1547J. CHEM. SOC. FARADAY TRANS., 1994, 90(11), 1547-1551 Two Members of the ABC=DGR Family of Zeolites: Zeolite Phi and Linde D Karl Petter Lillerud" and Rosemarie Szostakt Department of Chemistry, University of Oslo, P.O.Box 1033 Blindern, N-0315 Oslo, Norway Alicia Long Georgia Tech Research Institute , Georgia Institute of Technology, Atlanta , Georgia 30332,USA Linde D and phi, members of the ABC-double six-ring (ABC-DGR) family of zeolites, have been studied by X-ray diffraction. Broadening of specific lines in the X-ray powder diffraction pattern is observed which is consistent with the existence of faulting within the crystals. Both X-ray powder diffraction patterns can be simulated by introducing double six-member-ring sheet structural faults into the AABBCC stacking sequence of chabazite.These zeolites can be synthesised from a potassium-sodium-containing gel system in the absence of organic ami nes. A chabazite-like material identified as Linde D was first reported by Breck and Acara in 1961.' This zeolite crys- tallises from silica-rich gels in the presence of both sodium and potassium. Though the reported adsorption properties of Linde D differed from that of pure samples of chabazite, no additional studies on the nature of this phase have been undertaken. s2 Phi is the name given to another patented aluminosilicate phase. The X-ray powder diffraction pattern, chemical com- position and adsorption properties were reported by Grose and Flanigen in 1978.3 This zeolite was prepared from sodium aluminosilicate gels in the presence of tetra-methylammonium cations.A similar phase was prepared by Jacobs and Martens, and Franco et al., using tetra-ethylammonium instead of tetramethylammonium cation^.^,^ In addition to the organic amine cations, potassium cations were also present in their synthesis Franco et al. further characterised this material. They reported the nature of the diffraction lines. Both broad and sharp peaks were found to be present, an indication that this material may contain a significant amount of structural fault^.^ They acknowledged the similarity between the X-ray powder dif- fraction patterns of Phi and chabazite and suggested that phi contains offretite-like zeolite intergrown with the chabazite structure.Recently, Lob0 et al. prepared several materials fol- lowing altered recipes of both Franco et al. and Jacobs and Martens. They conclude that their material, which they also called phi, did not represent a pure phase but a physical mixture of the two zeolites offretite and chabazite.6 We have synthesised materials from inorganic gel systems with chemical composition, X-ray powder diffraction pattern and IR spectra matching those reported in the literature for zeolites D and phi.'*',* In order to identify the structural phases that these materials represent, we have simulated their diffraction patterns using the DIFFaX program.' High-temperature X-rai diffraction, 29Si NMR and IR spectros- copy were also used to characterise these materials.From our studies, we conclude that neither Linde D nor phi represent physical mixtures of phases or intergrowths of offretite and chabazite. These materials constitute faulted members of the ABC-D6R or chabazite family of zeolites. Experimental Si02/A1203= 6-28; Na20/Na20 + K20 = 0.8-0.95; Na20 + K20/Si02= 0.4-0.55; and H20/A1203 = 250-400. While Ludox HS from DuPont and sodium metasilicates from Fisher Scientific or Sigma Chemicals have been used suc-cessfully as a silica source; aluminium hydroxide from Pfaltz and Bauer was the preferred source of alumina for producing the pure phi-type phase. KOH and NaOH were both from Fisher Scientific. Crystallisation occurs within 3 days at 100°C. Chemical analysis of crystals obtained from gels with Si02/A1203 of 15 indicate a crystal composition with Si02/A1203= 4.46, Na/K = 2.1 and Na + K/AI = 1.For samples prepared in a more aluminium-rich system (Si02/A1203= 10) the Si02/A1203 ratio determined from 29Si NMR is found to be 4.0. The range over which phi-type materials crystallise is compared with those reported in the literature for this zeolite and summarised in Fig. 1. Linde D can be prepared using the following batch compositions as claimed in the patent:'26.18NaOH :4.62KOH :28Si02 :2Al(OH), :250H20 and L 1 25 --/Jacobs' phi -'phi' ,20 -this work a-9 '15-v) phi patent 10 --/France's phi Linde D -Breck -I 5-F 13 The phi-type zeolite discussed in this study crystallises from 0.4 0.5 0.6 0.7 reactive gel mixtures within the following composition range : R20/Si0 Fig.1 Ranges of SiOJAI,O, and hydroxide content over which Present address : Center for Catalysis and Separation Science, zeolites Linde D and phi have been claimed to crystallise (seetext for Clark Atlanta University, Atlanta, Georgia, USA. references) 24.7NaOH :2.74KOH : 28SiO, : 2Al(OH), : 250H,O. The source of silica was Ludox LS40 (DuPont) and the source of alumina was aluminium hydroxide from Aldrich. Reagent grade sodium and potassium hydroxide were purchased from Eka Nobel. The Si02/A1,03 ratio determined from NMR is found to be 4.4, falling within the range claimed in the patent. A typical synthesis procedure for both phi and Linde D is as follows: sodium hydroxide and potassium hydroxide were weighed and dissolved in water, the aluminium hydroxide powder was dissolved in this solution and Ludox was added to the resulting basic solution.Stirring was brief and only used to ensure adequate mixing. 30 ml capacity Teflon-lined autoclaves were charged with the resulting gel to three-quarters of their capacity. The autoclaves were sealed and heated without stirring for 1 to 3 days. After crystallisation was complete, the autoclaves were rapidly quenched by placing them under flowing cold water until they attained room temperature. The solid was filtered and washed with copious amounts of deionized water before X-ray analysis. The highly crystalline chabazite (zeolite K-G) used for comparison was prepared by optimising the procedures ini- tially reported by Barrer and Baynhum." The batch com- position that produced the best fault-free chabazite was: 5K20 : 5Si0, : Al,O, :600H,O.Crystallisation was com-plete after 11 days: 5 days at 90°C and 6 days at 150°C. Offretite was prepared from the batch composition :0.48KOH : 4.32NaOH :0.48TMAOH :6Si0, :2Al(OH), :400H,O using the same reagents and methods as the phi synthesis reported above. The tetramethylammonium hydroxide used was from Fluka. A crystallisation temperature of 110°C was used and crystalline product results after 3 days. A trace amount of erionite intergrowth was observed in this sample, but, it did not influence the results of this study.Physical mixtures of offretite and chabazite were prepared by carefully weighing out the desired amount of each phase and mixing the phases by grinding them together. The XRD pattern was run several times using different sample holders and packing methods. Differences between spectra were minimal. Analytical Methods A Siemens D500 powder diffractometer equipped with a ger- manium primary monochromator to ensure strictly Cu-Ka, radiation, an automatic aperture slit and a Bhuler high- temperature sample holder for measurements under con-trolled atmosphere and temperature were used in this study. The identification of faulted phases from the powder XRD may be even more difficult than identifying physical mixtures. Treacy, Newsam and co-workersg have developed a com-puter program (DIFFaX) to simulate diffraction patterns of such phases.Their work shows that the pattern derived from faulted materials is not a linear combination of the patterns of the end members. Their methodology was used in this work. IR spectra were recorded on a Perkin-Elmer model 225 grating spectrometer and on a Bruker model 88 FTIR spec- trometer covering the regions 4O00 to 350 cm-'. Pellets were prepared using 10 wt.