Facilitated transport of ethene through Nafion membranes. Part I. Water swollen membranes

Facilitated transport of ethene through Nafion membranes. Part I. Water swollen membranes

Journal of Membrane Science, 85 (1993) 89-97 Elsevier Science Publishers B.V., Amsterdam 89 Facilitated transport of ethene through Nafion membranes...

763KB Sizes 6 Downloads 48 Views

Journal of Membrane Science, 85 (1993) 89-97 Elsevier Science Publishers B.V., Amsterdam


Facilitated transport of ethene through Nafion membranes. Part I. Water swollen membranes Odd I. Eriksen*, Elin Aksnes and Ivar M. Dahl SINTEF SI, P.O.Box 124, Blindern, 0314 Oslo3 (Norway) (Received July 27,1992; accepted in revised form June 23,1993)

Abstract Poly (pertluorosulfonic)ionomer membranes (Nafion N-l 17 ) were studied for the separation of ethene from ethane. Membranes in the sodium form were ion-exchanged in an aqueous 2 M solution of Ag+, and the flux of ethene was measured as a function of the partial pressure of ethene in the humidified feed gas. The experimental data were consistent with a carrier transport mechanism for a diffusion limited analytical model. The plot of the inverse facilitated ethene flux versus the inverse partial feed pressure was used to calculate the equilibrium constant for complexation of silver and ethene within the membrane. Effective diffusion coefficients for ethene were also calculated from the flux data. The facilitation factor for ethene with a humidified feed mixture of 50 mole% ethane and 50 mole% ethene at total atmospheric pressure was 170. The separation factor was 470. The permeability coefficient for ethene corresponded to 1000 barrer. Key words: facilitated transport; separation of ethene/ethane;

Introduction In a facilitated transport process [ 1,2] carriers reversibly react or coordinate with a solute which is transported through a membrane. Liquid membranes using Ag+ as carrier for olefins have been studied extensively by Steigelmann and Hughes [ 3,4]. Aqueous solutions of Ag+ , in particular AgNOB, were immobilized in porous supports. The supports are often porous cellulose acetate filters. A pilot plant study, using aqueous solutions of Ag+ immobilized in hollow fibres has also been published [ 51. Le Blanc et al. [6] were the first to demonstrate the facilitated transport of CO, and ethYTowhom correspondence should be addressed.


Nafion membranes

ene in an ion-exchange membrane (IEM). A cation-exchange membrane of sulfonated polyphenylene oxide with an exchange capacity of 2 meq/g and a thickness of 25 p was converted to the Ag+ form by soaking in aqueous AgN03. The ethene permeability at 25°C was 2.3 x lo-’ cm3-cm-1-sec-‘-cmHg-’ using ethene saturated with water at atmospheric pressure as the feed gas. The permeability of ethane saturated with water was only 0.8~ lo-’ cm3cm-’ -set-‘-cmHg_‘. From the data of the pure gases, this membrane should readily separate ethene from ethane. In a recent work it was shown that gas permeabilities of ethene and propene in dry Nafion membranes could be enhanced 2-3 fold by ion-exchange with Ag+ into the ionic do-

0 1993 Elsevier Science Publishers B.V. All rights reserved.


main of the membrane [ 71. The authors concluded that the silver ions work as carriers even in the dry membrane. The authors did not report the fluxes obtained, but selectivity ratios of ethene and propene to nitrogen were in the range of l-7depending on the temperature. A patent by Kraus assigned to Monsanto Company describes water-free Nafion membranes in which the silver-exchanged Nafion membrane was swollen in glycerine before testing with a dry feed [8]. The permeability of ethene was 7 barrers with a corresponding separation factor of 10 for a dry mixture of 50 mole% ethene and 50 mole% ethane at ambient conditions. Facilitated transport using silver-exchanged Nafion membranes has also been studied for separation of liquid phase olefins. Koval et al. reported the separation of 1-hexene and 1,5hexadiene [ 91 from saturated liquid hydrocarbons, and the separation of styrene from ethylbenzene [lo]. Facilitated transport of COZ in Nafion membranes has been extensively studied by Way et al. [ 111 and Noble et al. [ 121 using mono protonated ethene diamine as carrier. Noble et al. have also reported an analytical model for facilitated transport in which the influence of the mass transfer resistance and the diffusion coefficient of the solute complex can be determined if reaction equilibrium is obtained and the system is diffusion limited [ 131. To our knowledge there are no reports on the flux and selectivity of ethene through water swollen Nafion membranes with Ag+ as carrier, although this could be a very promising system for the separation of gaseous olefins from saturated hydrocarbons. This paper presents the results from using Ag+ exchanged Nafion-N117 membranes for the separation of a humidified feed of ethene and ethane. The purpose of this study was to characterize the performance of the system with respect to the flux of ethene versus the feed pressure and the

