摘要:

J.C.S. Faraday I, 1980,76,2448-2456Distribution of Cobalt Ions among Octahedral and TetrahedralSites in C0Ga,Al2-~0~ Spinel Solid SolutionsBY PIERO PORTA* AND ANNA ANICHINICentro di Studio su Struttura ed Attivita Catalitica di Sistemi di Ossidi del C.N.R.,Istituto di Chimica Generale ed Inorganica, Universiti di Roma, Roma, ItalyReceived 6th August, 1979The cation distribution in several CoGgA1,_,O4 solid solutions with x ranging from 0.00 to 2.00has been studied by magnetic susceptibility, reflectance spectroscopy, lattice parameter variation andanalysis of some reflections whose intensities are particularly sensitive to variation in the cationpositions.The results show that there is a small but definite change towards a random cation distributionwhen the preparation temperature is increased from 873 to 1473 K.The X-ray and magnetic results show, moreover, that the fraction of Co2+ ions in octahedralsites varies with gallium and aluminium contents for the whole temperature series ; this variation is,however, not linear.For samples quenched from 873 K, CoA1,04 is 16 % inverse ; with increasingx , cobalt octahedral occupation first decreases up to x w 0.25 and then continuously increasesreaching an inversion of 78 % for the pure CoGa204 spinel.The influence of x on cation distribution is explained in terms of anion polarization and/orcrystal field stabilization energy effects.A large number of compounds having the general formula XY,04 are known tocrystallise in the spinel structure which may simultaneously accommodate metal ions,X and Y, among the octahedral, Oh, and tetrahedral, Td, sites available in the close-packed oxygen framework.Studies of cation distribution in spinels are of considerable interest in solid statechemistry because they may allow investigation of the relative stabilities of metal ionsin Oh and Td coordination and better understanding of the correlations betweenstructure and properties such as colour, diffusivity, magnetic behaviour, conductivity,catalytic activity, etc., which are well known to be quite dependent on the relative Ohand Td occupancy by transition metal ions.This paper is one of a series which describes an investigation of the relation betweenstructure and variables such as temperature and composition of spinel solidsolutions.1-4 In this work the cation distribution in CoGaxA12-x04 solid solutionshas been studied by magnetic susceptibility, reflectance spectroscopy, lattice parametervariation and, mainly, by analysis of some X-ray reflections whose intensities areparticularly sensitive to variation in cation positions. The present system is ofinterest because the end members of the solid solution series, CoAl,04 and CoGa,O,,have previously been reported close to " normal " and " inverse " spinels, respec-tively ;5-6 by varying x one might thus expect, other than variation of cation distribu-tion with the temperature of preparation, gradual and substantial changes in thestructure and in the magnetic properties. We describe the combined use of the fourtechniques to evaluate the cobalt ion distribution and its dependence on temperatureand composition.244P .PORTA A N D A . ANICHINI 2449EXPERIMENTALPREPARATIONPure CoA1204, pure CoGa204 and 7 C O G ~ , A ~ ~ - ~ O ~ solid solutions with x = 0.25,0.50, 0.75, 1.00, 2.25, 1.50 and 1.75 were prepared by soaking A1203 and Ga203 with cobaltnitrate in stoichionietric quantities ; the soaked mass for each composition was dried at393 K and ground, heated in air at 873 K for 2 h in order to decompose the nitrates, carefullyre-ground and pressed into pellets (at a pressure of w 8000 kg cm-2). The compoundswere sintered in air at 1473 K for 100 h ; then for each composition one batch of pellets wasquenched in water from 1473 K, a second batch was cooled to 1073 K and equilibrated atthis temperature for 50 h before quenching in water and a third batch was cooled to 873 Kfor 50 h and then quenched in water. This procedure enabled the preparation of specimensof each of the 9 compositions at three different temperatures.The colour of the samplesranged from intense blue to deep green with increasing gallium content.X-RAY INVESTIGATIONS AND CALCULATION OF INTENSITIESIron-filtered Co Kg radiation was used to investigate all the compounds, both for thelattice parameters and for the intensity measurements. For all specimens X-ray diffractionpatterns showed no lines other than those belonging to the cubic spinel structure. For allsamples the reflections were very sharp.The details of the experimental methods used for precise determination of lattice para-meter and careful analysis of some X-ray reflections whose intensities are particularly sensitiveto variation in cation positions are reported el~ewhere.~ The method used for the calcula-tion of the intensities has already been described in previous papers 39 and is an extensionto solid solutions of the method applied by Bertaut to pure spinels.'.By considering that the cation distribution in the present CoGaXAl2-,O4 solid solutionsis characterized by the formula :where a, 18 and y are the parameters describing the fraction of Co2+, Ga3+ and A13+ ions onTd sites, respectively, the intensities of the 220,400 and 422 reflections were computed in 0.1intervals of both p and y, with the following data also taken into consideration : (i) the xmolar composition of the spinel ; (ii) the value of u, the oxygen parameter, which was takenas 0.387 for all solid solutions ; 5 7 (iii) the scattering factors for Co2+, Ga3+, A13+ and 02-,the real part of anomalous dispersion for cobalt and the Lorentz-polarization correction(International Crystallographic Tables).The 400/220 and 400/422 calculated reflection intensity ratios were displayed for eachsolid solution by constructing a family of curves of intensity ratio against fi with y alsovarying ; the experimental intensity ratios were then compared with the calculated valuesand the cation distribution for each spinel was determined.Co,GaaxAl,(2-x)[Co(l-a)Ga(l -p)xAl<~-y) (2-x)104 (1)MAGNETIC SUSCEPTIBILITYThe Gouy method over the temperature range 100-295 K was used.Correction wasmade for the diamagnetism of the sample. A check was made that the susceptibilities wereindependent of magnetic field strength. The values of the Curie constant C and hence ofthe magnetic moment ,u were taken for all specimens from the slopes of 1 /xat against T plots ;the values of the Weiss constant 8 were taken from the intercepts of the l/xat lines on theT-axi s .RELECTANCE SPECTRAReflectance spectra in the range 350-2500 nm (28 500-4000 cm-') were recorded on aBeckmann DK-1 spectrophotometer with a standard reflectance attachment, using MgO asreference.1-72450 Co" I N CoGa,Al,-,O, SPINEL SOLID SOLUTIONSRESULTSLATTICE PARAMETERSThe results for the lattice parameter Q of the cubic cell, reported in table 1 andin fig.1, show that : (i) there is an increase in lattice parameter with increasinggallium content; (ii) the variation of a with gallium content is not linear and anegative deviation from both Vegard's law and Zen's relationship is observed ; (iii) thevalues of a for specimens with the same x vary slightly with firing temperature and itis noted that for small gallium contents a decreases with increasing firing temperaturewhereas for high gallium concentrations the lattice parameter increases with quenchingtemperature.TABLE 1 .-CoGa,Al,~,O, SPECIMENS WITH GALLIUM CONTENTS x, LATTICE PARAMETERS a,WEISS CONSTANTS 8, CURIE CONSTANTS c AND MAGNETIC MOMENTS p873 K 1073 K~~ ~1473 KX - 0 C p1B.M.- 0 C p1B.M. - e c p / ~ . ~ .0.000.250.500.751 .oo1.251.501.752.008.10508.12698.14478.17748.19818.22978.25758.29928.3250105105908060605060702.71 4.662.67 4.622.76 4.702.83 4.762.88 4.802.97 4.883.08 4.973.08 4.973.10 4.988.10588.12648.14448.17648.19858.23278.25838.30098.3243110 2.83 4.7685 2.69 4.6480 2.75 4.6980 2.86 4.7880 2.96 4.8770 3.01 4.9160 3.08 4.9760 3.08 4.9770 3.08 4.978.10518.12618.14368.17588.20148.23858.26018.30088.3268110 2.75 4.69100 2.71 4.6680 2.74 4.6880 2.88 4.8070 2.96 4.8765 3.01 4.9165 3.01 4.9170 3.04 4.9370 3.06 4.95X-RAY DIFFRACTION INTENSITIESFrom the comparison of the experimental 400/220 and 400/422 intensity ratiosand those calculated for various cation distribution, the best values of /? and y wereobtained for each specimen.where [Px+y(2-x)] = 21 is the total amount of both Ga3+ and A13+ ions in Tdsites, the values of a, the fraction of Co2+ ions on T,, were also derived.Table 2reports the mean values of Co2+, Ga3+ and A13+ octahedral occupations. The valuesof the fraction of Co2+ ions in Oh sites are also shown in fig. 2.The main conclusions which can be drawn are as follows : (i) for a given con-centration in the solid solution, the octahedral cobalt increases with increasing firingtemperature from x = 0.00 to x M 1.25, whereas the inverse trend occurs betweenx M 1.25 and x = 2.00; (ii) for a given temperature the Co2+ cation distributionvaries with x ; the octahedral cobalt firstly decreases with increasing x (up tox M 0.25) and then there is a continuous increase of Co2+ octahedral occupation.These results imply that there is a minimum Oh cobalt occupation at x x 0.25.Since the following relation is valid :a = 1-21 = l-[Px+y(2-x)] (2)MAGNETIC SUSCEPTIBILITYMagnetic measurements can also be used to distinguish between Oh and T dcobalt(11).9, O For high-spin octahedral CO" the ground level (4T1,) is orbitallydegenerate and there is an appreciable orbital contribution to the magnetic moment ;observed p are, for Co2+ in Oh coordination, in the range 4.7-5.3 B.M., but most oP . PORTA AND A . ANICHINI 245 18.107 I I I I0 0.5 1.0 1.5 2.0XFIG.1 .-Lattice parameters a/A plotted against gallium content x in CoGa,A12-,04 solid solutions :(A) 873, (0) 1073 and (0) 1473 K ; (- - -) line corresponding to Vegard’s law.TABLE 2.-cATION DISTRIBUTION IN CoGaxA12-x04 SOLID SOLUTIONS WITH GALLIUM CONTENTx AND VALUES OF Co2+, Ga3+ AND A13+ OCTAHEDRAL OCCUPATION0.00 0.16 - 1.84 0.18 - 1.82 0.23 - 1.770.25 0.14 0.24 1.62 0.15 0.20 1.65 0.20 0.18 1.620.50 0.23 0.28 1.49 0.22 0.31 1.47 0.26 0.28 1.460.75 0.34 0.41 1.25 0.36 0.41 1.23 0.38 0.41 1.211.00 0.46 0.54 1.00 0.49 0.53 0.98 0.50 0.54 0.961.25 0.56 0.72 0.72 0.59 0.68 0.73 0.61 0.68 0.711.50 0.70 0.85 0.45 0.67 0.85 0.48 0.65 0.90 0.451.75 0.77 1.01 0.22 0.72 1.05 0.23 0.70 1.07 0.232.00 0.78 1.22 - 0.74 1.26 - 0.72 1.28 2452 col* IN CoGa,Al,-,O, SPINEL SOLID SOLUTIONSthem are % 5.1 B.M.ll For Td cobalt(I1) the ground term is , A 2 ; spin-orbitcoupling thus causes some mixing-in of the ,TI and ,T2 terms to the ground term 4A2and an effective magnetic moment peff.= pss0. (1 -4L/lODq), where il is the spin-orbit coupling constant, Dq is the crystal field strength and p-ls.o. is the spin-only valueof the magnetic moment (for Co" in both 0, and Td coordination pse0. = 3.9 B.M.).12Observed moments for Co" in tetrahedral configuration are in the range 4.2-4.7 B.M.I10 0.5 1.0 1.5 2.0XFXG. 2.---Fraction of Co2+ in octahedral sites plotted against gallium content x in CoGa,A1,-,04solid solutions : (A) 873, (0) 1073 and (0) 1473 K ; (- - .) and (- - -) lines passing through thepoints 873 and 1473 K, respectively.The results of the magnetic measurements are presented in table 1 ; the values ofthe magnetic moments p are also shown in fig.3. The complete set of results showsthat the magnetic moments are all in the anticipated range for a cobalt distributionintermediate between Oh and Td sites. A slight decrease in the magnetic moment forsmall additions of Ga3+ ions to cobalt aluminate and then a continuous increase inp, up to the value of = 5.0 B.M., are observed for each series at constant quenchingtemperature. Small differences in p for specimens at equal x but treated at differenttemperatures are also observed.From the intercepts with the T-axis of the linear plots of 1 /xat against T, the Weissconstants 0, of the Curie-Weiss law xat = C/(T+8), were derived and found to be inthe range - 110 to - 50°C, as reported in table 1 .Therc is a continuous decrease in8 up to x =5: 1.50 and then an increase from x = 1.50 to pure cobalt gallateP . PORTA AND A . ANICHINI4.52453l$(Co)tet ------- -I I -p (CO) oc t 5.1 ,-.--.-.-.-.-.-.-.-.-.-..- I-XFIG. 3.-Magnetic moments p/B.M. plotted against gallium content x in CoGa,A1,_,O4 solidsolutions : (A) 873, (0) 1073 and (n) 1473 K ; (- - -) and (- - - .) lines passing through the points873 and 1473 K, respectively ; the expected values of p for CoII in octahedral and tetrahedral co-ordination are also shown.REFLECTANCE SPECTRAReflectance spectra are another tool for detecting octahedrally and tetrahedrallycoordinated cobalt ions in solids, although when both coordinations are present, asin our case, the strongest intensities of Td bands with respect to those of Oh bandsdo not help much on the quantitative determination of the concentration of bothspecies.Only an indication of the presence of these species may be drawn from thereflectance spectra in the present case.The Td bands dominate the spectra of all our samples. We observed the occur-rence of the following bands: (i) a broad band centred around 4200-4650cm-l,(ii) a group of 3 bands at 6000-7400 cm-l, (iii) a shoulder at x 9200 cm-l, (iv) 3bands at 15 3000-18 000 cm-I and (v) a group of 3 bands in the region 20 300-24 400 cm-l.Our spectra accord very closely with those reported in the 1iterat~re.l~. l4 Band(i) is an envelope of the unresolved bands corresponding to the Td4A2(F) -+ 4T2(F)transition ; bands of group (ii) are associated with the Td4A,(F) -+ "T,(F)transition ;bands (iv) are the Td4A2(F) -+ 4T1(P) transitions and the group (v) correspond to theTd spin-forbidden 4A2(F) -+ 2T(G) transitions. The shoulder at 9200 cm-' can beattributed to the octahedral 4T1,(F) -+ 4T2,(F) transition.All bands moved to lower energies with increasing gallium content.Inspectionof the relative intensities of T d bands with respect to the Oh one (shoulder at 9200 cm-l)revealed that the shoulder becomes more and more pronounced with increasinggallium content.No significant variation of the relative intensities of Td and O h bands has beenobserved for samples at equal x prepared at different temperatures2454 CO" IN CoGa,Al,-,O, SPINEL SOLID SOLUTIONSDISCUSSIONIn the spinel structure the close-packed array of negative ions can accommodatecations either in Td or in O h interstices. The cation distribution in spinels is deter-mined by various energy terms, such as the Madelung energy, the Born repulsiveenergy, the electrostatic ordering energy, anion polarization and the individualcrystal field stabilization energy (c.f.s.e.).For spinels containing divalent and tri-valent cations, in the absence of site-preference energy, the " normal " distribution,i.e., that structure in which all trivalent cations are in the Oh sites and all divalentcations in the Td interstices, is favoured by the Madelung energy and anion polar-ization ; when the octahedral-c.f.s.e.of divalent cations makes a significant con-tribution to the total lattice energy or when trivalent cations exhibit a strong preferencefor Td coordination, the " inverse " distribution is favoured. C.f.s.e., when high,has thus been used to predict the cation distribution in binary spinels and the observeddistributions of metal cations between Oh and T d sites in spinels have so far beenregarded as signifying the influence of this energy l6 It has, however, beenargued that any given structure cannot be predicted by the c.f.s.e. alone, even whenits value is non-zero, and that a quantitative calculation of the change in lattice energyon inversion, mainly in the Madelung term, must be made as the most important testto verify the occurrence of any observed structure.17In the present CoGa,Al,-,O, system the CoAI,O, and CoGa,O, end membershave been observed close to normal and inverse spinels, respectively ; 5 9 the occur-rence of normality in CoAl,O, is easily understood since all the energy terms tend tofavour this structure and no pronounced difference in site-preference energy ispresent between Co2+ and A13+ ions.On the other hand, since Ga3+ and Co2+ ionsshow Td and Oh site-preference energies, respectively, it may easily be realized whyCoGa20, has experimentally been found a nearly inverse spinel. CoGa,Al,-,O,is therefore a very useful system for investigating the variation, if any, of the cationdistribution with thermal treatment as well as with composition.A definite trend towards more random cobalt distribution with increasingquenching temperature has firstly been evidenced from our investigation. Consider-ing that the " random " cation distribution in a 2-3 spinel corresponds to two-thirdsof divalent cations in Oh sites, i.e., [Co],, = 0.67, analysis of the X-ray intensitymethod (table 2 and fig.2) gives an explanation of most of the observed results. Infact, from x = 0.00 up to x = 1.25 all spinels have been found more normal thaninverse, [Co],, < 0.67, whereas above x = 1.25 the spinels are more inverse,[Co],, > 0.67 ; the increase of quenching temperature then favours, as expected,the randomization process.The same effect is observed from magnetic measure-ments, as shown in table 1 and in fig. 3. Reflectance spectra and lattice parametersdo not reveal any detectable variation of cation distribution with thermal treatment,probably due to the smallness of the effect.Let us now consider the relation between cation distribution and x . Thereflectance spectra show a shift of all the observed bands towards lower energies withincreasing gallium content, in agreement with the expansion of the unit cell volume,i.e., with increasing anion-cation distances ; moreover, the shoulder observed at9200 cm-l, attributable to the octahedral Co" ,TIg(F) 3 ,T2JF) transition, becomesmore and more pronounced (with respect to the relative intensities of T d bands)with increasing gallium content, thus indicating an increase of octahedral cobaltoccupation.The expansion of the lattice on going from CoA1204 to CoGa204 (table 1 andfig.1) is explained in terms of substitution of smaller A13+ ions (Goldschmidt ioniP. PORTA AND A . ANICHINI 2455radius = 0.57A) with larger Ga3+ ions (Goldschmidt ionic radius = 0.62A). Ithas already been said that the lattice parameter is also affected by the cation distribu-tion (inversion is well-known to decrease its value)12 and we think that the non-linearity in the increase of a with x is an indication of the dependence of cationdistribution on composition.The influence of composition on cobalt distribution may be derived from themagnetic and X-ray intensity results which both show, for all series of compounds, afirst slight decrease in cobalt octahedral occupation and then a continuous increaseof [CO]~,, by progressive addition of gallium.We suggest that the presence of sucha minimum in [CO]O, at a given value of x (= 0.25) is brought about in this case bypolarization effects. The trend may be explained as follows : on going from pureCoA1204 to gallium spinels, the Ga3+ ions, due to their preference for Td coordination,tend to substitute A13+ ions in the Td sites. A lower positive charge density thusresults in the Td sites, since the ionic radii of tetrahedral Ga3+ and A13+ ions are0.47 and 0.39 A, respectively.' As a consequence, oxygens are more polarisedtowards the octahedral sites favouring a major tendency for trivalent ions (namelyA13+) to occupy the Oh sites and for divalent ions (Co2+) to occupy the T d sites.Thisoccurs for relatively small gallium content. With increasing gallium content, nomore A13+ ions are available to change their coordination from tetrahedral to octa-hedral (see table 2) and the Ga3+ ions are forced to occupy the Td sites with the spinelsbecoming more and more inverse.A variation of cation distribution with composition has also been observed forother spinel solid solutions studied by us, such as Ni,Mgl-,A1204,1 CU,M~,-,A~,O,,~COG~,R~,-,O,,~ CoxMgl-,A1204 and Ni,Znl-,A1204.4It should also be pointed out that an effect similar to that observed by us, i.e., adiscontinuous dependence of cation distribution on composition, has indeed beenfound by Delorme in the CuA1,Fe2-,04 system l 9 and by Bracconi et al.in theCo2+Cr: +Coi ZxO4 spinel solid solutions.20 We believe that the results found bythese authors are covered by our explanation.Finally, the negative values of the Weiss constant 8 (table 1) obtained from theplots of 1 /xat against T confirm the predominance of antiferromagnetic interactionsfor all the samples. A decrease in 0 is observed up to z 67 % of [ C O ] ~ , , with asubsequent increase of 8. The variation in 6 values with [Co],, has also beenevidenced in a more detailed magnetic investigation (in the range 4.2-300 K) per-formed by other authors 21 on the CoGaxA12-,04 series prepared by us at 1073 Kand it has been suggested that the variation is indicative of the presence of anti-ferromagnetic and/or ferromagnetic interactions.We think that spin-orbit coupling,which modifies the values of p and also contributes to 8, should also be considered,especially when varying amounts of Co2+ ions in Td coordination are present in thesolid solutions.We thank Mr. G. Minelli for technical assistance and Mr. M. Inversi for thedrawings.P. Porta, F. S. Stone and R. G. Turner, J. Solid State Chem., 1974, 11, 135.F. Pepe, P. Porta and M. Schiavello, Proc. 8th Znt. Symp. Reactivity of SoZids (Gothenburg,Sweden, June 1976), 1P27.C. Angeletti, F. Pepe and P. Porta, J.C.S. Faraday I, 1977, 73, 1972.P. Porta, A. Anichini and U. Bucciarelli, J.C.S. Furaday I, 1979, 75, 1876.H. Furuhashi, M. Inagaki and S. Naka, J. Inorg. Nuclear Chem., 1973,35, 3009.M. Lensen and A. Michel, Compt. rend., 1958,246, 1977. ' E. F. Bertaut, Compt. rend., 1950, 230, 2132456 Col* I N CoGa,Al,-,O, SPINEL SOLID SOLUTIONSL. Weil, E. F. Bertaut and L. Bochirol, J. Phys. Radium, 1950, 11, 208.C. J. Ballhausen, Introduction to Ligand Field Theory (McGraw Hill, New York, 1962).lo B. N. Figgis, Introduction to Ligand Field (J. Wiley, New York, 1966).l 2 E. J. W. Verwey and E. L. Heilmann, J. Chem. Phys., 1947, 15,174.I3 M. Drifford and P. Rigny, Compt. rend., 1966,263, 180.l4 0. Schmitz DuMont, H. Brokopf and K. Burhardt, 2. anorg. Chem., 1958,295, 7.D. S. McClure, J. Phys. and Chem. Solids, 1957, 3, 311.l6 J. D. Dunitz and L. E. Orgel, J . Phys. and Chem. Solids, 1957, 3, 318.C. Glidewell, Inorg. Chim. Acta, 1976, 19, 445.R. D. Shannon and C . T. Prewitt, Acta Cryst., 1969, B25, 926.' M. M. Schieber, Experimental Magnetochemistry (North Holland, Amsterdam, 1967).l9 C . Delorme, Bull. SOC. Franc. Miner. Crist., 1958, 81, 79.2o P. Bracconi, L. Berthod and L. C . Dufour, Proc. 8th Int. Symp. Reactivity of Solids (Gothen-'' D. Fiorani and S . Viticoli, J. Solid State Chem., 1978, 26, 107.burg, Sweden, 1976), 2Mll.(PAPER 9/1240

