I n d . Eng. Chem. Res. 1988,27, 1264-1269
1264
= holdup of coarse particles in dense bed
D = diameter of the bed, cm
cs
df, = interaction force due to collosion between fine particle and dense particle dfd = drag force on fine particles due to air df, = frictional force due to bed wall and dense particles df, = net gravitational force dm, = mass of fine particles f k = interaction coefficient defined by eq 8 f k k = particle-particle interaction coefficient F, = friction factor in the bed with dense particles g = gravitational acceleration, cm/sz L = height of bed, cm m = loading ratio W,/ W , P,,,= average root mean square of pressure fluctuations Re, = Reynolds defined as dp(Ua- Up)pa/wa t = time, s U , = linear air velocity, cm/s U, = air velocity at which pressure fluctuations reach a maximum, m/s Uk = air velocity beyond U, at which pressure fluctuations level off, m/s U, = superficial air velocity, m/s U = linear particle velocity, cm/s $, = flow rate of circulating fine particles, g/s W , = mass flow rate of air, g/s W , = mass of dense particles, g
density of air, g/cm3 density of fine particles, g/cm3 ps = density of coarse particles, g/cm3 pa = viscosity of air, g/cm/s pa =
pp =
Literature Cited Fan, L.-S.; Toda, M.; Satija, S. Powder Technol. 1983,36, 107. Fan, L.-S.; Toda, M.; Satija, S. Chem. Eng. Sci. 1985,40, 809. Johnson, E. P. Chem. Eng. 1982,89(7), 39. Liu, K. T.; Martin, J. C.; Nack, H. Presented at the AIChE 86th National Meeting, Houston, TX, April 1-5, 1979. Nack, H.; Felton, G. W.; Liu, K. T. Proceedings of the 5th International Conference on Fluidized Bed Combustion, Washington, D.C., 1977; Vol. 3, p 223. Satiia. S.: Fan. L A . Chem. Enp. Sci. 19858.40. 259. SatGa; S.; Fan, L.-S. AIChE J.-1985b,31, 1554.' Satija, S.;Young, J. B.; Fan, L.-S. Powder Technol. 1985,43, 257. Schiller, L.; Naumann, N. VDI 2 (1857-1968) 1933,77, 318. Toda, M.; Satija, S.; Fan, L.-S. Proceedings of the IVth International Conference on Fluidization, Kashikojima, Japan, May 29-June 3, 1983; paper 2-7-1. Welschof, G. VDZ Forschungsheft 1962,492. Wen, C. Y.; Yu, Y. H. Chem. Eng. Symp. Ser. 1966,62(62), 100. Wisecarver, K.; Kitano, K.; Fan, L.-S. In Circulating Fluidized Bed Technology; Basu, P., Ed.; Pergamon: New York, 1986; p 145. Yang, W. C. AIChE J. 1984,30,1025. Yerushalmi, J.; Cankurt, N. T. Powder Technol. 1979,24,187. ,
I
Greek Symbols = void fraction in dense bed tp = holdup of fine particles in dense bed
Received for review March 25, 1986 Revised manuscript received August 24, 1987 Accepted December 18, 1987
t
Experimental Study of the Production of NO, N20, and O3 in a Simulated Atmospheric Corona R o b e r t D. Hill, Iraj Rahmim, and Robert G. R i n k e r * Department of Chemical and Nuclear Engineering, University of California, Santa Barbara, California 93106
T h e production of NO, N20,and O3 in atmospheric coronas was experimentally investigated using a specially designed plasma reactor operated a t a pressure of 0.1 MPa and a temperature of 300 K. T h e yields of NO, NzO, and O3 were found t o be (1.4 f 0.7) X (1f 0.5) X 1017,and (4 f 2) X 1017molecules/ J, respectively. There was also formation of NO2which was attributed mainly t o the rapid reaction between NO and 03.T h e results from this study confirm the predictions of mathematical simulations of corona chemistry. The Lin shock-coolingtheory (Chameides, 1979) predicts measurable production of N20 by lightning in the heated channel. On the other hand, the thermal convection cooling theory, proposed by Hill (1971, 1977) and confirmed experimentally by the Naval Research Laboratory (Picone et al., 1981),predicts no significant production of N20 in the heated channel (Hill et al., 1980; Hill and Rinker, 1981). Since enhanced concentrations of N,O have been observed in thunderstorms (Levine and Shaw, 1983) and since a fraction of the lightning and thunderstorm energy is actually expended in the corona (Hill et ai.,1984), Hill and Rinker (1981) proposed that N20 is produced in the lightning corona rather than in the heated channel. The distinction between trace gas production of NO and N20 by hot filamentary electrical channels and by electrical corona (or brush) discharges in air was ably demonstrated by Donohoe et al. (1977) at Caltech. This work, however, was apparently overlooked by the atmospheric pollution investigators of the early 1980s until it was brought to the attention of the investigators of lightning 0888-5885/88/2627-1264$01.50/0