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97

degree of separation which could be obtained for a 1: 1 mixture of ethene and ethane as feed gas. We also wanted to see whether the ethene flux obtained at increasing partial feed pressure was consistent with the analytical model derived by Noble et al. [ 131. Experimental Membrane preparation NafionN117 membranes were supplied by Aldrich. These membranes are ion-exchange membranes of poly (perfluorosulfonic acid ) with a nominal thickness of 170 pm and an exchange capacity of 0.91 meq/g (equivalent weight = 1100). The membranes were received in the H+-form. The membranes were ion-exchanged with sodium by refhrxing 4 hr in 0.2 M aqueous NaOH solution and immersed in water for 24 hr. Excess water on the membranes was removed with a paper towel before the membranes were placed in the permeation cell. After diffusive flux measurements, the sodium membrane was immersed in aqueous 2 M AgBF4 for 48 hr and then in water for another 24 hr. The content of water in the membranes was obtained by the weight difference between dry and wet membranes. The wet membranes were weighed by immersing them in a beaker with water immediately after removing the membranes from the permeation cell. The weight of the membrane was obtained by the weight difference of the beaker with the water before and after immersing the membranes. The weight of the dry membranes was obtained after drying in vacuum at 120’ C for at least 48 hr. The concentration of Ag+ in the transport active membrane was estimated from the ion-exchange capacity of the polymer, assuming 100% ion exchange, and the water content of the membrane. Thickness measurements for the membranes with a precision micrometer were generally repeatable to + 2%. The thickness and

0.1. Eriksen et al. /J. Membrane Ski. 85 (1993) 89-97

water content of the sodium-exchanged Nafion membrane was in agreement with values previously reported by Way et al. [ 111. Flux measurements The permeation cell and the method used for measuring fluxes of mixed gases through membranes were based on previous literature [ 14,151. The gas permeability equipment used was based on the principle of flowing a feed gas on one side of the membrane, and allowing the permeated gas to be picked up by an inert carrier gas flowing on the other side of the membrane. The permeate gas mixture was then transported to a sample loop for GC analysis, controlled by an Apple II computer. The membrane was clamped and sealed between two Orings in a ring-shaped test cell made of stainless steel. Above and below the film, there were 2 mm thin cavities for feed and purge gases. The effective membrane area was 12.6 to.2 cm’. The small volume in which the carrier gas flowed past the membrane was kept low to minimize residence and mixing times. Each compartment of the cell was equipped with a metallic mesh support for the membrane and two Swagelock (l/8” ) connections for inlet and outlet of the feed and the purge gases. Helium (99.995 mole% ) was used as purge gas in all the tests. During testing and calibration, the purge, feed and custom-mixed calibration gases were bubbled through distilled water before entering the test cell. The relative humidity of the feed and the purge gases to the permeation cell was not measured, but due to the steady-state flux conditions, a relative humidity was assumed of both the purge and the feed gases close to 100%. A Carlo Erba gas chromatograph with a PLOT Fused Silica capillary column from Chrompack and a FI detec tor with a Milton Roy integrator was used for analysing the concentrations of ethane and ethene in the purge gas from the test cell. The


