|
11. |
Synthesis and characterization of II–VI semiconductor nanoparticulates by the reaction of a metal alkyl polymer adduct with hydrogen sulfide |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1771-1780
Stephen W. Haggata,
Preview
|
PDF (1561KB)
|
|
摘要:
Synthesis and characterization of 11-VI semiconductor nanoparticulates by the reaction of a metal alkyl polymer adduct with hydrogen sulfide Stephen W. Haggata," Xiaochang Li,'? David J. Cole-Hamilton*' and John R. Fryerb "School of Chemistry, University of St. Andrews, St. Andrews, Fife, Scotland, UK K Y16 9ST bDepartment of Chemistry, Glasgow University, Glasgow, Scotland, UK G12 8QQ A new method for the preparation of nanosized semiconductor particles such as ZnS and CdS in the quantum confinement diameter region (2-5 nm) is described that utilizes the reaction of polymer adducts (between polymers that contain N and group 12 metal alkyls) with H2S in solution. The preparation conditions (metal/N ratio, concentration, solvent and temperature) were each found to have different effects on the sizes of the particles.TEM studies reveal that the particles are evenly distributed in the polymer matrix (film or resin) with a relatively narrow dispersity in size. The 11-VI particles exhibit size quantization and the change in the band-edge position is observed by photoacoustic spectroscopy. Semiconductor nanosized particles (or quantum dots) have attracted considerable research interest during the last several years because of their unique electronic, optical and catalytic properties that differ from those of the corresponding bulk macrocrystallites.1-7 These differences in properties are attri- buted to (i) the quantum size effect and (ii) the increased ratio of surface atoms to bulk atoms as the diameter of the semi- conductor crystallite approaches nanosized dimensions (2-20 nm).3 These materials have possible future applications as photocatalysts in photo reaction^,^^^ in opto-electronic and all-optical device^^,^^ for e.g. the development of flat-bed luminescent displays and erasable optical data storage; and in electronic devices as a replacement for transistors." In order to achieve a quantitative understanding of such a new class of material, the preparation of a high concentration of mono- disperse particles evenly distributed within a suitable medium with chemically well controlled surfaces is vital.Since the colour of the particles and the light emitted by them can be altered by altering the particle size (a change in size of 3-1.5 nm can change the bandgap from 0.5 to 4eV in Cd3P25), narrow size distributions are also essential if pure colours are to be obtained in luminescence.There are many synthetic methods reported which target these goals. These include syntheses in colloidal s~spensions,~~~~~ solutions of single-molecule precur- sor~,~~ zeolites,16LB films,17 micelle~~~'~ sol-gel~,~~ and polymer film~.l~*~~Of these methods, the formation of semiconductor nanosized particles in dielectric polymer media has received growing interest due to the ready processability of the polymer films, a good spatial distribution of particles and possible future application in device str~ctures.~~-~~ The synthesis of nanosized semiconductor particles in a polymer matrix usually includes two processes.First, polymer blend or copolymer films that contain organometallic blocks26 or metal salts coordinated with polymer ligand~~~ formed. Secondly, are hydrogen sulfide is passed over the surface of the polymer film to form metal sulfide particles distributed evenly within the polymer. Using these methods, the particle dispersity and distribution can be controlled by the microphase separation domains in the block copolymer. However, these methods are carried out between the solid and gas (H2S) phase and unless the pre-formed ultrathin polymer films have a well controlled morphology, a very low loading of the semiconductor particles and a varying metal/S ratio are observed. For the organometal- lic routes, the polymers involved are often complex and require 7 Present address: University Chemical Laboratories, Lensfield Rd., Cambridge, UK.e.g. the ring-opening metathesis polymerization of norbornenes bearing highly sensitive alkylated metal substituents.26 We have been interested in the functionalization of readily available polybutadiene by homogeneous catalysi~~~-~~ to form polymer ligands and thence to form soluble polymeric adducts with group 12 metal alkyl~.~~ These adducts may be used both in a method to purify metal alkyls through the adduct purifi- cation process32 or as safe metal alkyl sources for MOCVD because they are non-pyrophoric and thermally dis~ociable.~~ Alternatively, we have recently found34 and will report further here that the adducts between polymer ligands and metal alkyls can readily react with hydrogen sulfide to form metal sulfide semiconductor colloids.Polymer films embedded with nanosized semiconductor particles can be obtained after removing the solvent under vacuum. The particle sizes are in the region of size quantization (2-5 nm) and for both ZnS and CdS, the band edge is blue shifted from the bulk value. In this paper it will be shown that by simply changing the synthetic conditions such as polymer concentration, metal/N ratio, solvent, temperature and in some cases the choice of metal alkyl-polymer adduct, the particle size and the chemical com- position of the 11-VI material can be controlled. Experimenta1 All the experiments were carried out under dry oxygen-free argon purified by passing through a series of columns con- sisting of Cr2+ on silica and dry molecular sieves.Greaseless joints and taps were employed and manipulations were carried out using standard Schlenk-line and catheter tubing techniques. All the solvents were carefully dried by distillation from sodium diphenylketyl. 2-Methylpyridine was purchased from Aldrich and was distilled prior to use. Butyllithium (1.6 mol dm-3 in hexane) and dimethylzinc (2.0 mol dm-3 in toluene) were purchased from Aldrich and used as received. Polybutadiene (83% pendant, 17% trans-1,4, M,=3,000) was a commercial product (Nippon Soda Company) and was used after pumping for 2 h. Me2Cd was prepared by the standard literature method.35 Powder X-ray diffraction (PXRD) patterns were recorded on a Stoe STADI/P diffractometer using Cu-Ka radiation.Data were collected in transmission mode with a sample mounted in vaseline on a rotating disc and compared with the standard pattern obtained from the JCPDS database or the PXRD pattern of the commercial (e.g. wurtzite) sample. All the features of the experimental powder patterns are broad and therefore, the particle size can be estimated using a J. Mater. Chern., 1996,6(11), 1771-1780 1771 m~dification~~(see below) of the formula devised by S~herrer,~~ Synthesis of nanosized semiconductor particles where D=particle diameter, A= 1.5406, 8 is the angle of the reflection and B =J(b2-bo2)where bo=FWHM of a theoreti- cally infinitely narrow line, b =2 x [width of the left-hand side of the (110) reflection at half its maximum]; this approximation is preferred to the FWHM as the (103) peak overlaps with the right-hand side of the (110) reflection and gives an incorrect indication of its broadened profile.The modified equation is D=1.2A/B cos 8 (1) Transmission electron micrographs (TEMs) were obtained using a Phillips EM 301 microscope at 80 keV. All samples were embedded in an epoxy resin and the sections were then cut on a microtome with a diamond knife. The dried specimen sections were put onto a copper grid which had a carbon support film present. Another layer of carbon was then evapor- ated onto the sample in order to prevent specimen charging.High-resolution TEMs (HRTEMs) were obtained by using an ABT 002B microscope at 200 keV. The samples were either prepared as previously described or suspended in acetone and then ultrasonically dispersed and lifted off onto a graphite grid. Photoacoustic spectra (PAS) were obtained using an OAS 400 photoacoustic spectrometer. The sample absorbs light generated from a xenon lamp and the absorbed energy is converted into heat. The photoacoustic signal is produced when the heat-induced expansion in the sample is recorded. All samples were scanned from the UV region to the near IR (200-800 nm). Photoacoustic spectroscopy measures the band edge of the material and from this value the bandgap is calculated using E(eV) =hc/A. The wavelength (A) is measured by the position of the knee of the band edge.38 Synthesis of polymer adducts with metal alkyls The hydrosilated polybutadiene was functionalized as in the procedures previously described3' using 2-methylpyridyl or 4- dimethylaminophenyl Lewis-base groups shown in Fig.1, thereby obtaining polymeric Lewis bases denoted as 2PySiPB and 4DMASiPB respectively. The polymer adduct solution was obtained by adding metal alkyl (or its solution) to a stirred polymer solution (e.g. 3% g~m-~) at room temperature. The mole ratio of metal alkyl to the coordinating atom (N) was usually stoichiometric, i.e. M/N= 0.5 for ZnMe, and CdMe,.30 w0.17 > 2PySiPB CH3--i-CH3 (b) -pVq0.17 Q CH3~Nh-13 Fig. 1 Polymer ligands synthesized through the functionalization of polybutadiene: (a) 2-methylpyridine derivative and (b)dimethylamino-phenyl derivative 1772 J.Muter. Chem., 1996, 6(ll), 1771-1780 A polymer/metal alkyl solution contained in a flask (100cm3) under nitrogen was gradually exposed to a hydrogen sulfide atmosphere (as seen in the reaction scheme for ZnS formation, Fig. 2) for cu. 20 min while stirring. The transparent yellow polymer adduct solution quickly changed to an opaque milk- like suspension (ZnS) or a yellow-orange suspension (CdS). Specific ZnS suspensions could be stabilized as long as several weeks without any apparent precipitate although instant pre- cipitate was obtained on adding some non-solvent, e.g. light petroleum. Depending on the M/N ratio and solvent used, a yellow-orange powder (ZnS, yellow only) or a film composite of metal sulfide/polymer could be obtained after pump drying the sample under vacuum.Results Synthesis of nanosized ZnS particles In the absence of polymer. Experiments to react ZnMe, solution (5% in toluene) with H,S have been carried out for comparison with ZnS/polymer nanocomposite material. When H,S is passed through or exposed to a solution of ZnMe,, white ZnS particles form instantly as a precipitate. PXRD measurements show that the ZnS probably has a wurtzite (hexagonal) structure although the line pattern is very broad, [Fig. 3(u)] corresponding to an estimated average crystal diam- eter of 4.1 nm calculated using an adaptation of Scherrer's formula (see Experimental section).36 After heat treatment of the sample at 80°C for 4 days, the PXRD pattern shows a slight narrowing which suggests that the average particle size has increased.TEM studies reveal that the formed crystallites agglomerate to form bigger particles or grains with a diameter of 50-150 nm, (Fig. 4). At higher magnification, the crystallites are seen to have a size distribution of 3-10 nm, the majority of particles being cu. 5 nm in diameter. The observed lattice spacings are ZnMe2 I Me-Si-Me -Me-Zn-Mesolvent I 2PySiPB Polymer adduct I ZnS particle Fig. 2 Formation of the 2PySiPB adduct and the reaction of polymer adduct with hydrogen sulfide Fig. 3 PXRD patterns of (a) ZnS prepared in toluene in the absence of polymer; (b) ZnS prepared in toluene in the presence of 2PySiPB, run Zn2; (c) ZnS prepared in diethyl ether (fast bubbling) in the presence of 2PySiPB, run Zn7; (d) commercial wurtzite ZnS and (e)JCPDS pattern for cubic ZnS (sphalerite) consistent with the particles having the hexagonal (wurtzite) structure.These results suggest that the reaction between dimethylzinc and hydrogen sulfide leads to the formation of nanocrystalline particles which agglomerate to form larger crystallites. In the presence of polymer. When a polymer adduct between 2PySiPB and dimethylzinc was exposed to hydrogen sulfide, an opaque, milk-like suspension formed within 5-30 min. The suspension was dried under vacuum in order to isolate the polymer.Analysis of the PXRD pattern [Fig. 3(c)] and com- parison with the PXRD pattern of the commercial wurtzite ZnS [Fig. 3(d)] tentatively confirms that the three broad hkl reflections at 28=28.5" (002), 47.5" (110) and 56.4" (112) as well as the very broad (103) reflection at 51.7" are indicative of the wurtzite (hexagonal) phase and not the sphalerite (cubic) phase of ZnS which is shown as the standard pattern [Fig. 3(e)]. The broad peak at 17" corresponds to the polymer. The hexagonal phase is also confirmed by the lattice spacings in the HRTEM images. Rough estimates of the particle sizes of all ZnS runs are based on width measurements of the single reflection observed at 47.5'. When this sample is heat treated at 80°C for 3 days there is no noticeable sharpening in the PXRD pattern which suggests that the ZnS-polymer nano-composite material has good thermal stability and therefore good control in maintaining the particle size.In order to correlate the synthetic conditions with the size of the ZnS particle several series of experiments have been carried out and the results are summarized in Table 1. It was observed that the polymer adduct solution takes longer to Fig. 4 Low-resolution TEM of ZnS particles (magnification of 115 000) become opaque if the Zn/N ratio of the Polymer adduct is formed in the absence of polymer decreased (runs Znl-Zn3, Table 1). In fact, an almost trans- J. Mater. Chern., 1996, 6(11), 1771-1780 1773 Table 1 Dependence of ZnS particle size on various synthesis conditions" av.TEM or calc.' polymer particle run no. Zn/py ratio solvent conc. (%) size/nm -CZnl 0.5 toluene 5 Zn2 0.3 toluene 5 3.7 C-Zn3 0.2 toluene 5 Zn4 0.5 t hf 3.3 2.6' Zn5 0.5 thf 5 3.3b Zn6d 0.5 t hf 5 1.8' Zn7' 0.5 diethyl ether 5 4.1 4.6' " At room temperature unless otherwise stated. 'Calculated particle size from PXRD pattern. 'Diffraction patterns are too broad to obtain a reliable particle size, but the size increases in the order Zn3<Zn2<Znl. At -20°C. 'Fast bubbling. parent yellow ZnS-polymer suspension can be obtained when the Zn/N ratio is <0.2. It is noticeable that the PXRD patterns of runs Znl-Zn3 broaden with decreasing Zn/N ratio although any estimations of the particle size are extremely inaccurate as there are no isolated single reflections of which linewidths can be measured.The PXRD pattern of run Zn2 in toluene is shown in Fig. 3(b) and is certainly broad enough to suggest particle sizes of <2 nm. However, the measurement of particle size from the TEM of the same sample indicates a much larger value of 3.7 nm (see below). ZnS runs in various other solvents show measurements from PXRD and TEMs that have better correlation and this can be seen for particles prepared in diethyl ether (run Zn7, Table 1). A tendency to form larger particle sizes when increasing the concentration of the polymer adduct is seen (see runs Zn4 and Zn5, Table 1). Changes in temperature have the largest effect on the average particle size (see runs Zn5 and Zn6, Table 1).PXRD patterns obtained from these samples where the temperature has been varied from room temperature to -20°C (Fig. 5) show significant broadening at the lower temperature, i.e. a decrease in particle 200-h .