quite a number of larger irregularly shaped agglomerates. Although it seems likely that some carbon may still he present which partially ohscures the micrograph, the absence of a large number of spheres or particles significantly different in size from the feed material suggests that (a) either the rock did not completely melt or (h) that it did melt and had enough time to recrystalize into irregularly shaped crystals. In comparison with the results of Figure 4 for larger particle ore feed where many spheres were observed at lower power levels, alternative (h) appears most likely. Figure 4 shows micrographs of the larger -200 mesh ore feed and corresponding product solids. The average particle size (by volume) is about 60 pm. A similar size was found for the product solids. However, many more spheres are seen in the product than in that for the smaller jet-milled feed. Also, it is interesting to note that some of the spheres appear to he hollow or to consist of two layers such as CaO encapsulated ore. A dark interior of processed solids when viewed in transmitted light usually indicates a large difference in refractive index such as produced hy gas cavity. In the present case the cavity would have been phosphorus vapor which was encapsulated when the molten particle solidified. This encapsulated gas is observed only in spheres which were spheroidized and were molten in the plasma and has insufficient time to complete the reduction in the reactor. Presumably reduction was complete in the smaller particles. The net power level (P,) = 1.9 kWh/lh in run 70 rather than 3.5 kWh/lh in run 69 shown in Figure 3, so the reduction would he less likely to advance as far with the larger rock in the former case. The percent reduction in runs 69 and 70 determined from gas analysis was 81 and 55%, respectively. C. Effect of Experimental Variables on Reduction. Data for four typical higher power runs are given in Table 11. Torch gross power is the product of dc current and voltage a t the torch. Torch net power is power in the plasma gas leaving the torch and is the gross power minus power loss to the cooling water circulated through the torch. The main reason for the poor torch efficiencies obtained is the high proportion of the monatomic argon plasma gas used. Another reason is that the torch efficiency (net power/gross power) may appear lower than it is, since hack radiation from the adjacent hot reactor will appear as heat removed in the torch cooling water.
conditions
63
68
69
70
torch gross power, kW torch net power, kW argon plasma gas flow, g-mollmin total argon gas flow, g-mol/min % H, in plasma gas ore feed rate, glmin ore particle size, vol. av, Mm plasma voltage, V de reducing gas, 13% excess reactor diameter, in. i.d. % reduction by gas analysis %reduction by slag analysis energy available above 1400 K, kWh1lb of rock heat loss toreactor cooling water, kW calcd mixed reactor inlet temp, K calcd mixed reactor outlet temp, K reactor temperature (av), K gas residence time in reactor, s
16.9 3.5 1.51
19.2 4.2 1.45
36.6 7.4 1.51
20.4 4.0 1.48
2.37
2.10
1.91
1.88
7 20 25
15 10 25
30 16 25
15 16 60
40 CH, 1 44
40 C&, 3 38
62 CH, 3 81
CH,
51 3 56
I1 1.15
2.22
2.77
1.27
1.56
1.94
3.84
2.98
2125
3080
3325
2725
1005
1835
2005
330
1565
2460
2665
1530
0.035
0.25
0.21
0.38
Reactor inlet temperature is calculated by assuming plasma gas, ore, and reducing gas are perfectly mixed and that the heat in the plasma gas entering the reactor heats the reactants and provides the reaction endotherm and the endotherm to decompose the excess hydrocarbon present. Reactor outlet temperature is calculated by repeating the above calculation after deducting the heat loss to the reactor housing cooling water. Since considerable hydrogen was often present and significant hydrogen dissociation occurs at the reactor inlet temperature, the heat absorbed in dissociating the hydrogen was accounted for in the calculation. For the highest power run (no. 69), the heat required to dissociate hydrogen was even greater than the reaction endotherm. On the assumption that the reduction reaction does not proceed below approximately 1400 K, the energy availahle at the reactor inlet per pound of rock fed was