ISSN:0300-9599    

DOI:10.1039/F19807602448

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Farahy I, 1980,76,2457-2471Study of Molecular Sieve CarbonsPart 1.-Pore Structure, Gradual Pore Opening and Mechanism ofMolecular SievingBY JACOB KORESH AND ABRAHAM SOFFER"Atomic Energy Commission, Nuclear Research Center-Negev,P.O. Box 9001, Beer-Sheva, IsraelReceived 30th August, 1979The pore structure of a fibrous carbon molecular sieve has been studied by adsorption of COz,0 2 , CzHz, H2, N2, CO, Xe and SF6 as molecular probes. Apart from the negligible outer surfaceof the fibres, all adsorption sites possess molecular sieving properties. Mild activation steps enablethe graduated opening of critical pore dimensions in the range 3.1-5.681, which keeps adsorptionselectivity between molecules varying by merely 0.2 8, in cross-section 9 100 : 1 .The pore openingis effected by removing surface groups as COz and CO due to degassing at temperatures from 100to 700°C and by burning off skeletal carbon atoms in air at 400-450°C. Degassing at temperatures> 800°C leads to pore closure due to sintering. Removal of surface atoms must result in porewidening by steps as large as a few Angstroms, in contradistinction to the observed graduated poreopening.It is anticipated, therefore, that the fine discrimination between molecules of similar dimensionsis of kinetic-statistical origin, so that molecular sieving by pores, substantially greater than the mole-cules considered, is possible. The detailed model is based on the existence of a few rate-determiningconstrictions close to the outer surface of the fibres and of wider pores composing the major part ofthe pore volume.How-ever, constriction structure allows only a finite width of slits, still within a few Angstroms.High adsorption stereospecificity over a wide pore dimension range has enabled the studiedadsorbates to be ordered in a sequence of increasing critical molecular dimensions, which does notalways correspond with estimates based on gas-phase collision diameter or on bond length andVan der Waals radii.Adsorption behaviour of the flat benzene molecule suggests slit-like pores.Molecular sieve carbons compose a relatively new area in surface chemistry.Clear evidence of this property was observed by Franklin who found that fluidswere excluded from heat-treated coals in the order of their molecular size.found a reversal of the normal thermodynamic temperature dependence of physicaladsorption of small gaseous molecules on coals, a phenomenon related to activatedadsorption into pores a few Angstroms in dimension. laterpublished a thorough study of the molecular sieve properties of polyvinylidenechloride (PVDC) charcoal.Since then the subject has constantly been investigatedand revie~ed.~? As stated by Walker et aZ.,4 carbon molecular sieves may bepreferred to inorganic molecular sieves because of their stability at higher tempera-tures and in acidic media and also because of their lower affinity to water. Molecularsieve carbons are also attractive from the practical point of view. Molecular sievecarbons can be easily obtained by pyrolysis of many thermosetting polymers such aspoly(viny1idene chloride) (PVDC),3 poly(furfury1 alcohol),h cellulose, cellulosetriacetate, * saran copolymer, aswell as various coals such as coconutshell.12Pore structure can be opened by slight or extensive activation (oxidativeburnoff),13* l4 or closed by high temperature treatment in vacuum or under an inert2457MaggsDacey and Thomaspolyacryloni trile,l O and phenol formaldehyde2458 MOLECULAR SIEVE CARBONSgas atm0sphere.l The latter means of pore dimension design was best demonstratedby Lamond et a1.l5 who were able to reduce the pore dimension of a PVDC carbonin a few steps to 4.3 A by progressively increasing the high temperature treatment.Thus the gradual changes in pore structure which can be effected on mineralmolecular sieves by partial cation exchange l6.l 7 are also obtainable on carbons bymodifying the high temperature treatment. Since molecular sieve carbons are non-crystalline materials, their pore structure below 5A cannot be studied by X-raydiffraction analysis as is the case with most mineral molecular sieves. Transmissionelectron microscopy, on the other hand, is still insufficiently established for suchsmall pore structure dimensions. Analysis of the adsorption isotherms of molecularprobes of different cross-sections remains the most effective method for studies ofmolecular sieve carbons. By this method molecules of smaller cross-section arephysically adsorbed in preference to Iarger molecuIes.This is a highly sensitivemethod as demonstrated by numerous data where great differences in adsorbabilityare exhibited by molecules of very similar dimensions.l49 18* In this work a newapproach ,towards ultramicropore development in carbon is presented, by which,judging from molecular probe adsorption behaviour, critical pore dimensions of asingle starting material can be gradually increased for quite a large span, while thebandwidth of the pore distribution function remains very sharp. Some questionsconcerning pore structure and geometry are answered. Part of these deals with theouter surface roughness, the shape of pore cross-section and the change of cross-section along the pores.EXPERIMENTALAPPARATUSA conventional volumetric high vacuum system was used.Pressures were measuredwith a BLH type DHF transducer (0-20 p.s.i.). The gauge was calibrated by a ConsolidatedVacuum Corp. Mcleod gauge and a mercury differential manometer to an accuracy and aresolution of kO.1 Torr. The volumetric part had only glass and Teflon high vacuumstopcocks in order to allow work with organic gases. The analogue output of the pressuregauge was recorded so that fast adsorption kinetics, not easily followed by adsorptionbalances, could be monitored.MATERIALSCarbon cloth TCM 128 was supplied by Carbone-Lorraine, France. The density ofTCM 128 was 1.3 g cm3. Nitrogen and argon were Matheson " prepurified " products.Carbon monoxide was a research grade lecture bottle from Matheson. N2, Ar and COwere passed through a liquid nitrogen trap.High purity hydrogen and oxygen were pro-duced by thermal decomposition of uranium hydride and potassium permanganate, respec-tively. Carbon dioxide, acetylene and sulphur hexafluoride from Matheson were passedthrough a trap at -80°C and then cooled down to liquid nitrogen temperature and freedfrom permanent gases by evacuation. Benzene from Merck, A.R. grade, was dried bysoaking it in an excess of P205 within the vacuum line.Since several kinds of heat treatments were applied the following nomenclature wasadopted. In the case of only high temperature evacuation. the designation C-"C was used,for instance (2-200 for a carbon evacuated at 200°C. Unless specified otherwise, evacuationlasted 17 h. The designation for air oxidation treatment is Ox-time of oxidation inhours - temperature of oxidation in "C; the oxidation was performed in a moderate airoxygen stream over a carbon sampled in a tubular furnace, thermostatted to +2"C.Thegas flow rate, which was limited to avoid any temperature gradient along the sample,exceeded by far any reasonable uptake. Considerable care and accuracy were necessaryduring the oxidative activation in order to obtain reproducible adsorption specificityJ . KORESH AND A . SOFFER 2459Accurate oxidation times and temperatures were secured by first passing deoxygenated argonover the sample until temperature control was achieved, and then switching to oxygen or airflow, as necessary. The initial rates Vo tabulated in this paper were calculated from accurateoriginal expanded scale records of pressure against time plots and not from the figurespresented.Precautions were taken to use samples from the same batch of a few hundred g,while making the comparative study of adsorbability of different molecules. Differentbatches vary slightly in the activation conditions necessary to obtain the same adsorptiveproperties.RESULTS AND DISCUSSIONGRADUAL OPENING OF PORE STRUCTUREOPENING TO COz ADSORPTIONThe original carbon was found to be filled with water up to a volume of 0.12 cm3per g carbon. By pumping out the water at room temperature for 17 h, a productdesignated C-25 is obtained which, of the gases mentioned in the experimental section,adsorbs only carbon dioxide at the extremely slow rate shown in fig.1. Water canbe readsorbed into this carbon up to the same volume filling as that of the originalcarbon. Progressively increasing rates of CO, adsorption are achieved after furtherevacuation steps at increasing temperatures, as shown for 3 steps in fig. 1. Thecomplete series of initial rates are summarised in table 1. The overall lo4 foldincrease in kinetics is noteworthy.0.50.4UI M0 c3 8 0.3 z0.20.1-00- 0- A- aA0 AAA-0AAAa A0 8 a .AA AAtlminFIG. 1.-Adsorption kinetics of C 0 2 at 195 K on TCM 128 carbon fibres evacuated at varioustemperatures. Initial pressures in all experiments were 68-70 Torr. 0, 200°C, Vo = 21.93 pmolmin-I ; A, 100°C, V, = 8.23 pmol min-' ; ., 25"C, Vo = 0.8 pmol min-'.OXYGEN ADSORPTIONUptake of oxygen commences from C-200 and above on the high temperaturetreatment scale.Oxygen adsorption on C-200 is shown in fig. 2, together with acurve for C 0 2 adsorption reproduced from fig. 1. The pore structure is more ope2460 MOLECULAR SIEVE CARBONSto COz, and this is shown by the greater adsorption kinetics. Selectivity in favourof C02 is much greater in the more closed C-100. In this case, only a maximum of0.03 mmol per g oxygen can be adsorbed at an equilibrium pressure of 40 Torr,*while C 0 2 adsorption (fig. 1) amounts to 0.5 mmol g-1 at a pressure of 1.5 Torr.TABLE IN INITIAL RATES OF C02 ADSORPTION ON TCM 128 CARBON FIBRES AFTER VARIOUSHIGH TEMPERATURE TREATMENTSdesignation ofhigh temperature initial ratetreatment /pmol min-' g-l25°C-17 h 8 x lo-'100°C-17 h 8.237OO0C-17 h 2 .1 9 ~ 10'250°C-4 h 2.22x lo2750°C-21 h 7.81 x lo23OO0C-17 h 8.1 x 103..+ *0.1500Oao6* 0.035 10 20 30 LO 50 60 M 80t/minFIG. 2.-Adsorption kinetics of O2 on C-200 carbon at 195 K. Initial pressure 70 Torr. A, COz,Vo = 21.93 pmol min-' ; a, 02, Vo = 9 pmol min-', Tevac = 200°C.ACETYLENE ADSORPTIONThe degree of adsorption of acetylene at 195 K into the ultramicropores wasnext to that of oxygen. Thus the C-200 carbon, which adsorbs considerable amountsof oxygen, again exhibits only a minute outer surface adsorption of acetylene( z 0.024 mmol g-I at 51 Torr). Upon a slight increase in high-temperature treat-ment, acetylene penetrates into the pores of the resulting C-250 and C-300 carbonsas shown in fig.3. As with C02, the measurements with C2H2 were quite detailed,demonstrating again the very fine pore opening upon slight increase in high tempera-ture treatment.* Even this amount is attributed to the outer surface as discussed laterJ . KORESH A N D A . SOFFER-0 -A@In1 1 1 1 1 1 l 1 1 1 1 1 1 1 1 124610.05s- AA0 1 I I I I I 1 I- I M -0.3100 AA 0. A00AI 8010 50 100 150tlminFIG. 3.-Adsorption kinetics of acetylene at 195 K on C-250 and C-300 carbons.70 Torr, the C-300 refer to the upper time scale.Initial pressure0, 250°C, 21 h, Vo = 21.7 pmol min-' ; A,300"C, Vo = 360 pmol min-' ; W, 250"C, 5 h, V, = 2.16 pmol min-'.NITROGEN AND ARGON ADSORPTIONAccording to criteria of adsorbability identical to those of COz, O2 and C2H2,the series of adsorption experiments at 195 K was extended to carbons after furtherincreasing the evacuation temperatures.Thus argon and nitrogen adsorptions at195 K commence on the C-300 carbon and not on the C-250 carbon which adsorbsacetylene. Argon and nitrogen are adsorbed at about the same rate (fig. 4), i.e., theyhave similar critical cross-section diameters.00 A0 AA 0O At/min70 Torr. 0 , N2, Vo = 7 pmol min-' ; A, Ar, Vo = 4 pmol min-'. Tad = -80°C.FIG. 4.-Adsorption kinetics of argon and nitrogen at 195 K on C-300 carbon. Initial pressur2462 MOLECULAR SIEVE CARBONSA m *- Y 0' AW* aI I I I 1 I I 1 I 1CRITICAL DIMENSION OF HYDROGEN AND CARBON MONOXIDEIn order to eliminate thermal activation effects on adsorbability of variousmolecules, isotherms must be taken at the same temperature. Due to its low heatof adsorption,20 hydrogen cannot be practically adsorbed at 195 K, whereas manygases which differ greatly in volatility can. A comparative study of the adsorptionkinetics of 02, H2, N2 and CO was therefore performed at 77 K in order to placeH2 and also CO on the scale of critical molecular dimensions.Results are given infig. 5. Here again, stereospecificity is very high. Nitrogen adsorption on C-300carbon was at least two orders of magnitude lower than that on C-400 carbon at anytime. The tiny amount that adsorbs on C-300 is attributed to adsorption on theoutersurface of the fibres.Adsorption of oxygen commences at 77 K on the C-200carbon but that of hydrogen only on C-250. One can confidently deduce fromthese results that hydrogen is a larger molecule than oxygen. Also CO is slightlysmaller than N2 so that the sequence O2 < H2 < CO < N2 is obtained. The place-ment of H2 after O2 first suggested here is surprising.0-0 *O.lt-0 ******AmArntlminFIG. 5.-Adsorption kinetics of 02, H2, CO and N2 on carbon samples. O2 and H2 results aregiven for carbons at lowest high temperature treatment which enable their adsorption. Initialpressures in all samples were 65-70 Torr. 0, CO (4OO0C, 4 h), Vo = 32 pmol min-' ; *, N2(400"C, 4 h), Vo = 5 pmol min-' ; A, H2 (250°C), Vo = 5.5 pmol min-' ; 1, O2 (20O0C), VO =1.8 pmol min-'.Tad = 77 K.MECHANISM OF PORE OPENING BY HIGH TEMPERATURE EVACUATIONHigh temperature evacuation was found to be effective for pore opening up to400°C where adsorption kinetics increase considerably with evacuation time (table 2).From this temperature up to 800°C only slight changes in pore dimensions occurred.Thus xenon, the next larger molecule which we tried, can penetrate at a very slowrate into carbon C-700 but not into C-400. Quicker rates cannot be achieved byhigher temperature evacuation. Furthermore, pore closure by sintering takes placJ . KORESH AND A . SOFFER 2463above 800"C, as shown in table 2. Complete pore closure towards nitrogen adsorp-tion at 77 K is effected after evacuation at 1200°C.One can conclude that the mildpore opening by high temperature evacuation is due to the removal of surface oxygengroups as CO, and CO. The fact that it essentially terminates at = 400°C suggeststhat, at higher temperatures where surface groups can still be degassed as CO,,lsintering may already be taking place. The processing of organic fibres to producethe TCM 128 cloth includes high temperature treatment at temperatures as high as1200°C.22 This implies that initially the carbon cloth has no oxygen groups, sincethese are completely degassed above 1000"C.21 Exposure of the carbon to air isresponsible for the pore closure by oxygen chemisorption. We have observed thiseffect and so exposure of treated samples to air was avoided.TABLE 2.--INITIAL ADSORPTION RATES OF NITROGEN AT 77 K SHOWING OPENING AND CLOSUREOF ULTRAMICROPORES OF ?'CM 128 CARBON AFTER VARIOUS HIGH TEMPERATURE EVACUATIONStemp / "C 400 400 400 1000 1100 1200duration/h 1 4 17 2 3 3adsorption rate/ m o I g-l min-l 0.7 5 3240 200 13 0PORE OPENING BY MILD OXIDATIONA question of interest is now whether the well known oxidative activation can beused for a further extension of pore critical dimensions to achieve the same finegradation as that obtained by the high temperature evacuation.In order to meetthis goal, we chose air oxidation at z 400°C rather than steam or carbon dioxideactivation which can be performed only at considerably higher temperatures. Theresults of oxidative activation are given by the adsorption isotherms of xenon andsulphur hexafluoride presented in fig.6. Activation for 1 h (Ox-1-400 carbon) allowsthe adsorption of xenon but not of SF6, which requires 2 h activation at 420°C(Ox-2-420 carbon). The spherical molecules which have been used so far are Ar,Xe and SF6. The diameters of these molecules calculated from their liquiddensities 23 are 3.6, 3.94 and 5.02 A, respectively. The results show that evacuationat 400°C never opens the pores for xenon adsorption. Thus, diameters of3.6 < d/A< 3.94, 3.94 < d/A< 5.02 and 5.02 < d/A are suitable for the C-400,Ox-1-400 and Ox-2-420 carbons, respectively. The oxidative activation stepsdescribed above are therefore not as finely graduated as those of evacuation.Webelieve, however, that oxidation at lower temperatures and for longer times wouldalso enable fine and stepwise pore opening as for high temperature evacuation.ORDERING OF MOLECULES BY THEIR CRITICAL DIMENSIONSWe have demonstrated the possibility of almost continuous pore opening so thathigh stereospecificity in the adsorption of molecules of critical cross-section diametersranging from 3.1 to 5.6A is achieved. To our knowledge such continuous enlarge-ment has not yet been produced for any molecular sieve starting material. Thisfinding makes it possible to construct a cross-section diameter sequence for greatlydiffering molecules. Thus we deduce from the expderimental results obtained sofar that the critical molecular dimensions of the molecules studied change as follows :H20 < C02 < 0, < C2H, < H, < CO < N2 = Ar < Xe < SF62464 MOLECULAR SIEVE CARBONSA 0l I 1 I I I I 1 1 1 1 1 1 1 1 1 1 1 1 11 5 10 15 20p /TorrFIG.6.-Adsorption isotherms of xenon on carbon Ox-1-400 (0) and of SF6 on carbon Ox-2-420 (A).As far as we could determine, all available data for adsorbability on molecular sievematerials correlate with the above sequence. For instance, the preference of oxygenand carbon dioxide over nitrogen has been found for both ultramicroporous carbon l4and zeolites.24 Also, preference for carbon dioxide over acetylene has been reportedfor zeolites. This ordering can in principle be extended to any other adsorbatewhich can ultimately be accommodated by the pore structure.These results do notthoroughly correspond with collision diameter data,2 nor with bond lengths andvan der Waals radii.26 Discrepancies particularly arise with H, and C 0 2 . Thesewill be discussed in a following paper, in addition to an assessment of the numericalvalues of critical molecular dimensions based on geometrical considerations.PORE STRUCTUREAfter showing in the last section that apparently continuous widening of criticalpore dimensions can be effected by mild steps of evacuation, the elucidation of thepore structure which gives such a capability becomes of profound interest. In thefollowing section, activation procedures and adsorption isotherms will be carefullyanalysed in order to clarify the pore geometry.CONSTRICTIONS A N D PORE OPENINGAmong the first questions arising about pore structure is whether the critical poredimension is constant across the entire depth or whether there are a few constrictionswhich determine the sieving properties and which are followed by wider dimensionsinto which the adsorbate molecule is free to move.A better understanding of thepore opening by high temperature treatment would be of help in this respect. Hightemperature evacuation opens the pores by the removal of surface groups, resultingin an enlargement of pore volume. Degassing performed at temperatures not higherthan 400°C would remove mainly surface groups in the form of CO,. The porevolume increase which causes pore opening should therefore be :where AW, is the weight loss per g of carbon and dcoz is the solid density of COzJ .KORESH AND A . SOFFER 2465In cases of oxidation activation followed by 1 h evacuation at 400°C, the weight lossis the sum of the loss of carbon due to burnoff and that of CO, due to evacuation.The total experimental pore volume increase will therefore be :AW, AW,AVp= -+----&Oz dgraphitkTable 3 gives the experimental relative pore volume increments AVp/V,, due to hightemperature treatment. The pore volume Vp in the case of only high temperaturetreatment degassing was taken as that of total pore volume filling obtained from theC02 isotherm on C-300 carbon. In the case of oxidative activation, Vp was the totalpore volume filling by nitrogen. Weight losses were based on the dehydrated C-25carbon. If a constant pore width was prevailing for the pore system and enlarge-ment of pores by activation occurred evenly, i.e., by peeling off a layer from the pores,pore opening from a dimension which hardly enables CO, adsorption (C-100) to onewhich hardly enables N, adsorption (C-300) would require a relative pore dimensionincrease given by :for slit-shaped pores, and by :which is of higher values, for cylindrical pores.The magnitudes of 0 are the cor-responding critical molecular diameters. The values of (A Vp) /( V,), calculated accord-ing to eqn (3) are also given in table 3. The molecular diameters (a) which servedfor this calculation were estimated as 3.1 8, for C 0 2 , 3.28 A for 02, 3.44 8, for H2,3.59A for N2, 3.9411, for Xe and 5.0211, for SF,.23*Table 3 shows that the calculated values of (AVp)/(Vp), are z 3.5-6 times greaterthan the experimental ones, which is in contradiction to the model of homogeneouspore dimensions.This behaviour, on the other hand, can satisfactorily be explainedby assuming that, apart from few constrictions responsible for the molecular sievingeffect, the pore structure is relatively wide. Hence, a widening of the constrictionsonly (by high temperature treatment), and not of the whole pore depth, is necessaryto introduce increasingly larger molecules, so that experimental relative pore volumeincrements can be considerably below the values given in eqn (3) and (4). This isthe case encountered here. The portion of the total porosity occupied by thenarrower constriction may be at most equal to :f p = AVp -1- Q-Qcozv p Qcozprovided the whole relative volume increase is associated with it.Since in generalactivation is assumed to proceed over the whole pore walls, the above ratio, whichaccording to table 3 is z 17-28 %, is the ultimate value. The contribution of con-striction to the total porosity should therefore be considerably below that value.Once the constriction model is adopted, several unresolved peculiarities can beelucidated.* The diameters (T for the globular SF6 and Xe were calculated from molar densities of liquids.Shape correction factors were used for the other molecules2466 MOLECULAR SIEVE CARBONSTABLE 3 .-EFFECT OF HIGH TEMPERATURE TREATMENT ON RELATIVE PORE VOLUME EXPANSIONAND WEIGHT LOSS FROM TCM 128 CARBONtreatmentdegassing oxidation relative poreweight volume expansion largestmolecule/h /"C /h /"C (%) experimental calculated adsorbeddesignation time temp time temp loss ( %)c-100c-200C-250(2-250C-300c-400C-700OX-1 -400ox-2-420OX-1-450OX-2-4500x4450~ ~17 100 0.1 117 200 0.544 250 -17 250 0.7517 300 0.9717 400 1.192 700 1.31 400 2.52 420 5.51 450 11.32 450 18.64 450 29.40.351.742.423.133.834.17.812.715.221.426.6-co25.8 0 2C2H211.0 H215.8 N2N227 XeXe61.9 SF6SF6SF6SF6ABRUPT OR GRADUATED PORE OPENINGPore closure by sintering can basically be carried out to any extent.13 l 5 Unlikesintering, however, pore opening by degassing of surface groups or by burnoff ofsurface carbon atoms can be performed only uia abrupt steps, because of the atomocityof the material.Hence, removal of a surface group, or slight burnoff, must leavebehind a space of a few A units, similar to the dimension of the leaving group, whichresults in abrupt steps of local pore widening. This apparently contradicts theexperimental facts, where consecutive mild activation enables pore widening in stepsof at most 0.2A, as observed by our molecular probe adsorption experiments.Examples are the C-200 carbon which showed practically absolute selectivity for theadsorption of 0, in preference to N,, although both molecules differ by only 0.2Ain their van der Waals radii and by 0.3 A as calculated from both liquid density and 0values given above.Also C-400 carbon discriminates completely between N2 andXe while their diameters differ by only 0.1 or 0.33 A as calculated from liquid densityand Q values, respectively. The fact that molecular sieving carbons are quiteamorphous makes such high selectivity even more surprising since, unlike zeolite,their pore dimensions are not inherently determined by crystal structure. Further-more, and again because of abrupt abstraction of surface atoms, it is hard to imaginehow pore dimensions could be kept constant within +O.l A upon enlargement from3.1 A for CO, adsorption up to 5.02A for SF, adsorption. Presuming that thecritical pore dimensions' distribution function is indeed coarse, we anticipate that thehigh stereospecificity of molecular sieving carbons towards adsorbates, at least withinthe range studied (3.1-5.6 A), originates in kinetic-statistical effects rather than in thegreat homogeneity of pore dimensions.According to this approach there exists alayer composed of a series of a few critical passages or constrictions close to the outersurface of the adsorbent fibre. The contribution of this layer to the overall porosityis negligible, as is the amount adsorbed into it. The passage of a molecule througJ . KORESH AND A . SOFFER 2467this layer from the gaseous phase into the bulk of the porous adsorbent is the rate-determining step for adsorption. The probability for a molecule to pass throughsuch a layer is :NP(N) = Pi (6)i = 1where Pi is the probability of passing one constriction, i.For the sake of simplicityeqn (6) can be averaged to :P(N) E! pN. (7)In terms of the kinetic theorywhere Ea is the activation energy for passing one construction and p o is a frequencyfactor ; combining eqn (7) and (8) we obtainwhich shows that N constrictions in series produce the effect of multiplying N timesthe activation energy. Let us consider two adsorbate molecules A and B, which arevery close in their critical cross-section diameters. The larger one requires a slightlylarger activation energy for being " squeezed " through a constriction. However,although the difference AEAB can be quite small for molecules of similar dimensions,it is amplified N times due to the series of constrictions at the rate-determining layer.This relation is immediately obtained from eqn (9) written for each molecule.Hence,the ratio PA/PB between the adsorption rates of the two molecules assumes the form :P = PO exp -Ea/kT) (8)P(N) = p t exp (-NE,/kT) (9)Therefore, great sieving selectivity can be exhibited even if the molecular sieve has afairly wide pore distribution function.According to this model, criteria for adsorbability are kinetic in nature and do notoriginate from the true sieving effect. They should therefore be based on somepractical time of adsorption. Thus, a characteristic time of adsorption longer thansay 300 h determines a non-adsorbing molecule and a time shorter than 3 h determinesan adsorbing molecule.According to eqn (lo), N(EA-EB) = 1800 cal mol-1 at195K. For N = 10, EA-EB is merely 180calmol-l. Recognizing that E,originates in forcing a molecule to climb up the steep repulsive segment of its adsorp-tion potential, the value of 180 cal mol-1 is indeed a small difference, but neverthelessit suffices to provide great sieving selectivity owing to the multiplier N.Widening of the pores by a few A due to the abstraction of surface groups shouldby itself lead to an inhomogeneity of pore dimensions, i.e., to narrow and wide zones.However, only the constrictions in the layer close to the outer surface determine theadsorption kinetics. The molecules should be much more free to move deep intothe particle. Otherwise molecules of increasingly greater dimensions would still beadsorbed by passing fewer critical constrictions and the high stereospecificity wouldbe greatly impaired.The question of whether the characteristics of the outer layeroccur generally or are specific to the TCM 128 carbon are still under investigation.CONSTRICTION AS EVIDENCED BY THE ADSORPTION RATE OF co,It was shown above (fig. 1, tables 1 and 3) that following mild activations involvingonly 3.3 % of the relative pore volume expansion, the initial rate of adsorption of C 0 2could be increased by four orders of magnitude. Even more striking is the fact tha2468 MOLECULAR SIEVE CARBONSby evacuation for 4 h at 25"C, 12 wt % of adsorbed water are lost, whereas by evacuationfor a further 17 h, resulting in undetectable weight loss, C 0 2 adsorption is renderedpossible, as shown in fig.I. These observations can readily be explained on thebasis of the pore structure suggested above, namely a slight surface group removal issufficient to reduce N in eqn (9) to a value which allows a detectable adsorption rate.Homogeneous pore structure, on the other hand, would require much greater surfacegroup removal, which would be measurable, than could be obtained by prolonged0.40.3Lz-0.20 *I000*000I I I I I I I 1 1 I I l l t l l t l l l50 100 150 200t/minFIG. 7.-Adsorption rates of benzene on (2-300 and C-400 carbon at 25°C. Yo = 12 pmol min-'. n 1 TFIG. 8.-Dimensions of the benzene niolecule calculated from bond lengths and van der Waals radiiJ .KORESH AND A . SOFFER 2469evacuation at 25 or 100°C. A similar multifold is exhibited by acetylene (fig. 2) andby nitrogen (table 2). We believe that similar behaviour will be seen whenever anadsorbate is examined after several successive activation steps.157 10- M -- 5 - 8 k5 -GEOMETRY OF PORE CROSS-SECTION : ADSORPTION OF BENZENEA - - - a --a - A - AA0 -a0 - ---I 1 1 1 1 I I I I0.05 0.1 0.2 0.3 0.4The adsorption of the planar benzene molecule commences on the C-300 carbonat an extremely slow rate (fig. 7) and becomes faster and more significant on the C-400carbon. Benzene is considerably greater in width than the SF6 molecule (fig. 8)although its thickness is considerably smaller. Nevertheless, benzene is adsorbedon carbons from which sulphur hexafluoride is completely excluded.This behaviourindicates that the constrictions in the TCM 128 carbon are slit-shaped. Accordingto these results, the height of the slits in the C-300 carbon is < 5.02& which is thediameter of SF6 as calculated from its liquid density, and > 3.7& the thickness ofthe benzene molecule as obtained from Pauling’s van der Waals radii.AA0aFIG. 9.-Adsorption isotherms of nitrogen at the outer surface of the C-25 and C-200 carbons at77 K. a, Room temperature, S N ~ = 0.67 mZ g-1 ; A, 200°C, S N ~ = 0.86 m2 g-l.OUTER SURFACE POROSITYAn important characteristic of any molecular sieve adsorbent is the contributionof the more open porosity and of the outer, open surface, to the internal porosity.Such contributions reduce the overall stereospecificity of the adsorbent, and imposesome difficulties in analysing the behaviour of porous adsorbents. Selective adsorp-tion on such open surfaces can be performed using sufficiently large molecules to avoidpenetration into the ultramicropores.Thus, oxygen and larger molecules (in thesequence given above) could be selectively adsorbed on the outer surface of C-25carbon. A study of the adsorption on the outer surface of fibrous carbon would b2470 MOLECULAR SIEVE CARBONSadvantageous since the outer surface can be calculated from the fibre diameter providedthe diameter's distribution is not too large. The outer surface is given by :4-3 -rl bos , .CI P 2 -4 A = -Pd- 00- 0al ?-t l I 1 I I 1 1 1where p is the average porous solid density and d is the fibre diameter.For the carbonTCM 128 studied, A is 0.38 m2 g-l.Adsorption isotherms on large samples of TCM 128 carbon are given in fig. 8 fornitrogen on C-25 and C-200 carbons. Both are not penetrated by the adsorbate.Clear evidence of outer surface roughness and pore development is given by thegreater B.E.T. surfaces of 0.67 and 0.86 m2 g-l, respectively. This is confirmed byPIP01.2 m2 g-l.FIG. 11.-Adsorption isotherm at 77 K of O2 on the outer surface of a C-25 carbon. So2 =PIP0FIG. 12 -Adsorption isotherm of nitrogen on the C-400 carbon at 77 K. S N ~ = 1.2 m2 g-'FIG, 10.-S.E.M. micrograph of untreated TCM 128 carbon.[To face page 247J.KORESH AND A . SOFFER 2471the S.E.M. micrograph of the non-treated TCM 128 carbon shown in fig. 9. Anadsorption isotherm of O2 on the C-25 carbon is presented in fig. 10. Here thecorresponding B.E.T. surface, 1.2 m2 g-l, is almost twice that of N2. Thus molecularsieving effects already exist for the outer surface porosity. However, due to theshallow penetration, in our terms low N values, the adsorption rate is very fast andmolecular sieving selectivity appears to be poor. The ratio between moles of N2necessary for total pore volume filling shown in the isotherms of fig. 12 and the N2monolayer coverage of the outer surface is about three orders of magnitude, demon-strating the high degree of molecuiar sieving of the TCM 128 carbon.CONCLUDING REMARKSAdsorption experiments with molecular probes have been performed in order toobtain comprehensive information about pore structure and properties.A strikingresult of these experiments is that high stereospecificity between molecules of verysimilar dimensions can be achieved for the wide range 3.1-5.6 A developed by modify-ing the same starting material. To the advantages of carbon molecular sieves overzeolites given by Walker et aL4 one could add that, within a certain range, poredimensions may be tailored to any desirable value.We thank D. Rosen for technical assistance.R. E. Franklin, Trans. Faraday SOC., 1945, 45, 668.F. A. P. Maggs, Nature, 1952, 169, 793.J. R. Dacey and D. G. Thomas, Trans. Faraday SOC., 1954, 50, 740.P. L. Walker, L. G. Austin and S. P. Nandi, Chemistry and Physics of Carbon, ed. P. L. Walker(Marcel Dekker, N.Y., 1966), vol. 2, p. 257-371.D. H. T. Spencer, Porous Carbon Solids, ed. R. Bond (Academic Press, London, 1967),H. Marsh and W. F. K. Wynne-Jones, Carbon, 1964, 1, 269.J. J. Kipling and R. B. Wilson, Trans. Faraday SOC., 1960, 56, 557.L. B. Adams, E. A. Baucher and D. H. Everett, Carbon, 1970, 8, 761.Carbon and Graphite (Society for Chemical Industry, London, 1971), p. 467.Graphite (Society for Chemical Industry, London, 1970), p. 380.Chemical Industry, London, 1970), p. 7.Otdel. khim. Nauk, 1961, 1, 29.Akad. Nauk S.S.S.R., 1964, 157, 656.pp. 87-151.a B. McEnaney, Carbon, 1975, 13, 515.l o D. H. T. Spencer, M. A. Hooker, A. C. Thomas and B. A. Napier, Proc. 3rd Conf. Industrial1 1 P. L. Walker, T. G. Lamond and J. E. 111 Metcalfe, Proc. 2nd Con5 Industrial Carbon andl 2 J. J. Kipling and B. McEnaney, Proc. 2nd Con$ Industrial Carbon and Graphite (Society forl 3 M. M. Dubinin, E. D. Zaverina, A. T. Kaverov and Kasatochkin, Izvest. Acad. Nauk S.S.S.R.,l4 M. M. Dubinin, 0. Kadlec, I. Botlik, E. 0. Zaverina, A. Zukal and B. Sumec, Dokladyl 5 T. G. Lamond, J. E. Metcalf 111 and P. L. Walker, Carbon, 1965, 3, 59.l6 D. W. Breck, W. G. Eversole, R. M. Milton, T. B. Reed, and T. L. Thomas, J. Arner. Chem.l7 R. M. Barrer and D. W. Riley, Trans. Faraday Soc., 1950, 46, 853.l 9 D. W. Breck, J. Chem. Ed., 1964, 41, 678.2o R. W. Barrer, Proc. Roy. SOC. A, 1937, 161,476.21 S. S. Barton, D. Gillespie and B. H. Harrison, Carbon, 1973,11, 649 ; S. S. Barton and B. H.22 Special communication from the producer.23 J. Koresh and A. Soffer, to be published.24 D. W. Breck, Zeolite Molecular Sieces (John Wiley, N.Y., 1974), p. 635.2 5 G. L. Kington and A. C . Macleod, Trans. Faraday Soc., 1959,55, 1799.26 L. Pauling, Nature of the Chemical Bond (Cornell Univ. Press, Ithaca, N.Y., 3rd edn, 1960).SOC., 1956, 78, 5963.Y. Toda, N. Yuki and S. Toyoda, Carbon, 1972, 10, 13.Harrison, Carbon, 1975, 13, 47.(PAPER 911384