-rn+. c 3-4-([I size. The type of solvent chosen for the reaction also has a profound effect on the crystallite growth, and will be discussed in detail for CdS particles (see later). When light petroleum (bp 40-60°C) is used, the PXRD produces sharper patterns corresponding to an increase in particle diameter. Likewise, it is observed that when H,S is bubbled through the polymer solution the crystallites produced are slightly bigger than those formed when stirring under an atmosphere of H2S (see run Zn7, Table 1).Size distribution graphs of TEMs obtained from runs Zn2 and Zn7, in toluene (atmosphere of H2S) and diethyl ether (bubbling of H2S) respectively, are illustrated in Fig. 6. The HRTEM of run Zn7 (Fig. 7) shows an even distribution of particles within the polymer matrix. All ZnS particles have a narrow size distribution, primarily owing to the constraints of the polymer media. The narrowest range pertains to Zn2 which is a direct result of a smaller Zn/N ratio and a more gentle exposure to H2S. The average sizes for ZnS particles of runs Zn2 and Zn7 are calculated from TEM measurements to be cu. 3.7 and 4.1 nm respectively, according to d=Znidi/Zni where ni is the number of particles of diameter di.The average particle size for run Zn7 is similar to that which is calculated 50 (b) 40-30-20-10-Fig. 6 Size distribution graphs of ZnS particles prepared in the presence of 2PySiPB: (a) diethyl ether [fast bubbling of H2S (run Zn7)] and (b)toluene [atmosphere of H2S (run Zn2)] 10.0 20.0 30.0 40.0 2Bldegrees Fig. 5 PXRD patterns of ZnS-2PySiPB films prepared in a 5% solution in thf at (a)room temp. (run Zn5) and (b) -20 "C(run Zn6) 1774 J. Muter. Chew., 1996,6(ll), 1771-1780 Fig. 7 Low-resolution TEM (magnification of 600 000) of ZnS particles prepared in run Zn7 from the PXRD line broadening [Fig. 3(c)] which could indi- cate a fairly uniform crystallinity within and on the surface of the particle.In the reaction of the polymer adduct with hydrogen sulfide, the polymer precipitates simultaneously with the ZnS particles, suggesting that the polymer not only is a coordinating agent for dimethylzinc, but also coordinates to the ZnS thus terminat- ing and controlling the particle growth. We have recently demonstrated3' that in gas-phase reactions between Me2M (M =Zn or Cd) and H2S, nanoscale particles can be formed and that their size can be controlled by adding pyridine to the gas phase. We have shown that the major role of pyridine in the gas phase is to bind to surface atoms of the growing particle and prevent further reaction with H2S. It is worth noting that larger particle sizes can be obtained when a weaker polymer ligand donor is used, e.g.DMASiPB. When exposing the DMASiPB-ZnMe, adduct to H2S, the ZnS-polymer nano- composite formed is, as in its original state, a sticky resin which suggests that there is a weak interaction between the metal and the Lewis base and which also coincides with a sharpening in the PXRD patterns as compared to the patterns obtained from samples prepared using the 2PySiPB polymer; thus indicating an increase in particle size. The bandgap of commercial, wurtzite ZnS (3.62 eV) differs markedly from the ZnS-polymer composites. A change in the Zn/N ratio from 0.5 to 0.2 for runs Znl-Zn3 in toluene results in a blue shift in the absorption edge, an increase in bandgap from 3.97 to 4.05 eV and a decrease in particle size. A closer analysis of band-edge shift or 'size quantization' will be dis- cussed for CdS.The polymer itself absorbs at a higher energy, in the UV region, with a bandgap of 4.32 eV. Synthesis of nanosized CdS particles The procedure for the formation of CdS nanoparticles is identical to that for ZnS. Similarly, PXRD measurements of a commercial CdS sample containing particles >10 nm show the hexagonal or 'greenockite' phase. Some PXRD patterns have sharp peaks at 26=21.5 and 24" corresponding to impurities present in the vaseline used to mount the sample. All patterns for the polymer-passivated CdS are broad and indicate that the hexagonal phase is present independent of the synthesis conditions. The PXRD patterns of samples prepared by changing the temperature of the reaction (runs Cdl-Cd3, Table 2) show large variations in hkl-dependent line broadening (Fig.8). As found for ZnS, the broadening of the h01 reflections, in particular the (103) plane, indicates that stacking faults reside within and on the surface of the crystallite. Calculations of the coherence length of the crystallite using Scherrer's formula therefore give very poor estimations of crystallite size and are consequently omitted from Table 2. Instead, measurements of crystallite diameter are calculated from TEMs only. The PXRD patterns of cadmium sulfide- polymer samples prepared by varying the polymer concen-tration (runs Cd4-Cd5, Table 2) are shown in Fig. 9. The h01 reflection (103) is observed at 26=48" and the (002) plane at 28=26.5" is the most intense peak in all CdS patterns.TEM studies of samples prepared from runs in diethyl ether indicate that the particles exhibit stacking faults which will produce noticeable hkl-dependent line broadening. Electron diffraction of all CdS samples confirms the identity of the crystallite phase to be hexagonal. Size distribution graphs of the CdS runs Cd2 and Cd6 are shown in Fig. 10. The Cd:polymer ratios change from 0.5 to 0.2 and indicate average particle sizes of ca. 4 and 4.3 nm respectively. The average sizes recorded for these runs are larger than the average size distribution of the ZnS sample, run Cd2 [see earlier, Fig. 6(b)] prepared under similar conditions which gives an average value of ca.3.7 nm (see Discussion). Commercial (greenockite) CdS and other CdS bandgaps calculated from samples prepared under various synthesis conditions are shown in Table2. The blue shift of the band edge for samples prepared under different reaction tempera- tures can be seen in the PA spectra and are compared with Table 2 Dependence of CdS particle size on various synthesis conditions av. TEM run no. WPY ratio solvent T/"C polymer conc. (YO) bandgap/eV particle size (range)/nm bulk CdS 2.42 Cdl Cd2 0.5 0.5 toluene toluene 60 r.t." 3.0 3.O 2.48 2.50 5.0 (3-6) 4.0 (2-6) Cd3 0.5 toluene -78 3.O 2.93 2.5 (1-3) Cd4 0.5 diethyl ether 25 3.0 2.59 3.0 (2-5) Cd5 0.5 diethyl ether 25 10.0 2.61 3.5 (2-5) Cd6 0.2 toluene 25 3.0 2.67 4.3 (2-6) Cd7 Cd8 0.2 0.5 diethyl ether light petroleum 25 25 3.0 3.0 2.98 2.42 3.0 (2-6) 4.0 (2-10) a r.t.=room temperature. J. Mater. Chern., 1996,6( ll), 1771-1780 1775 Fig. 8 PXRD patterns of CdS samples prepared in toluene in the presence of 2PySiPB at different temperatures: (a) 60 "C (run Cdl), (b)room temp. (run Cd2) and (c) -78 "C (run Cd3). Impurities present within vaseline are denoted as 0. Fig. 9 PXRD patterns of CdS samples prepared in diethyl ether in the presence of 2PySiPB at different concentrations: (a) 10% (run Cd5) and (b)3% (run Cd4). Impurities present within vaseline are denoted as 0. the band-edge position of the commercial material [Fig. 11(a)]. These results indicate that the smallest particles are formed at the lower temperature of -78 "C [Fig.11(c)] and a general increase in bandgap is observed from the commercial value (2.42 eV) to 2.5 and 2.93 eV for runs at room temperature (run Cd2) and -78 "C (run Cd3) respectively. The more gradual slope of the band edge seen in Fig. ll(b) as compared to Fig. 11(c) is indicative of (i) a wider particle size distribution and (ii) transitions to lower energy bands or trapped states between the valence and conduction bands which are the result of crystallite defects. When the polymer concentration is varied, only a slight sharpening is observed in the PXRD on an increase in concen- tration and the bandgap appears to remain unchanged at ca. 1776 J. Muter.Chern., 1996, 6(11), 1771-1780 "" I I ow I I particle size/nm Fig. 10 Size distribution graphs of CdS samples prepared in toluene at different Cd:2PySiPB concentrations of (a) 0.5 (run Cd2) and (b)0.2 (run Cd6) 2 v)C a,c .-c 0 wavel engt h/nm Fig. 11 PA spectra of (a) commercial (greenockite) CdS, (b) sample prepared in the presence of 2PySiPB at room temp. (run Cd2) and (c) sample prepared in the presence of 2PySiPB at -78 "C (run Cd3) 2.6 eV (Fig. 12) although it is shifted from the bulk value. A change in Cd/N ratio for diethyl ether and toluene runs gives no change in particle size albeit a large shift in the band edge to higher energy when Me,Cd/( pyridine on the polymer) =0.2. We relate this curious behaviour to severe defects (multiple domains of no coherence) within the crystallites at lower ratios and the tendency of the band edge to be determined by the smallest dimension and not the diameter of the crystallite. Discussion There are several approaches to obtaining metal sulfide nano- crystallites, notably by the reaction of a metal source with H2S 0.7 0.6 0.5 0.4 0.3 0.21 0.1 0 wavelength/n m Fig.12 PA spectra of CdS samples prepared in diethyl ether in the presence of 2PySiPB at different concentrations of (a) 10% (run Cd5) and (b)3% (run Cd4) (or sometimes Na2S or S8).The metal sources mainly reported for the synthesis of ZnS particles are zinc salts.27 Recently, organometallic compounds such as Zn( SPh)240 and the organometallic block copolymers of methyltetracyclododecene (MTD) and (bTAN)(ZnPh), {bTAN=2,3-truns-bis[(tert-butylamido)methyl] norborn-5-ene} 26 have been used as more reactive Zn sources, while there are no reports using ZnMe, as the Zn source, perhaps because of the pyrophoric and air- sensitive nature of the compound.Polymer ligands (or Lewis bases) will react with Me2M (M=Zn or Cd) to give soluble ad duct^.^' The Zn-polymer adduct can be isolated and characterized by 'H NMR spectroscopy. The Cd-polymer adduct, however, is difficult to isolate mainly because of the weak Cd-N bond and any purification of the polymer film by removal of the solvent under vacuum results in the dissociation of the Me2Cd-polymer adduct to give the free polymer.Reaction of these polymer adducts, generally prepared in situ from the polymer and the metal alkyl, with H2S produces nanoparticles dispersed within the polymer film. The ZnS PXRD pattern of run Zn2 in toluene [Fig. 3(b)] is very broad (<2 nm particle size) suggesting the presence of a plethora of crystallite defects. This assumption can be made mainly because the average particle size calculated from the TEM of the same sample is 3.7nm. A much sharper PXRD pattern is expected for crystallites of this size which therefore suggests that defects drastically affect the hkl-dependent line broadening and the accuracy of particle size measurements. For similar reasons, CdS particle size measurements from PXRD are considerably inaccurate.ZnS runs in other solvents show a better comparison between PXRD and TEM which suggests the presence of less defects and a more uniform crystallinity. All the PXRD and TEM patterns for the ZnS and CdS samples are consistent with their being of the hexag- onal phase. It is a feature of all the PXRD patterns [excluding that of the CdS sample obtained from light petroleum, run Cd8, Fig. 13(u)], however, that the (101) and (103) reflections are severely broadened and the (102) reflection is practically unobserved. The suppression of the (103) reflection suggests that there is a certain degree of crystal imperfection (e.g. J. Muter. Chem., 1996, 6(11), 1771-1780 1777 s d 2 m H2 c4\2 I 1 -1. ,1 1""1".'1"'-1 30.0 40.0 50.0 60.0 2Bldegrees Fig.13 (a) PXRD pattern of CdS prepared in the presence of 2PySiPB in a 3% light petroleum solution (run CdS) and (b)JCPDS pattern for hexagonal CdS (greenockite). Impurities present within vaseline are denoted as 0. stacking faults, turbostratic distortions and twinning).'374f942 Similar line broadening has already been observed by Bawendi and co-w~rkers'~ in nanoparticles of hexagonal CdSe. However, in some cases others43 have assigned similar PXRD patterns arising as predominantly from the cubic phase where perhaps now the hexagonal phase with crystal imperfections would be a more accurate description. The broadening or sharpening of the (002) reflection is indicative of either partial registry in the ~rystallite~~ or some degree of preferred orien- tation along the direction of the plane.It is noticeable that the spherically shaped crystallites formed in light petroleum coincide with sharper h01 reflections and broader (002) reflec- tions. However, runs in light petroleum (4.0nm; range 2-10 nm) have much wider size distributions than analogous diethyl ether (3 nm; range 2-5 nm) or toluene (4.0 nm; range 2-6 nm) runs and a much lower polymer content. This arises because the polymer is of low solubility in light petroleum and, in effect, the reaction between Me2Zn and H2S occurs in the absence of the polymer. The particles and the polymer become mixed on evaporation to dryness but the particles are not dispersed in the polymer so they tend to cluster. The dependence of crystallite growth on variations in tem- perature is clearly observed when considering results for the formation of ZnS and CdS particles. In this system, it is probable that the polymer plays a variety of different roles.Thus, we have shown that the metal alkyl forms adducts with the pendant pyridyl moieties but that these interactions are weak and reversible [eqn. ( l),M =Zn or Cd]. Me,M +2PySiPB GMe2Zn* [P!.SiPB], (1) If we consider zinc only and since it can accommodate 18 electrons in its outer shell, the adduct on the right of the equilibrium in eqn. (1) will be unreactive towards H2S. At least partial dissociation of the adduct will be required for initial reaction to occur. Thus, anything 7n hich affects the position of equilibrium in eqn.(1) (e.g. temptrature) will also affect the rates of nucleation and of growth of the particles. At higher temperatures, less of the Me2Zn will be bound to the polymer and more Me2Zn molecules will be available for reaction with H2S [eqn. (2)]. This reaction suggests that larger particles should form which compares well with the experimen- tal observation. Me,Zn +H2S+(ZnS), +( ZnS), + +2CH4 (2) Formation of larger particles owing to the process of Ostwald ripening5 is also observed. The probability of larger seed growth, (ZnS),, at the expense of the smaller seed, (ZnS),, is increased at higher temperature. The destruction of the smaller seeds and the subsequent release of Zn and S ions as shown [eqn.