determined and included in Tahle 11. Also included in Tahle I1 is the gas residence time in the reactor, which is determined from the average reactor temperature and gas flow rate. The gas residence time will he the lower limit on particle residence times, and these will he discussed in the next section. In Figure 5 results are plotted as percent conversion vs. power level per unit of rock, i.e. (net kWhjlh of rock). Since the relatively small torches such as used in this work are much less efficient than the large megawatt torches suitable for scale-up, net power is the relevant parameter. The calculated curve in Figure 2 includes sensihle heat in a 30% Hz in Ar plasma gas. A t the relative flow rates of rock (16 g/min) and plasma gas (1.9 g-mol/min) used, about half the sensible heat is in the plasma gas, and this accounts for the high slope of this curve. It is seen that at least for the higher degrees of reduction with the present equipment, considerahly more heat than theoretical is required. Although some scatter exists, a simple straight line can he drawn through the experimental points for
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979 265
The reaction endotherm which must be provided per particle, Qz, is
CALCULATED TO 14W"KI
Q2
80
I 1 z L 2 6 0
I I
= n/6D3pAHX
(9)
where p is the density of the rock or Ca3(P04)2= 3.14 g/cm3, AH is 1090 cal/g of ore reaction endotherm, and X is fractional conversion. Evaluating eq 9 for 70% conversion for a 50-pm particle
/
Qz = n / 6 (0.125) X 10-6(3.14)(1.09)X 103(0.70) = 0.156
X
cal
The time required or residence time required to transfer heat from 2000 K plasma gas to a 1400 K particle to provide the reaction endotherm is therefore e JET MILLED ROCK, 3 IN.REACTOR 0 JET MILLED ROCK, 1 N R E A C T O R
A
REDUCING GAS
1
7
-200 MESH NON-JET MILLED ROCK
0 KWH (NET1 LE. ROCK
Figure 5. Reduction of phosphate rock in hydrogen-argon plasma.
methane reducing gas for all reactor diameters and both particle sizes of ore. If heat transfer to the particles from the plasma gas were limiting the reduction reaction, as is so often the case in high-temperature plasma processing, one would not expect results for different particle sizes to be correlated on the same line, since heat transfer is so dependent on particle size. In one run (no. 68) cyclohexane rather than methane was used as the reducing gas and as can be seen from Figure 2 for a given power level the percentage reduction is considerably (nearly 40%) less than for methane. Cyclohexane was used to simulate a naphtha-type hydrocarbon and the results from a single experiment indicate that type of feed stock would be less satisfactory than natural gas. D. Residence Time Considerations. About four times as much heat is required to provide the reaction endotherm (1.09 kcal/g of ore) as is needed to heat the ore to the temperature where the reduction will proceed so a heat transfer calculation will be done first for the former case. For an ore particle suspended in the plasma gas the rate of steady-state transfer by convection from the plasma gas to the particle is given in eq 8 xD2h (Tg- T,) =
Q1
(8)
where D is particle diameter, h is convective heat transfer coefficient, and T and T are average gas and particle temperatures at wkich reJuction proceeds. Consider an average reactor gas temperature of 2000 K (see Table 11) and a particle temperature at 1400 K where the reduction is initiated. For small particles in the 50-pm range suspended in a plasma gas, Johnston (1972) has found that the Nusselt number is close to 2. That is, hD/k = 2, where k is the thermal conductivity of the gas at the film temperature, T where T = (T + T,)/2 = 1700 K. For a 30% H2-70% Ar plasma gas, kfmixture), was determined from K(H,) and k(Ar) by the method of Reid and Sherwood (1966). The thermal conductivities of hydrogen and argon were evaluated from Lennard-Jones potentials (Bird et al., 1960). The radiative heat flux can be neglected at 1400 K in comparison with heat transfer by convection (Johnston, 1972). Equation 8 can now be evaluated: D = 50 pm = 0.5 X lo-' cm; k = 0.39 X cal/cm s K; h = 0.155 cal/cm2 s K; and nD2h(T, - Tp)= Q1 = n[0.25 x 10-4][0.155(2000 - 1400)l = 0.73 X cal/s.