% zeolite in dry KBr and pressed to obtain thin transparent wafers. The ,'Si NMR spectra were recorded on a Bruker CXP-200 pulse Fourier-transform NMR spectrometer oper- ating at 39.7 MHz. The spectrometer was fitted with a magic- angle spinning probe with a D-poly(methylmethacry1ate) rotor spinning at 3.0 kHz.Using a 30 degree pulse angle (4.1 p) with 6.5 s repetition time, 8 x lo3 data points were recorded over a spectral width of 20 kHz. 3000 scans were recorded per spectrum and 10 Hz line broadening was J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 applied. Chemical shifts were determined relative to external Si(CH3),, . Results and Discussion The major difference between the synthesis of a phi-type zeolite and Linde D appears to be the SiO,/Al,O, ratio of the gel. The crystallisation of Linde D is claimed for potassium/sodium-containing batch compositions of SiO,/Al,O, ratio of 28. By lowering this ratio and lowering the hydroxide content of the gel, we have observed the forma- tion of a phi-type phase.Phi-type zeolites readily crystallise from inorganic-only reaction mixtures over a wide range of Si02/A1,03 ratios from 6 to 28. This differs from phi prepared in the presence of organic amines.' This can be seen in Fig. 1. The Na,O/Na,O + K,O and Na,O + K20/Si02 ratios in the inorganic system border the crystallisation fields identified for Linde D.',, Though Breck reports the synthesis of D at gel Si02/A1203 ratios of 7, few further details are provided., Both ranges reported for Linde D are shown in Fig, 1 for completeness. Crystallisation of both phi and Linde D occurs from 24 and 72 h, respectively, at 100°C. Crystallisation of phi-type materials from organic containing systems as described by Franco et al. requires 8 to 13 days at similar crystallisation temperatures.' By adjusting the synthesis conditions, other phases are observed.These include offretite and phillipsite in both inorganic and organic synthesis systems. A more detailed summary of batch compositions examined and crys- talline products resulting from the template-free system are provided in Table 1. At a SiO2/A1,O, ratio of 28 and low potassium content in the gel, Linde D is formed exclusively. In the absence of pot- assium cations, faujasite readily crystallises. Increasing the amount of potassium in the gel results in the crystallisation of an offretite-type phase (Linde T) along with Linde D. The X-ray powder diffraction pattern of this mixture of Linde T and Linde D is similar to the diffraction pattern reported by Lob0 et al.for the material they claim as phi.6 Lowering the SiO2/A1,O3 ratio of the gel from what is claimed in the Linde D patent produces the phi-type phase. At lower SiO,/Al,O, Table 1 Batch compositions and products at crystallisation tem- perature of 100-110°Cafter 1 to 3 days, H,O/AI,O, = 250-400 SiO,/AI,O, Na/Na + K R,O/SiO, product 28 0.8 0.49 Linde T, Linde D 28 0.85 0.46 Linde D 28 0.9 0.5 Linde D 15 0.9 0.4 phi15 0.9 0.4 phi, faujasite 15 1 0.47 faujasite15 0.9 0.3 amorphous12 0.9 0.52 phi12 0.9 0.45 phi12 0.9 0.4 phi12 0.8 0.46 phi, (phillipsite) 12 0.8 0.4 phi, (phillipsite) 12 0.9 0.3 amorphous10 0.9 0.52 phi, (trace of phillipsite) lo" 0.9 0.6 phi, (faujasite) 6 0.9 0.52 phi6" 0.9 0.52 faujasite6b 0.9 0.51 phillipsite6 0.7 0.45 phillipsite Sodium aluminate as alumina source.Catapal B as alumina source. J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 ratios, phillipsite is occasionally observed as a secondary phase. At Si02/A1203 ratios of 10, long crystallisation times (2 weeks) do not result in the conversion of phi to phillipsite or to any other phases indicating no interdependence between these phases. The source of alumina is critical for the formation of the phi-type phase in this system. Successful crystallisation occurs when highly soluble aluminium hydroxide is used as the aluminium source. Using sodium aluminate encourages the formation of faujasite and catapal B results only in the crystallisation of phillipsite. With time, any low-crystalline faujasite formed does disappear.Low hydroxide contents fail to produce any products within 3 to 5 days. Longer crys- tallisation times were not explored. Elevated temperatures generally result in producing phillipsite. Offretite, or Linde T, forms in potassium-rich systems at the higher temperatures. The X-ray powder diffraction pattern for Linde D is shown in Fig. 2(a)and is compared with that of template-free phi in Fig. 2(b).The d-spacings (indicated as solid lines) claimed in the respective patents are included to confirm the identity of the materials reported here. The presence of the low-angle reflection around 11.6 A in the X-ray powder diffraction pattern represents a characteristic of zeolite phi that is present to a lesser extent in the diffraction pattern of Linde D. This low-angle diffraction peak in