(3)] is followed consecutively by the growth of ZnS on the larger (ZnS), seeds in a similar manner to the reaction previously described in eqn. (2). (ZnS), +( ZnS), -+Zn2+ +S2-(3) It is also expected, however, [and we have shown this in gas- phase reactions of Me2M (M=Zn or Cd) with H2S in the presence of pyridine3'] that the polymer bound pyridine units will also bind to the surface of the particles [eqn. (4)] ZnS +PySiPB S-Zn PySiPB (4) Once again, the position of the equilibrium in eqn. (4) is important since the 2-methylpyridine bound zinc atoms will not be available for reaction with H2S and particle growth will be inhibited. Lower temperatures will favour the species on the right of eqn. (4)and enhance the possibility of early growth termination and the formation of smaller particles.Since this is observed, it seems probable that an important role of the polymeric Lewis base is to control the particle size by termination cf particle growth as well as by control of particle nucleation. A decrease in the Zn/N ratio results in a decrease in particle size due to the same effects described for changes in tempera- ture. When decreasing the Zn/N ratio, the polymer chain has more coordinating sites available in the reaction medium. 1778 J. Muter. Chern., 1996, 6(ll),1771-780 These free bases are able to terminate growth at an early stage by binding to the surface of the growing particle, and thereby stabilize the formation of smaller particles26 by preventing further growth.The bases also provide steric hindrance which prohibits growth due to the blocking of potential sites for the reaction with H2S.Increasing the concentration of the polymer in solution also appears to increase the crystallite size. This change in particle size could possibly be due to the nucleation of particles in close proximity to each other resulting in the formation of agglomerates and larger particle growth. An alternative and more plausible view could be that the effective- ness of the metal-polymer Lewis-base interaction in promoting early growth termination is less than that of the solvent. Therefore, fewer solvate molecules, i.e. less Zn-0 bonds, and the increased presence of the Zn -N metal-polymer interaction would result in an increase in the growth of the particle.In the case of CdS particle growth, the dependence of crystallite size on changing Cd/N ratio and polymer concen- tration is less than that for ZnS. The main reason for this is related to the extremely weak Cd-N bond. In Table 2, for the series of runs in diethyl ether, the particle sizes are smaller than analogous runs in less polar solvents suggesting that the polar solvent interaction will always be stronger than in runs where the polymer interacts with the metal. At this level of solvent saturation, any variation in solvent volume will have only a small effect on a change in particle size. In toluene, the polymer appears to exert a greater control on particle size than the solvent as it alone is now able to block the surface sites and as a result, larger particle sizes are recorded.In the case of the non-solvent, light petroleum, the particle size varies considerably when changing the Cd/N ratio which is probably related to the insolubility of the polymer in the solvent and not a consequence of metal-polymer interaction. The effects of the solvent on ZnS and CdS particle growth are generally determined by the polarity of the solvent. As shown earlier in the size distribution graphs of CdS (runs Cd2 and Cd6, Fig. 10) and ZnS [run 2112,Fig. 6(b)] samples prepared in toluene at fairly similar concentrations, Cd/N ratio and temperature, the average particle sizes for CdS (cu. 4 and 4.3 nm for Cd2 and Cd6 respectively) are larger than ZnS (3.7 nm).In such non-polar solvents, the strength of the 2PySiPB interaction with the metal on the surface of the growing particle enables it to exert control over termination of particle growth. For ZnS, termination is observed at an earlier stage due to the stronger interaction of Zn-N as compared to Cd-N and a greater efficiency in blocking of metal surface sites prior to reaction with H2S. Polar solvents such as diethyl ether and thf, tend to mimic the polymer, pushing the equilibrium of eqn. (4) over to the right thereby terminating growth early on. The particle sizes of CdS are generally larger for reactions in toluene as compared to reactions in diethyl ether indicating that the Cd-N interaction is weaker than the combined Cd-N and Cd-0 interactions.Comparing this case with the formation of ZnS particles where the reactions in toluene give crystallite sizes comparable to reactions in thf, clearly, the interaction between the metal and the polymer is stronger in the case of Zn and is comparable to the effect of the solvent in stabilizing the particle size. Narrow size distributions are observed for CdS nanoparticul- ates formed in both diethyl ether and toluene solvents. Reactions in the non-solvent, light petroleum, give wider size distributions indicating that the solubility of the polymer in the solvent is a major factor in particle size control. Conclusions The reaction of polymer adduct solutions with H,S provides a simple and reliable route to synthesizing ZnS and CdS semiconductor, nanosized particles of a relatively narrow size distribution in the size range 2-5 nm.On changing the reaction conditions, i.e. concentration of polymer, metal-polymer ratio, solvent, temperature and choice of polymer ligand, the size of the growing crystallite can be controlled. ZnS nanoparticulates are probably wurtzite (hexagonal) in phase and in some cases appear to exhibit partial registry. Likewise, for CdS particles, PXRD patterns suggest that the hexagonal phase, greenockite, is present and exhibit hkl-dependent line broadening. All prepared nanoparticulate materials show typical size quantiz- ation effects corresponding to an increase in bandgap values as the particles decrease in size.We are currently developing a series of polymer encapsulated quantum dot materials and are also testing their optical properties. We thank John Mackie at the Bute, St. Andrews, for low- resolution TEM; Dr. P. Lightfoot for XRD; the EPSRC for funding (S.H.) and the Royal Society for a K. C. Wong Fellowship (X.L.). We are also grateful to the Nippon Soda Company for generous gifts of polybutadienes. References 1 M. G. Bawendi, M. L. Steigerwald and L. E. Brus, Annu. Rev. Phys. Chem., 1990,41,477. 2 A. Henglein, Chem. Rev., 1989,89, 1861. 3 Y. Wang and N. Herron, J. Phys. Chem., 1991,95,525. 4 H. Weller, Adv. Muter., 1993,5,88. 5 H. Weller, Angew. Chem., Int. Ed. Engl., 1993,32,41. 6 J. H. Fendler and F. C. Meldrum, Adv.Muter., 1995,7,607. 7 A. R. Kortan, R. Hull, R. L. Opila, M. G. Bawendi, M. L. Stiegerwald, P. J. Carroll and L. E. Brus, J. Am. Chem. SOC., 1990, 112,1327. 8 K. Kalyanasundaram, in Energy Resources by Photochemistry and Catalysis, ed. M. Gratzel, Academic Press, London, 1983, p, 217. 9 C. Flytzanis, F. Hache, M. C. Klein, D. Ricard and P. Roussignol, Progress in Optics, 1991,29, 321. 10 S. H. Risbud, The Encyclopedia of Advanced Materials, ed. R. W. Cahn, Cambridge University Press, Cambridge, 1994, pp. 2115-2121. 11 E. Corcoran, Sci. Am., 1990,263,74. 12 N. Herron, Y. Wang and H. Eckert, J. Am. Chem. SOC., 1990, 112,1322. 13 C. B. Murray, D. J. Norris and M. G. Bawendi, J. Am. Chem. SOC., 1993,115,8706. 14 J. G. Brennan, T.Siegrist, P. J. Carroll, M. Stuczynski, L. E. Brus and M. L. Steigerwald, J. Am. Chem. SOC., 1989,111,4141. 15 B. Breitscheidel, J. Zieder and U. Schubert, Chem. Muter., 1991, 3, 559. 16 Y. Wang and N. Herron, J. Phys. Chem., 1987,91,257. 17 X. K. Zhao, L. McCormick and J. H. Fendler, Chem. Muter., 1991, 3, 922. 18 J. N. Robinson and D. J. Cole-Hamilton, Chem. SOC. Rev., 1991, 20,49. 19 Y. Wang, A. Suna, M. Mahler and R. Kasowski, J. Phys. Chem., 1987,87,7315. 20 M. E. Wozniak, A. Sen and A. L. Rheingold, Chem. Muter., 1992, 4,753. 21 J. P. Kuczynski, B. H. Milosavljevic and J. K. Thomas, J. Am. Chem. SOC., 1986,108,2513. 22 Y. Wang and W. Mahler, Opt. Commun., 1987,61,233. 23 E. Hilinski, P. Lucas and Y. Wang, J. Chem. Phys., 1988,89,3435.24 Y. Wang, A. Suna, J. McHugh, E. Hilinski, P. Lucas and R. D. Johnson, J. Chem. Phys., 1990,92,6927. 25 S. Yanagida, T. Enokida, A. Shihdo, T. Shiragami, T. Ogata,T. Fukumi, T. Sakaguchi, M. Mori and T. Sakata, Chem. Lett., 1990,1773. 26 V. Sankaran, J. Yue, R. E. Cohen, R. R. Schrock and R. J. Silbey, Chem. Muter., 1993,5, 1133. 27 Y. Yuan, J. H. Fendler and I. Cabasso, Chem. Muter., 1992,4,312. 28 A. Iraqi and D. J. Cole-Hamilton, J. Muter. Chem., 1992, 2, 183 and references therein. 29 P. Narayanan, B. Kaye and D. J. Cole-Hamilton, J. Muter. Chem., 1993,3, 19. 30 A. Iraqi, S. Seth, C. A. Vincent, D. J. Cole-Hamilton, M. D. Watkinson, I. M. Graham and D. Jeffrey, J. Muter. Chem., 1992, 2, 1057. 31 X. Li, C. M. Lindall, D. F. Foster and D. J. Cole-Hamilton, J. Muter. Chem., 1994,4, 657. J. Muter. Chem., 1996, 6( ll), 1771-1780 1779 32 D. J. Cole-Hamilton, Chem. Br., 1990,26, 852. 39 N. L. Pickett, D. F. Foster and D. J. Cole-Hamilton, J. Muter. 33 34 X. Li, D. F. Foster and D. J. Cole-Hamilton, Polym. Ado. Technol., 1994,5, 541. X. Li, J. R. Fryer and D. J. Cole-Hamilton, J. Chem. SOC., Chem. Commun., 1994,1715. 40 Chem., 1996,6,507. J. G. Brennan, T. Siegrist, P. J. Carrol, S. M. Stuczynski, P. Reynders, L. E. Brus and M. L. Steigerwald, Chem. Muter., 1990, 2,403. 35 36 37 38 D. F. Foster and D. J. Cole-Hamilton, Inorg. Synth., 1996,in press. M. G. Bawendi, A. R. Kortan, M. L. Steigerwald and L. E. Brus, J. Chem. Phys., 1989,91,7282. A. R. West, Solid State Chemistry and its Applications, J. Wiley & Sons, Chichester, 1992, p. 174. A. Rosencwaig, Optoacoustic Spectroscopy and Detection, ed. Y-H. Pao, Academic Press, London, 1977,p. 217. 41 42 43 Y. G. Andreev and T. Lundstrom, J. Appl. Crystallogr., 1995, 28, 534. Y. G. Andreev and T. Lundstrom, J. Appl. Crystallogr., 1994, 27, 767. G. Krauter and W. S. Rees, J. Muter. Chem., 1995,5,1265. Paper 6/03795J; Received 31st May, 1996 1780 J. Mater. Chem., 1996, 6(11), 1771-1780
ISSN:0959-9428
DOI:10.1039/JM9960601771
出版商:RSC
年代:1996
数据来源: RSC
|
12. |
Crystal structure and magnetic properties of Ba10(MnFeF11–xClx)3FxCl2 –x(x= 0.85). Structural relationships with the apatite-type structure |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1781-1784
Jacques Darriet,
Preview
|
PDF (922KB)
|
|
摘要:
Crystal structure and magnetic properties of Bale( MnFeFll -xC1,)3F,C12 -(x =0.85). Structural relationships with the apatite-type structure Jacques Darriet,* Virginie Nazabal and Jean Fompeyrine Institut de Chimie de la MatiBre Condensde de Bordeaux (ICMCB), UPR9048, Chateau Brivazac, Avenue du docteur A. Schweitzer, 33608 Pessac Cedex, France The crystal structure of the chlorofluoride Ba,,( MpFeF,, -xClx)3FxC12 -x (x=0.85) has been determined from single-crystal X-ray data with the following parameters: a= 11.075(2) A and c= 8.173(2) A,and the trigonal P31m space group. The residual values (R values) are R1= 0.0275 (based on F,) and wR2 =0.0584 (based on FO2)for 1384 unique reflections and 103 parameters. The structure can be described in terms of infinite chains of face-sharing MBa6 (M =F or Cl) octahedra, located at the origin of the unit cell and parallel to [OOl].Two other Ba atoms occupy the (1/3,2/3,0) and (2/3,1/3, ca. 0.5) positions. This network is similar to that encountered in the apatite-type structure [Ca,(PO,),(F,OH,Cl)]. The structure is highly disordered with the Mn and Fe atoms forming either isolated dimeric [MnFeF,, -xClx] groups of corner-sharing octahedra (85%) or as two isolated units (15%). This result has been confirmed by the magnetic properties and a theoretical model is proposed to fit the thermal variation of the magnetic susceptibility. The homolog9us phase with Fe2+ and Cr3 has been prepared. The corresponding unit-cell parameters + are: a= 11.033(2) A and c=8.180(2) A. The structural relationships with the apatite-type structure and related phases are discussed.The literature on the crystal chemistry of chlorofluorides is in ochromated Mo-Ka radiation. Crystal data are given in fact not very important. The most recent structural investi- Table 1. From the extinction conditions, the possible space gations involving 3d cations or A1 are K2Cr3F6C1,,1 groups are P3m1, P31m, Pjrnl, P?Jlrn, P321 and P312. Sr,oA12F25C1,2 K,Ba7A16F,,C12,3 the series Ba2MM'F7C14 (M, Intensities were corrected for Lorentz and polarization effects M' =3d ions), Ba6Mn2ZnFl2c1~ and B~,CU,F,,C~.