concentration of ethane in the purge gas was calculated from the calibration curve obtained from custom-mixed feed gases of 50, 100 and 150 ppm of ethane in helium. The concentration of ethene in the purge gas was calculated from the calibration curve obtained from custom-mixed feed gases of 1,5 and 20 mole% of ethene in helium. During calibration, the integration values obtained were reproducable to 2 0.2%. Constant carrier flow at a known rate was essential, since the flux was calculated from the permeate gas concentration in the carrier stream and the carrier gas flow rate. Accurate flow rate of the carrier gas was controlled by using Hi-Tee F 201C O-100 cm3(STP)-min-’ flow controllers calibrated for helium. A 2% error in flow rate was used for the flux calculations. The feed and purge gases were introduced to the test-cell at atmospheric pressure with a cocurrent flow. For each permeation test, the concentrations of ethane and ethene in mole% in the purge gas from the test cell were monitored by GC-analysis until constant values were obtained. During the flux measurements, the standard deviation for the integration areas obtained were below 2% for at least five measurements at steady-state conditions. For all membranes, the fraction of ethene in the purge gas was below 5% of the fraction of ethene in the feed gas. For calculation of the permeability coefficients, the driving pressure therefore was assumed equal to the partial pressure of ethene in the feed gas. During testing, the flow of helium through the membranes was checked by closing the feed inlet and connecting a soap-bubble meter to the feed outlet. No flow could be detected, and the flux of helium purge gas through the membranes was therefore assumed ~0.1 ml/min and was disregarded in the calculations. The helium purge flow rate in all tests was 10 cm3 (STP) -min-l. The flux Ji in cm3(STP)-cm’-set-’ of ethene or ethane through the total membrane area A


0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97

was calculated from the mole% (Xi) of component i in the purge gas at the purge flow rate @ to the test cell by eqn. (1). [email protected]/[(l-Xi)A]


The standard deviations for the calculated fluxes in cm3 (STP) -cm-2-set-’ were below 4%. All gases were supplied by Hydro Gas.

feed side was calculated from eqn. (2) in which pi is the partial pressure of ethene in atm, V, is the molar gas volume at 25’ C, and m is the partition coefficient or Henry’s law constant, i.e. the ratio of the molar concentration of ethene in the feed gas to the molar concentration of ethene in the membrane at the feed side.



Results and discussion The measured fluxes of ethene gave a Fickian type of diffusion through the Na-membrane as shown in Fig. 1. The linear regression correlation on the data is 0.99 with a slope of (117 + 8) X lo-l2 mo1e-cm-2-sec-’ atm-l for a 95% confidence limit. The calculated permeabilities of ethene and ethane were about 7 and 2 barrers respectively. The transport of a solute through Nafion is assumed to take place mainly through the ionic domains of the polymer in which the water is immobilized [ 111. For a solution-diffusion mechanism the separation factor for ethene and ethane should mainly be determined by the solubility ratio and not the diffisivity because of the minor difference in molecular size. The concentration of ethene in the membrane at the

2 E f







For ethene in water at 25°C a value of l/ m= 0.107 has been reported in the literature [ 161. A corresponding partition coefficient m for ethene of 9.35 was used in the calculations. The observed separation factor of about 3 for the Na-membrane agrees with the ratio of the reported solubilities of ethene and ethane in water [ 171 of 0.13 and 0.05 g/L respectively at room temperature. The effective diffusion coefficient for ethene was calculated from the slope of the line, the measured film thickness & and the fraction of water fN in the membrane according to Fick’s law as described in eqns. (3) and (4). A possible heterogeneity of the distributed water in the membranes was not taken into account. Hence the use of the term effective diffusion coefficient. It was assumed that the resistances in the


, 0


0,4 Ethene p&al

Fig. 1. Linear regression and measured ethene fluxes (lo-”




pressure in feed btm.)


through the Na-membrane.

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97


whole process of mass transfer of ethene and ethane between the gas phases and the membrane phases did not influence the rate and therefore could be disregarded in the calculations.

J&N =fNDACA DA = slope LN VAm/fN

(3) (4)

The facilitated transport of ethene through the membrane was modelled by assuming a reversible coordination of ethene to silver ions in the aqueous phase of the membrane [ 18,191. Ag+ (aq) + ethene (aq) = complex+ (aq)


If the facilitated transport of ethene is diffusion limited, and the coordination of ethene to Ag+ in the membrane is at equilibrium, the facilitated flux will be proportional to the concentration gradient of the silver-ethene complex in the membrane. We assume that the starting concentration of Ag+ is Cr, the concentration of ethene at the feed side of the membrane is CA, the concentration of the silver-ethene complex is X at the feed side and zero at the permeate side. The equilibrium constant K for eqn. (5) will be given as follows: K=X/(