ISSN:0300-9599    

DOI:10.1039/F19807602457

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I , 1980,76,2472-2485Study of Molecular Sieve CarbonsPart 2.-Estimation of Cross-sectional Diameters of Non-sphericalMoleculesBY JACOB KORESH AND ABRAHAM SOFFER*Atomic Energy Commission, Nuclear Research Center-Negev,P.O. Box 9001, Beer-Sheva, IsraelReceived 19th November, 1979The estimation of molecular dimensions by various experimental methods is discussed and theirrelevance to non-spherical molecules is evaluated. Liquid molar densities provide an average dimensionwhich is adequate for spherical molecules but completely insensitive to molecular shape. The combina-tion of bond lengths and van der Waals radii enables one to estimate satisfactorily the length of linearmolecules but not their width. The kinetic diameters calculated from different physical properties ofgases diverge significantly and are insensitive to molecular shape.Adsorption in molecular sieve(MS)solids exhibit high sensitivity to the width or smallest dimension of the molecule. Molecular sievecarbons (MSC) seem promising in this respect since their avera e pore diameter can be tailored to theexact critical dimension of any molecule in the range 3-5.51 studied so far. The combination ofadsorption stereospecificity data and liquid molar volumes provides reasonable numerical estimates ofthe width of non-spherical molecules. Polar molecules may have different dimensions depending onwhether the carbon surface is polar (oxidised) or non-polar. Hydrogen acquires a surprisingly largewidth which is in accordance with its high liquid molar volume.Adsorbent-adsorbate interactions playa crucial role in determining molecular dimensions and serve to elucidate the unusual behaviour of bothhydrogen and polar molecules. An overview of the concept molecular dimension is given in terms of theeffect of intermolecular forces.It has been shown previously' that: (1) The pore opening in molecular sievecarbon (MSC) derived from the same material can be adjusted to any desirablevalue from 3.1 8, to at least 5 8, by stepwise mild thermochemical treatments. (2) Ahigh adsorption stereospecificity, comparable with that of zeolite molecular sieves(MS), can be maintained over the entire pore dimension range. These two proper-ties enable one to sort various small molecules according to their molecular dimen-sions, by determining their penetration into an ultramicroporous structure.In thisrespect, MSC seem to have more advantages than crystalline MS, since in the lattercase the change of pore dimension involves essentially the synthesis of variouscrystalline materials. Accordingly pore dimension changes should still occur insteps while practically any value could be achieved by modifying carbons within thespecified range.In this work on the assessment of molecular dimensions, comparison with othermethods is made and any discrepancies are elucidated. Some questions, such as thatof the dependence of the results on specific adsorbent-adsorbate interactions andthe type of critical molecular dimension which determines absorbability, are thenconcerned.247J .KORESH A N D A . SOFFER 2473SOME COMMON METHODS OF ESTIMATINGMOLECULAR DIMENSIONSESTIMATION FROM MOLAR VOLUME OF THE CONDENSED PHASEThis is probably the most accepted approach to estimating molecular dimensionsin surface chemistry. It assumes a close packing of spheres2 so thatwhere d is the diameter, A4 the molecular weight, p the condensed phase densityand N is Avogadro’s number.To our knowledge, however, eqn (1) has not been applied in any systematic studyof MS absorbents, probably because it does not correlate well with the experi-mental molecular sieving sequence of non-spherical molecules. For non-sphericalmolecules, however, this approach must not be ruled out and some shape factorshave to be accounted for.ESTIMATION FROM B O N D LENGTH A N D V A N DER WAALS R A D I IIn order to obtain a molecular dimension by this method bond lengths have tobe completed by the dimensions of the non-bonding electron clouds of the outeratoms of a molecule, i.e.by the van der Waals radii3 These have been estimated forthe most abundant covalently bonded atoms and groups, as an average of values toan accuracy of kO.1 A. However, the van der Waals radii are only adequate in thedirection opposite to the covalent bond of the atom. At small angles from the bonddirection, the covalent atom radius may be as much as a few tenths of Angstromsmaller, as noted by P a ~ l i n g , ~ and it cannot then be estimated to any accuracy.This approach cannot therefore account for the width of linear molecules, althoughtheir length may be estimated from it quite successfully. This concept may alsoinclude estimations based on molecular model4ESTIMATION FROM COLLISION DIAMETERS I N THE G A S PHASEThis method is presently the most common one since it was found to correlatebetter than the former two with the sieving properties of mineral MS.5 The molecu-lar dimension is an essential parameter which appears whenever a gas propertywhich is sensitive to the “covolume” of the molecules (second virial coefficient,viscosity, diffusion) is calculated using a model of a potential energy function.Molecular dimensions considered by this approach relate to the average of allmutual orientations of the colliding molecules so that the molecules are consideredto be essentially spherical.Some studies by Kihara‘ and Corner7 allow for molecular shape in the calcula-tion of the second virial coefficient B(T) according to the Lennard-Jones potential.Kihara extended the work of Isihara and Hayashida,’ who provided a generalformula for the second virial coefficient of non-spherical rigid molecules, in the caseof “soft” interacting molecules. He obtained very satisfactory results of B against Tplots for both spherocylindrical and spheroidal H2, C02, N2 and C2H4 molecules.The obtained lengths and widths of these molecules were, however, different by afew tenths of 8, for the two models (table l), indicating that “from the second virialcoefficient alone it is impossible to assert generally which model represents rea2474 MOLECULAR SIEVE CARBONSTABLE 1 .-MOLECULAR DIMENSIONS (A) AS ESTIMATED BY VARIOUS METHODSobserved bond length and kinetic diameter (gas) liquid molar volumeincreasing van der Waals radii width and length’ average shape correctedsequence width length spherocyl ellipsoid average” (d) d/2’Ib width length3.7w 4.6W4.53 4.03 3.1 5.56 co2 2X5 5.38 5.9L 5.09L 4B00, 2.8 4.01 3.54 4.0 3.51 3.28 3.99C;H2 5.72 4.22 4.6 4.1 3.33 5.63H2 2.4516 3.15 4.06 3.62 3.44 3.852.81W 3.24W3.55L 3.32L 2.973.41 ;- 4.09W ;-N2;Ar 3.0;-’s6 4.1;- 4.57~;- 4.29~;- 3.15; 3.42 3.85; 3.6 3.59; 3.6 3.59; 3.6 4.2; 3.6SF6 4.53* 5.87* 5.51 5.63 5.02x, 4.05 4.42 3.94 3.94 3.94* Calculated from molecular model.molecules more closely”.It seems, therefore, that the second virial coefficient is, ashas been also stated by Hirshfelder et al.,9 quite indifferent to molecular shape, sothat the latter caniiot be assessed at the accuracy needed to account for the highadsorption stereospecificity of zeolites5 and MSC.’ Insensitivity to molecular shaperesults from the shape factor f of rigid non-spherical molecules derived by Isiharaand Hayashida.8 This value is obtained by integrating over all mutual positions oftwo colliding molecules allowing equal probability for all orientations.Thus, thestatistical contributions of the closest and farthest positions of non-spherical mol-ecules are only fractional and therefore a tendency for averaging should be expectedwhenever covolume dependent properties are studied.EXPERIMENTALAdsorption measurements were carried on in a volumetric high vacuum system.Pressureagainst time curves were recorded, from which adsorption against time curves were obtained.The starting amount of gas was so chosen that the initial pressure in the adsorption cell was6 M 5 Torr for all experiments. The starting material which had been treated to obtain differentpore openings was a fibrous carbon cloth TCM 128, a product of Carbon-Lorraine, France.Other experimental details were as described previously.’CRITICAL MOLECULAR DIMENSION CONTROLLINGMOLECULAR SIEVINGThe accumulated data on adsorption on solid MS indicate clearly that the mol-ecules occluded during adsorption are preferably oriented so that their passagethrough the pores is determined by their smallest dimension.For instance, pre-ferred adsorption of normal over branched paraffins occurs on both minerallo~’’and carbon MS.12313Also, Dubinin et al. found that adsorption of benzene was considerably betterthan that of cyclohexane on MSC.I4 This behaviour clearly indicates that thenarrower normal paraffin and the flat benzene molecule are aligned parallel to thepore during their migration into it. The same results are obtained from the adsorb-ability of the simpler axial molecules carbon dioxide and acetylene, which areadsorbed on NaA zeolite after partial exchange with potassi~m,~ while the“shorter” nitrogen molecule is not adsorbed.We have previously observed’ that these molecules behave similarly on our MSCand we found evidence of the parallel orientation of benzene in slit-like pores.Wehave also found that dinitrogen oxide is adsorbed in preference to nitrogen, asshown in fig. 1. In conclusion, the smallest dimension of non-spherical molecules iscritical in controlling their penetration into ultramicroporesJ . KORESH AND A . SOFFER 24750.50.44'w 0.3E-i2s 0.20.100000001 2 3 4 5 6tlminFIG. 1.-Adsorption kinetics of dinitrogen oxide on C-250-4.' In this case nitrogen as well as hydrogenare completely excluded from the pore system. V, is the initial adsorption rate, 220 pmol min- ; initialpressure 60 Torr. T,d = -80°C.OBSERVED SEQUENCE OF ADSORBABILITY I NRELATION TO MOLECULAR DIMENSIONThe experimental sequence of adsorbability is given in table 1 together withmolecular dimensions calculated by different common methods.In this table anytwo molecules are considered to have different dimensions if we could develop aMSC for which the ratio of their initial rates of adsorption into the pores was atleast one thousand.? Except for the last two columns on the right, which will bediscussed in the next section, no other columns correlate with the experimentalorder over the whole span of adsorbates studied. In the following, each columndealing with width or average dimension will be discussed in detail. The inadequacyof the van der Waals width frequently used as a measure of the smallest dimen-si0n~9~ has been shown above.By this approach carbon monoxide, carbon dioxideand oxygen would have the same dimension, namely that of a chemically boundoxygen atom (2.8 A). Also, N 2 0 would have the same width as nitrogen (3 A).However, a net preference of N 2 0 adsorption over N2 (fig. l), of C 0 2 adsorptionover O2 and of the latter over CO is manifest. An even more significant discrepancyis shown by hydrogen, which exhibits a surprisingly large dimension in completecontrast to its van der Waals radius. The last four molecules presented in table 1 fitthe experimental sequences according to all approaches presented, presumablyt Such high selectivities (hence sensitivity to molecular dimension) were obtained for a closed porestructure. Under such conditions the rate was so slow that frequently equilibrium values of adsorptioncould not be obtained after any reasonable time.Whenever these could be estimated we observed that atsimilar and fairly low pressures they differed by only a small amount, in contrast to the orders-of-magnitude ratios in the initial rate. This is in accordance with the view' that the pore volume iscomposed of relatively large, poorly stereoselective pores with few constrictions responsible for molecu-lar sieving, which is therefore mainly of kinetic origin2476 MOLECULAR SIEVE CARBONSbecause of their proximity to the spherical shape, and mainly because of the largedimensional differences between them. The kinetic diameter in the gas phasepresents a different sequence which deviates from the experimental one mainly withregard to elongated C02, C2H2 and N 2 0 (c = 3.8 A) molecules.This could resultfrom averaging the overall orientations of the linear molecules, and thereby greatlyreducing their sensitivity to shape. Beyond any shape considerations, the relativelysmall dimension of hydrogen is in complete disagreement with adsorption experi-ments. The hydrogen molecule exhibits an extraordinarily large dimension in theadsorbed state. The same behaviour is exhibited by the average dimension calcu-lated from the liquid molar density, according to eqn (1) (to be discussed later). Asin the case of the kinetic diameter the liquid molar volume sequence does notcorrespond to the observed one for the elongated C 0 2 and C2H2 molecules, whichsuggests the necessity of shape corrections.As a frame of reference for the average molecular dimension, we preferred dataobtained from liquid molar density rather than those from the kinetic theory ofgases because of (1) the better correspondence of the hydrogen dimension; (2) theadsorbed state being closer to a condensed state than to an ideal gas; (3) wideravailability of data on liquid density, especially in case of large organic moleculeswhich will be studied in the future.It is also a well accepted method of estimatingcross-sectional areas of adsorbates on open surfaces from liquid molar volumes.ESTIMATION OF MOLECULAR DIMENSIONAND GEOMETRYThe objective of this section is to assign simple, geometric shapes and numericalestimates of dimensions to adsorbate molecules in such a way that: (1) the widthcalculated will correspond to the experimental adsorbability sequence.In particu-lar, the width of the N2 molecule has to fit the argon atom dimension, since theiradsorbabilities are equal. (2) The average molecular dimension of a molecule justpassing a narrow constriction will be 21/6 smaller than its average dimension in theliquid phase.The reason for the factor 21/6 is given below. Taking the liquid phase dimensionof adsorbate as a reference, we recognise that the liquid molecular diameter standsfor the average dimension at the minimum of the intermolecular potential function.However, the critical dimension for the passage through the pores of a MS isreckoned at the zero value of the potential energy.’?Adopting the Lennard-Jones (6,12) potential for the non-polar molecules treatedin this section, the average critical dimension of molecules passing a pore would be0 = d/2Ii6.(2)We obviously attributed a spherical shape to the noble gas atoms and also to the“globular” SF6 molecule.To the linear molecules studied so far we attribute a spherocylindrical or aprolate ellipsoidal shape with the foci located at the nuclei of the edge atoms.Although arbitrary, these geometric shapes, considered also by Kihara,6 assume thegenerally accepted “shape” of linear molecules, which is sharpened at the edges andhas cylindrical symmetry. We will furthermore make use of this arbitrariness byt This implies, very reasonably, that the dimension of a molecule in contact with the walls of a narrowpassage corresponds to a nil interaction.In fact, thermal activation may enable the “squeezing” of evenlarger molecules through pores leading to a slight difference in their dimensionsJ . KORESH AND A. SOFFER 2477trying fractional contributions of spherocylindrical and ellipsoidal shapes so thatthe molecular width can fit the observed sequence of adsorbabilities. According torequirement (2) above and to eqn (2) we equalise the volume of the adsorbatemolecule to that of a sphere of diameter 0. Thus for a spherocylinder we have:where r is the radius, 1 is the sum of bond lengths and o is the average diameterfrom the liquid phase. For a prolage ellipsoidal molecule we have:a = JFZwhere b is the minor axis, a is the major axis of the ellipsoid and c is half of the sumof the bond lengths.The widths 2r and 2h can therefore be calculated from eqn (3) and (4).Thecorresponding values obtained for various molecules are given in table 2 in theobserved sequential order. The numerical width values do not agree with this order.We took, therefore, fractional contributions of spherocylinder and ellipsoid in theform of arithmetic averages of the widths and lengths of the two shapes, for alllinear molecules except C02 for which pure spherocylindrical geometry is retained.The results, which are given in the last two columns of table 1, indicate a fullcorrespondence of widths with the experimental sequence including the coincidenceof the N2 width with the diameter of argon.Another striking feature of theseaveraged values is that the lengths are similar to the van der Waals lengths within0.2 A or less, so that they are not too far from reality. The hydrogen molecule isexceptional in this respect.It appears from tables 1 and 2 that the eccentricity of the pure ellipsoid moleculeis too small to account for both (adsorption) widths and van der Waals lengths,whereas that of spherocylinder too large, so that the average has had to be used. Ifa linear molecule acquires a shape of revolution its volume is given by the followingequation, which is a generalization of the left-hand side of eqn (3) and (4).v, = 4nM3f(4 (6)where E is the eccentricity (i.e. the ratio of length to width) and f ( ~ ) is a functionwhich describes the shape of the molecule (equal to E for an ellipsoid and $ 6 - fora spherocylinder).The use of the arithmetic average is equivalent to the applicationTABLE 2.--DIMENSIONS (A) OF SPHEROCYLINDER AND ELLIPSOID MOLECULES AS OBTAINED FROMEQN (3) AND (4)ellipsoid spherocylinderwidth length width lengthadsorbate (2b) ( 2 4 (24 (2r + 1)co2 3.1 5.560 2 3.5 3.7 3.06 4.27CzH2 3.72 4.98 2.95 6.27H2 3.6 3.67 3.29 4.04N2 3.8 3.95 3.37 4.42478 MOLECULAR SIEVE CARBONS3.03'on 2.0E * *-;1 -1.0of the functionf(e), which in terms of eccentricity is somewhere between an ellip-soid and spherocylinder. The large adsorption stereospecificity of MS imposes strictlimitations with regard to the widths of the molecules but leaves freedom to choosethe lengths from considerations such as crude correspondence with the van derWaals radii.Accordingly significant freedom to choose the shape function f(c) isprovided so that the choice of a simple arithmetic averaging is as adequate as thatof more elaborate methods.I---* -1, -t -EFFECT OF THE INTERMOLECULAR POTENTIALFUNCTIONS ON MOLECULAR DIMENSIONSO F ADSORBATESThe width exhibited by a molecule passing through a pore of a MS is dependenton its interaction with the walls, so that the greater the attractive force the smalleris the apparent width sensed in an adsorption experiment. In view of the extremelysteep repulsion potential it may be possible to consider the molecule and adsorbentas rigid bodies as far as molecular dimensions are concerned.However, the tem-perature dependence of the stereospecificity of MS clearly indicates that this is notthe case since molecules can be thermally activated to overcome narrow pores, i.e.to climb up the repulsive portion of the potential. This behaviour has already beenobserved by Maggs," who showed the increasing adsorption of nitrogen on coalwhen increasing the temperature from 77 K. Dubinin et a1.,16.'7 showed that woodcharcoal MS treated to adsorb oxygen in preference to nitrogen and argon at 77 Kaccommodated nitrogen and argon at 195 K. A similar behaviour was observed forargon and nitrogen on inorganic MS.'* We have also observed thermally activatedadsorption. A carbon thermally treated at 400°C (C-400) whose pores are relativelyopen to nitrogen showed adsorption isotherms having a normal temperaturedependence, as shown in fig.2. On the other hand a carbon with a more closedIA0I 1 1 . 1 I 1 I I 1 I 1 1 . 1 I I I I 1 1 I I I10 50 100 150 200p/Torr**0 AAY aFIG. 2.-Adsorption isotherm of nitrogen at two temperatures and on two differently treated carbonsshowing the inversion of the normal dependence of isotherms for the more closed C-300 carbon due tothermal activation. *,- 196"C, C(400); U, - 196"C, C(300); A, -8O"C, C(300); @, -8O"C, C(400)J . KORESH AND A . SOFFER 2479structure (C-300) shows inversed temperature dependence so that a nitrogen iso-therm at 195 K is considerably higher than one taken at 77 K.A much more pronounced effect of the ability of molecules to be “compressed”through narrow pores is observed by comparing the behaviour of polar adsorbateson polar and non-polar carbon surfaces.As previously described’ the carbons usedso far had a substantial coverage of chemisorbed oxygen whose partial removalserved to enlarge the pore apertures. In most cases described, however, the removalof oxygen surface groups was far from complete. Degassing was mostly below500°C so that CO surface groups did not leave the surface. The various carbonsused so far should therefore be considered as having acquired polar surfaces andAAAAAa0 .AAa0aI l l l t l l l l l l l l l l ’ 1 5 10 15t/minFIG. 3.-Adsorption rates of A, HC1; I, HBr and 0, N2 on the non-polar carbon C-1100.K,,‘,, = 1100°C.will be termed polar carbons.A second way of adjusting pore dimensions is tosinter at temperatures above 800°C. Results indicating the possibility of gradualpore closure have been described previously by means of decreasing adsorptionrates of nitrogen.’ The sintered MSC differs from the polar carbon mainly in thealmost complete absence of oxygen surface groups, as is well known from otherstudies,” and it will be termed non-polar. The sequence of adsorbzbility of the fewnon-polar molecules which we examined was identical to that of the polar carbonpresented in table 1, namely C02 > O2 > H2 > N2.Highly polar adsorbates, however, behave very differently on polar and non-polar MSC.Thus, the adsorption of HBr (p = 0.8 D) on non-polar MSC is slightlyslower than that of N2, while the adsorption of HCl (p = 1.0 D) is faster, as shownin fig. 3. This indicates that the critical diameter of these two molecules is approxi-mately that of N2, i.e. slightly above and below 3.6 A, respectively (table l), inaccordance with the van der Waals (and ionic) radii of atomic chlorine (1.6 A) andbromine (1.85 A).3 These two molecules cannot be absorbed on a non-polar MSCwhich does not adsorb N2 but does adsorb H2 and C02. Polar MSC, on the otherhand, with narrow pores adsorbing only C02, does adsorb both of these gases, asshown in fig, 4, indicating that in this case a critical dimension of ~ 3 . 1 8, isexhibited by these molecules.The same effects of “shrinkage” in contact with pola2480 MOLECULAR SIEVE CARBONSMSC pores was found with methyl chloride ( p = 1.86 D). In fig. 5 the adsorptionrates of CH3C1 at 195 K and of CH3CN at room temperature are comparable withthat of H2 at 77 K, the rate ratio If,,/VcH,c, being 278. This ratio is inversed andbecomes 0.06 for polar carbon C-250-12 h* (fig. 6), indicating inversion of theadsorbability sequence of these two molecules. The sequence inversion is muchmore profound with CH3CN as observed by comparing fig. 5, where the CH3CNadsorption rate on C-1200-1/4 h is comparable with that of H2,Jf with fig.7, whereits adsorption rate on C-100-17h is faster than that of C02. In this case the“shrinkage” of CH3CN due to the polar surface may be estimated to be >0.3 A.In order to understand this behaviour, a deeper view of the role of intermolecularforces in determining the penetration of molecules into molecular sieving pores isA05 10 15 20tlminFIG.4.-Adsorption rates of A, HC1; 0, HBr and D, C 0 2 on polar carbon C-100.necessary. The adsorption energy of pores of molecular dimensions was recognizedto be considerably higher than the adsorption energy on a single plain surface.Calculations by Everett and Powel2’ based on a Lennard-Jones type potentialshow that the highest ratio of the minimum potential energy of an adsorbate withina slit shape pore to that of an open surface is 2. The maximum ratio occurs at ad/ro, for which 2d is the distance between the foci of nuclei on the opposite walls ofthe pores and ro is the regular zero-energy distance of the two-particle potentialfunction.We may generalize some of Everett’s potential-energy drawings to anyother potential-energy function, provided they exhibit a minimum and a re-pulsive portion. Fig. 8A describes various potential energy curves for a molecule* This designation implies that the high temperature treatment is degassing at 250°C for 12 h. t On a carbon with longer degassing time the C-1200-1 h adsorption of both CH,Cl and CH3CNceased completely while that of hydrogen continued. Prior to degassing at 1200°C both C-1200-1/4 hand C-1200-1 h were degassed at 1000°C for 1 h to assure the removal of surface oxygen groupsJ . KORESH AND A . SOFFER0.157 0.14n3 i2 E248 1--AA3I I I 1 1 1 1 I20 60 100 140 180t,lSFIG.5.-Adsorption rates of CH3C1, CH3CN and H2 on the non-polar carbon C-1200-1/4. @, H2,V, = 780pmol min ', rJld = -196 C; ., CH,CN, V, = 260,umol min-', T,d = 24'C; A, CH3C1,V, = 2.8 pmol min TId = -80 C.between two opposite walls of a pore. Our view is that the pores of an MSC aregenerally not dimensionally homogeneous so that constrictions are responsible formolecular sieving. Thus the distance between the walls in fig. 8A corresponds tothat existing between the constrictions in fig. 8B. The potential minima existingwhen the walls are far apart coincide with a lower minimum when they are close toeach other so that the attractive portions of the molecule for single wall interactions' AAAAAA810 30 50 701 :minFIG. 6.---Adsorption rates of CH3Cl and H2 on polar C-250-12h.A, CH3C1, V, = 31.7 pmol min-',7&,, = -80 C; @, H2, V, = 2.1 pmol min-l, Kd = - 196 C.1-72482 MOLECULAR SIEVE CARBONSAA5 10 15tlminFIG. 7.-Adsorption rates of CH3CN and COz on the polar carbon C-100. A, COz, Vo = 8.23pmolmin-', T,d = -80°C; 0, CH,CN, Vo-85.7 pmolmin-l, Kd = 24°C.overlap. As the walls come closer to each other, the minimum is raised due to theoverlap of the repulsive portions until the pore becomes impermeable. In contrastto fig. 8, fig. 9 describes the changes in potential energy when an adsorbate movesalong a pore with varying widths starting from a very wide aperture with twoseparate minima.The abscissa in fig. 9 is the adsorption coordinate which is moreor less parallel to the walls. In this figure two types of activation of passage throughconstrictions are evident : desorption activation [column (b)] and normal activationEI IIFIG. 8.-A, Potential energy E of a molecule of a diameter 2ro as a function of distance (d) between thewalls; E is the minimum potential for d 4 00. B, Lowest energy positions of the molecule in a pore withconstrictionJ. KORESH AND A . SOFFER 2483AE 0 -e2s -( a > ( b ) ( C ) ( d 1FIG. 9.-A, Potential energy E as a function of the adsorption coordinate into a pore of varyingdimensions. B, Corresponding lowest energy positions along the coordinate; dashed line, minimumenergy pathway (adsorption coordinate); (a) walls far apart compared with yo, unactivated adsorption(d > ro), (b) desorption activation from the pore (d = yo), (c) normal activation through a constriction(d < yo), (d) constriction practically impermeable (d < yo).[column (c)].In fig. 10(a) we added an attractive potential to the drawing offig. 9(d), which represents an impermeable pore, in order to account for a changefrom a non-polar to a polar surface. The resultant potential shown in fig. lO(c) has aconsiderably lower activation energy so that the adsorbate molecule can penetrateinto the pore. Such a change probably shifts the whole scale of adsorbabilitytowards smaller pores but does not eventually severely affect the sequence as longas the attractive potential added is of the same type for all adsorbates.This may bethe case if all adsorbates were non-polar and a dipole-induced-dipole attraction isadded. [Numerical estimates of widths and lengths as those given in the last twocolumns of table 1 also remain unaffected as long as the adsorbability sequence isnot changed. Moreover, molecular dimensions are scaled primarily by a property ofthe pure adsorbate (liquid molar volume) whereas the sequence is introduced onlyEI I IFIG. 10.-Interaction energy plotted against adsorption coordinate of an adsorbate in a pore near aconstriction: (a) for a non-polar surface as reproduced from fig. 9(d), (b) additional attraction due tosurface polarity, (c) resultant curve for polar surface2484 MOLECULAR SIEVE CARBONSthrough shape corrections.] If, however, a polar molecule is examined against adimension sequence of non-polar molecules, the dipole-dipole potential added isconsiderably greater and will therefore shift the location of the molecule towards asmaller dimension in the non-polar molecule sequence.This is eventually the casewith HCl, HBr, CH3C1 and CH3CN.The unexpectedly large dimension of the hydrogen molecule (table 1) demon-strates intermolecular forces behaving in a way opposite to that in polar molecules.Its interaction with the surface is very weak compared with other non-polar mol-ecules of considerably higher polarizabilit y. This molecule cannot, therefore, comeas close to the walls as other molecules do.SIGNIFICANCE OF MOLECULAR DIMENSIONSMolecular dimensions are not of a well defined magnitude.This is mainlybecause the electron distribution function +b2 ( X , Y, 2) is assymptotic and vanishesonly at infinity. The distance ro for which the intermolecular potential functionbecomes zero can provide an adequate definition. This value depends, however, onthe nature of the second body which interacts with the molecule studied and itsinfluence cannot be neglected, at least as far as polar molecules are concerned. Ourview is that the dependence of molecular dimensions on environment has to beconsidered a reality so that the above definition must be adopted. In this sense, thetransfer of adsorbate molecules through MS pores provides a means of arrangingmolecules according to their width, while the influence of environment (walls) isalready taken into consideration. The question now is whether different series ofmolecular dimensions have to be expected for each set of experimental conditions.This is not necessarily so, since it has been shown above that only the dimensionsof very polar molecules change significantly according to the polarity of the walls.One could therefore deduce with certainty that the sequence given in table 1 isgenerally adequate for non-polar molecules in either polar or non-polar environ-ments. The term “environment” may in turn be generalized to biological mem-branes, thin-layer films, diffusive solids, gas mixtures etc.CONCLUDING REMARKSThe ability to change continuously the average pore dimension of MSC enablesThe change of surface polarity of the carbon makes it possible to demonstrate theThe molecular dimensions of spherical molecules in the liquid phase as a stan-one to order different molecules according to their smallest dimension.influence of environment on molecular dimensions.dard can provide numerical values for molecular width.REFERENCESJ.Koresh and A. Soffer, J.C.S. Farnday I, 1980,76,2457.2S. Brunauer, P. H. Emmett and E. Teller, J . Amer. Chem. Soc., 1938, 60, 309.3L. Pauling, Nature of the Chemical Bond (Cornell Univ. Press, Ithaca N.Y., 3rd edn, 1960).4A. N. Ainscough, D. Dollimore and G. R. Heal, Carbon, 1973, 11, 189.6T. Kihara, J . Phys. SOC. Japan, 1951, 6, 297.7J. Corner, Proc. Roy. SOC. A, 1948, 19% 275.‘A. Isihara, J. Chem. Phys., 1950, 18, 1446; A. Isihara and T. Hayashida, J. Phys. SOC. Japan, 1951, 6,9J. 0. Hirshfelder, Molecular Theory of Gases and Liquids (John Wiley, N.Y., 1964), p. 206.D. W. Breck, Zeolite Molecular Sieves (John Wiley, N.Y., 1974), pp. 634-6.40, 46J . KORESH AND A . SOFFER 2485'OR. M. Barrer, Quart. Rev., 1949, 3, 293; T. G. Lamond, J. E. Metcalfe I11 and P. L. Walker. Carbon,l 1 D. W. Breck, Zeolite Molecular Seives (John Wiley, N.Y., 1974), p. 640.l 2 J. R. Dacey and D. G. Thomas, Trans Faraday SOC., 1954, 50, 740.I3S. S. Barton, M. J. B. Evans and B. H. Harrison, J . Colloid Interface Sci., 1974, 49, 462.14M. M. Dubinin, E. D. Zaverina, A. T. Kaverov and Kasatochkin, Izv. Akad, Nauk S.S.S.R., Otdel.15F. A. Maggs, Nature, 1952, 169, 793.I6M. M. Dubinin, 0. Kadlec, 1. Botlik, E. 0. Zaverina, A. Zukal and B. Sumec, Dokludy Adad. Nauk,17M. M. Dubinin, 0. Kadlec and A. Zukal, Nature, 1965, 207, 75.I'D. W . Breck, J . Chem. Ed., 1964, 41, 678."S. S. Barton, D. Gillespie and B. H. Harrison, Carbon, 1973, 11, 649; S. S. Barton and B. H. Harrison,'OD. H. Everett and J. C. Powel, Trans Faradaji SOC., 1976, 72, 619.1965, 3, 59.khim. Nuuk. V , 1961, 1, 29.S.S.S.R., 1964, 157, 656.Carbon 1975, 13,47.(PAPER 9/1843