~ and empirical absorption corrections were made using the psi- Generalizations in chlorofluoride crystal chemistry are not yet scan technique.Calculations were performed using possible since syntheses of chlorofluorides have not been SHELXS86, and SHELXL938 programs. Complete lists of investigated extensively, particularly when two 3d elements bond lengths and angles, anisotropic displacement coordinates are present in two oxidation states (2 + and 3 +). This paper and observed and calculated structure factors, are available deals with the crystal structure of the new chlorofluoride from the Institut fur Anorganische Chemie, Bonn (Germany). Balo(MnFeF,, -xC1x)3FxC12-x (x=0.85) belonging to the Atomic coordinates, thermal parameters, and bond lengths BaF,-BaC1,-MnF,-FeF, quaternary system. A homologous and angles have been deposited at the Cambridge phase has been observed in the BaF,-BaC1,-FeF2-CrF, system. Crystallographic Data Centre (CCDC).See Information for Authors, J. Mater. Chem., 1996, Issue 1. Any request to the CCDC for this material should quote the full literature citation Experimenta1 and the reference number 1145/13. . Various preparations of different compositions in the BaF2-BaC1,-MnF,-FeF, system showed the existence of a Table 1 Crystal data and structure refinement parameters of new phase close to the formula Bale( MnFeFloCl),FCl. The Ba,,,(MnFeF,, -xClx)3FxC12-x (x=O.85) corresponding X-ray pattern showed a mixture of this new Ba,,(MnFeF,, -xClx)3FxC12-xphase and traces of BaFCl as an impurity (~3%).The con- formula (X =0.85)ditions of preparation were as follows. The starting materials Mw 2431.64(BaF,, BaCl,, MnF, and FeF,) were weighed (ca.3g) and wavelength/A 0.71073 ground in an agate mortar, and placed in platinum tubes, the crystal system trigonal whole procedure being performed in a dry box.The platinum space group P31m (no. 157) tubes were sealed under argon and heated to 600 "C for 12 h unit-cell dimensions/A U= 11.075(2) and cooled slowly to room temperature (6 "C h-'). A treatment volume/A3 c =8.173( 2) at 750°C resulted in the formation of transparent light pink Z 868.22 1crystals of the phase. An examination of a well shaped needle- Dc/g cm3 4.651 like single crystal by the Buerger precession technique and absorption coefficient/mm -13.84 rotation method showed that theo symmetry is trigonal with F(OO0) 1057.6 the unit-cell parameters ax 11.08 A and cw8.18 A.No extinc- crystal size/mm3 0.2 x 0.25 x 0.2 tion conditions were found. These parameters allowed the 8 range for data collection/degrees 2.5-35" -17<h<17, -17<k <17,indexing of the X-ray powder diffraction data corresponding index ranges -0<1<13 to this new phase. An isostructural phase was also isolated in reflections collected 7139 the BaF,-BaC1,-FeF,-CrF, system under the same exper- independent reflections 1384 [R(int)=0.0387)] imental conditions. The correspondinog unit-cell parameters refinement method full-matrix least-squares on F2 were: a= 11.033(2) A and c=8.180(2) A. data; restraints; parameters 1384; 0; 103 R1= 0.0275, wR2 =0.0584A selected single crystal of Ba,,( MnFeF,,C1)3FCl was final R indices [I >3o(I)] absolute structure parameter 0.02(7)mounted on a four-circle automatic diffractometer CAD4 largest diff.peak; hole/e A-3(Enraf Nonius) and data were collected using graphite-mon- 1.636; -1.686 J. Mater. Chem., 1996, 6(11), 1781-1784 1781 Table 2 Atomic coordinates for Ba,,( MnFeF,, -xC1,)3F,C12-, (X =0.85) atom site X Y Z B,/A2 2b 213 113 0 1.39(1) 2b 213 1/3 0.5024( 2) 2.69(2) 3c 0.7322( 1) 0 0.7835( 2) S.36(1) 3~(8 5 Yo) 0.2325( 1) 0 0.2198( 3) 1.75(2) 3c( 15%) 0.2448 (6) 0 0.2844( 12) 1.75(2) 3c 0.6071(1) 0 0.261 1 (3) 1.20(2) 3~(8 5 Yo) 0.4068 (2) 0 0.8244(4) 1.39( 4) 3c( 15%) 0.3 8 60 ( 14) 0 0.74 14 (7) 1.61(8) 6d 0.8487(5) 0.4488 5) 0.7414(7) 1.61(8) 6d 0.8547( 6) 0.2428 6) 0.9514(7) 1.63(8) 3c 0.5376( 12) 0 0.0409( 13) 4.16( 25) 6d 0.7976( 6) 0.5123 6) 0.2592( 10) 3.00( 13) 3c 0.6806( 17) 0 0.4711( 15) 6.33(50) 3c 0.76 13 (8) 0 0.1557( 13) 2.02( 15) 6d( 50%) 0.4379( 14) 0.9653 18) 0.3556(20) 3.50( 43) 3c(85%) 0.2501 (6) 0 0.6090( 7) 3.22(9) 3c( 15%) 0.2501 (6) 0 0.6090( 7) 3.22( 9) la 0 0 0.9319( 8) 1.72(7) 1 a( 85 YO) 0 0 0.3214( 33) 3.10( 37) la( 15%) 0 0 0.4549( 66) 3.10( 37) Magnetic susceptibility and magnetization were measured using a Quantum Design MPMS 7s SQUID susceptometer.Data were recorded at a field of 0.25 T for which the magnetiz- ation is linear with the field in the whole temperature range (5-300 K). Experimental values were corrected for the core diamagnetism of the sample using Pascal's constants.Results and Discussion Crystal structure The structure was solved successfully in the space group P31m. The positional parameters for the barium, manganese and iron atoms were determined from the SHELXLS86 program and chlorine and fluorine atoms were found in subsequent differ- ence Fourier maps. Refinements were carried out by the full- matrix least-squares method. In a first stage all the atoms corresponding to the formula Ba,,( MnFeF,,Cl),FCl were located and the corresponding R factors with anisotropic thermal displacement parameters for all the atoms decreased to R1= 0.0440 (based on F,) and wR2 =0.1058 (based on Fo2). All the positions are fully occupied except the F(7) site which is shifted to a half-occupied general position near to the xOz ideal position (Table 2).However, a difference Fourier synthesis showed maxima near the Ba(4), the Mn and the F(8) positions. These atoms were delocalized onto Ba(4) and Ba(4'), Mn and Mn' and F(8) and Cl(8') sites, respectively (Table 2). Owing to the high correlations between occupancy and temperature factors, simultaneous refinements were difficult, therefore the anisotropic thermal parameters for each pair of atoms were refined with the same values (Table 2). The refinements show clearly that the occupancy rate of the pairs of positions are the same (85%, 15%; Table2). The corresponding R factors are R1=0.0280 and wR2=0.0624. At this stage, an analysis of the bond distances spows that the Ba(4)-C1(1) (2.653 A) and Mn'-C1( 1) (1.868 A) are too short, therefore a disordering between chlorine C1( 1) and fluorine F(1') in the same position was considered. The refinement led to the same ratio of disordering as found previously, with a repartition of 85% of chlorine and 15% of fluorine atoms (Table2). The final R values decreased to R1=0.0275 and wR2 =0.0584 {w= 11 [a2F02+(O.O203P), + 5.04P1 with P = [max.(0 or Fo2)+ 2FC2]/3}. The electron density in the Fourierodifference syn- thesis at the final refinement cycle was 1.6 e A-3 (max.) and -1.7 e A-3 (min.). The structure is highly disordered and it could not be solved routinely. Several hypotheses have been formulated, and we have admitted as evidence both the decreas- ing of the R values and the observation of satisfying bond distances.The main interatomic distances are given in Table 3. The structural formula of the studied single crystal is Balo(MnFeFl1 -xC1x)3FxC12-x with x=0.85. It can be thought of as a 'mixture' of two configurations, one being Balo(MnFeFl,C1),FC1 (85%) and the other one Balo(MnFeF,,),C12 (15%). A strong similarity is found with the chlorofluoride K,Ba7Al,F,,C1,,3 which can also be formu- lated (K,Ba,)(Al,F,,),Cl, and for which a crystallographic disordering has also been observed. Views of the structure for the two limiting values of x are given in Fig. 1 and 2, and the corresponding projections along the [OOl] direction in Fig. 3 and 4. In both cases, the structure consists of Ba, octahedra sharing faces running along the three-fold axis located at the origin of the unit cell (Fig. 3 and 4).The two other barium atoms [Ba(l) and Ba(2)] are located on the other two ternary axes (Fig. 3 and 4). The framework of the barium atoms is very similar to that of calcium encountered in the well known apatite-type structure [Ca,(P0,)3(C1,0H,F)],9 (compare Fig. 5 and 3 or 4).The main difference between the hydroxy; fluoro- and chloro-apatites is the position of the anions in the face- sharing Ca, octahedra chains (Fig. 6). The C1 atom is located at the centre of the octahedron (z= 1/2) [Fig. 6(a)] whereas the fluorine is displaced in one face in z= 1/4 [Fig. 6(b)]. For the hydroxyapatite the oxygen atom occupies an intermediate position [Fig. 6(c)].The same phenomena are observed in the Table 3 Selected bond lengths/A for Ba,,(MnFeF,, -xClx)3FxC12-x (x=0.85) Fe-F(7) Fe-F( 5) Fe-F( 6) Fe-F(4) 2 x Fe-F(3) 1.880( 11) 1.900( 13) 1.913( 8) 1.943(6) 1.957 (9) Mn-F(2)"2 x Mn-F( 1)"2 x Mn-F( 3)" Mn-C1( 1)" 2.010( 6) 2.066( 5) 2.286( 10) 2.472( 6) Mn'-F( lob Mn'-F( l)b2x Mn'-F(2)b2x 1.868( 18) 2.327( 15) 2.112( 12) C1(2)-Ba(3)3 x C1(2)-Ba(4)"3 x C1(2)-Ba(4')b3 x 3.204( 3) 3.489( 5) 3.955( 10) F(8)-Ba(4)"3 x 2.706( 8) Cl(8')-Ba(4')b3 x C1( 8')-Ba( 3)b3 x 3.048( 26) 4.003(24) Ba( 1)-F( 1)3 x Ba(1)-F(2)3 x Ba(1)-F(4)3 x Ba(1)-F( 3)3 x Ba(4)-F(6)"2 x Ba( 4)- F(8)" Ba( 4)- F(7)" Ba(4)-F(2)"2 x Ba(4)-F(4)"2 x Ba( 4)- C1( 1 )" 2.755( 6) 2.755( 6) 2.765( 7) 3.241(2) 2.662( 5) 2.706( 8) 2.725( 14) 2.757( 6) 3.024( 7) 3.187( 6) Ba(2)-F(1)3 x Ba(2)-F(4)3 x Ba(2)-F(7)3 x Ba( 4')-F( 7)b Ba(4'-F( Ba(4'-F(6)b2 x Ba(4')-F(4)b2 x Ba(4')-Cl( 8')* Ba(4')-F(2)b2 x 2.634( 5) 2.666( 7) 3.145( 17) 2.426( 15) 2.653( 11) 2.877( 8) 2.960( 8) 3.048( 26) 3.1 56 ( 10) Ba(3)-F(5) Ba(3)-F(2)2 x Ba(3)-F( 1)2 x Ba(3)-F( 3) Ba( 3)-F( 6) Ba( 3)- C1( 2) Ba(3)-C1(1)"2 x Ba(3)-F(l')b2 x 2.61 6 ( 12) 2.703 (6) 2.741 (5) 3.0S2( 14) 3.059( 11) 3.2O4( 3) 3.208( 4) 3.208 (4) a 85% in this configuration. 15% in this configuration.1782 J. Muter. Chem., 1996,6( ll), 1781-1784 Fig. 1 View of the structure of Ba1,(MnFeF1,~,C1,),F,C1,~, for x= 1 Fig. 2 View of the structure of Balo( MnFeF,, -xClx)3FxCl~ -,for x =0 Fig. 3 Projection of the Balo(MnFeF,, -xClx)3FxCl~~x structure (x = 1) along the [OOl] direction Fig.5 Projection of the apatite-type structure along the [OOl] direction Fig.6 Positions of the C1-, F-and OH-in the Ca, octahedron of the apatite-type network structure of the chlorofluoride: the Cl(2) atom is located near the centre of the octahedra formed by the Ba(4) and Ba(3) atoms (85%) or the couple Ba(4')Ba(3) (15%); the fluorine F(8) is displaced near a face and only linked to the Ba(4) atoms, whereas the C1( 8') chlorine position is shifted towards the centre of the octahedra formed by the Ba(3) and Ba(4') atoms (Table 3). The displacement between the F(8) and Cl(8') positions along the c axis is 1.09 A.It is satisfying to observe ihe same type-of anionic disordering in the structure of the chlorofluoride as those observed in the apatite-type structures.The main difference between the structure of Balo(MnFeF1, -xC1,)3F,C12-, (x =0.85) and the apatite model concerns the species inserted between the Ba6 octahedra chains running along the [OOl] direction. In the apatite structure, the PO, tetrahedra are isolated from each other (Fig. 5). In contrast, for the main configuration (85%) of the structure of the chlorofluoride, the octahedra share a corner, forming isolated dimeric units [Fig. 7(a)]. These octahedra pairs are constituted by FeF6 and MnF,Cl octahedra with the chlorine atom in the trans position with respect to the common fluorine atom F(3) [Fig. 7(a)]. The FeF6 octahedra are quite regular with Fe-F $stances ranging from 1.88 to 1.94A; the mean value [1.92 A] is in agreement with Fe3+ located in these Fig.4 Projection of the Balo(MnFeFl,-,Cl,),F,C1,-, structure (x= Fig. 7 Perspective view of the Mn and Fe environments in the 0) along the [OOl] direction Ba1,(MnFeFl,~,C1,),F,C1,~, with x= 1 (a) and x=O (b) J. Muter. Chem., 1996, 6(11), 1781-1784 1783 sites.” The MnF5C1 octahedra are elongated along the F(3)-Cl(l) axis (Table 3). For the less occupied situation (15%), the environment of the Mn2+ ions [Mn’ position] is so distorted that it becomes a psFudo-square pyramid with a mean Mn2+ -F distance of 2.15 A (Table 3). The result of the displacement of the manganese atom is that the MnF, square pyramids and the FeF, octahedra are now isolated [Fig.7(b)]. This situation can be compared to that observed in the structure of Bas( Re05)3C111’12 where the rhenium atoms have square-pyramidal oxygen coordination (Fig. 8). To conclude with the structural aspects, we note the great flexibility of the apatite framework which can contain different species as two isolated tetrahedra, two isolated square pyra- mids, pairs of octahedra sharing a corner or one square pyramid and one octahedron. Magnetic properties The magnetic susceptibility of a polycrystalline sample (selected small crystals, m= 7.42 mg) fitted from 135-300 K (Fig. 9) conforms closely to a Curie-Weiss law, with a Curie constant C=0.32 x lop3m3 mol-I K and Weiss constant @= -44 K. The negative value of 0 indicates antiferromagnetic near- neighbour exchange.Below 135 K, the inverse susceptibility deviates from Curie-Weiss behaviour with an abrupt decrease below 20 K (Fig. 9). The value of the Curie constant calculated for six S=5/2 ions per formula unit, each with g=2, is 0.33 x m3 mol-’ K and is in good agreement with the observed experimental value. The magnetic curve has been Fig. 8 Projection of the Ba,(ReO,),Cl structure along the [OOl]direction 11111111111111 40 80 120 160 200 240 280 TfK Fig. 9 Thermal variation of the molar reciprocal susceptibility of Ba,,( MnFeF,, -xClx)3FxC12 -x (x=0.85) [black circles (exptl.), solid line (calc.)] fitted by considering a mixture of two magnetic species, one being isolated S =5/2 ions, and the other one antiferromagnetic pairs of two spins S=5/2.In the latter, the magnetic exchange interaction is treated by a Heisenberg model with H = -2JS1S2 where J represents the exchange interaction within the mag- netic pairs. By a vectorial decomposition of the angular spin momenta, the following expression for the susceptibility of a pair of spins, S= 5/2 can be deduced: xdim =(Ng2p2/kT){ [1IO exp (304 +60 exp (20x1 +28 exp(12x)+lOexp(6x)+2 exp(2x)]/[1+3 exp(2x) +5 exp(6x)+7exp(l2x)+9exp(20x)+ll exp(30x)l) where x =J/kT. The total molar susceptibility is expressed as: xtotal=3(1 -a)xdim+ 6a~5/2, where a represents the proportion of isolated spins 5/2 with an atomic susceptibility x5/2. If the magnetic exchange interaction within the dimers is antiferromagnetic, the contribution of the isolated spins pre- dominates mainly at low temperature and then a Curie law is observed. When the temperature increases, there is a compen- sation between the susceptibilities of the monomers and the dimeric units, and then a plateau is observed. The temperature range of this plateau is related directly to the value of the intradimeric magnetic exchange interaction.At higher tempera- tures, both the monomer and dimer susceptibilities decrease and a Curie-Weiss law is observed. The best agreement between the calculated and observed values was obtained for J/k= -6.1 K, g=1.94 and a=0.135 (Fig. 9). This result agrees very well with the structural determi- nation where the rate of exchange was monomer refined to 0.15.Conclusion The crystal structure of the new chlorofluoride Ba,,(MnFeF,, -xC1,)3F,C12-x (x=0.85) shows strong anal- ogies with the network of the apatite-type structure. The structure is highly disordered, as observed in the hydroxy; fluoro-and chloro-apatites. The structural relationships between these phases and the related structure type Ba5(ReO5),C1 has been described. The main result is the great flexibility of the apatite framework which can contain various species as tetrahedra, square pyramids or octahedra. The bulk magnetic susceptibility of Ba,,( MnFeF,, -xClx)3FxC12 -, (x=0.85) has been interpreted and confirms the rate of cationic disordering observed from the structural determination. The drawings were produced with ATOMS (E.Dowty, 1994) References 1 J. C. Dewan, A. J. Edwards and J. J. Guy, J. Chem. SOC., Dalton Trans., 1986,2623. 2 A. Hemon and G. Courbion, J. Solid State Chem., 1989,81,293. 3 A. Le Bail, A. Hemon-Ribaud and G. Courbion, J. Solid State Chem., 1993,107,234. 4 J. J. Maguer, G. Courbion, M. S. Schriewer, J. Fompeyrine and J. Darriet, J. Solid State Chem., 1995, 115,98. 5 J. Darriet, M. Ducau, M. Feist and A. Tressaud, Eur. J. Solid State Inorg. Chem., 1992,29,435. 6 J. Fompeyrine, V. Nazabal, J. Darriet and G. Courbion, Eur. J. Solid State Inorg. Chem., 1995,32,977. 7 G. M. Sheldrick, Acta Crystallogr., Sect. A, 1990,46,467. 8 G. M. Sheldrick, SHELXL93, Programfor the rejnement of crystal structures, University of Gottingen, Germany. 9 K. Sudarsanan and R. A. Young, Acta Crystallogr., Sect. B, 1978, 34, 1401 and refs. therein. 10 R. D. Shannon, Acta Crystallogr., Sect. A, 1976,32,751. 11 J. P. Besse, G. Baud, G. Levasseur and R. Chevalier, Acta Crystallogr., Sect. B, 1979,35, 1756. 12 M. S. Schriewer and W. Jeitschko, J. Solid State Chem., 1993, 107, 1. Paper 6/02186G; Received 28th March, 1996 1784 J. Mater. Chem., 1996, 6(11), 1781-1784