phenomena by Hill and Rinker (1981) at UC, Santa Barbara. As already stated, the modeling of N 2 0 production by corona discharges in lightning was begun a t UC, Santa Barbara, in 1981, and these results were published by Hill et al. (1984). However, no further experimental confirmation of the production efficiencies of NO and N 2 0 determined by Donohoe et al. (1977) had yet been performed. The purpose of the present investigation (Rahmim, 1984) was to experimentally confirm and refine the original Caltech work, in light of the theoretical framework which had been established. It is noteworthy that the theoretical interpretation, offered by Donohoe et al. (1977) of the chemical reactions mainly responsible for NO and N 2 0 production in coronas, was found to be in disagreement with our interpretation (Hill et al., 1984). It therefore became important for the Santa Barbara group to independently investigate the experimental production of NO and NzO and, if possible, to measure directly the production of 03.Measurement of the latter was not reported 0 1988 American Chemical Society
/6 [q Spark
harg e
- -
..
Variac
Neon-tu be Transformer
Tesla Transformer
Figure 1. Schematic diagram of the discharge circuit. To Other End of Tesla Secondary (grounded) Gas I n
.
--
Teflon Tee Fitting
Central Electrode
Outer Electrode Secondary From Tesla
Lucite Discs
- 'I'
/
Gas Out
Figure 2. Plasma reactor.
but was certainly inferred by the Caltech group.
Apparatus and Procedure The apparatus used in our experiment was a steady-flow nonequilibrium plasma reactor which followed closely on the design given by Donohoe et al. (1977). The source of the corona discharge was a Tesla transformer capable of delivering a many-kilovolt pulse of radio-frequency oscillations across a reactor cell. Schematics of the discharge circuit and of the plasma reactor are given in Figures 1and 2. The discharge cell electrodes consist of an aluminum cylinder and a copper wire mounted along the cylinder's central axis. The cylinder is 0.18 m long with a 3 X m wall thickness and has an inside diameter of 2.35 X m. It is connected to one end of the Tesla transformer secondary, which is grounded. Two Plexiglas discs, which are 0.01 m thick with a diameter of 0.03 m, are used as endcaps for the reactor. They also serve as windows for viewing the plasma, using fiber optics connected to a photomultiplier and pulse counter in series. When the pulse counter gives zero signal, this criterion ensures that the reactor is operating with a pure corona plasma, the condition under which all the experiments in this paper were performed. Two Teflon tee fittings are used as the anchors of the central electrode on each end of the reador and as the inlet or outlet for the flow of gas. In addition, the inlet tee contains a feed nozzle which is concentric to the central electrode. The central electrode is a 3.5 X 10"' m diameter copper wire which was drawn tightly for an extended period of time before sealing. The two ends of the wire, one of which is connected to the Tesla secondary, are then held in place in the tee fittings using epoxy resin. The entire discharge apparatus is mounted on a Plexiglas base and enclosed in a grounded electrically shielded box. In order to determine the electrical power input to the reactor, it was necessary to measure both the voltage across and the current through the discharge cell. The voltage was measured by using a Tektronix high-voltage probe. The probe is rated at 40-kV peak with a 4 X 10" s rise time and a 1OOO:l attenuation. The current was measured by
Ind. Eng. Chem. Res., Vol. 27, No. 7 , 1988 1265 Table I. Specifications of the Gases Used name N2 air
grade specifications extra dry min purity, 99.7%; max moisture, 3 ppm zero max total H-C, 1 ppm; max moisture, 3 PPm 18 ppm NO, balance N2 NO/N2 custom 16 ppm NOz, balance N2 NOZ/N2 custom 29 ppm NzO, balance N2 N20/Nz custom