=
Q2/Q1=
0.156 o,73
X
= 0.214 s
Although this simplified calculation is somewhat crude since it assumes a steady-state whereas the temperature varies considerably axially along the reactor, the results indicate that the available residence times for the 3-in. reactor (0.221-0.38 s; see Table 11) are adequate. However, the residence time in the 1-in. reactor is not sufficient. It should be noted that the results of the above calculation are likely to be conservative (give too large 7) for two reasons. (1) The assumption is that the ore particles immediately reach the gas velocity whereas their residence time in the reactor will actually be somewhat larger than the gas. For example, 5.5 cm is required for a 35-pm particle to be accelerated by the surrounding plasma gas to reach 90% of the plasma gas velocity. (2) The second reason is that the Nusselt number will be greater than 2 at the point when the particle enters the reactor due to the difference in particle and plasma velocities and consequently h will be larger and the required residence time shorter. The time required to heat the ore particles from ambient to 1400 K, where reduction will begin, can be calculated using eq 10 given by Johnston (1972). The equation was derived by equating convective heat transfer rate to heat content of a particle
T, - TP/ D2C$, T,- T, 6kNu where Tg, TP/, and T, are plasma gas, particle inlet, and particle outlet temperatures. D, C,, and Ppare diameter, specific heat, and density of the particle, and k, Nu, and t are gas thermal conductivity, Nusselt number, and required time, respectively. Equation 10 was evaluated for D, = 50 pm, Tp', T, and T, = 300, 1400, and 2000 K for a 30% &-yo% Ar plasma, and the required residence time t was determined to be 0.006 s. As previously indicated, the time to heat up the rock is small in comparison with the time to transfer the much larger reaction endotherm. The importance of having finely divided ore is seen since t 0: D2. The residence time, Q2/Q1, is proportional to the first power of particle diameter. This dependence on particle size indicated by both correlations is not evident in the experimental results plotted in Figures 2 and 3, although very few experiments with large particle size rock (-200 mesh, av D, = 60 pm) at the interesting higher power levels were carried out. E. Process Considerations. A scaled-up plant based on the present process (American Cyanamid Co., 1971, 1972, 1974) would include the following features not in-
266
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979
cluded in small-scale experiments done so far: the plasma torches would be 3-10 MW, which would operate on a recycled mixture of product gas, 2 parts of H2to 1 part of CO. The torches would operate on 60-cycle ac power at about lo00 V at 80% efficiency producing a heated H2-CO gas mixture with an enthalpy of 13000 Btu/lb (Fey et al., 1970). The higher enthalpy plasma gas should result in a greater proportion of its heat in the useful temperature range above 1400 K, which would be expected to reduce the net power requirement for 80% reductions below the 3.5 kWh/lb of ore as shown in Figure 2. Probably the largest improvement on scale-up would be in torch efficiency = (net power in torch exit gas)/(electrical power input to torch) from 0.2 in the present experiments to 0.8 (Westinghouse, 1970). A byproduct credit could be obtained for the 2:l H2-CO mixture produced even after some of this gas is used as fuel to burn with air to preheat the ore to be fed to the reactor. Conclusions This work has demonstrated the following: (a) the feasibility of using methane reducing gas, (b) up to 80% yields of phosphorus by reduction of phosphate rock in a plasma reactor, and (c) continuous operation with a hot graphite reactor giving largely free-flowing particulate solids. Net power requirements on the present 3 lb/h scale were surprisingly large, with a minimum of 24 kWh/lb of phosphorus in comparison with the theoretical value of about 7 kWh/lb of phosphorus. The former value would be expected to decrease considerably on scale-up since heat losses decrease with decrease in surface/volume ratios. Also, envisaged large-scale operation would utilize higher (more than double) enthalpy H2-C0 plasma which has a higher proportion of its enthalpy in the useful temperature