Equation (8) states that the facilitated flux will be directly proportional to the concentration of the complex within the membrane; fF is the fraction of water and Lv is the thickness of the Ag-membrane. DAB is the effective diffusion coefficient for the facilitated transport of ethene. It follows that a plot of the inverse facilitated ethene flux versus the inverse partial pressure of ethene on the feed side of the membrane should give a linear correlation given by eqn. (9): l/J,=(l/cK’)(l/p)+l/c


c=fdm G/L,


Figure 2 shows the inverse facilitated ethene fluxes versus the inverse partial pressures of ethene for the Ag-membrane. The linear correlation on the data was 0.99. As the ethene partial pressure out of the test cell on the permeate side is 25% of the ethene partial pressure in the feed and the ethene partial pressure on the feed side decreases correspondingly, the error in the calculated ethene flux is 2-5% from what is expected for a zero permeate pressure of ethene and a constant ethene pressure on the feed side. The slope of the line in Fig. 2 is (3.05 + 0.03) x 10’ atm-set-cm2-mole-’ with intercept (5.2 + 0.29) x lo7 for 95% confidence limits. The equilibrium constant K’ for the silver-ethene coordination within the membrane was calculated directly from the slope and intercept of the line. A value of K’ = 1.7kO.l atm- ’ was obtained. The corresponding equilibrium constant K=390t23 M-’ with respect to the molar concentration of ethene in the aqueous phase was obtained from the partition coefficient m for ethene in water and the molar gas volume VA at 25 ‘C. The calculated K is significantly higher than the literature value of K = 98 M - ’ for an aqueous 1 M AgN03 bulk solution at 1 atm ethene pressure reported by Trueblood and Lucas [ 191. However, Featherstone and Sorrie [ 181 studied the solubility of ethene at 1 atm pressure and 20 ’ C in aqueous solutions of AgNOB and AgBF4 at several concentrations. They reported that the olefin to silver mole ratio was strongly dependent on the silver ion concentration and the nature of the counterion. At a silver ion concentration between 1.0 Nand 7.5 N, the ethene to silver mole ratio varied between 0.18 and 0.84. From these data, the range of K’ values is 0.2-5.2 atm-‘. It should be noted also that the aqueous environment in Nafion membranes is quite special. It is estimated [ 201 from solvent absorption studies that 6 and 14 water molecules are as-

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97

94 70 ~ 60 -50 -40 ~? 30 --

OJ 0






Fig. 2. Linear regression and measured inverse facilitated ethene fluxes ( lo7 cm*-set-mole-‘)

sociated with each ionic site in K+- and Na+exchanged swollen Nafion membranes, respectively, and a recent work reports on the silver ion coordination in Nation and other gel type membranes [ 211. The aqueous environment in these membranes was found to be sufficiently different from bulk water that it caused a change of the coordination of the Ag+ ions. The authors conclude that the coordination properties of carrier species in facilitated transport membranes can be affected by the membrane matrix, and that the strength of the interaction, and thus the coordination behavior, could also be concentration dependent. In view of the cited reports, we find that the calculated K within the water-swollen silver-exchanged Nafion membrane is reasonable. With the very high facilitation factors observed for ethene in the Ag-membrane, the flux is totally dominated by the facilitated transport, and the equilibrium constant for the silver ion-ethene coordination within the membrane becomes very important for predicting the facilitated flux. Figure 3 shows the total ethene fluxes obtained through the Ag-membrane and the fluxes predicted from the regression data in Fig. 2. The parameter values derived from eqns. (9) and (10)are summarized in Table 1.

through the Ag-membrane.