using a Pearson current monitor which has a 500-A peak current rating, a rise time of 10 x s, and an output of 1 VIA. Both the current and voltage traces were monitored on a Tektronix oscilloscope which could respond to a rise time of less than IO-' s. The duration of a corona pulse was determined by numerical analysis of pulse oscilloscope tracings and was found to be (1.3 f 0.2) X lO-'s. For the highest flow rates used (-50 X lo4 m3/s), typical voltage and current pulses averaged (630 f 30) V and (70 f 5) mA, respectively. The average power and energy per pulse, assuming that the voltage and current oscillations were in phase, were therefore (44 f 5 ) W and (5.7 f 1.5) X J, respectively. Flow System. All gas flow were controlled and monitored by a multichannel mass-flow controller. Each of the control modules was precalibrated for the individual species or mixtures to be monitored. The gases used were purchased in high-pressure gas cylinders. Purities (supplied by the manufacturer) are listed in Table I. Gas Analyses. An NO, analyzer was used for direct and continuous nitric oxide and nitrogen dioxide determinations. This analyzer has selectable full-scale ranges from 0.25 to 25 ppm. Ita maximum sensitivity is 5 ppb on the 0.25 ppm range. It was frequently calibrated using 18 ppm NO in nitrogen as the span gas. Determinations of N 2 0 were also performed continuously by the NO, analyzer by quantitatively first converting the NzO to NOz. The latter oxidation was conducted photochemically by UV irradiation of the sample stream. The resulting O(lD) produced from the irradiated ozone in the sample stream converted NzO to NO which was further oxidized to NOz by excess ozone. This was checked by standardized calibration mixtures. Unreacted O3in the reactor effluent was determined by first mixing it with a known concentration of NO and measuring the resulting NOz with the NO, analyzer. In the NOz analysis by the NO, analyzer, the error resulting from the levels of ozone concentration found in the sample stream did not exceed 5% and generally was less than that. The fluid dynamics of the reactor are important in interpreting our trace gas productions. Our reactor design is essentially the same as that used by Donohoe et al. (1977). They observed that the corona discharge was concentrated in a small axial region, estimated to have a radial width of 0.015 m surrounding the central wire of the reactor. The physical characteristics of the reactor discharge are complex, and they were not analyzed in either of the two laboratories. Our observations confiim Donohoe et al.'s findings concerning the extent of the brush (corona) discharge, and we have tentatively assumed that only approximately 0.1 of the reactor volume was directly involved with the corona plasma chemistry. In order to reduce possible dependences of trace-gas yields on the location of the corona region in the reactor, we attempted to achieve fluid dynamic conditions which would bring about essentially complete radial mixing of the reactor gas, i.e., plug flow. As seen in the experimental results, and as estimated in the following theoretical arguments, we were apparently unsuccessful in achieving plug flow at all gas flow rates, but by studying the trace
1266 Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 Exhaust Sample NOx analyzer Mixing
Air
A$
l l
=
Air-flow Control
Module
Reactor Figure 3. Flow system for the analysis of NO and NO, in dry air after passing through the plasma reactor.
yields as a function of flow rates, we were able to establish those flow conditions which approached plug flow. In our reactor, the feed gas was admitted to the 2.3 X m (i.d.) reactor tube via an inlet tee containing an axially located nozzle of 1.27 X m (i.d.) through which was also threaded the 3.5 X m (0.d.) wire electrode. For a typically accepted Reynold’s number of 2 X lo3,it is calculated that turbulence would be expected to develop at the inlet nozzle for a velocity larger than 11 m/s, and in the reactor cylinder for a velocity exceeding approximately 0.6 m/s. Therefore, turbulence at the reactor inlet was calculated to occur for a gas flow rate larger than approximately 50 X lo4 m3/s. In the larger reactor tube, turbulence was not expected until the flow reached 250 X IO4 m3/s. Since there was an abrupt change in diameter between the inlet nozzle and the wider reactor tube, however, it is expected that turbulence could more easily develop before the critical value of 50 X lo4 m3/s was reached a t the inlet. For our normal flow rates of between 10 and 50 X lo4 m3/s, this is typically the behavior that was observed.