ISSN:0300-9599    

DOI:10.1039/F19807602472

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980, 76,2486-2495Theory of Adsorption from Multicomponent Liquid Mixtureson Solid Surfaces and its Application toLiquid Adsorption ChromatographyBY MIECZYSLAW JARONIEC:: AND ANDRZEJ PATRYKIEJEWDepartment of Theoretical Chemistry, Institute of Chemistry,M. Curie-Sklodowska University, 2003 1 Lublin, PolandReceived 24th September, 1979The theory of adsorption from multicomponent liquid mixtures on energetically homogeneousand heterogeneous solid surfaces has been developed using the prevailing approaches to this problemand utilizing the theory of mixed-gas adsorption. The general isotherm equations have been derivedand applied to evaluate the distribution coefficient characterizing the process of liquid adsorptionchromatography with a multicomponent mobile phase.In contrast to the extensive literature concerning the adsorption of binary liquidmixtures on solid adsorption from multicomponent liquid mixtureshas been comparatively little studied.Theoretical studies of adsorption from multi-component liquid mixtures on solids have usually been related to the region of lowconcentrations. 2* Investigations of multicomponent liquid-solid systems, in thewhole concentration region, have been made by Oicik and Minka and Myers8They considered the adsorption on energetically homogeneous surfaces only.Recently, their results have been extended to adsorption on heterogeneous solidsurfaces 9 9 lo by applying the thermodynamic treatment of Jar0niec.l'. l 2In this paper some aspects of adsorption from multicomponent liquid mixtures onhomogeneous and heterogeneous solid surfaces are discussed.Different adsorptionmodels assuming ideality as well as non-ideality of both phases have been considered.The results of the theoretical considerations are analytical equations for the molefractions of components in the surface phase. These equations have been appliedto evaluate the distribution coefficient of a chromatographed substance in a multi-component mobile phase.ADSORPTION FROM MULTICOMPONENT LIQUID MIXTURESHOMOGENEOUS SURFACESLet us consider adsorption from an n-component liquid mixture. The followingassumptions are made : (a) the surface phase is supposed to be a monolayer, (b) bothsurface and bulk phases will be considered as ideal or non-ideal, (c) the molecularsizes of all components are identical, (d) the total number of molecules in the surfacephase is constant and (e) the adsorbent surface is energetically homogeneous.Theabove assumptions are frequently used in the theory of adsorption from solutions onsolids.Let the adsorption energy of the nth component be smallest. Then the adsorptionmechanism may be represented by the series of quasichemical reactions between a248M. JARONIEC AND A . PATRYKIEJEW 2487molecule of the ith component (i = 1,2, . . ., n- 1) and a molecule of the nth com-ponent :(i)' + (n)s + (i)s + (n)' (1)where the symbols i and n denote molecules of components i and n, respectively, andthe superscripts 1 and s refer to the bulk and surface phases.The equilibrium constantfor reaction (1) isa;.:Kin = - for i = 1 , 2 , . . .,n-1anaiwhere a] and a; are the activities of the ith component in the bulk and surface phases,respectively. The activities a: and a: are defined as follows :and(3)(4)wheref: andfi are the activity coefficients of the ith component in the bulk and surfacephases and x] and xi' are the mole fractions of the ith component in the bulk andsurface phases, respectively. Eqn (2) may be rewritten in a slightly different form,i.e.,wherex:/x,S = Kinflinx)x:, for i = 1 , 2 , . . ., n-1 ( 5 )Summing the mole fractions xi, xi, . . ., xi-l, the following equation may beobtained :, for i = 1,2 ,..., n-1 KinPinxVxX1 + C KjnBjnxf/xtn - 1 x; =j = 1and(7)Evidently the parameter P i n is a function of the composition of both phases.Eqn (7)with pin = 1 has been derived by Minka and Myers * for the description of theadsorption from liquid mixtures on homogeneous surfaces, when both phases areideal. Next, eqn (7) with Pin = 1 was derived by Jaroniec et aL4 in terms of thestatistical thermodynamics.CONCEPTION OF SURFACE HETEROGENEITY I N ADSORPTIONFROM LIQUID MIXTURESGeneralizing Jaroniec's treatment of the adsorption from multicomponent gasmixtures on heterogeneous solid surfaces " 9 l 2 to liquid adsorption we obtain :nxtt = x:(x', E ) F(E) dE, for i = I, 2, . . ., nJ A(92488 ADSORPTION FROM MULTICOMPONENT LIQUID MIXTURESwhere E = (El, E2, . . ., En) and x 1 = (xi, x$, . . ., xL- 1) are vectors, El is the adsorp-tion energy of the ith component for a heterogeneous surface, A is an n-dimensionalintegration region of E and F(E) is an n-dimensional distribution function of E,which is normalized to unity, i.e.,l A F ( E ) dE = 1.(10)The symbol $(xl, E ) denotes the mole fraction of the ith component in the surfacephase for adsorption sites characterized by E. In this formulation, each adsorptionsite is characterized by an n-dimensional vector E. The adsorbent surface is ener-getically heterogeneous if it contains adsorption sites of different values of E.In eqn (9) the mole fraction $(XI, E ) may be evaluated according to eqn (7). Theconstant Ki, is equal t o :where A i , is the entropy factor. The mole fraction xf, expressed by eqn (7), is thefunction of differences of adsorption energies Ei, (i = 1, 2, , .., n - 1). This factenables us to introduce a new distribution function characterizing the surface hetero-geneity in adsorption from solutions.Let us characterize each adsorption site by an ( n - 1)-dimensional vector E* =(El,, E2,, . . ., The surface is energetically heterogeneous if it containssites of different values of E*. Similarly as in the case of eqn (9), an (n- 1)-dimen-sional distribution function G(E*) has been introduced to characterize the energeticheterogeneity of the adsorbent surface. Thus, the mole fraction x$ may be expressedas follows :x!,~ = xi(xl, E*) G(E*) dE*, for i = 1 , 2 , . . ., n (12)G(E*) dE* = 1. (13)s. whereIn the above, R is an (n - 1)-dimensional integration region of E*.Eqn (12) maybe obtained from eqn (9) by assuming the following definition of the function G(E*) :G(E*) = JF(E)dE,. (14)These conceptions of surface heterogeneity in adsorption from solutions are notequivalent. For illustrative purposes we consider the adsorption from a binaryliquid mixture. Let us assume a homogeneous surface characterized by the functionwhere 6 is the Dirac function. According to the second conception, this surface isalso characterized by the Dirac function :F(E1, E2) = F&?31)-F&) = 6(E1 -ET)*6(E, - E ; ) (1 5 )W 1 2 ) = 6(&2 -EL). (16)In the case when the functions Fl(El) and F2(E2) have identical shape but areshifted on the energy axis, the difference of adsorption energies, E12, may be identicalfor all adsorption sites.13 It might even happen that the adsorbent surface appearsto be heterogeneous according to the first conception and quite homogeneous accord-ing to the secondM.JARONIEC AND A . PATRYKIEJEW 2489In adsorption of binary liquid mixtures on heterogeneous solid surfaces, eqn (12)has usually been 4 9 ' 9 lo The cited papers deal with the theoretical considera-tions of adsorption models assuming ideality of both phases or ideality of the surfacephase and non-ideality of the bulk phase. Then the topography of adsorption siteson the surface is not important.12 However, in the case of a non-ideal surfacephase the distribution of adsorption sites on the surface should be taken into con-sideration.As with the case of mixed-gas ad~orption,~. l2 with adsorption fromliquid mixtures two models of hererogeneous surface are considered : (a) a modelwith a patchwise distribution of adsorption sites and (b) a model with a randomdistribution of sites on the surface. For patchwise surfaces the activity coefficientf:is a function of the composition of the solution adsorbed on a given patch and thecomposition of the bulk liquid, i.e.,j f = f i ( x s , x') where xs = (x;, x i , . . ., x:-~). (17)However, for surfaces with a random distribution of adsorption sites the activitycoefficientf: is a function of the composition of the solution contained in the wholesurface phase and the composition of the bulk liquid, i.e.,Thus, for random surfaces the parameter Pin, appearing in eqn (7), is not dependenton E*.The integration in eqn (12) and (7) is formally identical for adsorption modelsassuming: (a) ideality of both phases, (b) ideality of the surface phase and non-ideality of the bulk phase and (c) non-ideality of both phases and an adsorbent surfacewith random distribution of adsorption sites. In the case of patchwise surfaces eqn(7) is a complex function of xi and integration of eqn (12) is very difficult.f: = f f ( x i , x') where xi = (xi,,, xi,,, . . ., xi- l,t). (1 8)ANALYTICAL EQUATIONS FOR MOLE FRACTION,Let us consider a liquid mixture in which (n- 1) components show similar inter-actions with the adsorbent surface and the behaviour of the nth component is com-pletely different.Such a model of adsorption has been considered for adsorptionfrom gaseous l4 and liquid mixtures.l* Then, the diffrence of adsorption energies,Eil = &-El, = ui = constant (i = 1, 2, . . ., n - l), is identical for the wholesurface. Taking into account this relationship and assuming the random distributionof adsorption sites on the surface, we can transform eqn (7) into the following form :wheren- 1andwher2490 ADSORPTION FROM MULTICOMPONENT LIQUID MIXTURESEqn (22) is formally identical with the integral equation used in the adsorption ofgases l 5 and binary liquid mixtures 3* on solids. Thus, the solution of eqn (22)is analogous to the solutions of the integral equations used in gas and liquid adsorp-Now, we derive equations for x$ by means of eqn (22) for two quasi-gaussianenergy distributions GI n(Eln). These distributions give equations for ~ f , ~ of thetype of the T6th's l 6 and Sips' l 7 isotherms.In the case of T6th's distribution thefollowing equation is obtained :tion.3. 4 9 15Eqn (24) leads to the following expression for x& :The constant c in the above equations denotes the heterogeneity parameter character-izing the shape of T6th's distribution function and the parameter KTn is connectedwith the average difference of adsorption energies, Eln.For the distribution function obtained by Sips l7 we obtainandwhere d is the heterogeneity parameter. Eqn (24)-(27) have been derived by assuminga random distribution of adsorption sites on the surface and constancy of the differ-ences of adsorption energies, ie.,and for all types of adsorption sites.to adsorp-tion from multicomponent solutions on solid surfaces of random distribution ofadsorption sites gives :Eil = ui = constant, for i = 1,2,.. ., n-1 (28)Application of the approximation of Crickmore and Wojciechowskiwhere w is the heterogeneity parameter. This equation is an extension of the adsorp-tion isotherm discussed in ref. (9). Eqn (25) and (29) will be used to derive thedistribution coefficient for liquid adsorption chromatography with a multicomponentmobile phase.LIQUID ADSORPTION CHROMATOGRAPHYThe chromatographic process for the rth substance in the n-component mobilephase relates to the adsorption from the (n+ 1)-component liquid mixture, when theconcentration of the rth substance is infinitely low.Let us assume that the 1stcomponent of a mobile phase is a most efficient eluting solvent. The chromato-graphic process may be represented by the following exchange reaction :( 1 y + ( r y + (1)' (30M. JARONIEC AND A . PATRYKIEJEW 249 1and the reactions represented by eqn (1). The equilibrium constant Krl for theexchange reaction (30) is expressed bywhere Prl is defined analogously as in eqn (6) ; however,n ni = 2 i = 2andn nSince the mole fractions x: and x: are infinitely low, the factor Prl is a function ofx1 and xs only, i.e., it is independent of x: and x:.According to the theory of adsorption chromatography,l the ratio of the molefractions x; and x: is equal to the distribution coefficient k, :k, = x:/x:.(34)(35)A simple combination of eqn (31) and (34) gives :Eqn (35) is a general expression defining the distribution coefficient of the rth sub-stance in the n-component mobile phase when non-ideality of both mobile and surfacephases is assumed. The mole fraction xs, appearing in eqn (35), may be calculatedaccording to eqn (7).kr = Kr 1 x; /(xi P r 1 )*For Prl = 1, eqn (35) reduces to the following expression :20kr = Kr1x;/x: (36)and is related to the model based on the assumption of ideal surface and mobilephases. Then, assuming that Pln = P2n = . , . = Pn-l,n = 1, the mole fraction x;,evaluated from eqn (7), becomesNow, we shall define the equilibrium constant Kin by means of the distributionFollowing Minka and Myers * we can write the relationship : coefficients kri and krn.Kin = KirIKnrwhereBecause the concentration of the rth substance is infinitely low, the mole fractionsxl and xf are very close to unity.Then,where kri is the distribution coefficient in the ith solvent. Analogously,Ki, = I / & = l/kri (40)Knr = l/Krn = 1 /krn.Eqn (38), (40) and (41) give :Kin = krdkri2492 ADSORPTION FROM MULTICOMPONENT LIQUID MIXTURESSubstituting eqn (42) into eqn (37) we obtainHowever, combining eqn (36), (41) and (43) we have :nIlkr = C x;/kri.i = 1(44)Eqn (44) may be also obtained in terms of Snyder's theory of adsorption chromato-graphy.,ISnyder,l Soczewinski 22 and Jandera and Churacek 23 considered mobilephases in which the 1st solvent was considerably more polar than other solvents.Then, for higher values of xi the adsorbent surface is practically occupied by the 1stsolvent, i.e.,Taking into account condition (49, eqn (26) becomesxi % 1.(45)orIn k, = In krl -In x i .I n kr = In K , , -In u: +In f:.(47)However, applying condition (45) to eqn (35) and assuming ideality of the surfacephase, we get :Eqn (48) has been also derived by Slaats et aZ.24 by using another theoretical approach.Eqn (35) is the most general equation in our formulation of liquid adsorptionchromatography with an n-component mobile phase and with a homogeneous solidsurface. The main relationships used in a liquid adsorption chromatography maybe obtained from eqn (35).Moreover, applying analytical expressions for the activitycoefficientsf: andf: * * 2 5 in eqn (39, new equations for the distribution coefficient krcan be obtained.(48)HETEROGENEOUS SURFACESLet us consider a heterogeneous surface showing rn types of adsorption sites.Let g p be the ratio of the number of adsorption sites of thepth type to the total numberof adsorption sites. The ratios g , , g,, . . ., g , satisfy the conditionf g p = l .p = l(49)First, we shall discuss the distribution coefficient k, for a heterogeneous surfaceand ideal surface phase (the mobile phase may be either ideal or non-ideal). Inthis case, the topography of adsorption sites on the surface is not important.12According to eqn ( 3 9 , the distribution coefficient of the rth substance on the pthtype of adsorption sites is given by :wherM.JARONIEC AND A . PATRYKIEJEW 2493The distribution coefficient of the rth substance on the entire heterogeneous surface,kr,t, is defined as follows :Using the approximation discussed by Jaroniec et ~ 1 . ~ ~ in eqn (52) we obtainwhere K:1 is an averaged distribution coefficient for the rth substances in the 1stsolvent, referring to the entire adsorbing surface, w is the heterogeneity parameter,analogous to that introduced in eqn (29).Now, we consider a special case of eqn (53), related to ideal surface and mobilephases, i.e.,k , , = ccKrolx~,t)"wlix: -In this case eqn (29) may be rewritten in the following form :(54)(x,'/k,",)"i = 1when the relationship [see eqn (42)] :is satisfied.Substituting eqn (55) into eqn (54) and remembering eqn (56), we obtainFor w = I, eqn (57) reduces to eqn (44), which was obtained assuming energetichomogeneity of the solid surface.Now, we consider the adsorption model with a non-ideal surface phase. Accord-ing to eqn (17) and (18), the variable Prl depends on the properties of an adsorbentsurface.For a patchwise distribution of adsorption sites this variable is a function of thecomposition of the bulk solution and molecules adsorbed on the pth surface patch.However, for a random distribution of adsorption sites on the surface, the variableP r l is a function of the composition of the bulk and surface solutions. Thus, thedistribution coefficient kr,t is given by :mkr,t = (l/xi) gpKrl,pxF;,p/Pr,,p for a patchwise surface (58)p = landmp = lkr,t = (l/x:flr1) gpKrl,px(il,p for a random surface.(59)Using the approximation used in eqn (53) with the sum appearing in eqn (59), weobtain :Eqn (60) describes the distribution coefficient of the rth substance in the n-componentnon-ideal mobile phase and for a non-ideal surface phase formed on a heterogeneoussurface with a random distribution of adsorption sites2494 ADSORPTION FROM MULTICOMPONENT LIQUID MIXTURESCONCLUSIONSThe equations used in adsorption chromatography refer to very simple adsorptionmodels. These equations are special cases of more general expressions eqn (35),(52), (58) and (59), which have been derived in terms of the theory of adsorption frommulticomponent liquid mixtures on either energetically homogeneous or heterogeneoussolid surfaces.Equations defining the distribution coefficient of the rth substance in the n-component mobile phase contain the mole fractions of solvents in the surface phase.These mole fractions may be determined : (a) from analytical equations correspondingto a given adsorption model or (6) directly from experimental excess adsorption data,using the following relationship :where ns is a total number of moles in the surface phase and ni is an adsorption excessof the ith solvent.The first procedure is useful for determining the influence of the adsorption modelon the distribution coefficient.For some adsorption models the equations for thedistribution coefficient are very simple.The other procedure, using the excessadsorption data, is more useful for the description of experimental data obtainedfrom liquid adsorption chromatography.x: = nr/ns++! (61)LIST OF PRINCIPAL SYMBOLSactivityentropy factor in eqn (1 1)heterogeneity parameter in eqn (24)heterogeneity parameter in eqn (26)activity coefficientdistribution function of Eadsorption energydifference of adsorption energies Ei and Enfraction of adsorption sites of the pth typedistribution function of E*distribution coefficient of the rth substance in a mixed mobile phase on ahomogeneous surfacedistribution coefficient of the rth substance for a heterogeneous surfacedistribution coefficient of the rth substance in the ith solventequilibrium constant for a quasi-chemical reaction (1)number of types of adsorption sitesexcess adsorption isotherm of the 1st solventtotal number of moles in the surface phasemole fractionheterogeneity parameter in eqn (29)parameter defined by eqn (6)VECTORM.JARONIEC AND A. PATRYKIEJEW 2495SUBSCRIPTSi the ith componentn the nth componentp thepth type of adsorption siter the rth chromatographed substancet refers to a heterogeneous surfaceSUPERSCRIPTS1 mobile (bulk) phases surface phase(a) D. H. Everett, Trans. Faraday Soc., 1964, 60, 1803 ; (b) D. H. Everett, in Colloid Scienceed. D. H. Everett (Specialist Periodical Report, The Chemical Society, London, 1973), vol. 1,chap. 2.C. E. Brown and D.H. Everett, in Colloid Science, ed. D. H. Everett (Specialist PeriodicalReports, The Chemical Society, London, 1975), vol. 2, pp. 52-100.(a) J. OScik, A. Dqbrowski, M. Jaroniev and W. Rudzinski, J. Colloid Interface Sci., 1976,56,403 ; (b) A. Dqbrowski, J. OScik, W. Rudzifiski and M. Jaroniec, J. Colloid Interface Sci.,1979, 69, 287.M. Jaroniec, A. Patrykiejew and M. Borowko, in Progress in Surface and Membrane Science(Academic Press, New York, 1980), vol. 14.C. J. Radke and J. M. Prausnitz, Amer. Inst. Chem. Eng., 1972, 18, 761.L. Jossens, J. M. Prausnitz, W. Fritz, E. U. Schlunder and A. L. Myers, Chem. Eng. Sci.,1978,33, 1097.J. OScik, Bull. Acad. Pol. Sci., Cl. 3, 1961, 9, 23, 29.C. Minka and A. L. Myers, Amer. Inst. Chem. Eng., 1973, 19,453.M. Jaroniec, J. Res. Inst. Catalysis, Hokkaido Univ., 1978, 26, 155.lo M. Borowko, M. Jaroniec, J. OScik and R. Kusak, J. Colloid Interface Sci., 1979, 69, 311.l1 (a) M. Jaroniec, J.C.S. Faraday 11, 1977, 73, 933 ; 1978, 74, 1292 ; J. Colloid Interface Sci.,1975, 53, 422 ; 1977, 59, 230, 371 ; (b) M. Jaroniec and W. Rudzinski, J. Res. Inst. Catalysis,Hokkaido Univ., 1977, 25, 197.l2 M. Jaroniec, Thin Solid Films, 1978, 50, 163.l3 M. Jaroniec and J. Toth, Colloid and Polymer Sci., 1976, 254, 643.l4 M. Jaroniec, J. Narkiewicz and W. Rudzifiski, J. Colloid Interface Sci., 1978, 65, 9.l5 M. Jaroniec, Surface Sci., 1975, 50, 553.l6 J. Tbth, W. Rudzinski, A. Waksmundzki, M. Jaroniec and S. Sokolowski, Acta Chim. Acad.l7 R. Sips, J. Chem. Phys., 1950, 18, 1024.l8 P. J. Crickmore and B. W. Wojciechowski, J.C.S. Faraday I, 1977, 73, 1216.l9 L. R. Snyder, Principles of Adsorption Chromatography (Marcel Dekker, New York, 1968).2o M. Jaroniec, J. K. R6iylo and B. OScik-Mendyk, J. Chromatog., 1979, 179, 237.21 M. Jaroniec, J. Narkiewicz and M. Borowko, Chromatographia, 1978, 11, 581.22 E. Soczewinski, J. Chromatog., 1977, 130, 23.23 P. Jandera and J. Churacek, J. Chromatog., 1974, 91, 207.24 E. H. Slaats, J. C. Kraak, W. J. T. Brugman and H. Pope, J. Chromatog., 1978,149,255.25 A. S. Jordan, J. Electrochem. Soc., 1972, 119, 123.26 M. Jaroniec, J. K. R6zylo and W. Golkiewicz, J. Chromatog., 1979, 178, 27.Sci. Hung., 1974, 82, 11.(PAPER 9/1514