ISSN:0959-9428
DOI:10.1039/JM9960601781
出版商:RSC
年代:1996
数据来源: RSC
|
13. |
The electronic and magnetic structures of stoichiometric SrCoO3: ASW calculations |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1785-1788
Samir F. Matar,
Preview
|
PDF (602KB)
|
|
摘要:
The electronic and magnetic structures of stoichiometric SrCoO,: ASW calculations Samir F. Matar," Antoine Villesuzanne and Michael Uhl Institut de Chimie de la Matikre Condenske de Bordeaux ICMCB-CNRS, Chateau Brivazac, Avenue du Docteur Schweitzer, F33608 Pessac Cedex, France The electronic and magnetic structures of the stoichiometric ferromagnetic oxide SrCoO, are self-consistently calculated within the local spin-density functional theory using the augmented spherical wave (ASW) method. The influence of hybridization between the different l-states on the chemical bonding and the onset of the local magnetic moments are discussed from the site-projected densities of states (DOS) as well as from the modulation of the DOS by the sign and magnitude of the overlap integral, i.e.with the so-called crystal orbital overlap population (COOP). Agreement is found with the experimental value of magnetization per unit cell of 2.1 pBwhere moments are carried by both Co and 0.In addition the magnetic exchange properties are analysed. Few ferromagnetic oxides exist. This arises from the antiparallel spin alignment between the magnetic sublattices in the majority of cases leading to ferri-or antiferro-magnetic ordering. Stoichiometric SrCoO, was prepared under high pressures by Watanabe as early as 1957,' but recently Bezdicka et al. have used an original method to prepare it by electrochemical oxidation of SrCoO,., brownmillerite-type oxide.2 The resulting phase is a cubic perovskite (a=383.5 pm) having a ferromagnetic order with a relatively high Curie temperature for an oxide: Tc=280 K.The ordered magnetic moment of 2.1 pugis assumed to mainly arise from Co 3d and to a lesser extent from 0 2p. Qualitatively, the moment carried by oxygen is explained by the closeness of Co 3d to 0 2p states resulting into a partial electron transfer from oxygen to cobalt by virtue of covalency. This transfer, involving a finite spin density on oxygen, was evidenced by neutron diffraction3 and investigated by us theoretically4 in isostructural antiferromagnetic SrFeO, . The purpose of this work is to address the features of the chemical bonding relevant to the occurrence of the magnetic moments carried by the different species in SrCoO,. This will be carried out by use of first-principles self-consistent calcu- lations as a part of systematic studies of magnetically ordered oxides prepared and investigated in our The ASW calculation method and a new implementation As in former studies, we used the augmented spherical wave (ASW) method7 to calculate the electronic and magnetic properties of SrCoO,.This method is based on the local spin- density functional (LSDF) theory in which the exchange and correlation effects are included using the scheme of von Barth and Hedin' and Janak.g For calculations assuming non-mag- netic structures, the total spin population is accounted for in the potential. A magnetic moment is allowed to appear on an atomic site when majority spin (7) and minority spin (4) electron populations are treated explicitly.The potential and its exchange-correlation part within the local density approxi- mation (LDA) are then made spin dependent (local spin- density approximation; LSDA). The ASW method uses the atomic sphere approximation (ASA) which assumes overlap- ping spheres centred on the atomic sites, and a spherically symmetric shape of the potential. The volume of the spheres has to be equal to the cell volume because the charge in the interstitial region is only accounted for by contributions from the tails of the spherical wavefunctions extending out of the atomic spheres; this is detailed in the original work by Williams et aL7 For poorly packed crystal structures such as the perovsk- ite one, the empty space is represented by pseudo-atoms with 2=0 called empty spheres (ES), whose introduction is neces- sary to avoid an otherwise too large overlap between the actual atomic spheres: Sr, Co and 0;moreover, ES can account for possible covalency effects. Within the ASA our non-unique choice of atomic sphere radii was subjected to the following ratios: rSr/rES= 1.44,rCo/rES=1.14 and rO/rES= 1.26, which mini- mize the overlap between the spheres.The Brillouin zone (BZ) integration was carried out for a sufficiently large number of k points in the irreducible wedge of the first zone of the simple cubic Bravais lattice. For the two calculations carried out in the non-magnetic and magnetic states, 84 and 165 irreducible k points were used respectively.Usually magnetic calculations require a larger number of k points to ensure reliable values of the moments carried by the different atoms. Self-consistency was obtained when no variation of the charge transfers (AQ <lo-*) and of the total variational energy, Eva, (AE < Rydberg) could be observed upon additional cycles. Such convergence criteria are more strenuous than in our former works due to the improved precision of the actual version of the program in use." The following ASW basis function sets were used for the different atomic species: Sr and Co, 1=0, 1, 2; lma,=3; 0 and ES, l=O, 1; lma,=2 (I is the orbital quantum number). The charges associated with l,, correspond to residues from all higher 1 states not explicitly accounted for in the secular matrix, but should always be <0.1 total electrons (non-spin-polarized calculations) or 0.05 electrons/spin (spin-polarized calculations) to ensure convergence of the limited ASW basis Low-lying 0 2s orbitals are omitted from the limited ASW valence basis set and replaced by empty 0 3s orbitals.With this procedure the valence basis set is more complete. A preliminary check for our system shows an energy stabilization when 0 3s is used as a valence wavefunction. Furthermore, in this work, the chemical bonding features are discussed based on the so-called COOP (pronounced CO- OP; crystal orbital overlap population), of which a comprehen- sive account was given by Hoffmann from the quantum chemistry standpoint (extended-Huckel calculations)." This allows for the density of states (DOS) features to be discussed on bases of chemical bonding criteria by weighting them with the sign and magnitude of the overlap integral between the relevant orbitals.We recently implemented the COOP in the ASW method12 with the objective of extracting further infor- mation on the chemical bonding from first principles. Finally we point out that all calculations are implicitly performed at 0 K, thus accounting for ground-state properties. J. Mater. Chem., 1996,6(1l), 1785-1788 1785 Crystal structure SrCo03 crystallizes with one formula unit in the simple cubic perovskite structure (space group Pm3m). Cobalt is in a regular octahedron formed from six oxygen atoms with COO, octahedra sharing corners in three dimensions.The large Sr atoms sit at the cube corners and empty spheres are at positions +,O,O; O,+,O and O,O,+. From this it appears that the shortest distances are for 42, i.e. for Co-0 within the octahedron on one hand, and Sr-ES along the cube edges on the other hand. It is expected that strong interactions in the lattices will be favoured along these directions, i.e. rather than along the cube diagonal between Co and Sr. This will be shown in next section. The experimentally derived lattice constant2 is used through- out the present calculations. Calculations and Discussion Non-spin-polarized (NSP) calculations Objectives of the NSP calculations. SrCoO, orders ferromag- netically with T,=280 K and an ordered moment of 2.1 pB.2 However, we start our investigations firstly for an assumed non-magnetic state, i.e.non-spin-polarized (NSP). From the results of such calculations a preliminary approach of the chemical bonding can be performed. We propose to do so through the assignment of the states which mix (hybridize) together as well as through the COOP. Moreover, the NSP density of states (DOS) at the Fermi level (EF), n(EF), can be used within the Stoner theory of band ferromagnetism to assign a role in the onset of the magnetic moments for each atom. Electronic configurations. In our choice of the atomic radii the following charge transfers could be observed between the different species. The change of configuration of the valence states us.the starting neutral atomic configuration is: Sr 5s2, 5p0, 4d0+5s0s06, 5po.O8, 4d0.22, #f O-04; Co 4s2, 4p0, 3d74s0*22, 4p0.30, 3d6.74 # 0.06. 0 3s0, 2p4+3s0.02 2 4.70 3d0.08. ES ls09P ,7.f ) 72p0+1s0.12 2 0.11 3d0.05 . Th e values in italics correspond to 9 7P 7 charges associated with l,,, higher terms. There is a small d contribution for Sr which should result from the small mixing with Co 3d, i.e.from tails of Co 3d wavefunctions. The resulting departures from neutrality are AQsr = -1.59; AQco = -1.68; AQo= +0.81 and AQES= +0.28. The overall charge transfer can be understood in the following way: electrons that depart from cobalt transfer to oxygen, as expected, so that the residual difference, 3AQo -AQ,, =2.43 -1.68=0.75, can only arise from electrons departing from Sr.This gives 1.59 -0.75 =0.84 elec- trons leaving Sr to ES, i.e. the total charge in 3 ES, 3AQES, is 0.84. From this it appears that charge transfer is not balanced in the lattice for reasons of covalency. Despite the exhibited trends in charge transfer, it needs to be stressed that in the framework of ASW calculations the formal ionicity expected by the chemist cannot be obtained although a correct represen- tation of the physical and magnetic properties can be described such as in pure and substituted Cr02.13914 Density of states (DOS). The site-projected DOS (in eV-l atom-') are shown in Fig. 1 for SrCoO,. The energy scale along the horizontal axis is taken with reference to EF.As expected the Sr and ES DOS show similar shapes; they peak at -4, -2 and EF, i.e. at 0 2p and Co 3d DOS, but their contribution is negligibly small. Major DOS are observed for Co and 0 which show similar shapes in the energy range -6 to -2eV, i.e. in the range where 0 2p states interact with Co 3d to ensure chemical bonding. Note that there is no energy separation between 0 2p and Co 3d states which look similar. This is another indication of the covalency of the chemical bond within the COO, octahedra. The Co states show 1786 J. Muter. Chem., 1996, 6(11), 1785-1788 7t -6 4 -2 0 2 4 (E-E,)/eV Fig. 1 Partial densities of states (DOS) per atom of non-magnetic SrCoO,; Sr (dotted), Co (solid), 0 (dashed) and ES (dash-dotted) a rather broad band ranging from -6 to 4 eV due to the covalency of the chemical bond, and a sharp peak around and centred upon EFwhich seems to present less bonding character if one considers the relative intensities of the DOS peaks.The nature of these states can be understood straightforwardly using crystal field considerations. In the octahedral crystal field of oxygen atoms, Co 3d split into t2g and eg manifolds with the corresponding orbitals pointing between and towards oxygen 'ligands' respectively. This is shown in Fig. 2 for the Co d DOS projected along t2g and e,; the oxygen DOS are shown as well. There is no definite separation between t2g and e, like in molecular complexes; instead, these states mix together.Like in the formerly studied iron compound SrFe03,4 eg states form a-type bonds with oxygen (see the dashed line peak at cu. -4.5 eV) whereas t2g orbitals which point between the oxygen atoms form weaker n-type bonds (see the lower intensity full line peak at -4eV). Thus it can be seen that Co 3d states show itinerant behaviour through the broad eg band and a localized one through the narrow t2, band. It is hence expected that the magnetic moment which will develop on Co will mainly arise from the exchange splitting of t2g lying at the Fermi level and showing little interaction with oxygen. The latter exhibits an interesting feature of mixing not only with e, but also with t2g, mainly shown by the peak at EF which has the same aspect as Co (t2g).Another important feature is the broadness of eg as opposed to the bandwidth of t2g. Since the a bonds are much stronger than the rc bonds, the e, states should split into bonding and antibonding parts at low and high energies respectively. We can now show this I81 I I -6 -4 -2 0 2 4 ( E-€,)lev Fig. 2 Non-magnetic SrCoO,; Co d-states projected along t,, (solid) and eg (dashed) symmetries of Ohcrystal field. Oxygen DOS are shown with dotted lines. I I 1 bonding 41 antibonding -4 -6 -4 -2 0 2 4 (€-€,)/eV Fig.3 Crystal orbital overlap population (COOP) of the Co-0 interaction in non-magnetic SrCoO, feature through an analysis of the crystal orbital overlap population (COOP). Fig. 3 shows the COOP of the interaction of one cobalt atom with one oxygen atom, plotted with the same energy window as Fig.1 and 2 to enable comparisons with the DOS. The COOP are bonding in the range -6 to -1 eV and antibonding above. Bonding and antibonding o(eg) and o*(eg) can be seen to dominate in the energy ranges -6 to -3 eV and above 0.5 eV respectively. The large energy splitting can be compared to the smaller one between n(t,,) and n*(tZg) at -4 eV and EF. It is relevant to discuss here the COOP intensities which show similar features to the above DOS except for the states lying on EF which are smaller than the large DOS peak. This is in agreement with the fact that the major part of this peak corresponds to non-bonding nnb(tZg)cobalt states. The consequences will be further dis- cussed below. Analysis of the NSP-DOS within Stoner theory.In as far as Co 3d and 0 2p states were treated as band states in the framework of our calculations, the Stoner theory of band ferromagnetism can be applied to address the gross features of spin polarization. At zero temperature the product I-n(EF) provides a criterion for the instability of the non-magnetic configuration towards intra-band spin-polarization if I.n(EF)> 1. Here I is the Stoner integral which is an atomic quantity that can be derived from spin-polarized calculations; n(EF) is the density of states at the Fermi level in the non- magnetic state. From our spin-polarized calculations, I(Co 3d) is found to be 1.06 eV and a Stoner product of 7.8.Moreover we find a Stoner product for 0 2p= 1.5. This leads to the suggestion that both Co and 0 will self-polarize, i.e. they will undergo intraband splitting when two spin populations are allowed for. We stress that this mean field analysis is to be considered qualitatively because the Stoner criterion rigorously applies when I and n(E,) refer to quantities for non-hybridized atomic species of different sorts, whereas in our calculations all the interactions are built in, so that our values of the Stoner product are likely to be larger than actual ones. Spin-polarized (SP) calculations Spin-polarized (SP) calculations were then carried out for the SrCoO, system. This was done by initially allowing for two spin occupations, i.e.majority (spin up,?) and minority (spin down, -1) spin directions for all atomic species, then self- consistently converging the charges and the magnetic moments for a sufficiently large number of k points to ensure for convergence. 