Results A. Nitric Oxide Production. Figure 3 shows the arrangement for measuring the NO plus NOz concentration in the reactor effluent. The 77 X lo4 m3 reactor volume was normally connected in series with a 200 X lo4 m3 mixing vessel located between the reactor and the NO, analyzer. The connections were made with polyethylene tubing, and the total volume of mixing vessel and connections between reactor and analyzer was 265 X lo4 m3. It was determined from the NO, analyzer that the effluent contained NOz but no NO exceeding 5 ppb, which was the limiting sensitivity of the NO, analyzer. At first glance, this result was surprising, since the theoretical simulation (Hillet al., 1984) showed that the concentration of NO in a single corona pulse was of the order of 105times larger than that of the NOz directly produced. It is clear, however, from both Donohoe et al.’s work (1977) and our simulation (Hill et al., 1984) that the absence of NO from a purely corona discharge in air is attributable to the quantitative conversion of NO into NO2 caused by the presence of excess O3which is also produced in a corona reactor. This conversion is considered later more closely when the observed NOz concentration curve is discussed. The theoretical simulation of trace gas production in air (Hill et al., 1984) showed that NO is primarily produced by the ‘D excited state of N atoms, according to the reaction NPD) + Oz NO + 0
-
K, = 6
m3/s (1) The N(2D) is produced with approximately 67% probability for each ion pair created in air by corona. Because X
air flow
reactor V=77mL
mixer X = 265 mL
NOx
analyzer
Ind. Eng. Chem. Res., Vol. 27, No. 7 , 1988 1267 Exhaust
The concentration of NOz leaving the reactor, CoNo2.,is therefore assumed as simply represented by the equation
Sample NOx
Analyzer Mixing Volume
where YNo (molecules of NO per pulse) is the production of NO (entirely converted to NOz) in the reactor, p , is the pulse rate in pulses per second, V is the volume of the reactor, and T is the residence time of an air sample passing through the reactor, i.e., T = V/v,where v (m3/s) is the air flow rate. In the mixing vessel, the concentration CNO2decays at a global rate given by k4CoO3CNo2, where Coo3is the ozone concentration leaving the reactor. Since the O3 concentration is relatively much larger than that of NOz, it can be assumed invariant. The concentration of NOz, emerging from the mixer of volume X , can therefore be obtained from
C N O=~ C 0 ~ oexp(-hCoo,t) 2
(6)
Since the residence time in the mixer is t = X / v eq 6 can alternatively be given by
The form of eq 7 is convenient for a comparison with observed CN02versus T as shown in Figure 4. The dotted curve in Figure 4 is merely the function T exp(T/r), where T has been chosen to have the value 3.3 s in order that the maxima of the theoretical and experimental curves coincide. The theoretical value of T (equal to V/k4Coo3X) is approximately 1.2 s. Although the finer details of the curvatures at the sides of the peaks are somewhat different from one another, the similarity of the curves indicates that reasonable agreement has been attained in accounting for the observed NOz concentrations in terms of reactions 3 and 4. The form of the CNOzvs T curve beyond a residence time of approximately 10 s is complicated by the chemistry of NOz and NO3 (and the combined products Nz05 and NzO,*) by the flow characteristics of the reactor and mixing vessel a t long residence time. Many of these problems were earlier addressed by Donohoe et al. (1977), but the problems of plasma operation and transport in the reactor were difficult and remained only partly resolved. Because of the limitations of our research resources, we did not pursue these problems further but have confined our analysis to only the short residence times shown in Figure 4. The NO yield can be derived from the observed rate of CNO2 increase in the initial part of Figure 4. This part of the curve corresponds to NO and NOz production in the reactor. The following equation is therefore used to evaluate YNO: YNO
= 2.69 x l o ' g v c N O / p
(8)
where CNO is assumed equal to CN02, the observed concentration of NOz at an air flow rate v (m3/s) and at a pulse rate p (E?). The numerical factor 2.69 X 1019converts CNO, (ppm) to concentrations of molecules by volume (m3) at STP. Thus, it is calculated from Figure 4,for CNOl equal to 0.65 ppm a t a flow rate of 38.5 X lo+ m3/s (corresponding to a residence time in the reactor of 2 s) and at a pulse rate of 820 pulses/s, that the yield YNo is approximately 8.2 X 10" molecules/pulse. This value of YNo,however, does not accurately represent the yield within the corona plasma, because of two
NO/N2
Air
-
I
Nitrogen Flow Control Module
Air-flow Control Module
Reactor
Figure 5. Flow system for the analysis of O3in dry air after passing through the plasma reactor.