ISSN:0300-9599    

DOI:10.1039/F19807602486

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980,76,2496-2506Thermodynamics of Liquid Mixtures of Nitrous Oxide and XenonBY JosB R. S. MACHADO AND KEITH E. GUBBINSSchool of Chemical Engineering, Cornell University,Ithaca, New York 14853, U.S.A.AND LELIO Q. LOBO AND LIONEL A. K. STAVELEY*Inorganic Chemistry Laboratory, University of Oxford, Oxford OX1 3QRReceived 1st November, 1979The total vapour pressure, excess volume and excess enthalpy of the system nitrous oxide + xenonhave been measured as a function of composition ; the vapour pressure and excess volume measure-ments are at 182.32 K (the triple-point of N20) and those for the excess enthalpy are at 184.05 K.The vapour pressure results have been used to estimate the excess Gibbs energy. The mixture exhibitsa positive azeotrope at a N,O mole fraction of = 0.08.The experimental results are compared withvalues calculated from perturbation theory for non-spherical molecules, using an intermolecular poten-tial model that includes dipolar and quadrupolar electrostatic terms. Agreement is good for the pro-perties of both the pure fluids and the mixture and is a substantial improvement over theories thatneglect the acentric nature of the intermolecular forces.This paper is one of a series devoted to an examination of the effects of electro-static forces and molecular shape on the thermodynamic properties of binary liquidmixtures of which one constituent is polar. The experimental values of the primaryexcess thermodynamic functions are compared with those calculated from the per-turbation theory developed by one of us (K.E. G.) and his collaborator^.^'^ Twoprevious papers dealt with the systems xenon + hydrogen chloride, xenon + hydrogenbromide and hydrogen chloride + hydrogen bromide 4 * and hence were largelyconcerned with the influence of a dipole in the polar species. Here we present theresults of an experimental and theoretical study of the system nitrous oxide(dinitrogen oxide, N,O)+xenon, selected to serve as a model for mixtures of thetype quadrupolar molecules + non-polar molecules. It was clearly desirable tochoose as the polar component a compound with a relatively large quadrupolemoment. A suitable choice might seem to be carbon dioxide, but the relatively hightriple-point temperature and pressure of this substance rule it out on practical grounds,at least as far as studies of a mixture with a rare gas are concerned.The moleculeof nitrous oxide has a quadrupole moment ( Q = -3.65 x e.s.u. cm2 =- 12.2 x( Q = -4.3 x C m2) and though it has a dipolemoment this is so small ( p = 0.166 D = 0.55 x C m) that one would expectthat its influence on the properties of the N,O + Xe system would be negligible.C m2) which is not much smaller than that of carbon dioxidee.s.u. cm2 = - 14.3 xEXPERIMENTALThe experimental techniques used to measure the excess Gibbs energy of mixing GE, thevolume of mixing VE and the enthalpy of mixing HE have already been described.8-10 Thevapour pressures were measured with a Texas Instruments quartz spiral gauge, which hadbeen calibrated against a wide bore mercury manometer and a dead-weight piston gauge.249MACHADO, GUBBINS, LOB0 A N D STAVELEY 2497The pyknometer used for the volume determinations had been calibrated using a 99.96 molpercent sample of ethane and the results of Haynes and Hiza." In the heat of mixingexperiments, temperature was measured with a copper resistance thermometer which hadbeen calibrated by measuring the vapour pressure of ethane, using the vapour pressureequation of Goodwin et u1.l' The total volume of the upper compartment of the calori-meter, where N20 was condensed, was 3.16 cm3 ; the lower chamber, into which the xenonwas condensed, had a capacity of 8.66 cm3.GE and VE were determined at the triple-point of N20 (182.32 K) and H E at a slightlyhigher temperature (184.05 K), since attempts to measure H E at the triple-point of one ofthe two components might present experimental difficulties owing to the partial solidificationof that component.The xenon used was research grade quality, of purity > 99.995 mol percent.Thenitrous oxide was taken from a cylinder (medical grade ; at least 99.0 mol percent) andfractionated in the laboratory low-temperature c o l ~ m n . ~ The purity of the sample usedwas checked by the constancy of its triple-point pressure, for which the value of (87.866f0.001) kPa was obtained (mean value of six determinations, as measured directly with amercury manometer ; cf. 87.864 l 3 and 87.853 kPa 1 4 9 15). Our measured value of thevapour pressure of liquid xenon at the triple-point of NzO was 247.742 kPa.The valuesinterpolated from the results of Michels and Wassenaar l 6 and Theeuwes and Bearman l 7are 247.434 and 247.306 kPa, respectively. The differences between these figures and oursare covered by a difference of w 0.03 K in the temperature scales.The average of six determinations of the molar volume of liquid N20 at its triple-pointwas (35.487k0.002) cm3 mol-1 (cf. 35.80 l 5 and 35.46 cm3 mol-l 18). Three determina-tions of the molar volume of xenon at the same temperature gave (46.453+0.010) cm3 mol-1(cf. 46.409 l7 and 46.534 cm3 mol-1 19).In the evaluation of C", VE and H E from the actual experimental measurements, valuesfor certain physical properties must be adopted and it will be convenient to summarize theseand to indicate their source.The value of 182.26 Kreported by Blue and Giauque l4 was estimated ' ' 9 'l to be equivalent to 182.32 K onIPTS-68. B forNzO at room temperature, which is required in assessing the quantity introduced into thepyknometer (or calorimeter), was taken from the work of Couch et dZ2 and Schampet ~ 1 .~ ~ The value so obtained for 298.15 K was -133.2 cm3 mol-I. B for N20 at182.32 K was calculated to be -400 c1n3 mol-1 from the equation of Pitzer andThe acentric factor co was estimated as 0.160 from the vapour pressure data of Couch et al."and the critical constants listed by mat hew^.'^ The necessary B values for xenon weretaken from Brewer's report.26 For mixtures of the two gases it was assumed that the crossvirial coefficient BIZ is the arithmetic mean of those of the two pure components.For HE (for the use made of the following quantities, see Lewis et ~ 1 .' ~ ) : (1) We usedour own values of the molar volumes of the pure liquids, of the volume change on mixingand of the vapour pressures of the pure components at the temperature of the H E determina-tion. The molar volume VG of gaseous NzO at this temperature was estimated by correctingfor gas imperfection as far as the second virial coefficient. VG for xenon was interpolatedfrom the results of Streett et ~ 1 . ~ ~ The composition of the vapour phase in the calorimeterafter mixing was calculated from our Redlich-Kister equation for GE (vide infva).(2) The molar enthalpy of vaporization AHv of NzO at 184.05 K was estimated as 16.59 kJmol-1 from the value of 16.56 kJ mol-1 found by Blue and Giauque l4 at the normal boilingpoint.AHv for xenon was interpolated from table 6 of the paper of Streett et ~ 1 . , ~ * thevalue so obtained being 11.93 kJ mol-l. (3) The coefficient of expansion a, of NzO at itssaturation vapour pressure was derived from the results of Leadbetter et u1.,l5 while thatfor xenon was interpolated from table 7, ref. (28). The values obtained were a,(NzO) =2.1 x K-l and a,(Xe) = 2.5 x lW3 K-l at 184.05 K. (4) The isothermal compressibilityKT of liquid NzO was calculated to be 0 . 9 ~ MPa-' from the approximate relation 29KT w aTV/(AH,-RRT). For Xe, a value of 2.3 x lop3 MPa-l was interpolated fromtable 7, ref.(28).For GE and VE : (1) The triple-point temperature of NzO.(2) Second virial coefficients (B), (also needed in the H E calculations)2498 THERMODYNAMICS OF N20+XeRESULTSThroughout this paper, nitrous oxide is designated 1 and xenon 2. Table 1 givesour results for the total vapour pressure P as a function of xl, the liquid mole fractionof N20. The system is markedly non-ideal, forming a positive azeotrope which isnot far from being a tangent azeotrope. GE was evaluated by Barker's method,30minimizing the pressure residuals RP = Pexp -Pcalc. The values in table 1 of yl, themole fraction of N20 in the vapour in equilibrium with the liquid mixture, are calcu-lated. A three-term Redlich-Kister equation was found to be adequate for GE,namelywith A = 1.1829 (aA = 0.0015) ; B = 0.0532 (a, = 0.0028) ; C = 0.0425 (ac =0.0056), the CJ being the standard deviations of the parameters.G: = (448.2k0.6)J mol-I.GE/RT = x,(l- xl)[A + B(2~1- 1) + C(2~1- 1)2] (1)TABLE ~.-VAPOUR PRESSURE AND EXCESS MOLAR GIBBS ENERGY OF THE SYSTEM NITROUSOXIDE XENON (2) AT 182.32 K. Rp = PeXp-~,,1,.X1 Y1 P/kPa Rp/Pa GE/J mol-'0.000 000.092 840.191 380.309 940.438 390.566 210.751 560.855 720.922 731.000 000.000 000.090 630.154 570.207.970.252 1 30.295 030.381 320.480 450.607 101 .ooo 00247.742249.539246.745240.267231.168218.731191.942163.939136.37987.866-841- 56172223- 284- 92- 17-0149.6274.3378.7436.3433.2335.9226.4113.70The molar volumes of mixtures of known composition and the derived values ofVE are recorded in table 2.These values of VE refer to mixing at the saturationvapour pressure. (The difference from the values at zero pressure is negligible.)They fit the equationwith D, E and F, respectively, equal to 2.617, -0.878 and 0,100 em3 mol-l, thestandard deviation ofthis fitting beinga = 0.006 cm3 mol-l. V t = 0.654 cm3 mol-I.The results for HE are represented in table 3 in the form adopted in a recentpublication on the methane + ethylene system.31 The HE values have been calculatedboth for mixing at the saturation vapour pressure, HE(Ps), and at zero pressure,HE(0). As the vapour pressures under the prevailing experimental conditions werecomparatively low, the differences between HE(P,) and HE(0) are very small.TheHE(0) values fit the equationwith G = 2.3574 ; H = -0.0273 ; J = 0.5827, the standard deviation beinga = 5.5 J mol-l. Hi(0) = 901.8 J mol-'.VE = x,(l -x,)[D+E(2x, - 1)+F(2x1 - 1)2] (2)HE(0)/RT = XI( 1 - x,)[G + H(2~1- 1) + J ( 2 ~ 1 - 1)2] (3MACHADO, GUBBINS, LOB0 A N D STAVELEY 2499TABLE 2.-MOLAR VOLUMES AND EXCESS MOLAR VOLUMES OF LIQUID MIXTURES OF NITROUSOXIDE XEN XENON (2) AT 182.32 K AND AT THE SATURATION VAPOUR PRESSURE^0.000 000.152 780.284 770.392 230.503 150.639 440.772 500.892 710.910 071 .ooo 0046.45345.20643.94242.82141.58939.98838.37136.85336.62435.48700.4280.61 10.6690.6530.5470.3900.1900.1510-0.004- 0.003- 0.0010.000- 0.0020.009- 0.000- 0.010IRV is the volume residual, = VE- VFalc, where VFalc is the excess molar volumecalculated from eqn (2).TABLE 3.-EXCESS MOLAR ENTHALPY OF THE SYSTEM NITROUS OXIDE (l)+XENON (2) AT(1 84.05 0.01) KaHE(Ps) HEW) RHn /mol n2 /mol X1 Q/J /J mol-1 /J mo1-l /J mol-10.020 52 0.053 47 0.2750 58.533 754.0 753.9 - 5.60.023 47 0.044 39 0.3442 59.847 842.8 842.7 6.00.040 12 0.048 36 0.4513 81.273 899.0 898.8 2.40.047 53 0.039 09 0.5455 78.297 890.8 890.7 - 4.60.043 01 0.022 87 0.6469 55.889 838.2 838.1 - 0.70.047 34 0.014 51 0.7562 43.114 706.2 706.1 1.8a Q is the energy supplied to the calorimeter to maintain it at the initial temperature.RH is the enthalpy residual, = H E - HFalc, where HFalc is the excess molar enthalpy calcu-lated from eqn (3).Finally, the values of SE at 182.32 K, derived on the assumption that HE at thistemperature has the same value as at 184.05 K, conform to the equationwith K = 1.1747; L = -0.0805; M = 0.5402.TS; = 453.6 J mol-l.In fig. 1, GE, HE and TSE are plotted against xl, the mole fraction of N20. Thedependence of VE on x1 is shown graphically in fig. 2. All four curves are fairlysymmetrical, the most skewed being that for VE. It will be noted that TSE and VEare both positive and relatively large. The total vapour pressure data are shown inSE/R = ~ l ( l -xl)[K+L(2~1- l)+M(2~1- 1)2] (4)fig. 3.COMPARISON WITH THEORYThe theoretical approach has been fully described in previous papers,4.3 2 * 33 sothat only a brief outline of the method is given here. The Helmholtz free energy Afor the mixture is expanded in powers of the anisotropic part of the intermolecula2500 THERMODYNAMICS OF N,O+Xepotential energy for a pair of molecules of species c1 and p, about the free energyA . for a reference mixture of spherical molecules. The reference potential z& isdefined to be an unweighted average over the orientations of the full potential uaB.800LIII 222 600aw" 400 uw"%2 00I I I I0.2 0 . 4 0.6 0.8x1FIG. 1.-Excess molar Gibbs free energy at 182.32 K and excess molar enthalpy at 184.05 K forN20+Xe, plotted against xl, the mole fraction of nitrous oxide. Points are experimental data,lines are from eqn (5).The dashed line is the experimental excess entropy, obtained from TSE =HE- GE.0.6r(.-. I 8 0.4EL.m20.20 0.2 0.4 0.6 0.8x1FIG. 2.-Excess molar volume YE for N20 + Xe at 182.32 K from experiment (points) and eqn ( 5 ) (line)MACHADO, GUBBINS, LOB0 AND STAVELEY300200 2 24 \ a,100250 1----I 1 I Ix, Y ( N 2 0 )FIG. 3.-Vapour-liquid equilibrium for N20+Xe at 182.32 K from experiment (points) and eqn (5)(solid line). The dashed line is the result calculated using van der Waals one-fluid theory withisotropic n,6 potentials for each of the pair interactions.With this choice of reference the first-order term A l vanishes and the series to third-order is used as the basis for a simple Pad6 approximantThis expression is in good agreement with computer simulation results for dipole-dipole and quadrupole-quadrupole potentials and for anisotropic overlap potentialsof the type used here.Comparisons of theory and experiment therefore provide atest of the intermolecular potential models used. The procedure for calculatingproperties of the reference system was as described by Gubbins and Twu 32 andinvolved an expansion of the n,6 fluid properties about those for a 12,6 fluid, togetherwith the use of van der Waals one-fluid theory to relate the properties of the 12,6mixture to those of a pure 12,6 fluid. The Gosman et al.34 equation of state wasused for the free energy of the pure fluid and the equations of Gubbins and Twu 32were used for the integrals J and K that arise in the A , and A 3 terms in eqn (5).Forthe J' integrals the equation of Nicolas et al.35 was used, since it is more accuratethan that previously given by Gubbins and Twu.For the xenon/xenon interaction the Lennard-Jones 12,6 potential model wasusedUXe/Xe = uowith the parameters given previ~usly.~ This gives an excellent fit to the data for thepure coexisting gas and l i q ~ i d . ~ For the N,O/N,O interaction we initially used thepotential modelwhere ugs6) is the (isotropic) n,6 p ~ t e n t i a l , ~ up, . . . uQQ are the dipole-dipole. . . quadrupole-quadrupole potentials and uov( 101 + 01 1) and Udis( I01 + 01 1) are theleading terms in a spherical harmonic expansion of the anisotropic overlap anddispersion terms ; here 101 and 01 1 are the values in the expansion.Detailed( 6 ) (12,6)UN~OIN~O = Ub",6) + u , p + upQ+UQp+UQQ+Uo,(lol+o1 I)+ UdiS(101+011) (72502 THERMODYNAMICS OF N,O+Xeexpressions for these potentials are given in ref. (4) and (32). Multipole momentsand the anisotropic polarizability value were taken from independent experimentalmeasurements and the remaining parameters (e, a and n in the n,6 potential and theparameter 6 that occurs in uOv) were obtained by fitting the theory to saturated liquiddata in the usual way.4 Eqn (7) was found to give an excellent fit to the data for puregaseous and liquid N20. However, an equally good fit was obtained by omittingthe anisotropic overlap and dispersion terms- U ( n , 6 )and this potential model was the one finally used.Values of the potential parametersfor this model are shown in table 4. The value of the quadrupole moment of- 3.65 x e.s.u. cm2 was obtained by the direct method of magnetic susceptibilityanisotropy and is estimated to be accurate to kO.25 x(8) UN20/N20 - 0 + + upQ + uQp + uQQe.s.u. cm2.TABLE 4.-POTENTIAL PARAMETERS ---P QXe+Xe 231.5 3.961 12 0 0N20+ N2O 261.9 3.771 15 0.166' - 3.656Xe+N20 243.0b 3.881b 13.4 -pair (&/k)/K" a/Aa nu e.s.u. cm e.s.u. cm2--a Like-pair parameters from orthobaric liquid density and pressure. NzO+ N20 para-metersfromeqn(8). Xe+N20parameters&andafrom GF and Vg, II fromeqn(l1). Thesevalues correspond to cXe/N20 = 0.987 and vXe/N20 = 1.004, where 5Xe/N20 a ~ ~ d l y x ~ l ~ ~ o arethe usual parameters in the modified Lorentz-Berthelot rules, &ab = <ab(&aa&bb)' and Gab =iqab(oaa+ ebb).TIKFIG.4.-Orthobaric liquid density of NzO from experiment (points, Couch et aE.22 and this work)and theory (lines). The solid line is based on the Pad6 approximant of eqn (5) with theanisotropic potential model, eqn (8) ; the dashed line is for the isotropic potential model, eqn (9)MACHADO, GUBBINS, L O B 0 AND STAVELEY 25037000500030001000200 240 280T/KFIG. S.-Vapour pressure of N20 from experiment (points, Couch et aLZ2 and this work) and theory(lines). Key as in fig. 4.Comparison of theory (solid line) and experiment for the orthobaric liquid densityand vapour pressure for N20 are shown in fig. 4 and 5. The average deviationbetween theory [eqn (5)] and experiment for the temperature range 182.32 K (triple-point) to 295 K was 1 % for pressure, 0.3 % for liquid density and 2 % for gasdensity.At the critical point itself (309.58 K) the errors in the predicted gas andliquid densities are larger ( z 10 %), since the predicted critical point lies x 0.5 Kabove the experimental value. The predictions of a simple isotropic potential, then,6 model,are also included in fig. 4 and 5 (dashed lines). These curves represent the bestpossible fit to the data using this simpler model. It should be stressed that the samethree adjustable parameters ( E , Q, n) are involved in both the isotropic model of eqn(9) and in the anisotropic model of eqn (8) and the procedure for obtaining theseparameters is identical in the two cases.The anisotropic model is seen to be insubstantially better agreement with experiment and eqn (8) is thus superior to eqn (9)as an effective potential for the pure fluid. Similar comparisons have been made forcarbon dioxide, ethane and ethylene by M a ~ h a d o , ~ ~ with the same conclusion in eachcase. A comparison of theory and experiment 2 2 for pressures of the compressedgas is made in table 5.(9) - ug'6)'N20IN20 -Agreement is within 1 % or better for most points.For the xenon/nitrous oxide interaction a simple n,6 model was usedUXe/N20 = (10)The addition of an anisotropic overlap term uOv( 101) and a quadrupole-induceddipole-quadrupole term uQindQ(O0O) to this potential model resulted in no improve-ment to the fit and these terms were therefore omitted. The value of nXe/N20 wasestimated from the geometric mean rule 37IZXe/N20 = (nXe/XenN20/Nz0)'* (1 1)The values of E ~ ~ / ~ ~ ~ and oXeINZ0 were obtained by requiring agreement betweentheory and experiment for GZ and V: and are included in table 42504 THERMODYNAMICS OF N,O+XeTABLE 5.-cOMPARISON OF THEORY AND EXPERIMENT FOR GASEOUS N20, SHOWN AS THEPRESSURE P AT WHICH THE GAS HAS THE DENSITY pP/kPaT K plrnol dm-3 expt 22 calc.243.15243.15258258.15273.15273.15288.15288.15288.15303.15303.15303.15303.15348.15348.15398.15398.15398.15423.15423.15423.150.32180.70290.29931.1940.48271.8520.451 11.2782.9780.42431.4633.0435.17415.0717.653.8959.4823.4477.94113.5211.7860812166082 02710143 04110142 5344 46010143 0415 0686 28420 27130 40710 13520 27130 40710 13520 27130 407612123761 12 06510193 09010172 5544 50410163 0625 0906 23020 68731 48110 08920 16030 70110 10020 11930 508A comparison of theory and experiment for the mixture data is shown in fig.1-3.Since the experimental values of GF and V: are used in fitting parameters, the com-parison tests the ability of the theory to correctly predict the HE curve and the shapesof GE and VE. Agreement between theory and experiment is excellent for GE andVE; for HE the theory predicts values that are z 4% too high for the equimolarmixture. Good agreement is obtained for the pressure values (fig.3) and the azeo-trope is correctly predicted. The pressure values predicted assuming simple n,6isotropic potentials for all three interactions are also included in fig. 3. In thesecalculations the properties of the n,6 mixture are related to those for a 12,6 mixturein the usual way 3 2 and the properties of the latter are calculated from the van derWaals one-fluid theory ;3 values of n, E and c for each pair interaction were obtainedby the same procedure as for the anisotropic potential models described above. Asseen from fig. 3 the anisotropic potential model gives considerably better agreementwith experiment than the isotropic n,6 model alone. The excess properties were alsocalculated using these simple isotropic potential models.Good results were obtainedfor VE and for H: (both G: and V: were again used to fit parameters, so that theoryand experiment must always agree for these properties). However, the predictedHE and GE curves were not of the correct shape.CONCLUSIONThe thermodynamic data reported here for N,O+Xe mixtures, as well as theexisting data for pure N 2 0 and Xe ,are in good agreement with theoretical predictionsusing simple intermolecular potential models. The calculations indicate that thMACHADO, GUBBINS, LOB0 AND STAVELEY 2505effect of electrostatic forces is significant, but that these may be successfully approxi-mated by a multipole series terminated at the quadrupole-quadrupole term.Sincemultibody potential terms are omitted in these calculations the potential modelsshould be regarded as effective pair potentials suitable for the liquid phase. More-over, potential terms for anisotropic overlap (shape), anisotropic dispersion andinduction forces, found to have a negligible effect in the calculations given here,may be significant in calculations over a wider range of temperature and pressure.Despite these reservations, the anisotropic potential models developed here are asubstantial improvement over isotropic potential models with the same number ofadjustable parameters.Xe + N 2 0 mixtures offer the simplifying features that one of the components (Xe)is spherical, while the other (N20) is linear and has only a weak dipole moment.Weare in the process of studying the mixtures N20+C2H, and N,O+HCI and shallreport results for these shortly. Ethylene is non-polar, but possesses a non-axialquadrupole moment (Le., the quadrupole moment has two independent components),while hydrogen chloride has a relatively large dipole moment.This work was supported by grant ENG 7682101 from the National ScienceFoundation. The stay at Oxford of one of us (L.Q.L.) was made possible by aFellowship from C.P. Invotan (J.N.I.C.T./Portugal).C. H. Twu, K. E. Gubbins and C. G. Gray, Mol. Phys., 1975,29,713.M. Flytzani-Stephanopoulos, K. E. Gubbins and C. G. Gray, Mol. Phys., 1975, 30, 1649.C. H. Twu, K. E. Gubbins and C. G. Gray, J. Chem. Phys., 1976,64,5186.J. C . G. Calado, C. G.Gray, K. E. Gubbins, A. M. F. Palavra, V. A. M. Soares, L. A. K.Staveley and C. H. Twu, J.C.S. Faraday I, 1978,74, 893.L. Q. Lobo, L. A. K. Staveley, P. Clancy and K. E. Gubbins, J.C.S. Faraday I, 1980, 76, 174.W. H. Flygare, Chem. Rev., 1974, 74, 653.D. E. Stogryn and A. P. Stogryn, Mol. Phys., 1966, 11, 371.R. H. Davies, A. G. Duncan, G. Saville and L. A. K. Staveley, Trans. Faraday SOC., 1967,63,855.J. C. G. Calado and L. A. K. Staveley, Trans. Faraday SOC., 1971, 67, 289.lo K. L. Lewis and L. A. K. Staveley, J. Chem. Thermodynamics, 1975, 7 , 855.l1 W. M. Haynes and M. J. Hiza, J. Chem. Thermodynamics, 1977, 9, 179.l2 R. D. Goodwin, H. M. Roder and G. C. Straty, Thermophysical Properties of Ethane, froml3 Y. Yato, M. W. Lee and J. Bigeleisen, J.Chem. Phys., 1975, 63, 1555.l4 R. W. Blue and W. F. Giauque, J. Amer. Chem. SOC., 1935,57,991.l5 K. Clusius, U. Piesbergen and E. Varde, Helv. Chim. Acta, 1960, 43, 1290.l 6 A. Michels and T. Wassenaar, Physicu, 1950, 16, 253.l7 F. Theeuwes and R. J. Bearman, J. Chem. Thermodynamics, 1970, 2, 507.l8 A. J. Leadbetter, D. J. Taylor and B. Vincent, Canad. J. Chem., 1964,42,2930.l9 M, J. Terry, J. T. Lynch, M. Bunclark, K. R. Mansell and L. A. K. Staveley, J. Chem. Thermo-2o J. G. Hust, Cryogenics, 1969, 9, 443.21 T. B. Douglas, J. Res. Nut. Bur. Stand., 1969, 73A, 451.22 E. J. Couch, L. J. Hirth and A. Kobe, J. Chem. and Eng. Data, 1961,6,229.23 W. H. Schamp, E. A. Mason and K. Su, Physics Fluids, 1962, 5, 769.24 K. S. Pitzer and R. F. Curl Jr, J. Amer. Chem. Sbc., 1957, 79, 2369.25 J. F. Mathews, Chem. Rev., 1972, 72,71.26 J. Brewer, Determination of Mixed Virial Coeficients (AFOSR No. 67-2795, 1967).27 K. L. Lewis, G. Saville and L. A. K. Staveley, J. Chem. Thermodynamics, 1975, 7, 389.28 W. B. Streett, L. S. Sagan and L. A. K. Staveley, J. Chem. Thermodynamics, 1973,5,633.29 J. H. Hildebrand and R. L. Scott, The Solubility of Non-Electrolytes (Reinhold, New York,30 J. A. Barker, Austral. J. Chem., 1953, 6, 207.31 L. Q. Lobo, J. C. G. Calado and L. A. K. Staveley, J. Chem. Thermodynamics, 1980,12,419.90 to 600 K at Pressures to 700 Bar (NBS TN 684, 1976).dynamics, 1969, 1, 413.3rd edn, 1950), p. 4242506 THERMODYNAMICS OF N,O+Xe32 K. E. Gubbins and C. H. Twu, Chem. Ens. Sci., 1978,33,863,879 ; C. G. Gray, K. E. Gubbins33 P. Clancy, K. E. Gubbins and C. G. Gray, Faraday Disc. Chem. Soc., 1978,66,116.34 A. L. Gosman, R. D. McCarty and J. G . Hust, Nat. Stand. Ref. Dara Ser. Nar. Bur. Stand.,3s J. J. Nicolas, K. E. Gubbins, W. B. Streett and D. J. Tildesley, Mol. Phys., 1979,37, 1429.36 J. R. S. Machado, MSc. Thesis (Cornell University, 1979).37 T. M. Reed and K. E. Gubbins, Applied Statistical Mechanics (McGraw Hill, New York,38 T. W. Leland, J. S. Rowlinson and G. A. Sather, Trans. Faraday Soc., 1968, 68, 1447.and C. H. Twu, J. Chem. Phys., 1978, 69, 182.1969, 27.1973), p. 131.(PAPER 9/1761

ISSN:0300-9599    

DOI:10.1039/F19807602496

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980,76,2507-2509Molecular Sieving Range of Pore Diameters of AdsorbentsBY JACOB KORESH AND ABRAHAM SOFFER*Atomic Energy Commission, Nuclear Research Center-Negev,P.O. Box 9001, Beer-Sheva, IsraelReceived 19th November, 1979The very sensitive molecular-dimension criterion of adsorption selectivity of molecular sieves changesrapidly to a molecular-mass criterion upon a slight enlargement of the adsorbent-pore dimension. Thisresulted in the inversion of the sequence of adsorbability of hydrogen and oxygen on carbons ofincreasingly wider pores. Considerable care must be taken in relating molecular-sieve effects on adsorp-tion to molecular dimensions of adsorbates.Adsorption on highly porous media may have molecular sieving effects. Accord-ingly, the adsorbability of molecules as well as some other adsorptive propertiesmay be different, and the question is how to discriminate molecular sieves (MS)from wide-pore adsorbents.Such a discrimination is of ultimate importance when-ever molecular dimensions are to be assessed from adsorbability on MS. MS areendowed with the unique property of discriminating very sharply between mol-ecules of similar width.’ Molecules which differ by merely 0.2-0.3 8, in width maybe adsorbed at rates which vary by several orders of magnitude.’ A difference of0.3 8, results in an apparently complete exclusion of the larger molecules; this isseen from adsorption isotherms as well as from studies of adsorption kinetics.l Thissensitivity is, however, rapidly lost once the pores are too open compared with themolecular width, but still in the ultramicroporous range ( < 7 8,).2-4 Under thesecircumstances, adsorbability as observed by both adsorption kinetics and isothermsmay be controlled by factors specific to open surfaces, in addition to moleculardimensions, and the assessment of the latter will consequently be impaired.The aimof this communication is to determine at what degree of pore opening does molecu-lar sieving change to “ordinary” adsorption on a microporous adsorbent.This work is essentially based on the ability to widen the ultramicroporousstructure to any desirable extent by mild activation steps. By such operations thecomplete closure of pores to a certain adsorbate may be gradually changed viamolecular sieving stages into the completely free, non-selective admission into widepores.EXPERIMENTALThe adsorption apparatus, methods of modification of pore dimensions and characteristics ofthe carbon starting materials have been described elsewhere.RESULTS AND DISCUSSIONSIn a previous study of the molecular dimensions of adsorbate molecules, hydro-gen molecules appeared to be wider than C02, acetylene and oxygen.’ This surpris-25025080.50.4iMI 03-E-ia 02-101-MOLECULAR SIEVE CARBONS------_ * A aa I 1 I I l I I l I I I I L I I I I I 1t/min1 2 3 4 5 6 7 8 9 1 0I I I 1 I I I I I IA A J AAAA ** A *A *aaaa*1A 0FIG.1.-Adsorption kinetics of oxygen on carbon C-200 and of oxygen and hydrogen on carbon C-300at 77K.Initial pressures 6&65Torr; sample weight 100mg. Tad = -196°C A, 02, 300"C,V, = 86.95 pmol min-'; *, Hz, 300"C, V, = 150 pmol min-l; 0, 02, 200"C, V, = 1.8 pmol min-'.ing result was in accordance with the average dimension calculated from the liquidmolar volume but not with the kinetic diameter calculated from gas-phase molecu-lar theory5 or from van der Walls radii.6 On the other hand, hydrogen has beencommonly considered smaller than the above higher-weight molecules, and studiesof the adsorption kinetics on mordenite MS presented by Barrer7 seem to supportthis view, since the sorption rate of hydrogen was far greater than that of oxygen,nitrogen and argon and close to that of helium.The following brief comparativestudy of the adsorption rates of hydrogen and oxygen on molecular sieve carbon(MSC) appears to settle this controversy. In fig. 1, the adsorption rate of oxygen ona MSC designated C-200 is plotted. On the same carbon and under the sameconditions, hydrogen is not adsorbed to any measurable amount (our experimentalresolution for adsorption was <0.1 ymol g-' for an adsorption rate <0.1 ymol g-'s-'). On the other hand, for the slightly wider pore, C-300 carbon, the rate ofhydrogen adsorption is at first faster than that of oxygen, and then becomes slowerthan it. The slowing down of H2 adsorption rate with elapsed time is a manifesta-tion of the much lower adsorption isotherm of H2 compared with that of 02.The high ratio between the adsorption rates for the C-200 carbon is undoubtedlytypical of the molecular sieving effect and shows repeatedly' that oxygen behaves asa thinner molecule than hydrogen. The inversed ratio of initial rates for the C-300carbon is not as great as that for the C-200 carbon.Recognizing furthermore thatthe C-300 carbon accomodates nitrogen,' which is a wider molecule than eitheroxygen or hydrogen,8 we are led to the conclusion that the pores of the C-300carbon are too wide to discriminate between oxygen and hydrogen. The MS selec-tivity of the C-200 carbon is, therefore, changed into a selectivity of wide pore solid,into which the light hydrogen molecule diffuses faster than oxygen. The largJ . KORESH AND A . SOFFER 2509decrease in the hydrogen adsorption rate with elapsed time is probably due to theproximity of the saturation of the adsorbent with hydrogen which should have amuch lower physisorption isotherm than that of oxygen.One should thereforeexamine the width of the C-300 carbon pores as compared with the dimensions ofoxygen and hydrogen, which are 3.28 and 3.44A, respectively.8 Xenon was com-pletely excluded from this carbon, indicating that its pores are narrower thanatomic diameter of xenon which is 3.94 A. Therefore an increase of only a fewtenths of Angstrom in pore dimension is sufficient to cancel the molecular sievingeffects of ultramicroporous solids. It is probably the large difference between themasses of hydrogen and oxygen which makes their adsorption selectivity rapidlyexceed the molecular sieving criteria. Nevertheless, great care must be taken incomparing the adsorbabilities of molecules of different dimensions on a single MSwhich can accomodate all of them. In case of MSC C-300 and C-200, it is theability to change the pore diameters very gradually which enable us to “meet” themolecular dimension very closely and observe the true molecular sieving sequenceof oxygen and hydrogen.J. Koresh and A. Soffer, J.C.S. Faraday 1, 1980,76,2457.*Y. Toda, N. Yuki and S. Toyoda, Carbon, 1972, 10, 13.3J. R. Dacey and D. G. Thomas, Trans. Faraday Soc., 1954, 50, 740.4S. S. Barton, M. J. B. Evans and B. H. Harrison, J . Colloid lnterface Sci., 1974, 49, 462.’G. L. Kington and A. C. Macleod, Trans. Faraday Soc., 1959, 55, 1799.’R. M. Barrer, Quart. Rev., 1949, 3, 293.* J. Koresh and A. Soffer, J.C.S. Faraday I, 1980, 76, 2472.L. Pauling, Nuture of the Chemical Bond (Cornell Univ. Press, Ithaca, N.Y., 3rd edn, 1960).(PAPER 9/1844