165 irreducible k points were used here in order to obtain accurate values of the magnetic moments. The present calculations, which account for only one magnetic/ crystallographic sublattice of Co and 0,should describe cor- rectly the proper ferromagnetic order. Magnetic moments. The total variational energy of the magnetic configuration is found to be 0.42eV per unit cell lower than the NSP one. This stabilization, which agrees with the fact that SrCoO, is a ferromagnet, is an expected result. This is because the large n(EF) values, mainly due to Co 3d, in the NSP calculations make the non-magnetic configuration unstable with respect to the onset of intra-band spin polarization.In as far as the overwhole features of charge transfer are as in the NSP calculations, we discuss the magnetic moments. There is a negligible polarization of Sr valence states and, a fortiori, of ES. A moment of 1.74 pB is found on cobalt. It mainly arises from the polarization of Co 3d states. It can be broken down into its crystal field components: M(tzg)= 1.23 pB and M(eg)=0.51 pB.The larger moment carried by t2g (2 x eg) is clearly the consequence of the localization of these states at the Fermi level (cf. NSP calculations). Its magnitude corre- sponds to a low-spin configuration of cobalt.However, this should lead to a negligible moment arising from e,, but the observed magnitude is not small. A strictly low-spin configur- ation is hence ruled out. This point will be made clear in the next section. The moment carried by oxygen is 0.15 pB.With the Stoner criterion analysis in mind, the magnitude and the sign of this moment are rather expected because of the large value of the Stoner criterion found for oxygen, i.e. it should self-polarize and carry its own magnetic moment which is then not induced by that of the transition element. The magnitude of the oxygen moment is yet small; this is due to its interaction with cobalt through eg whose magnetic moment is small as shown above. The resulting magnetization per unit cell is then 2.19 pB which is in reasonable agreement with the value found by Bezdicka et aL2 Density of states.In order to address the effect of intra-band polarization we show in Fig. 4 the site and spin-projected (SP) DOS of SrCoO,. Spin polarization causes Co 3d to split mainly for t2g states which carry the largest magnetic moment. Majority spin (t)states are stabilized in energy with respect to minority spin (J) states. Their electron filling is hence larger. As a result of this splitting, the majority spin states exhibit a sharp peak just below the Fermi level which is an indication of their localization, whereas eg(t) states crossed by EF can be assigned an itinerant character just like in the NSP DOS. Above EF,minority spin states exhibit a sharp peak for empty 4 c 2 -2 -4 I I I Fig.4 Site projected densities of states of magnetic SrCoO,; Sr (dotted), Co (solid), 0 (dashed) and ES (dash-dotted) J. Mater. Chem., 1996, 6(1l), 1785-1788 1787 t,, states. The lowest part below EF (sharp peak in majority spin states) ensures o-bonding with 0 2p. As expected the splitting causes the total DOS to be much lower in intensity at E, than in the NSP calculations. The similar shapes of cobalt and oxygen DOS and their different weights between the two spin directions are indications of covalent bonding and spin polarization for the two species, especially for oxygen. There is, however, a larger mixing between cobalt and oxygen for majority spin, twice as large for o-type bonding than for n-type bonding.We follow here our discussion of SrFe0,4 and that of Bezdicka et al., in stating that this interaction causes a reduced oxidation state of oxygen by a back transfer 0 2p+eg(T) leading to an intermediate valent cobalt. eg(L), being at higher energies, are less affected by the transfer. This explains the relatively large occupation of Co(eg) and the finite moment of eg states. Although a low-spin configuration could be expected from above, the ground magnetic state is an intermediate spin state. This agrees with the recent work of Potze et al. on atomic multiplet calculations of this oxide.'' A relevant aspect of the SP DOS is the minimum observed at EF for majority spin states whereas a relatively large intensity mainly due to Co[t,,(L)] is observed.This should allow the prediction of a nearly half-metallic behaviour for SrCoO, and encourage further investigation of its Fermi surface. Magnetic exchange properties. Finally, we investigated the magnetic exchange properties of SrCoO,. In this context the so-called 'non-collinear ASW' method16 was used to perform total energy calculations of non-collinear spin alignments. Applying the method of Uhl and Siberchicot17 the Heisenberg exchange constants J(R)were determined by the total energies of different non-collinear configurations E(q) as J(R)=cE(d*cos(q.R), 4 where R is a real-space difference vector. Here the value of the magnetic propagation vector 4 is homogeneously spread over the first Brillouin zone.We used 10 different configurations ranging from (O,O,O) (ferromagnetic order) to (3 3 3)(antiferro-magnetic order). As a result, a large value of total magnetic exchange contribution is obtained:Jo=0.146 eV atom-', which should be compared with the value of metallic cobalt {Jo= Jo=0.174 eV atom-l}. This large amount of magnetic exchange energy leads to the expectation of a large value of T,; in disagreement with the experimental value close to room temperature., But since, first, the magnetic moment of cobalt is not localized as in Fe,04 for instance,17 and second, the magnetic moments of oxygen atoms play a crucial role in the magnetism of SrCoO,, we can suggest that an estimation of thermal properties, in particular T,, given in terms of exchange constants is inappropriate here. We thank Professors Jean Etourneau, Head of ICMCB-CNRS and GCrard Demazeau, Head of the High Pressure Research Group of the ICMCB, as well as Dr Peter Mohn of the Technical University of Vienna for critical reading of the manuscript.Computational facilities provided by the computer center of the University Bordeaux 1(CRIBX1) within the MNI project are acknowledged. References 1 H. Watanabe, J.Phys. SOC.Jpn., 1957,12, 515. 2 P. Bezdicka, A. Wattiaux, J. C. Grenier, M. Pouchard and P. Hagenmuller, 2.Anorg. AZZg. Chem., 1993,619,7. 3 H. Oda, Y. Yamaguchi, H. Takei and H. Watanabe, J. Phys. SOC. Jpn., 1977,33,967. 4 S. F. Matar, G. Demazeau, P. Mohn, V. Eyert and S. Najm, Eur. J.Solid State Inorg. Chem., 1994,31,615. 5 G. Demazeau, B. Siberchicot and S. F. Matar, J.Appl. Phys., 1994, 75,4617. 6 S. F. Matar, P. Mohn and G. Demazeau, J. Magn. Muter., 1995, 140-144,169. 7 A. R. Williams, J. Kubler and C. D. Gelatt Jr., Phys. Rev. B, 1979, 19, 6094. 8 U. von Barth and L. Hedin, J. Phys. C, 1972,5, 53. 9 J. F. Janak, Solid State Commun., 1978,25, 53. 10 V. Eyert, unpublished new version of the ASW program, 1994. 11 R. Hoffmann, Angew. Chem., Int. Ed. Engl., 1987,26,846. 12 V. Eyert and S. F. Matar, 1994,unpublished work. 13 S. F. Matar, G. Demazeau, J. Sticht, V. Eyert and J. Kubler, J.Phys. (Paris), 1992, 2, 315. 14 S. F. Matar, V. Eyert, J. Sticht, J. Kubler and G. Demazeau, J. Phys. (Paris), 1994, 4, 1199. 15 R. H. Potze, G. Sawatzky and M. Abbate, Phys. Rev. B, 1995, 51,11501. 16 M. Uhl, L. M. Sansdratskii and J. Kubler, J. Magn. Magn. Muter., 1992,103,314; Phys. Rev. B, 1994,50,291. 17 M. Uhl and B. Siberchicot, J.Phys.: Condens. Matter, 1995,7,4227. Paper 6/02152B; Received 27th March, 1996 1788 J. Muter. Chem., 1996, 6(ll), 1785-1788
ISSN:0959-9428
DOI:10.1039/JM9960601785
出版商:RSC
年代:1996
数据来源: RSC
|
14. |
Synthesis and characterization of inorganic gels in a lyotropic liquid crystal medium. Part 2.—Synthesis of silica gels in lyotropic crystal phases obtained from cationic surfactants |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1789-1794
Thierry Dabadie,
Preview
|
PDF (1479KB)
|
|
摘要:
Synthesis and characterization of inorganic gels in a lyotropic liquid crystal medium Part 2?--Synthesis of silica gels in lyotropic crystal phases obtained from cationic surfactants Thierry Dabadie," Andrii Ayral,*" Christian Guizard," Louis Cot" and Pascale Lacanb "Laboratoire des Matkriaux et Prockdks Membranaires, UMR 5635, CNRS, ENSCM, UMII, 8, rue de I'Ecole Normale, 34053 Montpellier Cedex 1, France bInstitut de Ciencia de Materials de Barcelona, Campus de la UAB, 08193 Bellaterra, Spain Microporous silica materials with tailored porosity are synthesized by the sol-gel process using lyotropic liquid-crystal phases as templates. The starting isotropic sol is obtained by mixing tetramethoxysilane, water and an alkyltrimethylammonium bromide. The polymerization of the silica network and the formation of the amphiphilic mesophase are simultaneous and cooperative.After thermal elimination of the surfactant molecules, an ordered porous texture is maintained in the silica material. The pore size is related to the size of the templating unit and modulated by the length of the alkyl chain of the used surfactant. In this paper, the various steps of the synthesis are studied and the porous texture of the final material is characterized and discussed in terms of templating effects. The sol-gel process is a very attractive method for producing porous ceramic materials and more particularly thin layers for specific applications, e.g. separative membranes. From a start- ing fluid sol, it is possible to produce mesoporous colloidal films' or microporous polymeric films.2 In order to improve the layer properties like separative efficiency, it is very important to control the pore size of the final material and the monodispersity of the pore size distri- bution.Our suggestion is to use a lyotropic liquid-crystal phase as a gelation medium. This mesophase, acting as a template for the growing inorganic network, induces the porous texture of the final inorganic material. Two different strategies of synthesis are possible. The first is the initial structural organization of the gelation medium. This route has been described in a previous work3 dealing with the gelation of a silicon alkoxide in lamellar phases produced from non-ionic surfactants.However, in that case, thin films are difficult to produce due to the high viscosity of the initial sol. The second method studied in this paper consists of a structural organiz- ation of initially isotropic and fluid silica sols during the gelation process using cationic surfactants. Sakata and Kunitake4 have shown that it is possible to produce multi- layered siloxane films from clear solutions of a double-chain ammonium amphiphile and methyl trimethoxysilane. Experimenta1 Synthesis of the materials The molecular precursor of the silica network was tetramethoxysilane (TMOS, Fluka). Different sols were prepared using different alkyltrimethylammonium bromides, C,H2x+1(CH3)3NH4+[x =8,10,12 (Lancaster); 14, 16 (Fluka)]. The various samples exhibited the same mass percentages of surfactant, TMOS and water: 21, 12 and 67% respectively, associated with a hydrolysis ratio of 47.5 for TMOS.A sol without surfactant but with the same molar concentration of TMOS and the same hydrolysis ratio was also prepared in methanol. If we consider the water-surfactant binary diagrams, we can see that the chemical composition of the sols containing the cationic surfactant corresponds to an isotropic regi~n.~ t Part 1, ref. 3. This fact is consistent with the very fluid nature of the starting sol. The samples are labelled C, where x is the number of carbon atoms of the alkyl chain of the surfactant. The wet gels, from C8 to C14, were thermally treated up to 450 "C (heating rate: 0.5 "C min-l) under nitrogen to remove methanol which is a byproduct of the sol-gel reactions, water and surfactant.Characterization techniques The structural evolution of the samples from the starting sols to the thermally treated materials was studied by low-angle X-ray diffraction devices using Cu-Ka radiationo and allowing the determination of Bragg spacings up to 90A. The nature of the birefringent mesophases formed in the wet gel was also checked from polarized optical microscopy observations. Transmission electron microscopy (TEM) observations were performed on replicas of cryofractured wet gels. The sol-gel transition was followed from rheological measurements in steady shear flow and oscillatory flow modes. The experiments were carried out on a Couette viscosimeter with a Money-Ewart cell (exterior and interior diameters, 30 and 27 mm respectively).For the viscoelastic measurements, the amplitude and the frequency were 0.4" and 0.1 Hz respect- ively. The condensation of the silica network in the wet gel was estimated from 29Si NMR experiments at 59.62 MHz in a cross-polarization configuration with magic angle spinning (CP MAS). The thermal evolution of the gels previously dried at 100 "C was studied using a thermogravimetry (TG) apparatus coupled with a Fourier-transform IR spectrometer. The heating rate for the analysis was 0.5"C min-' and the atmosphere was nitrogen or air. The porous texture of thermally treated gels was measured from the nitrogen adsorption-desorption isotherms at 77 K.The micropore size distribution was analysed using two differ- ent calculation modes: M.P.6 and Horvath-Ka~azoe.