effects which dilute YNo by a factor f. Dilution occurs because the corona plasma is a fraction only of the reactor volume and also because the plasma is actuated for only a fraction of time that the gas flows through the reactor. Thus f is given by reactor vol 1 X (9) = corona vol corona on-time/residence time Following Donohoe et al. (1977), we used a corona volume equal to 0.1 of the reactor volume, and as indicated in the plasma reactor section, we estimated that the corona ontime was equal to 0.108 f 0.024 of the residence time. On the basis of E equal to 5.7 X J/pulse, a value of YNO equal to 8.2 X lo1' NO molecules/pulse, and a dilution factor o f f equal to loz, the yield P is therefore PNO
= - N 1.4 X 10l6 molecules/J
E
(10)
Because of the large uncertainty in f, this value of P is estimated to be (1.4 f 0.7) X 10l6NO molecules/J. This yield is in close accord with the theoretical value of 2.1 x 10l6NO molecules/J which was the yield estimated from simulations of corona-excited chemistry of Bhetanabhotla et al. (1985). B. Ozone Production. In order to measure the concentration of ozone produced, a gas "titration" technique was used. As already stated, ozone generated in the reactor quickly oxidizes the relatively small concentration of NO to NOz in accordance with reaction 3. In the apparatus shown in Figure 5, a second stream of nitrogen containing a known concentration of NO (designated CN?,$ is added to the product gas a t the junction of the mixing volume and the connector tube leaving the reactor. For a particular product flow rate, the particular NO/Nz flow rate is adjusted until the concentration of NO issuing from the mixing vessel is a t the limit of detectability by the NO, analyzer (5 ppb). The O3 concentration, Coos,leaving the reactor is therefore calculated from the expression
coo3 =
(CNO/N2)(UNO/N,)
(11) Ualr
where vNOIN1 and vair are the volumetric flow rates of the NO/NOZand air mixtures entering into the mixing vessel, respectively. A concentration of 18 ppm NO in Nz was used as the titrant. The tube volume between plasma reactor and mixing point was 15 X lo4 m3, and the volume between the mixing point and NO, analyzer was 2.65 X m3. Allowing for the other possible sinks of O3 generated in the reactor effluent, such as that used to oxidize the directly produced NO to NOz, and including any other direct
1268 Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988
' C
i
'7 3 1
>-
p 0
10
30 40 50 Flow Rate (m3/sec) x l o 6 20
60
't
1
1
0
0
10
Figure 6. Yield of O3per pulse as a function of air flow rate in the reactor.
20
30
40
50
60
1
Flow Rate (m3/sec) x l o 6
Figure 7. Yield of N20per pulse as a function of air flow rate in the reactor.
production of NO2 in the reactor, the total measured concentration, Co3, should be given by '0,
=
C03titration
+
CNOeffluent
+ CNOzdirect
(12)
The CNO2direct term, however, was assumed to be zero, since theory indicated that direct NO2production was less than lo4 NO effluent. The stated concentrations of ozone were therefore given by the sum of the first two terms in eq 12. Yields (Yo,) of O3 molecules per pulse, calculated from observed concentrations (Co,, ppm) and the equation
Yo, = 2.69
X
1019Co,~/p
(13)
are given as a function of air flow rates in Figure 6. As an order of magnitude indication of the O3 production, a t a flow rate of 40-50 X lo4 m3/s, it is seen that the yield of O3per pulse is approximately 30 times the NO yield per pulse. From the theoretical studies already referred to (Hill et al., 1984; Bhetanabhotla et al., 1985), it is indicated that there is only one reaction in the corona plasma which significantly produces 03.This reaction is the following three-body reaction:
+ o2+ M
o(3~)
k14 = 5.5 x
- o3+
m6/s
M (14)