Linde D is not always observable under normal conditions encountered for the X-ray data collection.The conditions used to obtain the XRD patterns of Linde D are critical in the characterisation of this material at the low angles. The intensity of the low- angle peaks in zeolite diffraction patterns is strongly depen- dent on the amount of water adsorbed in the zeolite. Also, when the purpose of the X-ray powder diffraction data is only to identify the presence of crystalline phases, it is common to use fixed slits that expose more than the mounted sample at low angles. This results in an underestimation of the intensity of the low-angle reflections.In this study, we have used an automatic divergence slit in recording the dif- 100 r 15 20 25 30 35 fraction patterns represented here and the observed inten- sities are corrected with the l/sin 8 function. The X-ray powder diffraction patterns shown in Fig. 2 were recorded at 200°C with minimal water content in the sample, as deter- mined from water loss experiments using thermal analysis techniques. The presence of two phases has been suspected by others to explain the XRD pattern of zeolite phi.6 The broadness of only 1 lines (hexagonal indexing) in combination with a single uniform morphology observed for the crystal agglomerates of these samples, makes this possibility unlikely.To illustrate the difference between physical mixtures of offretite and cha- bazite and zeolites phi and D, two physical mixtures contain- ing different proportions of pure offretite and chabazite were used in this study. The first mixture composition chosen con- tains 5 wt.% offretite. Such a composition would match the weak-intensity reflection appearing around 11.6 A in the powder diffraction pattern of Linde D. The X-ray powder dif- fraction pattern of Linde D would then be expected to match that of the major phase, chabazite, more closely since the most intense reflection found in offretite appears at 11.4 A. The second physical mixture for this study is conservatively chosen to be that of 30% offretite-70% chabazite. With this amount of offretite, the intensities of the first two reflections in the phi diffraction pattern would be approximated.The X-ray powder diffraction pattern of Linde D, template-free phi and the two physical mixtures at 5% and 30% offretite in chabazite are shown in Fig. 3 and 4 and the d-spacings for phi and the two physical mixtures are shown in Table 2. What is immediately obvious from the diffraction patterns is the variability in the FWHM in several of the reflections in the X-ray diffraction pattern of Linde D and zeolite phi. Broadening of certain reflections is generally indicative of materials containing faults. The zeolites beta and ZSM-20 are two well known examples of materials con- taining stacking faults and exhibiting X-ray powder diffrac- tion patterns consisting of combinations of broad and sharp I A 5 10 15 20 25 30 35 B 1,I I I (b! nw5 10 15 20 25 30 35 26/degrees 5 10 15 20 25 30 Fig.2 Comparison of the reported diffraction patterns for zeolites 28jdegrees (a) Linde D and (b) phi with the patterns obtained in this study. Fig. 3 Comparison of the diffraction patterns of A, (a) Linde D and Dashed lines indicate spacings and intensities reported in the patents. (b)5% OFF-95% CHA and B, (a) phi and (b)30%OFF-90% CHA 60 50 40 30 20 10 LA. A I 0 5 10 15 20 25 0 5 10 15 20 25 2O/degrees ZO/degrees Fig. 4 Simulated X-ray powder diffraction patterns for a series of randomly faulted chabazites. The numbers indicate O/O disorder.Experimentally determined patterns for (a)chabazite, (b) Linde-D, (c) phi-I and (a) phi-I1 are also shown. reflection^.