The analytical model using the value of K’ in Table 1 predicts that about 60% of the silver carriers are converted to a complex on the feed side of the membrane at a partial ethene pressure of 1 atm. At 0.02 atm partial pressure of ethene, only about 2% of the silver ions are converted to a silver-ethene complex. Disregarding the concentration of silver-ethene complexes on the permeate side of the membrane for a permeate concentration of ethene < 2 mole% is reasonable and will give a systematic error in the calculations < 5%. The separation of ethene from ethane with a silver ion-exchanged Nafion membrane was studied by using a custom mixture of 50 mole% ethane and 50 mole% ethene as feed gas at atmospheric pressure. The separation factor, defined as the ratio of ethene permeability to the ethane permeability, was 470. A permeability coefficient for ethene P= 1.0x 10m7cm3-cm-cm-2-sec-’ cmHg-’ was calculated from the measured average film thickness and the driving pressure. This corresponds to a permeability of 1000 barrer. This is a very high permeability for ethene and almost comparable to the ethene permeability of 2300 barrers reported by Le Blanc et al. [6], especially taken into account the lower ion-ex-

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97


14.00 T






p&al pressure infeed(atm.)


Fig. 3. Calculated (from data in Table 1) and measured ethene fluxes (10-O mole-cm-2-set-‘)

TABLE 1 The parameter values obtained for predicting the flux of ethene through silver ion-exchanged Nafion membrane Symbol




3.3 x 10-6 2.0x 10-7 1.7 0.01 0.15 0.18 0.02 0.02

cm2-see-1 cm’-set-* atm-’ mole-cm-3

DAB K’ :

fF 2

cm cm

change capacity of the Nafion membranes compared with the sulfonated polyphenylene oxide membranes. The facilitation factor, defined as the ratio of the ethene flux through the Ag-membrane to the ethene flux through the Na-membrane, was 170. For pure humidified ethene as feed gas at a pressure of 1 atm, a permeability coefficient of P= 7.9 x lo-* cm3cm-cm-2-set-’ -cmHg-’ or 790 barrer was measured. Conclusion

Nafion membranes which are ion-exchanged in an aqueous 2 M solution of Ag+ are highly

through the Ag-membrane.

selective for the separation of ethene from ethane when a humidified feed is applied. It was shown that the inverse facilitated ethene flux versus the inverse partial pressure of ethene for the silver-exchanged Nafion membrane gave a very good linear correlation. This supports the assumption that the ethene flux through the membrane is diffusion-limited and that the coordination of ethene to the silver ions in the membrane is at equilibrium. Thus the assumption of no mass transfer limitation for the ethene flux through the membrane is justified. The effective diffusion coefficient DAB was estimated from the slope and intercept of the linear regression, the measured film thickness and water fraction and the calculated carrier concentration assuming 100% ion-exchange with silver ions. The equilibrium constant K’ for the silver-ethene coordination within the membrane was calculated directly from the plot of measured ethene fluxes versus ethene partial feed pressures. The values obtained for the equilibrium constant and the effective diffusion coefficient for the facilitated transport of ethene are reasonable in view of previous published results in the literature for facilitated transport membranes. It is however important to note that from the measured flux data and

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97


the calculated equilibrium constant K’ the silver ions within the membrane will be saturated at much higher ethene pressure than was studied in this work. Acknowledgement We thank Phillips Petroleum Company Norway for financing this work and the permission to publish the results. List of symbols pi DA

DAB @ xi




r, VA




partial pressure of component i (atm) effective diffusion coefficient for ethene (cm2-sec-1, Fickian diffusion) effective diffusion coefficient for ethene ( cm2-sec- ‘, Facilitated diffusion) purge gas flow rate to the test cell [cm3(STP)-min-‘1 content of component i in purge gas from test cell (mole% ) flux of component i [ cm3 (STP ) -cmw2set-l] flux of ethene through Na-membrane (mole-cm-2-set-1) facilitated flux of ethene through Agmembrane (mole-cm-2-set-1) membrane area (cm”) film thickness of Na-membrane (cm) film thickness of Ag-membrane (cm) molar gas volume at 25 ‘C (24467 cm3atm-mole-’ ) partition coefficient (concentration in gas phase/concentration in liquid phase ) concentration of ethene in the membrane at the feed side (male-cm-3) carrier concentration (mole-cm-3 ) concentration of silver-ethene complex in Ag-membrane (male-cm-3) equilibrium constant, silver-ethene coordination ( atm- ’ )

K fN



equilibrium constant, silver-ethene coordination (M-l) (K=K’mVA) fraction of water in Na-membrane fraction of water in Ag-membrane [ (wet weight - dry weight) / dry weight] permeability coefficient [ cm3 (STP) cm-cm-2-sec-‘-cmHg]