It is noteworthy that the destruction reactions of 03,such as the NO oxidation reaction 3, which is minor because of the relatively small amount of NO produced, and the NO2 destruction reaction 4, which is relatively slow and also minor, would not be expected to strongly affect the O3yield curve in Figure 6. Nevertheless, the O3 curve is observed to fall off rather strongly as the flow increases through the reactor from a rate of 10 X to 60 X lo4 m/s. It is noted here that the air flow rate had a small effect on the forms of the voltage and current pulse envelopes transmitted through the reactor. The average power and energy input values per pulse given in the plasma reactor section were for a flow rate of 10 X lo4 m3/s; however, the lower O3yields at the higher flow rates could have been associated with lower effective energy expenditures per pulse in the plasma. If a yield of 2.3 X 1013 O3 molecules per pulse a t a flow rate of 10 X lo4 m3/s is assumed to be associated with a pulse input energy of 5.7 X J and if a dilution factor of lo2 is assumed, then the O3 production efficiency calculated from the equivalent eq 10 is equal to 4 x 1017 molecules/J. This value which is subject to approximately a 50% probable error is approximately a factor of 8 larger
than the theoretical value (Bhetanabhotla et al., 1985) obtained for lightning simulation, but is is also subject to some uncertainty because of the wide variation by a factor of more than 2 in the variation of O3 yield with air flow rate observed in Figure 6. C. Nitrous Oxide Production. Modeling of N 2 0 production by lightning corona was performed for dry air by Hill et al. (1984) and later, for moist air, by Bhetanabhotla et al. (1985). Although it has not yet been unequivocally demonstrated by laboratory experiment that Nz(A3Z)is the source of N 2 0 production, certain phenomenological experiments point to N2(A3Z)as being the active generating agent of N 2 0 in irradiated and ionized air (Hill et al., 1984). Yields of N20, using the same plasma generator system as shown in Figure 3 but with a N20 to NO2 converter before the NO, analyzer, were reduced to the yields of N20 molecules per pulse as a function of air flow rate shown in Figure 7. From the highest flow rates used down to approximately 10-20 X lo4 m3/s, the N 2 0 yields are approximately constant. A similar dependence of N20yield on flow rate was also found by Donohoe et al. (1977). Below a flow rate of approximately 10 X lo6 m3/s, the N20 yield falls off very rapidly. The depletion of N 2 0 at very low flow rates is attributed to two factors: (i) a tendency for poor mixing of the gaseous products of the corona discharge in the immediate vicinity of the wire, and (ii) the occurrence of the following NzO depletion reaction which becomes more significant in a region of N 2 0 concentration buildup where the more intense corona discharge is located, i.e., N(2D) + N20 NO + N2
-
k15 = 2 x
m3/s (15) In Figure 7, the reasonably flat region of the curve at 6 X 10l2molecules per pulse, for an energy of 5.7 X J/pulse and a depletion factor (f) equal to lo2,corresponds to a production efficiency of 1017N 2 0 molecules/ J. This value is found to agree precisely with the Bhetanabhotla et al. value of 1017N20 molecules/J in moist air, but the effect of water vapor on the N 2 0 production process was theoretically found to be very minor. The observed value of 1017 molecules/ J is subject to the same probable error of 50% attributable to the possible inaccuracy of the f factor. Acknowledgment This experiment was supported in part by the National Science Foundation under Grant ATM-9-8100164 and in part by the Office of Naval Research under Contract
I n d . Eng. Chem. Res. 1988,27, 1269-1277
N00014-804-0293 to R.D.H. Contributions to the experiment in the early stages by William Savage are gratefully acknowledged. Finally, valuable interaction with Professor Frederick Shair at Caltech is very much appreciated. Nomenclature C = concentration, ppm E = energy per corona pulse, J f = dilution factor (dimensionless) p = pulse rate, s-l P = production yield, molecules/ J v = flow rate, m3/s Y = yield, molecules/pulse Registry No. NO,10102-43-9;NzO, 10024-97-2;Os, 1002815-6.
1269
Chameides, W. L. Geophys. Res. Lett. 1979, 6, 287. Donohoe, K.G.; Shair, F. H.; Wulf, 0. W. Ind. Eng. Chem. Fundam. 1977, 18, 208.