^^"*'^ What is more striking is the number of major reflections in chabazite and offretite that are not present in either zeolite phi or Linde D and reflections observed in these materials that are absent in the physical mixtures. The most notable difference is the absence of a strong reflection at 28 = 25" that is a low-intensity broad band in phi (Fig. 3). The intense reflection at 28 = 26" in phi appears as a very weak intensity reflection in the physical mixture. The doublet at 28 = 31" in the mixture shown in Fig. 3 appears as one intense reflection in Linde D. This is a further indication that both zeolites are not simple physical mixtures of these two components.The IR stretching frequencies for all materials are com- pared in Table 3. Vibrations unique to offretite around 775 and 575 cm-I are present in physical mixtures containing offretite and in Linde T but are absent in the samples of Linde D and phi. Those vibrations more characteristic of chabazite appear at 760 and 510 cm-'. The 760 and 510 cm-' bands appear in similar positions in Linde D. The 760 cm-' band is absent in phi but the 510 em-' band is observed, an indication of some uniquely different character about the phi structure. Comparing the IR spectra of phi with offretite, chabazite and their mixtures further substan- tiates the absence of offretite in the material. A similarity does exist between phi, Linde D and chabazite.It can therefore be suggested that these materials are members of the ABC-D6R family of structures. The com- binations of broad and sharp reflections are a result of faults contained within the structure and would arise from stacking disorders in the double six-ring AABBCC sequence. Such a conclusion was also reached by Franco et al. though they considered the faults to be due to offretite phase^.^ The J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 2 Comparison of the X-ray powder diffraction data for phi and mixed phases if chabazite (CHA) and offretite (OFF) phi" 5% OFF/95% CHA 30% OFF/70% CHA 11.78 (mbr) 9.368 (msbr) 11.39 (vw) 9.33 (m) 11.40 (m) 9.339 (m) 8.076 (vw) 7.112 (w) 6.878 (vs) 6.903 (vw) 7.537 (w) 7.061 (vw) 6.906 (w) 6.598 (m) 6.293 (vw) 5.558 (wbr) 5.039 (vs) 5.555 (vw) 4.997 (vs) 4.677 (w) 5.717 (vw) 5.556 (vw) 5.321 (vw) 4.999 (vvs) 4.673 (vw) 4.319 (sbr) 4.330 (m) 4.555 (vw) 4.326 (m) 3.975 (m)3.893 (w) 3.988 (vw)3.873 (m) 3.989 (vw) 3.873 (w) 3.811 (vw) 3.609 (vwbr) 3.750 (w) 3.443 (s) 3.578 (s) 3.454 (vw) 3.578 (s) 3.455 (vw) 3.237 3.184 (vwbr) 3.118 2.932 (vs) 3.235 (vw) 3.178 (mw) 3.119 (vw) 2.934 (s) 3.301 (vw) 3.233 (vw) 3.176 (w) 3.150 (w) 2.934 (m) 2.888 (s) 2.889 (m) 2.604 (s) 2.845 (vw) 2.778 (vw) 2.682 (w) 2.612 (w) 2.843 (mw) 2.780 (vvw) 2.680 (w) 2.612 (vw) 2.579 (vw) 2.501 (mw) 2.352 (vw) 2.312 (vw) 2.579 (vw) 2.502 (w) 2.353 (vw) 2.296 (mw) 2.304 (vw) 2.303 (vw) ~~~ Phi is in NH, form, CHA and OFF in K and Na/TMA forms, respectively.Data collected at room temperature under similar levels of hydration. a This work. 29Si NMR of phi exhibits a classic pattern for a material con- taining equivalent tetrahedral (T) sites which further rules out the presence of an offretite intergrowth or second phase. Linde D also displays a similar pattern. Structures with single Table 3 Comparison of the IR vibrations between 900 and 300 cm-' of chabazite (K) and reported chabazite-like materials (R,D), zeolite phi and physical mixtures of offretitwhabazite and offretite-erionite (T) zeolite R" chabazi te-like G" chabazite D" chabazite-like T" offretite-erionite phib 30% OFF-70% CHAb 5% OFF-95% CHAb " Ref.6; 'this work. wavenumber/cm 738 (w) 678 (w) 625 (m) 508 (mw) 452 (m) 426 (m) 720 (w) 696 (wsh) 632 (m) 515(m) 460(m) 408 (m) 755 (wsh) 711 (w) 631 (m) 513 (m) 459 (m) 415 (m) 771 (w) 718 (w) 623 (mw) 575 (w) 467 (ms) 433 (ms) 410 (vwsh) 730 (w) 775 (w)760 (vwsh) 685 (w) 725 (w)720 (wbr) 630 (m) 630 (mw)630 (m) 575 (vw) 511 (m) 510 (w) 465 (ms)460 (m) 460-370 (mbr) 435 (ms)440-350 (mbr) 410 (wsh) J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 six-member-rings, such as offretite, would give rise to inequi- valent T sites within the structure changing the 29Si NMR pattern. An additional peak at low 6 would be observed. Merlino has addressed the stacking of the zeolites with six- member rings.' Stacking sequences for six-member rings can be ABC at one extreme and AABBCC at the other.Interme- diate between these are the AABAAB sequences of offretite, AABAAC of erionite and ABBACC of TMA-E. What remains unobserved is the mixture of two double six-member rings followed by a single ring. A stacking sequence such as AABBC appears to be an unlikely possibility." The intro- duction of the single six-ring in an AABBCC double six-ring stacking sequence introduces significant strain in the crystal lattice and therefore would not be expected