ISSN:0300-9599    

DOI:10.1039/F19807602507

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980,76,2510-2518Apparent Molal Volumes of some Highly ChargedElectrolytes in WaterBY FRANCESCO MALATESTA" AND ROBERTO ZAMBONIIstituto di Chimica Analitica ed Elettrochimica dell' Universiti,Via Risorgimento 35, 56100 Pisa, ItalyReceived 5th December, 1979The apparent molal volumes of Mg,Fe(CN),, Sr2Fe(CN)6 and K,Co(CN), were measured at 25'C bypyknometric and dilatometric methods. The 2 :4 salts appreciably deviate from the theoretical limitinglaw in a direction which is opposite, in the more dilute solutions, to that predicted by Debye-Huckeltheory (DH). For K,Co(CN), no appreciable deviations from the theoretical slope are seen below0.01 mol dm-3 and the deviations are in the same direction as DH predicts when the concentration isincreased. Such behaviour is similar to that of Ca,Fe(CN), and K,Fe(CN),.The results are analysed by means of the Mayer theory and by numerical integration of the Poisson-Boltzmann equation.Both treatments can justify, in terms of electrostatic interaction, volumetric behav-iour similar to those of Mg2Fe(CN)6 and Sr2Fe(CN)6 for the 2:4 salts and to that of K,CO(CN)6 for the1 : 3 salts.An enhanced, qualitative discrepancy from the predictions of Debye-Huckeltheory (DH) occurs for the apparent molal volumes (4") of some strong multivalentelectrolytes.'-6 According to DH, the 4v slopes should usually be less than thetheoretical limiting slope (DHLL) at real concentrations, and approach DHLL asthe concentration decreases. In Nd(N03)3,2 K4Fe(CN)6,3'4 Ca2Fe(CN)2 and someother salts,'*5 the slope of 4v becomes greater than DHLL at low concentrationand then increases as the concentration is further decreased. The limiting slope isnot approached in the experimental range of concentrations, and it is difficult toextrapolate the molal volume at infinite dilution ( V O ) .Deviations from DH, when observed in aqueous dilute solutions of 1:1 electro-lytes, are reliably assumed as evidence of non-electrostatic interactions.3 7 However,the positive deviations of highly charged electrolytes'-' (and the similar deviationsobserved in apparent relative molal enthalpies, 4L8) may perhaps be justified interms of fundamentally electrostatic interactions. In aqueous solutions of highlycharged electrolytes' (as well as in non-aqueous solutions of weakly charged elec-trolytes") the linearisation of the Poisson-Boltzmann exponential term is math-ematically unjustified at ionic strengths ( I ) which are much lower than in aqueoussolutions of 1:1 salts, and DH gives a misleading view of the electrostatic behav-iour.Indelli and De Santis' found that Mayer theory"?l2 was able to justifypositive deviations from DHLL in dilute solutions of 1 : 3 and 1 :4 salts. They usedthe so-called DHLL + B213 approximation of Mayer theory, i.e. the same physicalmodel of the Debye-Huckel theory (primitive model). DHLL + B2 and a numeri-cal integration of the Poisson-Boltzmann equation, IPBE93l 0,14-1 were also ableto explain the positive deviations in 4L of highly charged electrolyte^.'^.^^ (In spiteof the well known inconsistencies of the Poisson-Boltzmann equation,' IPBEagrees with the experimental results better than does DHLL + B2.16917)We have therefore studied other, multivalent electrolytes.In this paper, theresults obtained for the 2:4 salts Mg,Fe(CN)' and Sr2Fe(CN)', and for the 1:3 salt251F . MALATESTA AND R . ZAMBONI 251 1K,Co(CN),, are shown. The volumetric behaviour of other highly charged electro-lytes [in particular Ca2Fe(CN)z and K,Fe(CN);] are reexamined for comparison.EXPERIMENTALThe apparatus was the same as that used p r e v i ~ u s l y ~ . ~ and did not show any difference incalibration. The experimental technique was also the same as in previous ~ o r k . ~ , ~MgFe(CN)6 and Sr,Fe(CN), were prepared by ion exchange on a Dowex 50 wx8 resin in thehydrogen form, so as to give H4Fe(CN)6, which was immediately gathered in MgO or SrC03.Both salts were recrystallised several times from water + ethanol mixtures and then three timesfrom conductivity water.K,Co(CN), was obtained following the Benedetti-Pickler method"and purified by iterated crystallisations. All salts were air-dried for 1 week and stored inhermetic containers.In order to check compositions and water content, salts were analysed, so as to obtain theirequivalent weights with respect to at least one of the components. K&O(CN)6 was tested forCo by an electrogravimetric method.20,21 We found Co = 17.75 f 0.05% (theoretical value forthe anhydrous salt, 17.73%), which gives an apparent molecular weight of 332.0 & 0.9 (theoreti-cal, 332.35).In addition the salt was tested once for its Co(CN):- content: H,Co(CN), wasobtained by ion exchange, gathered in NaOH solution and then back titrated with HC1. Theresult was consistent with a molecular weight of 332.7.Mg,Fe(CN), and Sr,Fe(CN), were analysed by a similar method4 for H,Fe(CN), ; apparentmolecular weights of 475.5 k 0.8 [theoretical, 476.76 for Mg,Fe(CN), - 12H20] and 619.8 & 0.9[theoretical, 621.39 for Sr2Fe(CN)6 * 13H20] were obtained. Mg,Fe(CN), was also tested forMg by EDTA titration: an apparent molecular weight of 477.0 & 0.9 was found.we assumed that the salts had the exact compositions K,CO(CN),,Mg,Fe(CN), - 12H20 and Sr,Fe(CN), * 13H20.Slight differences in water content could beresponsible for the systematic errors in 4; (z0.9cm3 rno1-l for the Sr salt, for instance,resulting from the difference between theoretical molecular weight and apparent molecularweight obtained by the ion-exchange-acidimetric method).In calculatingCALCULATIONSAll calculations were performed using the same value for the distance of closestapproach, 4, between cation and cation, cation and anion and anion and anion.4DHLL + B2 formulae for the apparent molal volumes have already been de-The IPBE calculations for apparent molal volumes are similar to thosefor 4,,,16717 except that the numerical differentiation of the activity coefficients iscarried out against pressure ( p ) instead of temperature (7').The value of the dielec-tric constant of water and of its derivative with respect to pressure were taken fromOwen.22 The water compressibility coefficients, p, are those of Kell and W h a l l e ~ . ~ ~Of course, both DHLL + B2 and IPBE (as well as DH) cannot directly providethe 4" values, but 4; quantities which differ from q5v in the V" term (ie.,4; = theoretical 4v - I/" value).RESULTS AND DISCUSSIONThe experimental results are reported in tables 1 and 2. K,CO(CN)6 behaves in asimilar way to K3Fe(CN)6.4 Mg2Fe(CN)6 and Sr,Fe(CN), display very similarbehaviour to Ca2Fe(CN),.' At low concentrations (below z 3 x lo-, mol dm-,)the 2:4 salts show striking positive deviations from the limiting slope (DHLL)2512 APPARENT MOLAL VOLUMES OF STRONG ELECTROLYTESTABLE 1 .-DENSITIES AND APPARENT MOLAL VOLUMES OF CONCENTRATED SOLUTIONSsolution c/mol dm-3 a di5 4" cm3 mol-'0.3 1 5050.092950.171730.421 10.214030.214050.1 I7740.21 3290.0975 11.0996891.0673631 .0279631.0330791.021 9271.0721701 .0356491.014911.041 74661.6,58.9555.0152.025 1 .0549.61154.50151.92149.79a Based on the assumption of the exact compositions Sr,Fe(CN)6. 13H20,Mg2Fe(CN)6-12H20, and K3Co(CN), for solid salts.bFor water at 25 C we took di5 =0.997 07TABLE 2.-4., FROM DILATOMETRIC MEASUREMENTS. NUMBERS IN BRACKETS IDENTIFY THE START-ING SOLUTION IN TABLE 1.(1) 2.806(1) 2.341(2) 1.906(2) 1.590(3) o.8277(1) 0.7392( I ) 0.7392(3) 0.6907(1) 0.3400(2) 0.291 2(1)O.l6l7(2) 0.1099(2) 0.0495(3) 0.0335(1 ) 0.0324(2) 0.0220(3) 0.021 5(3) 0.021 551.3050.9050.2449.8448.2748.0,47.7946.9446.0045.32"44.9643.842.540.841.440.639.040.0~~(1) 3.05 1(1) 1.906(1) 1.590(2) 1.529(3) 1.049(3) 0.8749(1) 0.5022(3) 0.241 7(1) 0.1099(3) 0.0869(3) 0.0577(2) 0.025( I ) 0.031 3(1) 0.0220(1) 0.0220~~45.7044.6544.2044.1t3.4743.0241.8940.3838.93K336.036.535.537.635.2~~ ~(1) 3.751(1) 3.130(2) 1.899(2) 1.899(2) 1.585(3) 0.86g3(2) 0.6027(1)0.2162(1) 0.9882(2) 0.2407(3)0.1100(2)(1) 0.053(3) 0.050(1) 0.0974(2) 0.0493(1) 0.043 3~~147.76147.48146.85146.S4146.6,145.80145.95145.40144.7144.4144.6143.7144.2143.5144.5143.s144.3~ ~~"Erratic value.An anomalous, progressive descent of the level in the capillary of the dilat-ometer was observed, which did not correlate with temperature (a bubble of air'I4).unlike the 1:3 salts. This cannot be justified in terms of DH theory,* but it agreescomparison, the positive deviations in $v of Mg2Fe(CN),, Sr2Fe(CN)6 andwith previous results for the & values of multiply charged electrolyte^.^*'^^'^ BY* In principle, DH predicts positive deviations from DHLL for high enough, positive values ofd In B/dp. In practice, values of d In B/dp that are too high (of the order of 10' b) are required in order tojustify positive deviations such as those in 2 :4 salts, and they seem unreasonable (furthermore, theshapes of the DH curves calculated in this way are entirely different from experimental ones).For similarconsiderations about $,, see ref. (17)F . MALATESTA A N D R . ZAMBONI 25137-3I - zi 5 0a3-Ca2Fe(CN)6 occur over a wider range of I than the corresponding deviations in #L.(This cannot fully be explained at present; however, other salts behave ~imi1arly.l~In terms of the theories based on the primitive model, it may mean that thedistances of closest approach cannot be considered as independent from T andAlthough the 4" values of the 2:4 salts are not linear functions of I+ below0.4 mol dmP3 ionic strength, their differences, A#", are almost linear when plottedagainst I' (fig. 1). The A#" values were obtained by subtracting the experimentalvalue of 4" for Mg2Fe(CN), and Ca2Fe(CN),* from linearly interpolated values of# v for Sr2Fe(CN)6 (similar results are obtained from graphical interpolation).A 2:4p . I 7 ) .0 . /./- I .---~ ~ ~ o ~ - o - o0 .o 08a1 I 1salt should reasonably simulate another 2:4 salt at low concentration, and one canexpect that the almost linear trends of A$" are maintained up to the I = 0 limit(although fig. 1 does not illustrate this). This suggests a difference in vc of = 5 cm3mol- ' between Sr2Fe(CN)6 and Mg2Fe(CN)6 [1.5 between Ca2Fe(CN)6 andThe V o values of the 2:4 salts cannot be obtained from direct extrapolation ofthe experimental #v curves. An evaluation of the vo values is provided by theadditivity relationships from vJ of weekly charged electrolytes.The values 30.1,36.1 and 36.7 cm3 mol- ', for Mg2Fe(CN)6, Sr2Fe(CN)6 and Ca2Fe(CN)6, respect-ively, are calculafed from Millero's selected data' if one assumes 108S4 (instead ofl-10.083) as the V" values of K,Fe(CN),. The difference, 6 cm3 rnol-I between theI/" values of Sr2Fe(CN), and Mg,Fe(CN),, compares favourably with the valuesuggested by fig. 1 (= 5 cm3 mol-'); the small discrepancy may be due to system-atic errors in $" (water content uncertainties in the salts) or to inaccuracies in theliterature data. The vc values of 30.1 and 36.1 were used to calculate the values of4: for Mg2Fe(CN)6 and Sr,Fe(CN),, respectively, to be compared with theoretical4: (DHLL + B2 and IPBE). A discrepancy of a few cm3 mo1-I from true extrapo-lations cannot change the meaning of such comparisons.Mg 2 F e m 6 1 .* The basic experimental data of ref.( 5 ) were used for Ca2Fe(CN)6.1-82514 APPARENT MOLAL VOLUMES OF STRONG ELECTROLYTESThe evaluation of v' as 36.7 cm3 mol-' for Ca2Fe(CN)6 is far too high, in ouropinion. The 4" curve of this salt5 lies between those of MgzFe(CN)6 andSr,Fe(CN),. A value of 34 cm3 mol-' has been given in a previous paper.5 Fig. 1suggests a smaller value, perhaps 31.6 according to the value of 30.1 forMgzFe(CN)6 [or 32.6 according to 36.1 for SI-~F~(CN)~]. The ionic conventionalvolume of Ca2+ was assumed to be greater than that of Sr2+ at 25"C,' according tothe data of CaC1z,24 a very hygroscopic salt (systematic errors?).According to thepresent results, it would lie between Mg2+ and Sr2+ (in analogy to results reportedat higher temperatures ').0.2 0.4 0.6 0.84: IFIG. 2.-Plots of 4: against J'l for 1 : 3 salts. DHLL + B2 : dashed lines [i = 3,4 and 8 8, for (a), (b) and(c), respectively]. IPBE: full lines d In h/dp = 0 [h = 3 and 4 A for (d) and (e), respectively]; dotted line(f), d In i/dp = p, 6 = 5.5 A. As a comparison, K,Co(CN), (0) and K,Fe(CN)z (0) are given. Dash-and-dot line: DHLL.As for K,CO(CN)~ [and likewise K,Fe(CN),4], there seem to be no problems inthe extrapolation vo. 143.3 cm3 mol-' is obtained on the basis of DHLL C146.4 forK3 Fe(CN), "1.Fig. 2 shows some DHLL + B2 and IPBE curves for the 1:3 salts. This nowconfirms, in terms of IPBE, what had already been found by means of DHLL + B2calculations :6 i.e., positive deviations from DHLL are justified at low concen-trations, for sufficiently small values of ii [the volumetric behavour of Nd(N03)3Zmay also be explained in this way].The greatest positive deviations are expected inthe approximate concentration range 5 x 10-4-5 x 10- mol dm- ; an occasionaland misleading parallelism with DHLL must occur in such cases at higher concen-trations (and may lead to erroneous extrapolation from data obtained at far toohigh ionic strength). Greater ii values lead to negative deviations only, similar tothose the DH theory predicts. The inversion occurs at ii z 4 A (in the assumptionthat d In h/dp = 0) and, in consequence, a slope close enough to DHLL should bemaintained up to high concentrations (0.01 or 0.02mol drnp3).The data forK,Co(CN), do not deviate appreciably from DHLL below 0.01 mol dm-3, and theV" value of 143.3 cm3 mol- is probably accurate; however, minor positive deviF. MALATESTA A N D R. ZAMBONI 2515ations cannot be excluded, and a slightly different I/' value, perhaps 0.5 cm3 rno1-lless, is also possible.Both DHLL + B2 and IPBE predict very large, positive deviations from DHLLat low concentrations for the 2:4 salts (fig. 3), except at unusually high ii values.When reasonable ii values are used, the positive deviations are similar or greater tothe experimental values of Mg,Fe(CN),, Sr,Fe(CN), and Ca,Fe(CN),. Accordingto both treatments, the limiting slope cannot be reached for 2:4 salts within theexperimental range of concentrations.From a quantitative viewpoint, the agree-ment with the experimental data is very feeble in the case of the Mayer theory. TheIPBE curves appear more satisfactory and a remarkable improvement can be15.+I -10m-..5I I0.2 0.4 0.6JIFIG. 3.---Plots of 4; against ,'I for 2:4 salts. DHLL + BZ: dotted curves [i = 7.5 and 6 A for (a) and (b),respectively]. IPBE: dashed lines, d In i/dp = 0 [i = 5 and 4.5 A for (c) and (d), respectively]; full lines,d In &/dp = p [ii = 7 and 6 A for (e) and (f), respectively]. As a comparison, Mg,Fe(CN), (0) andSr,Fe(CN)6(0) are given. Dash-and-dot line : DHLL.obtained when using a positive value of d In iildp, of the same order as the com-pressibility coefficient of water, p.The theoretical meaning of this parameter is ratherdoubtful; however, a value of d In h/dp equal to p also improves the resemblance inthe cases of K3Co(CN), and K,Fe(CN), [fig. 2, curve ( f ) ] . Greater d values arecorrespondingly required, which agree better with the d values to be used in &.I7This agrees with previous results for K4Fe(CN)6.4917 A pressure rise could perhapsmodify the water structure around the hydrated ions (increasing the ion hydration),so as to require a positive value of dlni4/dT in terms of the primitive model.However, the inaccurate shapes obtained with d In d/dp = 0 might come from innerinconsistencies of the IPBE treatment. An analysis of the data in terms of someother treatment of the primitive model, for instance HNC,25 would be helpful.IPBE (and DHLL + B2) may be useful for highly charged electrolytes within theconcentration range where DH works for 1 : 1 electrolytes. By plotting the differ-ences between experimental 4v values and & values calculated from IPBE (o2516 APPARENT MOLAL VOLUMES OF STRONG ELECTROLYTESDHLL + B2) one might identify a convergence towards the I/' value, indepen-dently of the arbitrary choice of ii and d In iildp.This method is not effective for thepresent data for 2:4 salts; it seems impossible to obtain sufficiently precise 4vvalues at such high dilutions as the method requires [different extrapolations areapparently suggested for d In i/dp = 0; slightly better results are obtained ford In G/dp = p, but not so good as to provide a reliable value of T/'.This is shown infig. 4 for Mg,Fe(CN)6]. However, the same method can usefully be applied to other(less charged) multivalent electrolytes, when appreciable positive deviations occur32 1 M AA AAI I0.2 0.4\'IFIG. 4. Plots of 4L - 4: (IPBE) against 4'1 for Mg,Fe(CN)6. Upper section, d In G/dp = 0 [C; = 5 (A), 4.5(0) and 4 A (o)]. Lower section, d In i/dp = p [i = 8.5 (A), 7 (0) and 6 A (O)].(but not so high as in the 2:4 salts). Fig. 5 shows the results one obtains forK,Fe(CN),, which may confirm the previous Vo evaluation (108.5 cm3 mol-1)4within z +_ 0.5 cm3 mol-' [the usual Redlich or Owen extrapolationare ineffective for K,Fe(CN),].As for specific? non-electrostatic interactions in dilute and very dilute solutions ofhighly charged electrolytes,' one cannot at present reach any definitive conclu-sion. Remarkable negative deviations in 4L8.27 and very small positive deviations indV5 were noticed for LaFe(CN),, and they can hardly be justified in terms of merelyelectrostatic interactions (DHLL + B2 and IBBE predict very large positive devi-ations for 3:3 salts).On the other hand, the positive deviations observed at lowconcentrations in 4) and & of 2:4 salts (as well as in 4, of 2:317 and 2:216 salts)are satisfactorily justified in terms of the approximate treatments of the primitivemodel by DHLL + B2 and IPBE, and this suggests that interactions are fundamen-tally electrostatic for such salts, Further studies, for other highly charged electro-lytes, are necessary in order to provide a better understanding of such mattersF .MALATESTA A N D R . ZAMBONI 251711210600 o o 8 .-----?@---V+7~ v v v0+-V VI I 0.4,(I10.2FIG. 5.-Plots of $\. - (IPBE) against ,j;I for K,Fe(CN), [from the data of ref. (4)]: d In Gldp = 0,ii = 4 8, (0); d In ildp = ,8 [ A = 4.2 (e), i = 3.5 8, (O)]. Dash-and-dot line: 108.5 cm3 mol-'.The authors thank Prof. A. Indelli for many valuable discussions concerning thisresearch. This work was supported by the Consiglio Nazionale delle Ricerche(C.N.R.).F. Millero, Chem. Rev., 1971, 71, 147 and references therein.2F. H. Spedding, M. J. Pika1 and B. 0. Ayers, J . Phys. Chem., 1966, 70, 2440.3L.G. Hepler, J. H. Stokes and R. H. Stokes, Trans. Faraday Soc., 1965, 61, 20.4A. Billi, F. Malatesta, R. Zamboni and A. Indelli, J . Chem. Phvs., 1974, 61, 4787.5A. Indelli and R. Zamboni, J . C. S. Faraday I, 1972,68, 1831.,A. Indelli and R. De Santis, J . Chem. Phys., 1969, 51, 2782.J. E. Desnoyer, M. Arel, G. Perron and C. Jolicoeur, J . Phys. Chem., 1969, 73, 3346; G. Perron, N. R.Desroisiers and J. E. Desnoyer, Canad. J . Chem., 1976, 54, 2163; A. ROUX, G. M. Mustbally, G. Perronand J. E. Desnoyer, Canad. J . Chem., 1978, 56, 24.E. A. Guggenheim, and R. H. Stokes, Equilibrium Properties of Single Strong Electrolytes (Pergamon,London, 1969).'E. Lange, The Structure oj'Electrolyte Solutions, ed. W. Hamer (Wiley, New York, 1959), chap. 9.''A. Indelli and F. Malatesta, Gazzetru, 1973, 103, 421."E. Mayer, J . Chem. Phys., 1950, 18, 1426.12J. C. Poirier, J . Chem. Phys., 1953, 21, 965.13J. C. Rasaiah and H. L. Friedmann, J . Chem. Phys., 1968, 48, 2742.14H. Muller, Phys. Z., 1927, 28, 324.15E. A. Guggenheim, Truns. Furaday Soc., 1960, 56, 1152-1962, 58, 86.16A. Indelli and F. Malatesta, Gazzettu, 1973, 103, 435.18L. Onsager, Chem. Rev., 1933, 13, 73.l 9 Znorganic Syntheses, ed. W. C. Fernelius (McGraw-Hill, New York, 1946), vol. 2, g. 225.2oW. W. Scott, Standard Methods in Chemicul Analysis (Van Nostrand, New York, 1939).21D. Salyer and T. R. Sweet, Analyt. Chem., 1958, 30, 1632.22B. B. Owen, R. C. Miller, C. E. Millner and H. L. Cogan, J . Phys. Chem., 1961, 65, 2065.23K. S. Kell and E. Whalley, Phil. Truns. Roy. Soc. A, 1965, 258, 565.24L. A. Dunn, Trans. Faraday Soc., 1966,62, 2348.2 5 J. C. Rasaiah, J . Chem. Phys., 1972, 56, 3071.F. Malatesta, Gazzetta, 1979, 109, 3252518 APPARENT MOLAL VOLUMES OF STRONG ELECTROLYTES260. Redlich and P. Rosenfeld, 2. phys. Chem. (Leipzig), 1931,155,65; Z . Elektrochem., 1931,37, 705; 0.Redlich and D. M. Meyer, Chem. Rev., 1964,64,221; B. B. Owen and S. R. Brinkley, Proc. N . X Acad.Sci., 1949, 51, 753.27E. Lange and W. Miederer, Z. Elektrochem., 1956, 60, 362.(PAPER 9,4934