References 1

R.D. Noble and J.D. Way, Liquid Membranes, Theory

and Applications, ACS Symp. Ser. No. 347, American 2






8 9




Chemical Society, Washington, DC, 1987. R.D. Noble, C.A. Koval and J.J. Pellegrino, Facilitated transport membrane systems, Chem. Eng. Prog., (1989) 58-70. E.F. Steigehnenn and R.D. Hughes, Process for separation of unsaturated hydrocarbons, US Patent 3,758,603 (1973 ). E.F. Steigehuann and R.D. Hughes, Semi-permeable membrane compositions based on blends of polyamides and polyvinyl alcohols, US Patent 3,980,605 (1976). R.D. Hughes, J.A. Mahoney and E.F. Steigelmann, Olefin separation by facilitated transport membranes, in: N.N. Li and J.M. Caol (Eds.), Rec. Dev. Sep. Sci., Vol. 9, Chapter 9, CRC Press, Boca Raton, FL, 1986. O.H. Le Blanc, Jr., W.J. Ward, S.L. Matson and S.G. Kimura, Facilitated transport in ion-exchange membranes, J. Membrane Sci., 6 (1980) 339-343. T. Kanno, S. Fujita and M. Kobayashi, Enhancement of the permeabilities of ethylene and propylene gases in the Nafion Ag composite, J. Chem. Sot., Chem. Commun., (1989) 1854. M.A. Kraus, Water-free hydrocarbon separation membrane and process, US Patent 4,614,524 ( 1986 ) . CA. Koval and T. Spontarelli, Condensed phase facilitated transport of olefins through an ion exchange membrane, J. Am. Chem. Sot., 110 (1988) 293-295. C.A. Koval, T. Spontarelli and R.D. Noble, Styrenel ethylbenzene separation using facilitated transport through pertluorosulfonate ionomer membranes, Ind. Eng. Chem. Res., 28 (1989) 1020-1024. J.D. Way, R.D. Noble, D.L. Reed and G.M. Ginley, Facilitated transport of CO, in ion exchange membranes, AIChE J., 33 (1987) 480-487. R.D. Noble, J.J. Pellegrino, E. Grosgogeat, D. Sperry and J.D. Way, CO, separation using facilitated transport ion exchange membranes, Sep. Sci. Technol., 23 (1988) 1595-1609.

0.1. Eriksen et al. /J. Membrane Sci. 85 (1993) 89-97 13




R.D. Noble, J.D. Way and L.A. Powers, Effect of external mass-transfer resistance on facilitated transport, Ind. Eng. Chem. Fundam., 25 (1986)450-452. D.G. Pye, H.H. Hoehn and M. Panar, Measurement of gas permeability of polymers. II. Apparatus for determination of permeabilities of mixed gases and vapors, J. Appl. Polym. Sci., 20 (1976) 287-301. L. Zanderighi, B. Marcandalli, M.P. Faedda, I.R. Bellobono and M. Riva, The transport of gases in polymeric membranes: A simple cell for the determination of diffusion and permeability coefficients, Gaxz. Chim. Ital., 109 (1979) 505-509. S.P. Wasik and W. Tsang, Gas chromatographic determination of partition coefficients of some unsaturated hydrocarbons and their deuterated isomers in aqueous silver nitrate solutions, J. Phys. Chem., 74 (1970) 2970-2976.

97 17 18 19



G.H. Aylward and T.J.V. Findlay, SI Chemical Data, 2nd edn., John Wiley & Sons, New York, NY, 1974. W. Featherstone and J.S. Sorrie, Silver-hydrocarbon complexes, J. Chem. Sot., (1964) 5235. K.N. Trueblood and H.J. Lucas, Coordination of silver ion with unsaturated compounds. V. Ethylene and propene, J. Am. Chem. Sot., 74 (1952) 13381339. T.D. Gierke and W.Y. Hsu, The cluster-network model of ion clustering in pertluorosulfonated membranes, in: Perfluorinated Ionomer Membranes, ACS Symp. Ser. No. 180, American Chemical Society, Washington, DC, 1982, pp. 283-306. M.R. Antonio and D.T. Tsou, Silver ion coordination in membranes for facilitated transport, Ind. Eng. Chem. Res., 32 (1993) 273-278.