to occur. Calcu- lation of the a dimension for the unit cell of phi (a = 13.8 A, c = 5 A) confirms the lack of single rings as the unit-cell dimensions favour the double-ring type stacking AABB (a = 13.75 A) of gmelinite rather than the AAB of the mixed double-single ring systems (a = 13.29 A) of offretite.Thus it is highly unlikely that faults within the phi structure are due to the presence of single six-member rings. From our results we can conclude that Linde D and zeolite phi do not represent physical mixtures of offretite and chaba- zite but represent members of the ABC-D6R family of materials. These phases contain stacking faults more prob- ably due to random stacking of double six-ring units rather than the presence of single six-ring units as found in offretite. The unit-cell values, and the IR and NMR spectra further substantiate the absence of single six-member rings. The pres- ence of faults in an AABBCC-type structure gives rise to the combination of broad and sharp bands in the X-ray powder diffraction pattern and to the unique reflections not found in a pure fault-free chabazite.High-resolution electron micro- scope imaging of these materials is presently underway to increase understanding of the nature of the faulting in the phi and D phases. Linde D and zeolite phi differ from one another and from chabazite by the number of stacking faults they contain. The X-ray powder diffraction patterns of these materials compare well with the simulated diffraction patterns based on increas- ing amounts of faulting in the chabazite structure. This is shown in Fig. 4. One difference observed between the calcu- lated and the actual XRD pattern is that the 003 reflection in the simulated pattern, which is based on the c-axis for CHA, shifts to a higher angle in the actual material.Such compres- sions along the c direction are also observed in more heavily faulted ABC-D6R family materials. l4 Compression or relax- ation of the lattice around the fault in the structure is not taken into consideration in constructing the simulation. Modification of the synthesis conditions e.g. by changing the Si02/A1203 ratio or hydroxide content of the gel pro- duces materials that may contain the same d-spacing as reported for D and phi with lineshapes that differ signifi- ~ant1y.l~Because of the large variability which can occur from the presence of faulting in such materials, it becomes difficult to assign the names 'Linde D' or 'phi' to any highly faulted material based on reported d-spacing position and intensity alone.Such difficulties must be considered when characterizing fault-containing materials. In the case of Linde D and phi, we can conclude that both are presently ill- defined members of the chabazite family of struct~res.'~ The authors thank Ms. Anne Horn for obtaining the IR spectra of these materials, Dr. M. Stocker for obtaining the NMR spectra and Norsk Hydro for use of their Biosym licence. RS is grateful to the Royal Norwegian Council for Scientific and Industrial Research for financial support. References 1 Br. Pat., 868,846, 1958. 2 D. W. Breck, Zeolite Molecular Sieves, Wiley, New York, 1974, pp. 290,291. 3 R. W. Grose and E. M. Flanigen, US Pat., 4,124,686, 1978. 4 P. A. Jacobs and J. A. Martens, Stud. Surf. Sci. Catal., 1987, 33, 15. 5 M. J. Franco, J. Perez-Pariente and V. Fornes, Zeolites, 1991, 11, 349. 6 R. F. Lobo, M. J. Annen and M. E. Davis, J. Chem. SOC., Faraday Trans., 1992,88,2791. 7 J. A. Martens, M. Tielen, P. A. Jacobs and J. Weitkamp, Zeo-lites, 1984, 4, 98. 8 J. A. Martens and P. A. Jacobs, Zeolites, 1986,6,98. 9 J. M. Newsam, M. M. J. Treacy, W. T. Koetsier and G. B. Deruyter, Proc. R. SOC. London, A, 1988, 420, 375; M. M. J. Treacy, J. M. Newsam and M. W. Deem, 1991,433,499. 10 R. M. Barrer and G.W. Bayhnum, J. Chem. SOC.,1956,2882. 11 M. M. J. Treacy and J. M. Newsam, Nature (London),1988,332, 249. 12 J. M. Newsam, M. M. J. Treacy, D. E. W. Vaughan, K. G. Strohmaier and W. J. Mortier, J. Chem. SOC., Chem. Commun., 1989,493. 13 S. Merlino, Proceedings of the Sixth International Zeolite Con- ference, ed. D. Olson and A. Bisio, Butterworths, Boston, 1984, p. 747. 14 K. P. Lillerud and R. Szostak, in preparation. Paper 3/04912D; Received 13th August, 1993