ISSN:0300-9599    

DOI:10.1039/F19807602510

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraduy I, 1980,76,2519-2530Complexation and Chemisorption of Trimethylphosphineon Ni ZeolitesBY ROBERT A. SCHOONHEYDT," DIRK VAN WOUWE AND HUGo LEEMANCentrum voor Oppervlaktescheikunde en Collo'idale Scheikunde,De Croylaan 42, B-3030 Leuven (Heverlee), BelgiumReceived 8th February, 1980After room temperature saturation of dehydrated NiY with trimethylphosphine two complexes areformed in the supercages. They are identified and quantified by reflectance spectroscopy as0.9[Ni(PMe3)J2+ per unit cell and W[(0,)3-Ni-PMe3]2+ per unit cell (0, is a lattice oxygen). Theformer is diamagnetic and trigonal bipyramidnl. the latter is paramagnetic and compressed tetrahedral.[Ni(PMe&I2 +. is only stable in excess PMe,, while the mono-phosphine complex is stable to z 383 K inuucuo.On NIX only the paramagnetic compressed tetrahedral complex is formed. The ligand fieldparameters of this complex were calculated.Chemisorption on lattice and extralattice oxygens gives strongly held O=PMe,, O=P (OMe3)3 and arange of decomposition products such as CO, COz, H20, hydrocarbons and oxygenated P on thesurface. These products were qualitatively identified by i.r. and mass spectrometry.The coordination of 3d transition metal ions in the zeolite cavities primarilydepends on the ligand field strength of surface oxygens relative to that of theadsorbed coordinating molecules. With ligands such as H20, NH,, ethylene-diammine and methylisocyanide the zeolite acts as a solvent. With ligands such asNO, CO, acetylene and olefins zeolitic oxygens remain in the coordination sphere.In this way unusual complexes can be stabilized on the surface.'Y2 The ultimategoal of these studies is not only to characterize these complexes but also to applythis knowledge to develop so-called "heterogenized homogeneous catalysts".,4Usually, these catalysts are phosphine-based. On zeolites, only the smallest ter-tiary phosphines can be adsorbed. Recently it was reported that PMe, reducedCu2+ to Cu' in zeolite Y, whereas on COY the formation of a low spin complex,formulated as lattice-bonded [Co(PMe,),] +, was evidenced by reflectance spec-troscopy and e . ~ . r . ~ * ~ These data illustrate the versatility of transition metal ionzeolites towards PMe3. In this paper we present our results on the interaction ofPMe, with NIX and NiY zeolites.Ni2+ is an interesting cation because the electro-nic spectra of Ni(PMe3)xX2 (x = 2 4 ) and of [Ni(PMe&I2' are well described inthe literature.'EXPERIMENTALSAMPLESLinde NaY and NaX were stirred for 1.728 x lo6 s in 0.1 moldmP3 solutions of NaC1,washed until Cl--free, air-dried and stored in a desiccator over saturated NH,Cl. Nix and NiYwere prepared from these stock samples by ion-exchange at room temperature for 8.64 x lo5 sin 0.01 moldmP3 NiC12 solutions at a so1id:liquid ratio of 1 gdmP3. After exchange, the2512520 Ni2+-PMe3 COMPLEXES ON ZEOLITESsamples were washed until Cl--free, air-dried and stored over saturated NH,Cl solution in adessicator prior to analysis.Chemical analysis for the exchangeable cation content yielded forNiY17:1.33meq Na'g-l, 2.66meqNi2+ g-' and 0.3meqH+g-'; for NiX28:4.17meqNi2+ g- ' and 2.04 meq Na' 8-l. The numbers following the sample symbols are the number ofNi2+ per unit cell.TRIMETHYLPHOSPHINEAn ampoule of PMe, from Strem Chemicals was connected to a vacuum line and frozen inliquid air. PMe, was purified by evacuation at liquid air temperature and then at slightly belowits freezing point. The mass spectrum after these treatments gave no indication of componentsother than PMe,.PROCEDURES AND TECHNIQUESREFLECT A N CE SPECTROSCOPYThe samples were treated simultaneously in the reflectance cells and the McBain balances,both connected to the same vacuum line, in order to attribute specific weight changes tospectral variations.Prior to adsorption of PMe, two pretreatments were performed. The firstone consisted of heating NiY17 and Nix28 in uucuo at 713 and 733 K, respectively, untilconstant weight. In the second pretreatment both samples were similarly dehydrated, O2 wasallowed to adsorb at room temperature and it was then desorbed at 373K until constantweight.Reflectance spectra of the pretreated samples were recorded before and after O2 adsorptionand after O2 removal. PMe, was allowed to adsorb at room temperature. During adsorptionthe samples were kept at ambient temperature with a water jacket around the reflectance cells.Reflectance spectra of the samples saturated with PMe, and after evacuation of PMe, wererecorded at several temperatures between 293 and 573 K.Spectra were recorded on a Cary 17instrument in the type I reflectance mode. The reference was BaSO,. The spectra were tape-recorded, computer-processed and plotted as F(R,) against wavenumber (5000-50000 cm- ')after subtraction of the baseline.INFRARED SPECTROSCOPY AND MASS SPECTROMETRYThin self-supporting zeolite wafers (5-10 mg cm-2) were dehydrated in uucuo at 683 K for3600 s prior to saturation with PMe,. 1.r. spectra were recorded after the pretreatment, aftersaturation with PMe, and after evacuation up to 573 K on a Beckman IR12 double beamgrating instrument in the range 1200-3800cm-'. For the analysis of the desorption productsz 2 g zeolite were connected to the empty i.r. cell through a side arm, pretreated as describedfor the reflectance measurements and saturated with PMe,.The gaseous desorption productswere collected in the i.r. cell and their spectra recorded in the range 1000-3800cm-'. For themass spectrometric analysis of the desorption products z500mg zeolite were loaded in aU-shaped quartz reactor, dehydrated and saturated with PMe, as described for the reflectancemeasurements. During desorption the gaseous desorption products circulated through the zeo-lite bed in a closed circuit. Aliquots of gases were conducted to the Balzers quadrupole massspectrometer QMG 101 A for analysis. The mass range analysed was 0-100.RESULTSPretreatment results in a weight loss which corresponds to 295 and 256 H 2 0UC-' for Nix28 and NiYl7, respectively.At room temperature Nix28 adsorbs 9-8O2 molecules per unit cell and NiY17 8.4. In both cases 5 O2 UC-' remainadsorbed after desorption at 373 K.The reflectance spectra obtained are identical to published spectra and are notreproduced her^.^.^ O2 has no effect on the spectrum of dehydrated Nix28 buR . A , SCHOONHEYDT, D. VAN WOUWE AND H . LEEMAN 2521eliminates the 14 100 cm-' band in the reflectance spectrum of NiY17. This bandwas previously ascribed to Ni +. * y 9The adsorption of PMe, is a fast, exothermic process. The zeolites in the reflec-tance cells must be kept in a constant temperature water bath to avoid excessiveheating. Table 1 gives the amount adsorbed after saturation and desorption atdifferent temperatures in vacuum.Only a minor fraction can be desorbed, even at473 K. This is indicative of chemisorption, while the colour changes accompanyingadsorption are in indirect proof of complexation of Ni2+. The numbers in table 1TABLE AMOUNTS OF PMe, ADSORBED (MOLECULES PER UNIT CELL)saturation with P(CH3j3desorption of P(CH3 j3at 323 Kat 373 Kat 418 Kat 473 K47.5 42.343.4 37.936.7 34.636.7 31.531.7 27.8are calculated as if the residual molecules on the surface are PMe,. Experiments tobe described below indicate that this is not the case due to chemisorption anddecomposition of PMe, upon high temperature evacuation. The adsorption of O2in the pretreatment step had no effect on the subsequent PMe, adsorption.REFLECTANCE SPECTROSCOPYTwo types of complexes are formed as evidenced by the reflectance spectra offig.1 and 2. One complex (I), only formed on NiY17, is characterized by absorptionbands at 18 300 and 34250 cm- '. These bands are removed by room temperatureevacuation of excess PMe,. The second complex (11) is formed on NiX28 and onNiY17 and is thermally stable up to 381 K. It is characterized by 3 bands in then.i.r.-visible region. The position of these band maxima are: 860&8800, 22000 and26000 cm- ' for NiX28; 800&8400, 21 000 and 26500 cm- ' for NiY17. While thefirst two bands are broad and asymmetric, the third is sharp. A broad absorptionencompasses the U.V. region, from which 3 bands are resolved upon desorption ofPMe, above 381 K. They are located at 33500, 38000 and 45000cm-1.Above381 K the spectra of the dehydrated Ni-zeolites are almost completely recovered,although in the U.V. region the 33 500 cm- band remains very pronounced.Thc disappearance of complex (I) on NiY 17 is accompanied by an increase in the8400 and 26 500 cm- ' bands of complex (11). These bands further increase withheating in vacuo up to 381 K. This increase in the intensity of the band withtemperature also occurs on NiX28. These phenomena are indicative of the fact thatcomplex (I) is transformed to complex (11) and that not all the Ni2+ is complexedafter saturation at room temperature.INFRARED SPECTROSCOPY A N D MASS SPECTROMETRYThe regeneration of lattice-bonded Ni2 +, as evidenced by reflectance spectro-scopy, is in contrast to the large amounts of residual PMe, on the surface.Chemi-sorption on lattice oxygens was suspected and this is shown by the ix. spectra offig. 3. Intense absorptions are found in the regions were gaseous PMe, absorb2522 Ni2+-PMe3 COMPLEXES ON ZEOLITES10'1 05 14 23 32 41wavenumber/103 cm-FIG. 2.-Reflectance spectra of PMe, adsorbed on NiX28: (1) saturated; (2) evacuated at 323 K for6.05 x 10's; (3) evacuated at 381 K for 3.24 x 10'sR . A . SCHOONHEYDT, D. VAN WOUWE AND H . LEEMAN 25233100 2900 1750 1550 1350wavenumber/cm -FIG. 3.-1.r. spectra of PMe, on NiY17: (1) evacuated at 683 K for 2700 s; (2) saturated with PMe,; ( 3 )evacuated at room temperature; (4) evacuated at 593 K for 7200 s.(1 282-1348, 141 7-1440 and 285G2970 cm- I).However, the forms of the band sys-tems are distorted with respect to those of gaseous PMe, (see fig. 4). Additionally, abroad asymmetric band around 1660 cm- is generated which intensifies with timeof contact. This band can be removed by evacuation at 373 K but for the elimin-ation of the low frequency shoulder at 1605 cm-' temperatures above 573 K arenecessary. This evacuation procedure leads to 3 groups of bands: a triplet withabsorption maxima at 1305, 1318 and 1345 cm-', a doublet with maxima at 1422and 1432 cm-' and a doublet with bands centered at 2922 and 2998 cm-'. Thelatter is broader than the former due to a low frequency shoulder. The i.r. spectra ofthe gaseous desorption products are shown in fig. 4. PMe, is the only desorptionproduct at 383 K (spectrum 1).At 438 K supplementary bands are revealed at1080crn-' (a weak broad band at 1140cm-' accompanies the 1080cm-' band)and in the CH stretching region, but only a band at 2885 cm-' is clearly resolvedfrom the C-H stretching of PMe3. Above 473 K the rotation-vibration spectrumof CH4 is superposed on the PMe, spectrum. The band centres of the CH4 spec-trum are at 3010 and 1310cm-'.The mass spectral analysis (fig. 5) of the gaseous desorption products confirmsand complements the i.r. data in that PMe3 is the main detectable desorptionproduct below 423 K, but it remains visible in the gas phase at all desorptiontemperatures. At z 500 K CH4 is detected, also in agreement with the i.r. data. Notdetectable by i.r. spectroscopy but clearly visible in the mass spectra are (i) H2which starts to desorb at ~ 4 5 3 K and (ii) a range of products, all of which aredesorbed above ~4450 K.They have characteristic masses at 27, 28, 29, 30 and 322524 Ni2+-PMe3 COMPLEXES ON ZEOLITES3200 2800 1500 1300 1100wavenumber/cm -FIG. 4.--I.r. spectra of gaseous desorption products of PMe,-saturated NiY 17 : (1) after desorption at383 K, p = 20.79 Pa; (2) after desorption at 438 K, p = 48.52 Pa; (3) after desorption at 483 K,p = 138.63 Pa; (4) after desorption at 553 K, p = 97.04 Pa.at 41,42,43,44 and at 71, 72 and 73. There is considerable overlap with the PMe,spectrum, but at the highest temperatures (spectra 4 and 5 ) these masses are clearlyvisible. There is also an effect due to the pretreatment: NiY17 pretreated in O2gives greater quantities of masses 43 and 44 than 41 and 42.The reverse holds forpretreatment il.7 vacuo. Desorption from Nix28 gives the same products but theamount of PMe3 with respect to the other products i s higher than for NiY17.dFIG. 5.-Mass spectra of PMe, and of products desorbed from NiY17 after saturated with PMe3: (1)PMe,; (2) desorption at 453 K; ( 3 ) desorption at 473 K; (4) desorption at 507 K; (5) desorption at 638 KR . A . SCHOONHEYDT, D . VAN WOUWE AND H . LEEMAN 2525DISCUSSIONThe reflectance spectra of Ni-zeolites saturated with PMe, can be interpreted interms of 2 Ni2+-PMe3 complexes without recourse to reduction of Ni2+ as in thecase of C U ~ + . ~ This is in agreement with the fact that in dehydrated faujasite-typezeolites Cu2+ is more easily reducible than Ni2+, not only by H2 but also by othermolecules such as CO, NH3 and ethylenediammine.' O-' This difference in behav-ior conforms with the difference in electrochemical potential^'^ and with ligandfield calculations.'The spectrum of complex (I) is interpreted as that of a spin-paired[Ni(PMe3),I2+ complex.The spectrum of complex (11) is ascribed to a high spinpseudotetrahedral species [(O,),Ni--PMe,] + where 0, stands for lattice oxygen.Evidence for these interpretations is given below.The known complexes of Ni2+ with PMe3, Ni(PMe,),X, (x = 2, 3, 4) and[Ni(PMe3),I2 +, are diamagnetic. In solution bis complexes are the most stable andexcess phosphine is necessary to incorporate more than 2 PMe, molecules in thecoordination sphere of Ni2+.7316 A similar situation exists in the supercages of thezeolites where an equilibrium is established between the surface complexes anduncoordinated PMe, :(Z0-),Ni2+ 3 (ZO-),Ni2+-PMe3 + [Ni(PMe,),l2+ + 3ZO-.(1)The reaction is driven to the right on NiY17 but stops at the pseudotetrahedralcomplex formation on NiX28. This is due to the larger Ni2+ content and thuslower PMe3:Ni2+ ratio in the supercages of NiX28. Other factors affecting thecoordination are the chemisorption of PMe, on surface oxygens and the differencein lattice negative charge density between X and Y.The diamagnetic trigonal bipyramidal complex [Ni(PMe,),] + is proposed onthe basis of the similarity of its spectrum with that of the same complex in solu-tion.I6 In the latter case the ' A ; + 'E'(e''4e'4 -+ e"4 e ', a;') transition is at19200cm-' with a band width of 5000cm-'.The symmetry forbidden and there-fore weak transition ' A ; -+ 'E" (e"4e'4+ ~ , " ~ e ' ~ a ' : ) is at 27800 cm-' and thea(PMe,)+ d(Ni) charge transfer band is around 38000 cm-'. We have ' A ; + ' E 'at 18300cm-' with a bandwidth of SOOOcm-' and o(PMe3)+d(Ni) at34250cm-'. The weak ' A ; + 'E" transition is not seen and is probably hidden bythe 26 500 cm- ' band of the pseudotetrahedral complex. We have eliminated[Ni(PMe,),] +, [Ni(PMe,),] + and [Ni(PMe,),] + as possibilities. Indeed, thefirst 2 cases have, in the homogeneous phase, approximate C2" symmetry andtherefore a system of 3 d-d absorption bands.[Ni(PMe3),I2+ is planar with a verybroad absorption (half band width = 10000 cm- ') encompassing 3 component^.^None of these features is seen in our spectrum.The change of environment from EPA solution (5 : 5 : 2 mixture of diethyl ether,isopentane and ethanol) to the zeolitic supercage slightly affects the ' A; + 'E' tran-sition (red shift of 900 cm-') but effects the L.M.C.T. much more, as evidenced bythe 3750crn-' red shift. This is a solution effect, the zeolite acting as a non-coordinating, anionic solvent. We now analyse the effect in terms of the opticalelectronegativity parameters.' The change in spin pairing energy, ASPE, in goingfrom d8 to d9 during the charge transfer transition is -4/3 D with D = 7B.Then,for the L.M.C.T. transition we have:PMe?(11) (1)vzt = v,, - ASPE = 30000 [xopt(Ni) - xopt(PMe3)] = 30000 AXnpt2526 Ni2+-PMe, COMPLEXES ON ZEOLITESFor B = 500 cm-', Axopt. = 1.3 and 1.4 in the supercages and in EPA solution,respectively. With jlopt(Ni) = 2.0-2.1 this gives xopt(PMe,) = 0.7-0.8 and 0.6-0.7,respectively. There is therefore a slight increase, of 0.1, in optical electronegativityof PMe, in the zeolite. This means more electron-attracting power or less basiccharacter. In other words one could speak of less a-donor capacity in the zeolite.This should also be reflected in the redox potential for these complexes. The effectis too small and hardly out of the experimental accuracy range to attempt ameasurement or a calculation.Note, however, that this red shift of L.M.C.T. bandsfor immobilized complexes appears to be a general phenomenon, as we have de-scribed a similar situation for Cu(en)$+(en = ethylenediammine) on the surface ofclay minerals.The spectrum of complex (IT) with one band in the range 5000-10000cm-1 ischaracteristic of a high spin complex. As all the complexes Ni(PMe3)xX2 withx = 2--5 are diamagneti~,~ the only possibility is that our spectrum is that of apseudotetrahedral complex with 1 PMe, and 3 lattice oxygens in the coordinationsphere of Ni2+. PMe, is a stronger ligand than the lattice oxygens. The complex isthen compressed tetrahedral with an idealized C3" symmetry. The general C3" casewas treated by Klier et al.I5 We have applied their general ligand field potential toour case with the following 2 assumptions: (i) the angle p, 0-Ni-P, equals 100";(ii) G;/G: = Gy/Gy = 10, where P and 0 refer to PMe, and lattice oxygens, re-spectively.G2 and G4 denote the radial integrals:-Ze2 4nR5 9 for point charges. 9G4 = 740 (r4)ion with Y40 = ~ -The ligand field energy diagrams with which we were able to fit the experimentalband maxima are shown in fig. 6. The following assignments are made:,E(,F) --+ ,E(,F), ,A,(,F): 8000-8800 cm-',E(,F) -+ 3 ~ , ( 3 ~ ) , ,A,(,F): 21 000-22000 cm-',E(,F) + ,E(,P): 26500 cm?Calculations of the ligand field and interelectronic repulsion parameters, based onthese assignments, give physically meaningful values which are summarized intable 2. Thus, PMe, is 2.1-2.3 times stronger a ligand than the lattice oxygens andthe orbital reduction factor is in the range 0.88-0.57.This indicates appreciabledeviation from the ionic bonding model but this is expected with a strong electron-rich ligand such as PMe, in the coordination sphere. The absorptions in the U.V.region cannot assist our interpretation. In the temperature range of the thermalstability of the complexes there is only one broad and unresolved band. Above381 K, the complexes are destroyed and the U.V. absorption must be ascribed tocharge transfer bands from the modified surface (modified by chemisorption) tobare Ni2 ions.Before adsorption of PMe, all the Ni2+ is in the small cavities, at least forNiY 17.' Adsorption of PMe, induces migration of Ni2 + to the supercages, until anew equilibrium is attained with complexed Ni2 + in the supercages and residual,uncomplexed Ni2+ in the small cavities.An upper limit to the number of[Ni(PMe3),l2 + complexes can be estimated from the intensity changes of the 800R . A . SCHOONHEYDT, D . VAN WOUWE AND H. LEEMAN 2527/FIG. 6.-Energy level diagram of &orbitals of Ni2+ in compressed tetrahedral configuration[(0J3NiPMe3I2+; right hand side G:/Gz = 2.1; left hand side G:/G? = 2.3.and 26500 cm-' bands of complex (11) upon evacuation (fig. 1). The assumptionsare (i) equal scattering coefficients for all the spectra of fig. 1; (ii) all the Ni2+ ofNiY17 is in the form of complex (11) after evacuation at 381 K. The fact that theband intensity of complex (11) was increased by the 381 K evacuation with respectTABLE 2.-RANGE OF RACAH'S PARAMETER B (Cm-') AND LIGAND FIELD PARAMETERS (Cm-l) FOR(O,),-Ni-PMe, ON Nix28 AND NiY17GSIIB 5.5-7.7 6-8B 872-590 918-653GSI 43604131 5508-5224GP, 10 028-950 1 11 567-1097010 m P e t 7428-7038 8568-812610 D&t 323G3 1 11 408G3872528 Ni2+-PMe3 COMPLEXES ON ZEOLITESto the room temperature evacuation favours this assumption ; (iii) [Ni(PMe,),12 + iscompletely converted to [(O,),-Ni-PMe,] + upon room temperature evacua-tion.The increase in the bands of the latter complex favours this assumption.We have after saturation with PMe, 0.9[Ni(PMe3),l2’ and 6 8 pseudotetrahed-ral complexes. In this way only 10.5-12.5 PMe, molecules are complexed to Ni2+,going up to 17 at 381 K.Most of the adsorbed PMe, is then available for chemi-sorption as shown by the desorption data in table 1. The same holds for Nix28 butnot all the Ni2+ can be transformed into the pseudotetrahedral complex even at381 K and an estimate of the number of complexes present in the supercages aftersaturation is not possible.CHEMISORPTION OF PMe,The low temperature desorption of PMe, covers the decomposition temperatureof the complexes and is therefore due to desorption of physisorbed and coordin-ated PMe,. The species remaining on the surface can be divided in 2 groups: (i)chemisorbed species absorbing at 1305-1345, 1422-1432 and 2910-3000 cm-(spectrum 4 of fig. 4). The latter spectrum is in very good agreement with that ofO=PMe, (1292-1305-1 340; 1420-1437; 2923-2999 cm- 1).20 Therefore it rep-resents strongly adsorbed O-PMe, molecules according to the reactionZ-0 + PMe,’-Z--O====PMe,. (2)Additional evidence for the formation of chemisorbed O=PMe3 comes from theaverage bond energies M-C and M-0 with M = Si, A1 and P.21-23 These aresummarized in table 3.It shows that P has a stronger affinity for 0 than Si and A1and that the Si-C and Al-C bonds are stronger than the P-C bonds. However,the fact that PMe, is found in the gas phase even after high temperature desorptionindicates that at least on some lattice oxygens reaction (2) is reversible. The pres-ence of the 1600-1660 cm- ’ band system is indicative of a strong chemisorptionprocess resulting in destruction of PMe, molecules. This band was not found in thegas phase spectra.Therefore, either the partial pressure of the components was toolow or these species react upon desorption. Note that the gas phase spectraobtained after desorption in the range 423483 K contain, besides the PMe, bands,the 1080, 1140 and 2885 cm-’ bands. We suggest that these bands are due to thephosphorester O=P(PCH3),. Indeed, 1080 and 1140cm-1 are in the range offrequencies for a P-0-CH3 vibration (101G1088 cm-’) and a H3C-0-Pvibration (1 168-1200 cm- ’), r e s p e c t i ~ e l y . ~ ~ ~ ~ ~ Secondly, its boiling point underatmospheric pressure is 470.4 K but is 358 K at 319.97. Pa.21 This range includesthe desorption temperature of our experiments. Therefore, in the presence of “reac-tive” oxygens, reaction (2) proceeds further:\ -.(3)Si ,“O====P(CH,), O=P(OCH,),/// A1 ’and eventually complete oxydation occurs to C02, H 2 0 and oxygenated P.The1600-1660 cm- band includes the deformation band of water, which can subse-quently be incorporated in PMe, following reaction (3). We have not detected C 0 2on the solid phase, but according to the mass spectra C 0 2 is desorbed at hightemperatures. It can result from a thermal decomposition of PMe, analogous tR . A . SCHOONHEYDT, D. VAN WOUWE A N D H. LEEMAN 2529TABLE 3.-AVERAGE BOND ENERGIES (kJ mol- ’) OF M-C AND M-O (M = P, si, Al)P-C in P(CH3)3 :263 P-0 : 397 P=0:585Si-C in Si(CH3)4: 301 Si-0 : 207AI-C in Al(CH,), : 255 A1-0 : 272NMe, or from a carbonate- or formate-like material, formed together with H 2 0and absorbing in the 160&1660 cm-’ range.26-29 The thermal decomposition ofthe phosphine molecules also accounts for the presence of H2, CH4, CO, ethylene,ethane, propane, n-pentane, acetone and propanol, all characterized by peaks inmass spectra in the ranges 26-30 and 3943.CONCLUSIONSThis report shows that zeolites are versatile supports for the synthesis of phos-phine complexes.Thus, we have synthesized [Ni(PMe,),] + and [Ni(0,),PMe3] + by simple gas phase adsorption of PMe, an dehydrated Nix28 and NiY17 zeolites.[Ni(PMe,),]’+ is only stable in excess PMe, and is converted to [Ni(Ol),PMe3]2fby room temperature evacuation of PMe,. [Ni(0,),PMe3I2+ is stable to evacua-tion up to ~ 3 8 1 K.It is a compressed tetrahedral, paramagnetic complex, whosespectral properties can be explained in C3v symmetry. A similar complex has notyet been found in solution chemistry. The spectrum of [Ni(PMe,),l2+ agrees withthat of a diamagnetic, trigonal bipyramidal complex. The fact that it is only formedin the supercages of zeolite Y shows that the PMe,:Ni ratio in the supercages iscritical. However, extensive chemisorption occurs which hampers the complexation.Additional experiments at low Ni-levels are necessary in order to separate theinfluence of the Ni-content and the chemisorption on the complexation of Ni byPMe, in the supercages.This work was sponsored partially by the Petroleum Research Foundation (PRFnr. 10706-AC5) and partially by the Belgian Government (Ministerie voor Wetens-chapsbeleid). R.A.S.acknowledges a grant as “Onderzoeksleider” of the “NationaalFonds voor Wetenschappelijk Onderzoek”. The authors thank Prof. J. B. Uytter-hoeven for his interest in this work.J. H. Lunsford, A.C.S. Symp. Series. 1977, 40, 473.Catalysis, Helerogeneous and Homogeneous, ed. B. Delmon and G. Jannes (Elsevier, Amsterdam,1975).Relations entre Catalysr Homogene et Catalyse HitirogGrze (Editions du CNRS, Paris, 1978).2 C . Naccache and Y. Ben Taarit, Acta Phys. et Chem. Szeged, 1978, 24, 23.’R. G . Herman. Inorq. Chim. Acta, 1979. 34. 119.6 R . A. Schoonheydt, D. Van Wouwe, M. Van Hove, E. F. Vansant and J. H. Lunsford, J.C.S. Chem.’A. Merle, M. Dartigucnave and Y.Dartiguenake, J . Mol. Swuct., 1972. 13. 413.8E. Garbowski, M. Primet and M.-V. Mathieu, Adv. Chem. Ser., 1977, 40, 281.‘E, D. Garbowski, M.-V. Mathieu and M. Primet, Chcm. Phys. Lt)tters, 1977, 49, 247.loY. Y. Huang, J . Catalysis, 1973, 30, 187.“1. E. Maxwell and E. Drent, J. Catalysis, 1976, 41, 412.”P. Peigneur, J. H. Lunsford, W. de Wilde and R. A. Schoonheydt, J. Phys. Chem., 1977, 81, 1179.13P. A. Jacobs, H. Nijs, J. Verdonck, E. G. Derouane, J.-P. Gilson and A. J. Sirnoens, J.C.S. Faraday I,l4 J. B. Uytterhoeven, Acta Phys. et Chem. Szeged, 1978, 24, 53.Comm., 1980, 33.1979, 75. 1 1962530 NiZf-PMe3 COMPLEXES ON ZEOLITES"K. Klier, P. J. Hutta and R. Kellerman, A.C.S. Symp. Ser. 1977, 40, 108.16J. W. Dawson, T.J. McLennan, W. Robinson, A. Merle, M. Dartiguenave, Y. Dartiguenave and H. B.Gray, J . Amer. Chem. SOC., 1974, 96, 4428.A. B. P. Lever, Inorganic Electronic Spectroscopy (Elsevier, Amsterdam, 1968). ' R. A. Schoonheydt, A. Maes, A. Cremers and J. B. Uytterhoeven, J . Phys. Chem., in press.19P. Gallezot and €3. Imelik, J . Phys. Chem., 1973, 77, 652.'OL. W. Daasch and D. C. Smith, J . Chem. Phys., 1951, 19, 22.21 Handbook of Physics and Chemistry, ed. R. C. Weast, (The Chemical Rubber Co., Cleveland, Ohio,22L. Maier, Progr. Inorg. Chem., 1963, 5, 27.23D. J. C. Jates, G. W. Dembinski, W. R. Kroll and J. J. Elliott, J . Phys. Chem., 1969, 73, 91 I.24L. C. Thomas and R. A. Chittenden, Spectrochim. Acta, 1964, 20,489.52nd edn, 1971).L. C. Thomas, The Identification of Functional Groups in Organophosphorus Compounds (AcademicPress, New York, 1974), chap. 3.26P. A. Jacobs and J. B. Uytterhoeven, J . Catalysis, 1972, 26, 175.27P. A. Jacobs, B. K. G. Theng and J. B. Uytterhoeven, J. Catalysis, 1972, 2, 191.'*P. A. Jacobs, F. H. Van Cauwelaert, E. F. Vansant and J. B. Uytterhoeven, J.C.S. Faraduy I, 1973, 69,29P. A. Jacobs, F. H. Van Cauwelaert and E. F. Vansant, J.C.S. Faraday I , 1973,69, 2130.1056.(PAPER 0/232

ISSN:0300-9599    

DOI:10.1039/F19807602519

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980,76,2531-2541Determination of the Viscosity of Molten KN03with an Oscillating-cup ViscometerBY YOSHIYUKI ABE, OTOYA KOSUGIYAMA, HIROYUKI MIYAJIMAAND A m NAGASHIMA"Faculty of Engineering, Keio University,3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223, JapanReceived 29th August, 1979An oscillating-cup viscometer for studying high temperature melts was built and absolute measure-ments on the viscosity of molten KN03 along the saturation line have been performed.For an accurate numerical evaluation of the viscosity, an assessment on existing rigorous workingequations was made.The temperature range of the present measurements covered up to 973 K where no earlier experi-mental data are available. In the temperature range where earlier experimental data are available,the measured viscosity of the present study was in good agreement with those of recent studies.A correlation based on the present results and those of other selected studies was made and theequation is believed to be valid over the whole temperature range where molten KN03 is stable.Compared with a current recommendation for the viscosity of KN03 proposed by Janz ef al., thepresent equation gives more satisfactory results for the viscosity of molten KN03.Although a number of experimental studies on the viscosity of high temperaturemelts exist, few satisfactory studies have been performed because of experimentaldifficulties mainly due to elevated melting points and severe conditions of chemicalhandling.Hence, even if experimental data obtained by different authors are avail-able, a sizeable disagreement is commonly found.Since precise information onthe thermophysical properties of high temperature melts is urgently required inpower technology and nuclear power technology in particular, the present situationmust be improved.is based on theidea of establishing " calibration-quality " data for moderately high and high tem-perature physical property measurements; KN03 and NaCl were selected as thestandard salts. For the viscosity of KN03, several consistent data sets obtained bydifferent experimental methods are available and the equation recommended byJanz et aL2 gives a reliable viscosity value up to 750 K. However, taking into accountthe fact that the melting point of KN03 is 607.4 K and comparing with other physicalproperties such as the density and the electric conductivity whose values are establishedup to 880 K with even smaller ~ncertainties,~ further precise knowledge of theviscosity in the extended high temperature range is still required.On the other hand,the situation as regards the viscosity of NaCl is also unsatisfactory [a recommendationappearing in ref. (3) is definitely doubtful1.tThe present study is an attempt to construct a new apparatus and to performprecise measurements on a series of high temperature melts over a wide temperaturerange. For this purpose the oscillating-cup method was applied and the viscosityThe Molten Salts Standards Program initiated by Janz in 1974t We will publish new results of the viscosity of NaCI.2532532 VISCOSITY OF KNOjcalculation was carried out on an absolute base using a working equation which wasselected after an assessment of existing rigorous working equations.The viscosity measurements on molten KN03, for the first step, were performedin the temperature range 623-973 K with an estimated accuracy of 1.1 %.A newcorrelation for the viscosity of KN03 was formulated by evaluating the resultsobtained in the present study and those of some other earlier studies judged by theauthors to be reliable.EXPERIMENTALPRINCIPLEFor the viscosity measurements on high temperature melts which have relatively lowviscosities, two major types of experimental method are generally applicable, that is, thethe capillary method and the oscillation method.While the capillary method has mostlybeen used as a relative method in a number of studies on high temperature melts, difficultiesdue to the corrosion and the machine working mean that the temperature range of this methodis limited to w 1100 K as the highest temperature.The oscillation method, since it enjoys two different variations, can be applied in the mostpreferable way. One method has a suspended pendulum, generally a disc, cylinder orsphere, immersed in the fluid to be measured and the other method has the fluid containedin a suspended spherical or cylindrical hollow crucible. Using either method, the viscositycan be determined from the geometrical dimensions of the suspended system and the charac-teristics of its torsional oscillation.The second method, the so-called oscillating-cup viscometer, consists essentially ofan oscillating system with a thin suspension wire and a hollow cylindrical cup which containsfluid.Once a torsional oscillation is given to the system, the executed oscillation is graduallydamped with a constant period and a constant decrement and these parameters are charac-terised by the viscosity and density of the fluid, the moment of inertia of the system andthe dimensions of the cylindrical cup. In other words, the viscosity can be evaluated fromthese measurable parameters.The following features enabled us to believe that the oscillating-cup method would bethe most suitable one for the viscosity measurements on high temperature melts : (1) Onlya small amount of specimen is required for measurements (for instance, 14.666 g in the presentstudy).(2) The desired temperature condition can be easily obtained since the specimenis contained in the small cup. (3) Suitable material for the cup can be successfully contactedwith corrosive fluids and its cylindrical shape allows precise machine working.WORKING EQUATIONIn contrast with the inherent advantages of the oscillating-cup method over others, itsmathematical complexity has restrained its extensive application. Most earlier measure-ments, therefore, were performed on a relative base. Meanwhile, several efforts have beenmade to derive a rigorous solution for this viscometry and the solutions thus obtained canbe classified into two types.The first ~ n e , ~ - ~ which involves the Bessel functions of complexarguments, has not been able to be employed practically for viscosity calculations. Onlymanageable approximations from this rigorous solution allowed the viscosity calculationon the absolute base; however, the applicability of such approximations is dependent onthe characteristic dimensions of the cup and the kinematic viscosity of fluid under investiga-tion. Further details about such restrictions are mentioned else~here.~ The secondsolution lo was derived by introducing the Laplace transform so as to facilitate the compli-cated evaluation.In the present study the following assessment on these two rigorous solutions was madeand the numerical evaluation of the measured viscosity was carried out using a selectedworking equation.First, we briefly show how the rigorous solutions were derived in theprevious analysesY . ABE, 0. KOSUGIYAMA, H . MIYASIMA AND A. NAGASHIMA 2533The motion of the oscillating system is described by the following basic equationd2a dadt2 dtI -+L -+Ka = 0where I is the moment of inertia of the system, a is the azimuthal angle of the oscillation, Kis the torsional constant of the suspension wire and L(da/dt) denotes the damping torque onthe system. To obtain an expression for L, one has to solve the Navier-Stokes equationtogether with the equation of continuity under the following assumptions : (1) There is nomotion in the axial direction. (2) There are no body forces except gravity.(3) There is noslip at the wall of the cup. (4) The velocity is small enough to neglect the non-linear terms.A final form for L is I ,m h r J2( m r) a) tanh (Znh)+2m4r2 c = 2nV[ - Jl(mr) n = l lik:where Z i = k i - m2 and m2 = (p/qT)(S- 2ni), (i = , / T ) r is the inner radius of the cup, h theheight of the fluid in the cup, p the density, q the viscosity, kn the roots of Jl(knr) = 0 and J1and J2 denote the Bessel functions of order 1 and 2, respectively.Consequently, the rigorous working equation for the viscosity evaluation is expressed inthe formwhere A = 6/2n, 6 represents " actual " logarithmic decrement and Tand To are the periodsof oscillation for the presence and absence of fluid, respectively. The left-hand side of eqn (3)is a modified form of that appearing in ref.(6) where some errors are found in the calculation.As we have mentioned, no attempt has so far been made to solve eqn (3) directly since theBessel functions of complex arguments could not be generated. We have employed sub-routines, however, which could generate the Bessel functions of complex arguments with anaccuracy of 0.001 % and, therefore, make it possible to solve eqn (3) without any approxima-tion.From eqn (3) one can obtain two viscosity values independently by solving either theimaginary or real part. In the imaginary part the term T/To-1, which is nearly zero, isdominant. On the other hand, the value 6 is dominant in the real part, so that the realpart is considered to give a more accurate viscosity value.Alternatively, another approach to obtain an expression for L was developed by Kestinand Newell.lo They facilitated the complicated evaluation by introducing the Laplacetransform and a series of dimensionless space coordinates.Their analysis took account ofan intrinsic damping torque due to internal friction of the suspension wire in addition to thatdue to the fluid contained in the cup. The rigorous working equation they derived iswhere(S+Ao)2+ 1 + D(S) = 0, (4)I' is the moment of inertia of the fluid in the cup, a0 is the logarithmic decrement due to thesuspension wire, ,Un are the roots of J1(pn) = 0 and qo and to are the dimensionless spacecoordinates h(2np/qTo)* and r(2rp/qTo)*, respectively.Eqn (4) also gives two viscosityvalues for the same reason as eqn (3). However, the real part is considered to give a lessaccurate viscosity value in this case, since the relationship between the real and imaginaryparts held in eqn (3) reverses in eqn (4)2554 VISCOSITY OF KN03Prior to the numerical evaluation of the measured viscosity, convergence manners andestimated accuracies for these four equations, the imaginary and real parts of both eqn (3)and (4) were examined. Table 1 lists the viscosity values calculated from each equationwith summation terms up to 10, 30, 50 and 100; estimated accuracy for each case is alsogiven. The accuracy was estimated by evaluating all the uncertainties of the values appearingin the equation. The considerable difference between the period dominant equations,imaginary part of eqn (3) and real part of eqn (4), and the decrement dominant equations,real part of eqn (3) and imaginary part of eqn (4), indicates that an uncertainty of the periodsTand To due to conducted thermal effect on the suspension wire is possibly greater than weestimated.However, for the decrement dominant equations, their accuracy estimationsare believed to be reasonable since even such an uncertainty scarcely affects them.TABLE 1 .-COMPARISON OF THE RIGOROUS EQUATIONS~50 100n = l c c equation accuracy( %) n = 1 n= 1 n= 1estimatedreal part of eqn (3) * 1.1 2.104 2.104 2.104 2.104imaginary part of eqn (3) & 5.1 2.71 5 2.71 5 2.715 2.715real part of eqn (4) & 3.5 3.781 3.065 2.924 2.819imaginary part of eqn (4) & 1.1 2.123 2.098 2.099 2.099a First experimental value at 673 K was tested for these calculations.The slight difference between the values obtained from two decrement dominantequations may be attributed to the difference in evaluation methods for 6.In eqn (3) the" actual " decrement S is regarded as a value which is corrected by subtracting the intrinsicdecrement do from the measured one. On the other hand, both intrinsic and measureddecrements are individually taken into account in eqn (4).Although table 1 indicates that eqn (3) is converging with a smaller sum, eqn (4) canconverge more rapidly in the iteration process and takes less time. For these reasons theimaginary part of eqn (4) was selected as the present working equation, but both decrementdominant equations can still be regarded to give identical viscosity values within the accuracyof the present measurements.Note that the viscosity calculated from an approximation deduced by Roscoe,' whichhas been employed by most of authors who made absolute measurements by the oscillating-cup method, was always 0.6-1.5 % lower than the rigorous one, so that the approximationcannot be validly applied in the present situation.APPARATUSA schematic diagram of the main part of the apparatus is shown in fig.1. The oscillatingsystem, which consists of a Pt92-W8 suspension wire l1 of 0.2 mm diameter (2), a reflectionmirror with its holder (3), an inertial disk (4), a molybdenum connecting rod ( 5 ) and acylindrical cup (6), is suspended in a closed vessel.The vessel can be evacuated to w 10-3Paso that disturbances due to any remaining gas can be eliminated during the run. An initialtorsional oscillation is fed to the suspension system using an oscillation initiator (1) locatedon the top of the vessel.The specimen contained in the cup is heated with an electric furnace which has fourteenheating elements made of Sic and the temperature of the cup can thus ultimately reach1800 K. The temperature is measured by means of five Pt-Ptl3Rh thermocouples (10)arranged around the cup. Each thermocouple was previously calibrated at the meltingpoints of six extra pure metals, uiz., tin (231.968"C), zinc (419.58"C), aluminium (660.46"C)Y . ABE, 0.KOSUGIYAMA, H . MIYASIMA AND A . NAGASHIMA 2535VacuumFIG. 1.-High temperature oscillating-cup viscometer. (1) oscillation initiator ; (2) Pt92-W8suspension wire ; (3) mirror ; (4) inertial disc ; (5) molybdenum connecting rod ; (6) cylindricalcup ; (7) molybdenum radiation shields ; (8) AI2O3 pipe ; (9) Sic heating elements ; (10) Pt-Ptl3Rhthermocouples.RG. 2.-Cylindrical cup2536 VISCOSITY OF KN03silver (961.93"C), gold (1064.43"C) and palladium (1 554°C). Electromotive forces cot-responding to these melting points were determined by applying the wire method l2 and itsestimated uncertainty is believed to be < 0.5 K even at the highest temperature.The cylindrical cup is made of stainless steel and is shown in detail in fig. 2 with majorcharacteristic dimensions.In any kind of oscillation viscometry, experimental error is mostly governed by an un-certainty due to the decrement measurement.In the present study the measurements ofthe period and the logarithmic decrement were perfolmed using the optical measuringsystem shown in fig. 3. The motion of the torsional oscillation can be detected as themotion of a beam which is a reflection of a He-Ne laser from the mirror fixed to the sus-pension system. The laser and the detecting devices are placed w 2 m from the viscometer.Slit A has to be properly placed so that it is positioned at the centre of the oscillation andslit B is placed beside slit A approximately 15 cm from it. When the moving reflectioncomes to a photomultiplier through the slit A or B, a signal to trigger a digital counter isgenerated.Mirr Slit A MirrorHe-Ne Laser nSlit B 3 Photomuttiptier %: Counter (period)Counter (logarithmicPrinterFIG.3.-Optical measuring system.The period is determined from a successive passing of the reflection at the slit A. Thelogarithmic decrement is also optically determined from an increment of successive timeintervals of the reflection passing through between slits A and B,13 that is,where ti is the time interval at an arbitrary moment and ti+n is that of nth period later. Theseprocedures are also graphically shown in fig. 4. The optical measuring system thus allowedprecise measurements for the decrement and the period, with an accuracy of k0.3 and& 0.003 %, respectively.PROCEDUREBefore the viscosity measurements were made, the moment of inertia, the period and thedecrement were measured under conditions exclusive of the fluid.The moment of inertiaat 20°C was determined using two different brass rings whose moments of inertia wereaccurately known. The period and the decrement were measured in the temperature rangefrom 20°C to the highest temperature where the viscosity measurements would be performed.While the period was fairly well represented as a linear function of temperature, the decrementwas found to be constant throughout the whole temperature rangeY. ABE, 0. KOSUGIYAMA, H. MIYAJIMA AND A . NAGASHIMA 2537The specimen, KNO, of purity 99.99 %, was first dried at 200°C in vacuo for 20 h andplaced in the cup in a nitrogen atmosphere.With the aid of a nut and a copper gasket, thiswas finally sealed in a vacuum chamber. Consequently, the measured viscosity turns out tobe a value along the saturation line.AFIG. 4.-Damped oscillation and time measurements.RESULTS AND DISCUSSIONThe viscosity measurements on molten KN03 were then performed in the tem-perature range 623-973 K at intervals of z 50 K. At each temperature the measure-ment was repeated several times and a reproducibility of k0.5 % was attained.The accuracy was estimated to be & 1.1 %. The density was calculated from anequation recommended in ref. (3). Table 2 iists the experimental results and eachnominal viscosity corrected to the nominal temperature.At higher temperatures a problem due to decomposition may arise.In theatmosphere the decomposition of KN03 is said to begin near 920 K.I4 At least atthe highest temperature of this study, 973 K, it is likely that decomposition has takenplace to some extent. The equilibrium constant at the highest temperature for thefollowing reactionis calculated to be 1.43 x lO-I.l4 Furthermore, taking account of the volume of themelt and of the space above the melt inside the cup, 9.14 and 4.67 cm3, respectively,the mole fraction of KNOz produced is estimated to be 0.025 and therefore the partialpressure of the evolved oxygen in the closed cup will be z 30 bar when vaporizationof the melt is not considered. The decomposition is considered to have two effectson the measured viscosity. First, the effect due to the evolved oxygen, from a roughestimation of the effect on the viscosity of KN03, is shown to influence the measuredviscosity values by 0.6 % at the most, since the logarithmic decrement is approximatelyproportional to the square root of the product of the viscosity and the density of thesubstance.When the vaporization of KN03 is taken into account, the estimatedinfluence will be smaller. Secondly, the effect due to the produced KN02 is estimatedto be of the order of 0.5 % by assuming that the viscosity for the mixture of KN03and KN02 is approximately expressed by a linear interpolation of their viscosityvalues extrapolated to the referred temperature.KN03 = KN02+902 (62538 VISCOSITY OF KNOSFrom these estimations and there being no anomaly in the measured viscosity,we conjectured that the influence of the decomposition on the present measurementsshould have been negligibly small or comparable to the experimental error at the most.In the case of unassociated liquids, it is empirically known that the ArrheniusTABLE 2.-EXPERIMENTAL RESULTStemperature Y mom./K T/s 6/10-2 /10-3Pa s Tnom./K /10-3Pa s622.8623.3624.0672.8672.1672.3671.8723.7723.5723.3723.7724.3773-5772.9773.3773.7773.1813.18 12.7812.4812.3812.3873.1872.9873.2873.5923.5923.1923.1923.4923.2973.8973.8973.5973.1973.36.955 566.956 276.955 846.955 966.956 156.956 316.956 356.958 826.958 616.958 496.958 756.958 786.960 766.960 486.960 526.960 496.960 506.963 746.963 586.963 516.963 526.963 616.968 776.968 566.968 576.968 486.972 906.972 686.972 786.972 636.972 726.976 046.976 396.976 046.976 116.976 311.81491.81291.80831.69621.69581.69571.70001.60191.60001.60021.59951.59891.51 531.51511.51261 S1051.51261.45091 .a 6 71.45021.44961.44871.36961.36801.37061.371 51.31831.32281.31851.32131.31861.27551.28071.27571.27851.27582.7402.7382.7102.0992.0972.0972.1151.6971.6911.6921.6891.6851.3981.3981.3911.3851.3911.2091.1991.2071.2061.2040.9990.9951.0011.0020.8790.8880.8800.8850.8800.7860.7950.7870.7920.787623 .O673 .O723.0773.0813.0873.0923.0973.02.7372.7432.7262.0972.0882.0902.1011.7021.6941.6941.6941.6941.4011.3981.3931.3891.3921.2091.1981.2051.2031.2011 .0000.9951.0021.0030.8800.8880.8800.8860.8800.7870.7960.7880.7920.78Y .ABE, 0. KOSUGIYAMA, H. MIYAJIMA A N D A . NAGASHIMA 2539equation is valid to describe qualitatively their temperature dependence of theviscosity. A least-squares fit of the present results to the Arrhenius equation yieldswith q in Pa s and R in J K-l mol-l. Although this equation represents themeasured viscosity (see fig. 5), a slight systematic deviation has been observed athigher temperatures.At lower temperatures some reliable studies by other authors are available.Taking these selected earlier experimental data into account in addition to the presentq = 8.541 x exp (17 928/RT), (7)-0.51 I I I I I1.0 1.2 1.4 1.6 1.8103 KITFIG.5.-Natural logarithm of measured viscosity (9 in Pas) as a function of inverse temperature.experimental data, an attempt at establishing a better correlation was made. Apolynomial expression with an inverse power series was found to be quite adequateand the resultant correlation has the form6q = aiTmi, (8)i = lwith the coefficientsQ, = - 1.298 806 x lo4,a3 = -9.978 276 x lolo,a2 = 5.845 601 x lo7,a4 = 8.421 488 xa5 = -3.492 264 x 10l6, a6 = 5.770 989 xThe measured viscosity values are fairly well represented with a standard deviationof 0.69 %.While eqn (7) is a simpler form with only two disposable parameters,to describe the viscosity over a wide temperature range with sufficient accuracy eqn(8), which is also based on other selected earlier experimental data, is more adequate.Fig. 6 shows the deviation plots of all the available data which have been reportedso far in reference to eqn (8) and table 3 lists the details of these experiments and thestandard deviations from eqn (8). The last column of table 3 also lists standarddeviations but from the value in ref. (2) which, as we have already mentioned, giv2540 VISCOSITY OF KN03the viscosity up to 750 K. The calculation of the standard deviations, therefore,excludes the experimental data beyond this temperature limit.When theseeliminated data are also taken into account, the standard deviations turn out to bethe values appearing in the parentheses of the last column. Within the temperaturerange the values in ref. (2) certainly give the viscosity with nearly the same standard4.0-4.0 0-6.0 * 1. I I 1 1 I600 700 800 900 1000temperature /I<FIG. 6.-Deviation from qcalc [eqn (8)]. 0, ref. (15) ; 0, ref. (16) ; A, ref. (17) ; +, ref. (18) ;A, ref. (20); 0, ref. (21); V, ref. (22); V, ref. (23); c), ref. (24); A, ref. (25); 0, ref. (26);H, ref. (27) ; 0, ref. (28) ; a, ref. (29) ; *, present study.TABLE 3 .-SUMMARY OF PREVIOUS VISCOSITY MEASUREMENTS ON KNOjs.d. fromref. temperature stated s.d. from valuesno. method range/K accuracy( %) eqn (8)( %) in ref.(2)( %)151617181920212223242526272829capilIarycapillaryoscillating-discoscillating-ballcounterbalancedoscillating-balloscillating-cupcapillarycapillarycapillaryoscillating-balloscillating- balloscillating-balloscillating-cupcapiHarysphere606-686620.2-779.2622-680621-815620-700630.2-716.2632-764621-764616-765613.2-743.2615-764615-671625.6-685.2635-78761 3-7431- +22f l21.81 -221.2- + 1.51.91.53 .O0.81.20.50.61.80.71.63.60.62.50.72.01.6 (2.6)2.90.7 (2.8)0.90.4 (1.3)0.7 (0.8)1.9 (1.8)0.61.2 (1.6)3.11 .o2.0 (4.4)0.Y. ABE, 0. KOSUGIYAMA, H. MIYAJIMA AND A . NAGASHIMA 2541deviation as eqn (8), however, it is obvious from the last column of table 3 that theextrapolation of the values to higher temperatures is no longer valid.From thisstandpoint eqn (8) of the present study is expected to be more adequate as the viscosityequation of KN03 for calibration use. Eqn (8) can represent the viscosity valueswith an uncertainty comparable to the experimental errors and is believed to bevalid over the whole temperature range where molten KN03 is stable.We thank Prof. J. Kestin of Brown University for his helpful suggestions on theapparatus and Drs. K. Furukawa and H. Ohno of Japan Atomic Energy ResearchInstitute for their kind advice on the chemical preparation of the molten salts.G. J. Janz, personal communication.G. J. Janz, R. P.T. Tomkins, C. B. Allen, J: R. Downey Jr, G. L. Gardner, U. Krebs andS. K. Singer, J. Phys. Chem. Ref. Data, 1975, 4, 871.G. J. Janz, F. W. Dampier, G. R. Lakshminarayanan, P. K. Lorenz and R. P. T. Tomkins,Nat. Stand. Ref. Data Ser., Nat. Bur. Stand., 1968, 15.E. G. Shvidkovskii, Uch. Zap., Mosk. Gos. Uniu., 1944, 74, 135.L. S. Priss, Zh. Tekh. Fiz., 1952, 21, 1050.M. R. Hopkins and T. C. Toye, Proc. Phys. SOC., 1950, €363,773.R. Roscoe, Proc. Phys. Soc., 1958, 72, 576.R. D. Reeves and G. J. Janz, Trans. Faraday SOC., 1965, 61,2300.J. M. Grouvel, Ph.D. Thesis (Brown University, Providence, R.I., 1974).J. Kestin and J. R. Moszinskii, Brown University Report, AF891/11, 1958.lo J. Kestin and G. F. Newell, 2 . Angew. Math. Phys., 1957, 8,433.l2 T. Shimotsuma, J. Nishikawa, Y. Sato, M. Awano and K. Sato, Keisoku To Seigyo, 1965,4,848.l3 J. Kestin and H. E. Khalifa, Appl. Sci. Res., 1976, 32, 483.l4 K. H. Stern, J. Phys. Chem. Ref. Data, 1972, 1, 747.l5 R. Lorenz and H. T. Kalmus, 2 . phys. Chem., 1907, 59,244.l6 H. M. Goodwin and R. D. Mailey, Phys. Rev., 1908,26,28.l 7 C . E. Fawsitt, J. Chem. SOC., 1908, 93, 1299.R. S. Dantuma, 2. Anorg. Allg. Chem., 1928, 175, 1.l9 K. Ogawa, Nippon Kinzoku Gakkaishi, 1950, 14B-2, 49.2o I. G. Murgulescu and S. Zuca, 2 . phys. Chem., 1961, 218, 379.22 P. I. Protsenko and 0. N. Razumovskaya, Zh. Prikl. Khim., 1965, 38, 2355.23 R. E. Wellman, R. DeWitt and R. B. Ellis, J. Chem. Eng. Data, 1966, 11, 156.24 A. Timidei, G. Lederman and G. J. Jam, Chem. Imtr., 1970, 2, 309.25 S. Zuca, Rev. Roum. Chim., 1970, 15, 1277.26 D. Dumas, K. Grjotheim, B. Hogdahl and H. A. (aye, Acta. Chem. Scand., 1970, 24, 570.27 T. Ohta, 0. Borgen, W. Brockner, D. Fremstad, K. Grjotheim, K. Trarklep and H. A. aye,28 T. Yokoo, M. Saito, Y. Kato and T. Ejima, Nippon Kinzokugakkai Shunkitaikai Koen Gaiyo,29 G. J. Janz, S. W. Lurie and G. L. Gardner, J. Chem. Eng. Data, 1978,23, 14.G. J. Janz and F. Saegusa, J. Electrochem. SOC., 1963, 110,452.Ber. Bunsenges. Phys. Chem., 1975, 79, 335.1978, 157.(PAPER 9/1377