~ The average hydraulic radius, Thy and the average Horwath-Kawazoe diameter, dHK,were determined from the resulting cumulative pore volume curves. It must be noted that the hydraulic radius deduced from the M.P. analysis can be easily related to the size of pores exhibiting a defined shape: the width of a slit-like pore, wp, is equal to 2rh and the diameter of a cylindrical pore, d,, is equal to 4rh. J. Mater. Chem., 1996, 6(11), 1789-1794 1789 Results and Discussion Study of the gel formation Rheological evolution of the sols. In contrast to the viscous lamellar sols previously obtained using non-ionic s~rfactants,~ the studied sols are initially fluid and the gelation time t, can be estimated macroscopically at about 3 h whatever the surfac- tant.This value must be compared to the longer gelation time observed for the gel synthesized in methanol: about 24 h. From viscosimetric measurements, we can observe the typical evolution of the sol from Newtonian to thixotropic flow behaviour' [Fig. 1 (a)]. The tenuous nature of the gels makes the viscoelastic measurements difficult. Nevertheless, it is poss- ible to see the increase of the storage modulus G and the loss modulus G"near the gelation point [Fig. l(b)]. Moreover, a reproducible decrease of G is observed for the sol C14 before the appearance of the mesophase structure in the gel.Structural evolution of the systems: low-angle X-ray diffraction study. We can see that, for the different surfactants used, the formation of the inorganic network, i.e. the polymerization of the silica, and the formation of the mesophase are simultaneous (Fig. 2). However the mesophases appear at different times after the introduction of the alkoxide in accordance with the used surfactant, sometimes before gelation (Fig. 2). Sol C,, exhibits a mesophase structure already 9 min after the introduc- tion of the alkoxide. In the case of sol C8, the mesophase appears a long time after the macroscopic formation of the gel. The diffractograms of the wet gels present at least a main diffraction peak. Its position corresponds to a large Bragg spacing related to the existence of an ordered amphiphile structure and depends on the nature of the used tenside.The diffractograms can be assigned to lamellar or hexagonal phases. For a lamellar liquid crystal phase, the main diffraction peak, dloo,is related directly to the lattice parameter, i.e. the thickness of the lamella, and a diffraction peak at dlo0/2 is usually ob~erved.~For a hexagonal mesophase, the cell parameter of the hexagonal lattice, a, corresponding to the distance between the centres of two adjacent micellar cylinders, is equal to 2dloO/,/3, where dloo is the position of the main diffraction 0 k 0 20 40 YIs-' collapse of the network under shear 150 -( b) mesophaseappearance 1 Ill 180 210 240 tlmin Fig.1 Rheological measurements on sol C14: (a)steady shear flow; (b) oscillatory flow 1790 J. Mater. Chern., 1996, 6(ll), 1789-1794 I , .I, $2 I6' 4 8 8 !3 4 5 6 .-E toC 0,+c.--2Wegrees Fig. 2 Structural evolution of the various sols by X-ray diffraction: c8; (b)clO; (c) c12; (dl c14 peak, and two peaks of weaker intensity, at d100/,/3 and dlO0/2, are usually observed. In contrast to isotropic organized amphi- philic structures like cubic or sponge mesophases, the lamellar and hexagonal mesophases exhibit an anisotropic structure inducing optical birefringence. The assignment of the diffractograms of the wet gels to lamellar or hexagonal structures is based on the good agree- ment between the measured Bragg spacings and the values observed previously for other materials produced from the same surfactant molecules, i.e.pure liquid-crystal mesophases (without alkoxide)," or on silicates and aluminosilicate mate- rials with similar ordered porous textures obtained by hydro- thermal synthesis." The presence of these two anisotropic phases is also confirmed experimentally by the observation of birefringence features by polarized optical microscopy. Diffractograms of gels C14 [Fig. 2(d)] and C16 show clearly the presence of a lamellar phase with two diffraction peaks at dloo and dloO/2. When the mesophase appears, gels Clo and C12 exhibit a diffractogram corresponding to a lamellar meso- phase [Fig. 2(b) and 2(c)]. After an ageing time of a few days at room temperature, an evolution of the structure of these two gels can be observed.One additional peak appears at smaller angles which can be assigned to a hexagonal phase, although the peaks at dlO0/2 and d100/,/3 cannot be detected. The main mesophases are the lamellar phase for aged gel C12 and the hexagonal phase for aged gel Clo (Fig. 3). Gel c8 exhibits only one diffraction peak assigned to a hexagonal phase. The increase of dlooo with the alkyl chain length of the surfactant (Fig. 4) is ca. 3 A per carbon atom for the lamellar phase which is in agreement with the literature." More surpris- ing is the sharp increase (5 A per carbon atom) for the hexagonal phase. The mechanisms of the silica polymerization and the meso- hex. lam.lam. 0 a I I I I 4 a 2eldegrees Fig. 3 X-Ray diffractogram of aged gel Clo 3.5 I I 3.1 2.7 / / /\ 0 -0’ 2.3 I.9 1.5 6 8 I0 12 14 16 18 X Fig.4 Evolution of d,,, us. x, the number of carbons of the alkyl chain of the surfactant: (+) hexagonal wet gels; (B)lamellar wet gels; (A)hexagonal thermally treated gels; (x) lamellar thermally treated gels Table 1 29Si NMR data on wet gels QJQ3 Q2/Q4 43/44 isotropic wet gel wet gel c8 0.29 0.11 2.03 0.05 7.00 0.48 phase formation are cooperative. This phenomenon of cooper- ative formation of organic/inorganic interfaces was recently studied by Monnier et aZ.12 The formation of a crystalline liquid phase could be explained in terms of evolution of the coulombic type interactions which exist between the polar head of the surfactants molecules and the growing silicate oligomers.Moreover, Monnier et a1.” observed the existence of a lamellar-hexagonal transformation during the formation of the inorganic network. Our own observations for the gel Clo and C12are in good agreement with this result. The final structure of the aged gels depends essentially on the surfactant. With hexadecyltrimethylammoniumbromide (x= 16), Beck et al.” evidenced the effect of the surfactant/silicon molar ratio on the nature of the resulting mesophase. This mesophase changes from hexagonal to cubic to lamellar when this ratio is >1 in agreement with the succession of the mesophases observed on the water-surfactant binary diagrams when the surfactant percentage increases.In our case, the surfactant/ silicon ratio increases from 0.7 for C16to 1.1 for c8 and we observe the reverse evolution. However, two parameters, the alkyl chain length and the chemical composition, are simul- taneously modified. Moreover, for gels C12, we studied the effect of a decrease of the surfactant/silicon ratio from 0.9 to 0.5. This variation induces only a weakening of the hexagonal peak intensity. Study of the silica network formation by 29Si NMR spec-troscopy. The experiments were carried out on wet gels because the gelation time is too short compared to the analysis time (several hours) to consider a study of the sol-gel transition. The Qn/Qn+lratios reported in Table 1 correspond to the ratios of the peaks assigned to the various configuration of Si, n being the number of bridging oxygen atoms of Si04 tetra- hedra.Although the CP MAS NMR method does not allow a strictly quantitative analysis of the Qn/Qn+lratios, these ratios are always higher in the case of the isotropic gel, as previously observed for lamellar silica gels.3 This result is linked to a higher level of polymerization of the silica network in the case of the structured gels. In the present study, it can be explained in terms of polymerization in the confined volume corresponding to the aqueous phase of the mesophase. This results are in agreement with those obtained by Monnier et a1.12 Structural and textural characterization of the wet gels By polarized optical microscopy on wet gel c8 [Fig.5(a)], typical fan-like birefringence features exhibited by hexagonal phases12 can be observed. The TEM micrographs show domains with polygonal forms with a mean size of ca. 40nm [Fig. 5(b)].Using a higher magnification, we can distinguish the internal structure of these polygonal grains [Fig. 5(c)]. In some areas, it seems possible to discern alignments of points and a hexagonal arrangement corresponding to the ordered alignment of the micellar cylinders. It should be noted that the mean distance between two spots is about 2.5 nm in agreement with the structural spacings measured by X-ray diffraction. On gel CI2,the TEM micrograph [Fig. 5(d)] depicts a large- scale texturation of the material with step-like patterns.At higher magnifications, a finer texture is evident. We can observe a stacking of dark and light layers which should correspond to the lamellar structure deduced from X-ray diffraction. Structural and textural study of thermally treated materials Study of the gel to oxide transition by thermogravimetric analysis. The curves of TG under nitrogen for the pure sur- factant tetradecylammonium bromide and for the dried gel C,, are shown in Fig. 6(a) and (b) respectively. It appears that the departure of the surfactant is not influenced by the presence of the silica network. The final mass of silicon oxide corre- sponds to ca. 18 mass% of the starting dried material in good agreement with the initial silicon alkoxide amount.By IR spectroscopy on the emitted gases during the thermogravi- metric measurements under nitrogen, we observed the same absorption bands as those existing on the IR spectrum of the pure surfactant.14 Under air, the IR spectra indicated partial combustion of the surfactant during the thermal treatment.14 For the other surfactants, the mass loss occurs also between 200 and 400°C and the shape of the curve does not depend on the nature of the surfactant. Structure of the thermally treated materials After the thermal treatment under nitrogen at 450 “C,a remark- able result is the retention of the main diffraction peak Fig. 5 Microscopy observations on wet gels: (a) polarized optical microscopy observation of gel c8;(b)TEM image of gel c8, polygonal grains; (c) TEM image of gel c8, detail of a polygonal grain; (d) TEM image of gel C,, J.Muter. Chem., 1996, 6(11), 1789-1794 1791 0 E" + -0.5E Q -1 Fig. 6 TG curves under nitrogen: (a) tetradecyltrimethylammonium bromide (x= 14); (b)dried gel C14 previously observed for the wet gels (Fig. 7). This result confirms clearly that the inorganic network is directly implied in the structure of the mesophase which must so define the organization of the pores around the amorphous silica network. The measured values of dloo after thermal treatment are reported in Fig. 4. The low decrease of dloo from the wet to the thermally treated gel can be associated both to a shrinkage of the material during the drying step and to a slight sintering at low temperature by the condensation of reactive groups of the silica network.15 Moreover a important broadening of the diffraction peak is observed. Assuming a monodisperse distribution for the size of the ordered domains, the order of magnitude of this size was estimated by extension of the Scherrer relation16 to our materials.The calculated value is ca. 20nm. Taking into account the sharpness of the diffraction peaks for the wet gels and the size of the polygonal domains observed on the wet gel c8 [Fig. 5(b)],i.e. ca. 40 nm, we can deduce that the size of the ordered domains is reduced during the transition from the wet to the thermally treated gel. The final size of these domains is small compared to the size of the liquid-crystal cell.Porous texture of the thermally treated materials. The skeletal density of the oxide network, ps,measured by helium pycnome- try is equal to that of amorphous silica, i.e. 2.2 g cm-3.14 Nitrogen adsorption-desorption isotherms for the thermally treated materials are reported in Fig. 8. Whatever the gel, the BET specific surface area, SBET,and the total porosity are very large, >1000 m2 8-l and around 60% respectively (Table 2). The c8 isotherm [Fig. S(a)] corresponds to a type I isotherm associated with a microporous material17 (pore size <2 nm). Further experimental confirmation of the microporous nature of c8 was obtained from dynamic characterizations on thin 5:O 0:2 0.15dlnm Fig.7 Diffractogram of gel c8 after thermal treatment under nitrogen at 450 "C 1792 J. Muter. Chem., 1996,6(11), 1789-1794 400m200 t I Fig. 8 Nitrogen adsorption-desorption isotherms for the thermally treated gels: (a) c8;(b)Clo; (c) C12;(d) C14 Table 2 Textural properties of the thermally treated gels measured by nitrogen adsorption thermally treated gel S,,,/ m2 g-' pore volume/ cm3 g-' porosity P ("/.I microporosity (Yo of P) c8 1260 0.67 60 82 CIO 1040 0.60 57 88 c12 1100 0.74 62 86 c14 1090 0.78 63 88 layers deposited on porous s~pports.~~,~~ For Cl0, a small hysteresis loop appears for a relative pressure PIP, of 0.5 and the area of the loop increases for materials C12 and C14.This loop is indicative of the appearance of larger pores, i.e. mesopores (2-50 nm) for which a capillary condensation phenomenon occurs. Typical shapes of hysteresis loops can be related to given mesopore shapes.17 However, in our case, it seems difficult to associate unambiguously the experimental hysteresis to one of the classified types. As a matter of fact, the micropore volume determined using the 't-plot' method'' includes in our case the contribution of the small mesopores inducing the hysteresis loop for the gels Cl0, C12 and C14. This micropore volume corresponds to the main part of the total porosity (Table2). It represents about 50% of the total volume of the thermally treated material. This value can be compared to the volume fraction of the amphiphilic part in the wet gels, i.e.around 20%, assuming a bulk density of unity for the surfactant assemblies. Taking into account the small decrease of the liquid-crystal parameter after the thermal treatment (Fig. 4), it appears that the densification of the inorganic network is combined with a increase of the size of pores associated with domains initially defined by the templating units, i.e. the surfactant bilayers or the micellar cylinders. The size of the micropores was experimentally determined by the M.P.6 and Horvath-Kawazoe7 methods (Fig. 9). The average hydraulic diameter, 2rh, and the average Horvath- Kawazoe diameter, dHK,are roughly proportional to x, the number of carbons of the alkyl chain of the surfactant and have the same order of magnitude than the length of the surfactant molecules, l,, calculated using the Tanford relation" (Appendix).The pore size can also be estimated from calculations on the lamellar or hexagonal mesophase structure using the measured structural parameter d,,, and the pore volume fraction in the mesophase structure, $p, which corresponds to the microporous volume determined by the 't-plot' method. From the experimental data reported in Table2, it can be easily calculated that (bP is ca. 0.55 for gels c8 and cloand ca. 0.58 and 0.60 for gels C12 and C14. The diameter of the cylindrical pore, d,, for the hexagonal gels (28 and Clo and the width of the slit-like pore, wp, for the lamellar gels C12 and 3 -1 1 OS5 t 6 7 8 9 I0 11 12 13 14 15 16 X Fig.9 Measured and calculated average size of the pores vs.x, the number