range above 1400 K than the H2-Ar plasma used in the present work. Acknowledgments The contributions of Mr. Vincent Di Stefan0 in the early stage of the project and the skillful experimental assitance provided by Mr. Howard Shaw throughout the project are gratefully acknowledged. Literature Cited American Cyanamid Co.. French Patent 7 038968 (1971). American Cyanamid Co., U.S. Patent 3679 363 (1972). American Cyanamid Co., US. Patent 3 832 448 (1974). Bird, R. B., Stewart, W. E., Lightfoot, E. N., "Transport Phenomena", p 744, Wiley, New Yolk, N.Y., 1960. Fastovskii, V. T., Methane", Gostoptekhizdat, Moscow-Leningrad, 1947. Fey, M. G., Hirayama, C., paper presented at 160th National Meeting of American Chemical Society, Chicago, Ill., Sept 14, 1970. Foex, M., Dumon, R., Bot-Langet, Y., Verouchalmi, D., French Patent 2 079 746 (1971). Gilles, H. L., Clump, C. W., Ind. Eng. Chem. Process Des. Dev., 9, 194 (1970). Goldberger, W. M., Baroch, C. J., US. Patent 3247014 (1968). "JANAF Thermochemical Tables", 2nd ed.,U.S.Department of Commerce, NSRDS-NBS 37, 1970. Johnston, P. D., Combust. &me, 18, 373 (1972). Kelley, K. K., et ai. U.S.Bureau of Mines, Contributions to the Data on Theoretical Metallurgy, Bulletins 303, 476, and 477 (1950). Klimovich, A. I., Izv. Vuzov, Khim. Khim. Tekhnoi., 8 , 71 (1958). Kushnir, S. V., Klimovich, A. I., Zh. Prik. Khim., 41(1), 106 (1968). Mosse, A. L.,Pechkovskii, V. V., Menkh, V. A,, Dvindenko, I. A,, J . Eng. phys. (USSR), 15, 6 (1968). "Selected Values of Chemical Thermodynamic Properties", Natl. Bur. Stand. (U.S.) Circ., No. 500 (1952). Reid, R. C., Sherwood, T. K., "Properties of Gases and Liquids: Their E s t i i t b n and Correlation", 2nd ed, p 205, McGraw-Hill, New York, N.Y., 1966. Westinghouse Research Laboratories, Pittsburgh, Pa., private communication, 1970.
Receiued for review December 30, 1977 Accepted October 24, 1978 Presented at the Symposium on Potential Plasma Processes, 168th National Meeting of the American Chemical Society, Atlantic City, N.J., Sept 8, 1974.
An Assessment of Distillation Packing Performance from Wetting Data Obtained under Equilibrium Conditions Sebastian Fabre and Anthony B. Ponter" Chemical €ngineering Institute, Swiss Federal Institute
of Technology, Lausanne, Switzerland
Max Huber Process Engineering Division, Sulzers Bros., Winterthur, S witzerland
Although it is generally assumed when selecting a distillation packing that the wetting of a solid surface by a liquid without mass transfer will indicate its ability to wet under distillation conditions, for systems whose surface tensions change considerably with composition this postulate is shown to be invalid. The underlying reasons for this are analyzed and illustrated by considering the wetting characteristics of aqueous 1-propanol films flowing down a vertical copper surface both under equilibrium conditions and at total reflux.
Introduction The efficient separation of components by distillation depends among other things on the ability of the liquid to spread over the packing to give the maximum surface *Address correspondence t o this author at the Chemical Engineering Department, University of Aston in Birmingham, Birmingham, U.K. 0019-7882/79/1118-0266$01.00/0
area for mass transfer to take place. This spreading characteristic depends both on the bulk physical properties such as liquid density and viscosity and also upon the surface tension and the contact angle which the liquid exhibits to the vapor-solid interface. Surface roughness also Plays a Peier et al. (1977) have recently presented a model which includes these parameters to describe the wetting of a 0 1979 American Chemical Society