ISSN:0300-9599    

DOI:10.1039/F19807602531

出版商:RSC    

年代:1980

数据来源: RSC

摘要:

J.C.S. Faraday I, 1980,76,2542-2551New Zealand Allophanes: A Structural StudyBY RITA VANDICKELEN, GILBERT DE ROY AND ETIENNE F. VANSANT*Department of Chemistry, Universitaire Instelling Antwerpen,Universiteitsplein, 1, B-2610 Wilrijk, BelgiumReceived 18th October, 1979Three natural allophane samples from New Zealand, the Egmont, Waxy-Pan and iron-richRuapehu, were investigated by infrared spectroscopy, X-ray diffraction and Mossbauer spectroscopy.The specific surface area and the pore sizes of these samples increase as the pretreatment temperatureis raised, up to a temperature where a decrease of the surface area and a narrowing of the pores isobserved. This can be explained by two competing phenomena : the release of water from the poresand the collapse of other pores.In comparison with the Egmont and Waxy-Pan allophanes, theiron-rich Ruapehu sample contains considerably less pores of sufficient size to accommodate water.Its exceptional chemical composition and the infrared spectra, X-ray diffraction patterns andMossbauer spectra suggest the presence of a separate, amorphous iron (111) oxide phase in the poresof this allophane.Amorphous minerals are very important components of the fertile soils of NewZealand. They mainly consist of allophanes and oxides of silicon, aluminium andiron of volcanic origin. The presence of allophanes is essential to the soil fertilitybecause they are responsible for the high natural water content of the soil.The allophane minerals are composed of an aluminosilicate and a hydroxy-polyalumina phase.l The aluminosilicate phase is formed of randomly linked Si-and Al-tetrahedra and possesses a permanent negative charge. The octahedralpolyalumina phase contains a pH-dependent positive charge caused by broken bonds.A lot of plant nutrients including humic substances can be adsorbed on these chargedsites.The allophanes are present in the clay fraction (< 2 pm) of the soil in aflocculated form that is stable over a wide range of pH values (3-7) and allows a gooddrainage.andpoor infrared Nevertheless, it has been observed that these allophanesexhibit an equivalent of internal surface.6 In order to explain this, a structuralmodel was proposed by Kitagawa on the basis of electron microscopic data. Thismodel is still tentative and some differences in properties can even be explained byvariations of the chemical comp~sition.~This study is focused on the active surface of three New Zealand allophanes:the Waxy-Pan, the Egmont and the iron-rich Ruapehu.Their X-ray powderdiffraction patterns and infrared spectra will be examined and the coordination andvalency state of the iron atoms present will be investigated using Mossbauer spectro-scopic techniques.The randomness of the allophanes causes insignificant diffraction patternsEXPERIMENTALMATERIALSThe allophane samples originated from New Zealand (Waxy-Pan, Egmont and Ruapehu)and were kindly supplied by Dr. K. G. B. Theng, Soil Bureau, Lower Hutt, New Zealand.254R . VANDICKELEN, G . DE ROY A N D E.F . VANSANT 2543They were chemically analysed and their compositions are summarized in table 1. Thenitrogen gas used in the sorption experiments was from Matheson Gas Products (certifiedpurity > 99.99 %) and was not further purified before use.TABLE THE CHEMICAL COMPOSITIONS OF THE NEW ZEALAND ALLOPHANE SAMPLES (INWEIGHT %)iron-richWaxy-Pan Egmont RuapehuSi02A1203Fe203H20aNa20K2OCaOTi02p205total29.5239.494.3923.651.960.040.150.500.30100.0032.0932.887.0116.109.630.510.410.800.6011.341.1785.780.330.010.010.480.070.82100.03 100.01Q Obtained by calcination at 383 K.ADSORPTION METHODSThe adsorption experiments were carried out in a conventional constant-volume B.E.T.apparatus.The allophane samples (0.5-1 .O g) were fist outgassed Pa) at varioustemperatures (table 2). Because of their lack of thermal stability the heating was carriedout stepwise with increments of 50 K.Subsequently the adsorption-desorption isotherms of N2 were set up while cooling thesample to liquid nitrogen temperature (77 K) in order to determine the active surface areaof the allophanes.8 Furthermore, the pore size distribution was obtained from the hysteresisregion of this isotherm. For medium-sized and large pores the corrected modelless methodof Brunauer et a1.' was used. It is based on the capillary condensation phenomena. Themicropore size distribution was obtained by the m.p. (micropore) method,1° an extensionof the so-called t-method of De Boer.'l Both methods use the desorption branch of theis0 t herm.X-RAY DIFFRACTIONThe diffraction patterns of th? allophane samples were recorded on a Scintag PAD-11automated diffractometer at a rate of 0.4"(28) min-l, using Cu-Ka radiation from a fine-focus X-ray tube.A Ni-filter was mounted between the flat powder specimen and theproportional counter in order to avoid Cu-Kp radiation as well as Fe-fluorescence.INFRARED SPECTROSCOPYInfrared spectra were obtained using a Beckmann 4240 spectrometer equiped with avariable reference beam attenuator. The allophane samples were pressed in a 2 % KBrdisc and recorded in air. The scanning rate was 600 cm-l min-l using the conventional slitprogram.MOSSBAUER SPECTROSCOPYThe Mossbauer spectra were recorded using a Canberra Quanta automated spectrometer.The source consisted of approximately 5 mCi of 57C0 in Pd as prepared by New Englan2544 NEW ZEALAND ALLOPHANES: A STRUCTURAL STUDYNuclear Corp., Boston, Mass.All data were obtained with the source and the absorberat room temperature. The spectrometer was calibrated using natural iron foil and sodiumnitroprusside. Isomeric shift data are expressed relative to the nitroprusside standard.The allophane samples were pressed in 150 mg 2 cm2 discs. Spectra were collected untilthe statistical count rate error became less than 3 :! of the adsorption peak height. Thespectra were subjected to a least-squares fit to a Lorentzian line shape and the Mossbauerhyperfine parameters were calculated from the fit assuming doublets.c I1 .o0.40.2 :'= 01.010 0.5 0 0.5 0 0.5 1r/nmFIG. 1.-Summary of the pore size distributions in New Zealand allophane samples pretreated atdifferent temperaturesR . VANDICKELEN, G . DE ROY AND E . F . VANSANT 2545TABLE 2.-sUMMARY OF PRETREATMENT TEMPERATURES AND SURFACE AREAS OF ALLOPHANESAMPLESpretreatmentsample temperature /K SBET/~' g-'Waxy-PanEgmontiron-rich Ruapehu29238848257829338047158229838347356347858160751727329 1350309473539449381I I I I I I I I I I20 10 8 6 4 2d/nmFIG. 2.-Powder diffraction patterns of (a) Waxy-Pan, (b) Egmont and (c) iron-rich Ruapehuallophanes. Traces of quartz (Q), feldspars (F) and mica-type minerals (M) are detected."1-8 2546 NEW ZEALAND ALLOPHANES: A STRUCTURAL STUDYRESULTSThe active surface area (SBET) of the Waxy-Pan, Egmont and iron-rich Ruapehuallophane samples, pretreated at different temperatures, was determined from thenitrogen adsorption-desorption isotherms (77 K) using the B-point method.* Thesedata are collected in table 2. From the desorption isotherms and the correspondingI I I I I4000 3000 2000 1500 1000 500klcm-'FIG. 3.4nfrared spectra of allophane samples: (a) Waxy-Pan, (6) Egmont and (c) iron-richRuapehu. The regions of physically as well as chemically adsorbed water vibrations (W) as well asthe " lattice " vibrations (L) are indicated.S,,, the pore size distributions were computed and correlated with the pretreatmenttemperatures (fig.1). These data reveal that both the surface area and the porewidths increase with increasing pretreatment temperatures, up to a certain temperaturewhere a collapse of pores is responsible for a lower S,,,R . VANDICKELEN, G . DE ROY AND E. F . VANSANT 2547In order to characterize the mineral composition and structure of the allophanesthe powder diffraction patterns and infrared spectra of the three New Zealandallophane samples were recorded and are shown in fig. 2 and 3, respectively.The iron phase, present in the allophane structure, was investigated from theMossbauer spectra (fig. 4). The fitted Lorentzian profiles are shown and the statisticalcount rate error is indicated by error flags. The hyperfine parameters are listed intable 3.These results indicate two iron centres in the Waxy-Pan and Egmontallophanes. However, in the iron-rich Ruapehu sample three different iron locationsin the mineral structure are observed.TABLE 3.-MOSSBAUER HYPERFINE PARAMETERS FOR THE NEW ZEALAND ALLOPHANE SAMPLEStype of I.S. Q.S. width occurrencesample site /mm s-1 /m s-l /mm s-l 1%Waxy-Pan I 0.672(2) 0.64(5) 0.342(6) 91 (2)I1 0.65(2) 1.09(4) 0.33(5)Egmont I 0.708(3) 0.616(9) 0.36(1) 9a3)I1 0.68( 1) 1.07(3) 0.29( 5) 8(3)iron-rich I 0.762(7) 0.71(1) 0.44(2) 59(3)Ruapehu I1 0.693(6) 1.36(2) 0.35(3) 23(3)111 0.55(1) 0.69(2) 0.34(3) 1N3)2.0 1.0 ox) -1.0(a)See caption overlea2548 NEW ZEALAND ALLOPHANES: A STRUCTURAL STUDYIII I i-, l ~ l l l l t f ~ ~ l ~ l l l l l l l l , l , l l ~ l t l l l , , ~ l l t l l l l l f l f ~ ~ , , ~ ~ l l I3.0 2 .o 1.0 0 -1.0 -2.0(c)FIG.4.-Mossbauer spectra of allophane samples : (a) Waxy-Pan, (b) Egmont and (c) iron-richRuapehu. The sites are named as follows : site I, Fe3+ in a symmetric octahedral surrounding ;site 11, Fe3+ in a cis-dihydroxyoctahedron ; site 111, a separate Fe203 phase. For more information,see textR . VANDICKELEN, G . DE ROY AND E . F. VANSANT 2549DISCUSSIONThe chemical analysis (table I) shows a marked difference in the composition of,on the one hand, the Egmont and Waxy-Pan allophanes and, on the other hand,the iron-rich Ruapehu sample concerning the SO2, Al,03, Fe,03 and H,O content.Although the allophanes mainly consist of Si02 and A1203 they can exhibit alarger surface area as compared with the composition-weight median of amorphoussilica and alumina (385 m2 g-l and 140 m2 g-l after 383 K pretreatment, respectively).A more symmetric arrangement of the Si- and Al-units must be responsible for thisbehaviour, so that more and wider pores can be formed.In order to investigate the allophane " framework ", the structural model asproposed by Kitagawa was tested with the experimental adsorption, pore sizedistribution, infrared, X-ray and Mossbauer data.According to the proposedmodel, the allophanes are composed of 5.5 nm spherical particles, surrounded by amonolayer of water. The allophane microaggregates consist of these particles in aclose-packed arrangement. In an aqueous environment all space between theseparticles is filled with water, but in the air-dry allophane two types of cavities appearbecause there is not enough water left to fill all the free space.Upon heating to383 K, loss of the adsorbed water results in the formation of microaggregates due tostrong physical interactions between the unit particles.The observed variations of the specific surface area (SBET) of the allophane samplesinvestigated as a function of the pretreatment temperature and the correspondingpore size distributions (table 2, fig. 1) fitted well in the proposed allophane model.The increase of SBET on pretreatments from 298 to 383 K indicates the removal ofphysically adsorbed water from the pores between the unit particles. The calculatedpore widths indeed become larger.Pretreatments from 383 to 473 K cause a furtherincrease in the surface area of the Waxy-Pan and the Egmont samples due to therelease of physically as well as chemically adsorbed water. The increasing pore sizesreflect this trend. Eventually, as the samples are pretreated at higher temperatures,SBET declines and the pore sizes are reduced by the removal of structural water and acollapse of the internal pores.The iron-rich Ruapehu sample, however, displays different behaviour in thetemperature range 383-473 K. Its surface area declines while the pore sizes are stillincreasing. This can be explained by the collapse of an important amount of smallpores. This behaviour is probably related to the exceptional chemical compositionof this sample and its extremely low water content.The absence of pores of sufficientsize can account for this phenomenon, but the elevated iron content suggests thepresence of a separate iron oxide phase in this allophane.The X-ray diffraction patterns of all three samples show only very weak intensities(fig. 2). No evidence is found for any crystalline phase, except for the Egmontallophane sample, where small amounts of quartz (Q), feldspar (F) and mica-typeminerals fM) were detected.13 The patterns show a striking similarity to thediffractograms of Japanese allophanes,14 with broad bands at d = 0.8-1.0 nm andd = 0.30-0.35 nm. The iron oxide in the iron-rich Kuapehu sample should be eitherin a highly dispersed crystalline form or in a completely amorphous form, present as aseparate phase or randomly distributed in the allophane phase.The infrared spectra of these samples are also comparable to those of Japaneseallophanes 4p l4 showing as main features intense bands at 3400-3500 and 1650-1630 cm-1 due to water vibrations and two bands near 1000 and 600 cm-1 charac-teristic of a l l o p h a n e ~ .~ ~ - ~ The water vibrations are due to both physically adsorbe2550 NEW ZEALAND ALLOPHANES: A STRUCTURAL STUDYwater and to structural hydroxyl groups. The large width of the 3400-3500cm-lband suggests complex hydrogen bonding.Vibrations in the SO4- and Al0,-tetrahedra occur in the region of 1250-850 cm-l l 7 and all allophane samples possess a broad and intense band in thisregion.A second band can be distinguished in the 750-400cm-l region, due tovibrations between adjacent Si- and/or Al-tetrahedra.l The 473,550 and 600 cm-labsorptions suggest the presence of iron in an octahedral oxygen surrounding.A weak but significant peak at 780 cm-I is observed in the Egmont sample. Itcan be attributed to quartz. This is in perfect agreement with the observed X-raydiffraction pattern of this sample.The infrared technique gave no answer on the state of the iron oxide in the iron-rich Ruapehu, Waxy-Pan and Egmont allophanes. hlossbauer spectra were thereforerecorded and analysed. Three different iron sites are detected in the allophanes.The relative abundances of iron atoms in these sites are plotted as a function of theiron content of the samples (fig.5). The iron atoms in site I are surrounded by oxygenatoms in a regular octahedral way. They are probably isomorphically substitutedfor aluminium in the octahedral polyalumina phase of the a1lophane.l The ironatoms in site I1 are octahedrally surrounded by oxygen ligands, two of them being0 0.2 0.4 0.6 0.8Fe203/(Si02 + AI2O3 + Fe203)FIG. 5.-The relative abundances of allophane iron atoms in the different sites as a function of thehydroxyl groups (in cis position) caused by bond cleavage. The resulting decreaseof symmetry causes the drastic rise of the quadrupole splitting. As more iron ispresent in the allophane, the number of broken bonds will increase, so more sites Iwill be converted into sites 11. In allophanes with a high iron content a third site(111) appears. Its parameters can be explained in two different ways: either as afree Fe,03 phase or as Fe3+ in a tetrahedral environment.lg~ 2o If the latter explana-tion is assumed, the iron would substitute for alumina in the aluminosilicate phase,causing the iron-rich Ruapehu to have nearly the same pore size distribution as theother samples.This is in conflict with our observations (fig. 1). A separate,total iron content. The sites are named as in fig. 4R. VANDICKELEN, G. DE ROY AND E. F. VANSANT 255 1amorphous Fe,O, phase thus has to be implemented. The absence of pores of0.6-0.7 nm, as compared with the Egmont and Waxy-Pan samples, suggests that thisphase acts as a binding agent, filling the pores between the allophane unit particles,thereby excluding the water molecules.In general, we can conclude that the New Zealand Waxy-Pan and Egmontallophanes behave in a very similar way to their Japanese analogues. The iron-richRuapehu sample, however, has a different chemical composition, which is reflectedin its structure as an amorphous, pore-filling iron oxide phase.E.F. V. thanks the National Science Foundation (Belgium) and the BelgianGovernment for their financial support and the Soil Bureau, D.S.I.R., Lower Hutt(New Zealand) for the research facilities. G. De R. thanks I.W.O.N.L. Belgiumfor a research grant. The powder diffraction patterns were recorded on the PAD-I1diffractometer of the " Laboratorium voor Oppervlaktescheikunde ", CatholicUniversity of Louvain (K.U.L.), Leuven (Belgium).The authors thank Prof. J. B.Uytterhoeven and W. J. Mortier for their kind collaboration.L. P. Van Reeuwijk and J. M. De Villiers, Agrochemophysica, 1970, 2, 77.K. S. Birrell and M. Fieldes, J. Soil Sci., 1952, 3, 156.M. Fieldes, New Zealand J. Sci., 1966, 9, 591.M. Fieldes and R. J. Furkert, New Zealand J. Sci., 1966, 9, 608.H. H. Adler, Prelim. Rep. No. 6, Amer. Petroleum Inst., Proj. 49 (Columbia University, NewYork, 1951).H. D. Orchiston, Soil Sci., 1959, 88, 159.Y. Kitagawa, Amer. Mineral., 1970, 56, 465.S. J. Gregg and K. S. W. Sing, Adsorption, Surface Area and Porosity (Academic Press, London,1967), chap. 2.S. Brunauer, R. Sh. Mikhail and E. E. Bodor, J. Colloid Interface Sci., 1967, 24, 451.B. C. Lippens, B. G. Linsen and J. H. De Boer, J. Catalysis, 1964, 3, 32.l o R. Sh. Mikhail, S. Brunauer and E. E. Bodor, J. Colloid Interface Sci., 1968, 26, 45.l 2 S. Brunauer, R. Sh. Mikhail and E. E. Bodor, J. Colloid Interface Sci., 1967, 25, 353.l 3 D. Carroll, Clay Minerals, a Guide to their X-ray identification (The Geology Society ofl4 N. Yoshinaga and S. Aomine, Soil Sci. Plant Nutr., 1962, 8, 6.l 5 J. D. Russell, W. J. McHardy and A. R. Fraser, Cfay Miner., 1969, 8, 87.l7 A. V. Kiselev and V. I. Lygin, Infrared Spectra of Surface Compounds (J. Wiley, New York,l 8 K. Nakamoto, Infrared Spectra of Inorganic and Coordination Compounds (Wiley Interscience,London, 1970).l9 G. M. Bancroft, Mossbauer Spectroscopy: an Introduction for Inorganic Chemists and Geo-chemists (McGraw-Hill, London, 1973).2o Chemical Applications of Mossbauer Spectroscopy, ed. V. I. Goldanskii and R. H. Herber(Academic Press, London, 1968).America, Special Paper 126, 1970).M. Fieldes, New Zealand J. Sci. Technol., 1955, B37, 336.1975), pp. 292-295.(PAPER 9/1657

ISSN:0300-9599    

DOI:10.1039/F19807602542

出版商:RSC    

年代:1980

数据来源: RSC