Hill, R. D. J. Geophys. Res. 1971, 76, 637. Hill, R. D. J. Geophys. Res. 1977, 82, 4967. Hill, R. D.;Rinker, R. G. J . Geophys. Res. 1981, 86, 3203. Hill, R.D.;Rinker, R. G.; Coucouvinos, A. J. Geophys. Res. 1984,89, 1411. Hill, R. D.; Rinker, R. G.; Wilson, H. D. J. Atm. Sci. 1980,37, 179. Iannuzzi, M. P.; Jeffries, J. B.; Kaufman, F. Chem. Phys. Lett. 1982, 87,570. Levine, J. S.; Shaw, E. F. Nature (London) 1983, 303, 312. Picone, J. M.; Boris, P. P.; Greig, J. R.; Raleigh, M., Fernsler, R. F. J. Atm. Sci. 1981, 38, 2056. Piper, L. G.; Caledonia, G. E. J . Chem. Phys. 1981, 74, 2888. Rahmim, I. M.S. Thesis, University of California, Santa Barbara, 1984. Swider, W. Geophys. Res. Lett. 1976,3, 335.
Literature Cited
Received for review May 19, 1986 Revised manuscript received July 29, 1987 Accepted February 29, 1988
Bhetanabhotla, M. N.; Crowell, B. A.; Coucouvinos, A.; Hill, R. D.; Rinker, R. G. Atm. Enuiron. 1985, 19, 1391.
Infinite Dilution Activity Coefficients Predicted by UNIFAC Group Contribution J. C.Bastos, M. E. S o a r e s , and A. G. Medina* Centro de Engenharia Quimica, Faculdade de Engenharia, Universidade do Porto, Rua dos Bragas, 4099 Porto Codex, Portugal
A UNIFAC parameter table specially suited for the prediction of infinite dilution activity coefficients (7") is proposed. T h e present table is exclusively based on experimental 7- data reported in the literature and correlates about 70% of the 11500 data points with a n average relative error of 20%. The 190 pairs of parameters of 40 different groups were estimated in order to reproduce the 7-data as accurately as possible. Infinite dilution activity coefficients play an important role in chemical technology, namely in qualitative and quantitative analysis of separation processes such as extractive and azeotropic distillation and liquid-liquid extraction. This justifies the considerable efforts dedicated to the development and improvement of experimental techniques and to the establishment of accurate correlation and prediction methods. A considerable amount of experimental information on ym is available in the literature and has been collected (Bastos e t al., 1984, 1985). The determination of the present UNIFAC parameter table is based on the experimental information contained in a data bank that was set up as a result of a joint project involving also the Universities of Dortmund and Trieste. Despite the existence of the vapor-liquid equilibrium (VLE) (Gmehling et al., 1982; Macedo et al., 1983) and liquid-liquid equilibrium (LLE) (Magnussen et al., 1981) parameter tables and the previous works of Zarkarian et al. (1979) and Alessi et al. (1982), who have studied the use of ymto obtain UNIFAC parameters, the determination of a new y mparameter table was encouraged by the improvement of accuracy and range of applicability of the UNIFAC method, as far as the calculation of ymvalues is concerned. Range of Applicability of the 7" P a r a m e t e r Table The present ymparameter table allows the calculation of 8300 data points of the data bank (72.5% of all data points) with a relative mean deviation of 20.2%. These 0888-5885/88/2627-1269$01.50/0
results mean a significant improvement in both range of applicability and accuracy of the UNIFAC model when compared with those obtained with the previous VLE and LLE parameter tables (Table I). The groups in the ymparameter table are fundamentally those defined in the previous VLE and LLE tables. With respect to the VLE table, the improvement on the range of applicability of the present y mtable is a result of the introduction of 68 new interaction parameters (out of a total of 190 determined) which reflect the distinct characteristics of the systems included in the VLE and y" data bases. The differences between the range of applicability of both VLE and y mtables are shown schematically in Figure 1. This figure indicates that the parameter table does not include parameters for dialkylamines (CNH), carboxylic acids (COOH), carbon disulfide (CS,), thiols (CH3SH),alkynes (CEC), and fluoro compounds (ACF and CFJ. On the other hand, it includes eight new groups: N-methylpyrrolidone (NMP), sulfolane, esters of benzoic and phthalic acids (ACCOO),sulfides (CH,S), dimethylacetamide (DMA), diethylene glycol (DEG), N-formylmorpholine (NFM), and triethylene glycol (TEG). Although most of the new groups are in fact molecules, their introduction can be justified by the importance of these compounds in chemical technology. Besides, the difinition of groups contemplating these compounds, and their derivatives or families, could not be performed with acceptable results due to the lack of experimental information. In this context the most promising situation would be the definition of three main groups for amides, alkyl0 1988 American Chemical Society