of carbons of the alkyl chain of the surfactant: (a)2rh, where rh is the measured average hydraulic radius; (b)4r,; (c) dHK, measured average Horvath-Kawazoe diameter; (d) I,, calculated length of the surfactant molecule; (e) d,, calculated diameter of the cylindrical pores of a hexagonal structure; (f)wp, calculated width of the slit-like pores of a lamellar structure CI4 are reported in Table 3 and in Fig. 9. The details of the calculation methods are given in the Appendix. These calcu- lated values are in good agreement with the average pore sizes deduced from the M.P. method, 4rh and 2Yh, respectively.However, the discontinuous evolution of the pore size associ- ated with a change in the nature of the mesophase structure between C,, and C,, is not observed for the average Horvath- Kawazoe diameter, dHK. The specific surface area, SM,developed by the mesophase structures, can also be calculated from d,,, and 9, (Appendix). The calculated values are also reported in Table 3. We note that SMis lower than the measured value, SBET.In order to explain this difference, we can consider the fact that the material is polycrystalline (composed of structurally ordered grains). The excess specific surface area, AS, can be assigned to the existence of an interface between these ordered domains. This can be summed up by eqn. (1) In a first approximation, we can suppose that the ordered domains are spherical with a diameter D.Then we can write eqn. (2) where pMis the density of the mesophase structure (Appendix). The calculated values of D are reported in Table 3. For the hexagonal material they have the same order of magnitude as the sizes determined from the width of the diffraction peaks using the Scherrer relation. For the lamellar materials, the mean calculated values are larger but in that case D depends strongly on the dlooand 4, values. Mechanisms The experiments show that it is possible to freeze a mesophase in the final gel from a starting isotropic sol. In order to understand the mechanisms of the formation of ordered wet gels, we can consider the models proposed to explain the synthesis of zeolites which are also ordered microporous materials.Among the models which explain the formation of zeolites, the 'can and cement' model proposed by Brunner" seems to be the most apt to describe the structure of the wet gel. Nevertheless, we should not consider the structure of the solid in terms of individual molecules [as for instance tetrapro- pylammonium, (C3H7)4NH4+] but as surfactant bilayers or micellar cylinders which are aggregates of amphiphilic mol- ecules. The main phenomena which could induce mesophase formation in the silica sol are the formation of silicate oligomers by hydrolysis and condensation of the silicon alkoxide and a preferential polymerization of the silicate species near the cationic surfactant head groups with substitution of the mobile counterions, i.e.Br-. This substitution must involve a change in the aggregation parameter of the surfactant which controls the structure of amphiphilic aggregates." During the formation of the inorganic network, both the available volume per amphiphilic molecule and the surface available per surfactant head are modified. The optimal value of the head-group area which determines the nature of the mesophase depends strongly on electrostatic and steric interactions between silica species and the surfactant.', It should be noted that this approach is confirmed by the analogy between the obtained mesophases and those existing in water-surfactant binary diagrams. We observed experimentally that the cooperative formation of the mesophase occurs on a different timescale to that of the gelation of the silica network.As a matter of fact, we should note that the gelation point corresponds to the percolation threshold of the silica network but does not mark in any case the end of the polymerization process. Moreover, it was demonstrated previously that well defined smectic clays could be transformed into hexagonal porous materials using ammonium-type cationic surf act ant^.^^ Conclusion In this paper, we have described the synthesis of new micropo- rous silica materials with a tailored porosity by the sol-gel process. Ordered wet gels are produced at room temperature from hydrolysis and condensation of silicon alkoxide in the presence of ammonium-type cationic surfactants.The starting Table 3 Textural properties of the thermally treated gels calculated assuming given mesophase structures (bold numbers correspond to mean values) 2.1 2.4 1.9 1177 ~~ 83 73 C8 0.55 2.3 2.7 2.1 1074 1260 186 33 2.5 2.9 2.2 988 272 22 2.6 3.0 2.3 950 90 68 ClO 0.55 2.8 3.2 2.5 883 1040 157 39 3.O 3.5 2.7 824 216 28 1.8 1.0" 1203 <O - C12 0.58 2.0 1.2" 1082 1100 18 366 2.2 1.3" 984 116 56 2.0 1.2" 1136 <O - CI4 0.60 2.2 1.3" 1033 1090 57 120 2.4 1.4" 947 133 48 a w,/nm. J. Mater. Chern., 1996, 6(11), 1789-1794 1793 sols are initially isotropic and fluid. The cooperative formation of the silica network and of the liquid-crystal mesophases is observed and can be explained in terms of interactions between the growing silicate oligomers and the polar head of the surfactants which modify the aggregation parameter of the amphiphilic molecules.A remarkable result is the retention, after the elimination of the surfactant, of a print of the organized mesophases inside the thermally treated inorganic gels. The obtained silica network is amorphous but the material presents an ordered microporosity. The micropore size can be modulated by the length of the alkyl chain of the surfactant used. The different experimental values which characterize the porous texture of the final material can be correlated to calculated values assuming a polycrystalline material com-posed of hexagonal or lamellar domains. We have to underline here the advantages of the sol-gel technology which enables us, not only to synthesize these materials at low temperature, but also to produce homo- geneous and continuous microporous thin film~.'~*~* These materials are very attractive candidates for applications in separative processes which require microporous membranes, such as nanofiltration, pervaporation or gas separation.Studies are now required for a better understanding of the mechanisms of formation of these new materials. They concern more particularly the processes of nucleation and growth in order to increase the size of the ordered domains and also the means of preferential orientation of the ordered domains to improve the permeability of the final oxide materials.Moreover, the increase of the size of the ordered domains will enable us to confirm our assignments of mesophase structures and also the validity of our porous texture modelling. T. D. thanks Kodak European Research, Centre de Recherche et Technologies, Chalon-sur-Sahe, France for financial sup- port and for low-angle X-ray diffraction facilities. P. L. received her financial support from EEC as a Human Capital and Mobility fellowship. She thanks Pr. C. Miravitlles for welcom- ing her to ICMAB and David Bellido (Serveis Cintifico- Tecnics, Universitat de Barcelona) for his technical assistance in the cryofracture and replica processes and for helpful discussions. Appendix Length of the alkyltrimethylammonium bromide molecule I, is given by the Tanford relation:20 Is(nm)=0.1265x+O.15 (All where x is the number of carbons of the alkyl chain of the surfac tan t.Hexagonal structure Relation between the Bragg spacing, dloo, and the cell param- eter of the hexagonal lattice, a: Relation between a and the diameter of the cylindrical pores, d, (from Luzzati et al."): (A3 where 4p is the volume fraction of the pores in the hexagonal structure. Expression for the specific surface area, SM: where ps is the skeletal density of the oxide network. 1794 J. Mater. Chem., 1996, 6(11), 1789-1794 Expression for the apparent bulk density, pM: PM =(1-4P)PS (A51 Lamellar structure Relation between the Bragg spacing, dlO0,and the width of the slit-like pores, wp: wp =dlOO4p (A6) where is the volume fraction of the pores in the lamellar structure. Expression for the specific surface area, SM: SM =2/CPsd100(1-4p)l (A7 where ps is the skeletal density of the oxide network.Expression for the apparent bulk density, pM: PM =( -4p)Ps (A81 References 1 A. Larbot, A. Julbe, C. Guizard and L. Cot, J. Membr. Sci., 1989, 44,289. 2 C. Guizard, A. Julbe, A. Ayral, J. Ramsay and A. Larbot, in Advances in Science and Technology 3: Ceramics-charting the future, vol. D, ed. P. Vincenzini, Techna, Faenza, Italy, 1995. 3 T. Dabadie, A. Ayral, C. Guizard, L. Cot, C. Lurin, W. Nie and D. Rioult, J. Sol-GelSci. Technol., 1995,4, 1.4 K. Sakata and T. Kunitake, J. Chem. Soc., Chem. Commun., 1990, 504. 5 T. Warnheim and A. Jonsson, J. Colloid Interface Sci., 1988, 125, 627. 6 R. H. Mikhail, S. Brunauer and E. E. Bodor, J. Colloid InterJ Sci., 1968,26,45. 7 G. Horvath and K. Kawazoe, J. Chem. Eng. Jpn., 1983,16,470. 8 M. D. Sacks and R. S. Sheu, J. Non-Cryst. Solids, 1987,92,383. 9 J. F. Sadoc and J. Charvolin, J. Phys., 1986,47,683. 10 V. Luzzati, H. Mustacchi, A. Skoulios and F. Husson, Acta Crystallogr., 1960, 13,660. 11 J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge, K. D. Schmitt, C. T-W. Chu, D. H. Olson, E. W. Sheppard, S. B. McCullen, J. B. Higgins and J. L. Schlenker, J. Am. Chem. SOC.,1992, 114, 10834. 12 A. Monnier, F. Schuth, Q. Huo, D. Kumar, D. Margolese, R. S. Maxwell, G. D. Stucky, M. Krishnamurty, P. Petroff, A. Firouzy, M. Janicke and B. F. Chmelka, Science, 1993, 261, 1299. 13 F. B. Roosevear, J. Am. Oil. Chem. SOC., 1954,31,628. 14 T. Dabadie, Thesis, University of Montpellier 11, France, 1994. 15 C. J. Brinker and G. W. Scherrer, in Sol-Gel Science, The Physics and Chemistry of Sol-Gel Processing, Academic Press, San Diego, 1990. 16 P. Scherrer, Nachr. Gottinger Gesel. Dtsch., 1918,2,98. 17 IUPAC report: Reporting Physisorption Data for GaslSolid Systems, Pure Appl. Chem., 1985,57,603. 18 T. Dabadie, A. Ayral, C. Guizard, L. Cot, J. C. Robert and 0. Poncelet, in Inorganic Membranes ICIM,'94, Proc. ICIM3, Worcester (USA), July 10-14, 1994, ed. Y. H. Ma, Worcester Polytechnic Institute, Worcester, USA, 1995, p. 41 1. 19 B. C. Lippens, B. G. Linsens and J. H. De Boer, J. Catal., 1964, 3, 32. 20 C. Tanford, in The Hydrophobic Eflect, 2nd edn., Wiley-Interscience, New York, 1980. 21 G. 0.Brunner, Zeolites, 1992, 12,428. 22 D. J. Mitchell and B. W. Ninham, J. Chem. SOC., Faraday Trans. 2, 1981,77, 601. 23 T. Yanagisawa, T. Shimizu, K. Kuroda and C. Kato, Bull. Chem. SOC.Jpn., 1990,63,988. Paper 6103437C; Received 16th May, 1996
ISSN:0959-9428
DOI:10.1039/JM9960601789
出版商:RSC
年代:1996
数据来源: RSC
|
15. |
Electrochemical synthesis and thermal decomposition of zirconia gels containing various metal ions |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1795-1797
Kenichiro Nakajima,
Preview
|
PDF (395KB)
|
|
摘要:
Electrochemical synthesis and thermal decomposition of zirconia gels containing various metal ions Kenichiro Nakajima, Shiro Shimada and Michio Inagaki Department of Applied Chemistry, Faculty of Engineering, Hokkaido University, Nishi-8, Kita-13, Kita-ku, Sapporo, 060 Japan Zirconia gels containing various metal ions have been obtained by cathodic electrolysis at -1.8 V (us. Ag/AgCl) of an aqueous solution of zirconyl chloride containing 10 mol% Ca, Mg, Y or Ce chlorides. The electrochemically produced gels were deposited on a Pt cathode as a viscous gel or solid. The thermal decomposition of the freeze-dried gels was followed by simultaneous thermogravimetryydifferential thermal analysis-mass spectroanalysis (TG-DTA-MS) and X-ray diffraction (XRD). It was found that the gels decompose in three steps with evolution of water and hydrogen chloride.The gels were amorphous at room temperature, but changed to cubic or tetragonal phases at 600 "C. Transformation to the monoclinic or mixed tetragonal and monoclinic phases occurred at 1000 "C in all materials except the Ce-containing samples which formed only a tetragonal phase. An electrolytic process for low-temperature synthesis of various ceramic powders and films has been introduced by Switzer.' Since then, various dielectric powders and films, such as BaTi03,2-4 have been prepared by anodic electrolysis of aque- ous solutions of metal salts under hydrothermal conditions or at temperatures below 100 "C. A cathodic process involving electrochemical reduction of water and subsequent deposition of hydroxides has also been utilized by several for producing oxide films, such as zirconia, PZT and alumina/ zirconia.The electrochemical technique using cathodic electrolysis of an aqueous solution may provide an interesting method for preparing various oxide powders, in addition to preparation of films with the advantage that the precipitation is controlled by the electrode potential or current density. In our previous paper,8 we reported that zirconia gels were produced by electrochemical reduction of zirconyl nitrate and chloride in their aqueous solutions, and were converted to tetragonal zirconia by heat treatment at 600°C. It is known that oxides such as CaO, MgO, Y203, etc. can stabilize the tetragonal or cubic phases of zirconia by the formation of solid solutions. Although there are many reports of the preparation of zirconia powders containing various oxides by co-precipitation and sol-gel technique^,^ electrochemical synthesis of zirconia gels containing metal cations from a zirconyl solution has not yet been reported.This paper describes the synthesis of zirconia gels containing various metal cations by cathodic process and the thermal decomposition behaviour of the zirconia gels formed. Experimenta1 The working and counter electrodes were platinum plates (10 x 10 x 0.5 mm3), and were connected to a potentiostat (HOKUTO HAB-151) through a platinum lead. The reference electrode was Ag/AgCl, connected to the cell through a salt bridge saturated with KCl solution.The salt bridge was separated from the cell by a glass-platinum separator" in which the wire was not completely enclosed by the glass, making an opening through which ionic migration was main- tained; the CV curve obtained by this method was the same as that using a Luggin capillary. Water was circulated to keep the cell temperature at 25 "C. The electrolyte was a 0.1 mol dmW3 zirconyl chloride solution containing 10 mol% of Ca2+, Mg2+,Y3+ or Ce3+ chloride. The pH values of the electrolytes were cu. 2.0. Electrolytic reductions of the chloride solutions were carried out without stirring at a potential of -1.8 V (us. Ag/AgCl) over 24 h. The gels, which were deposited on the platinum plate and scratched from it, were washed with distilled water several times to remove any remaining chlorides and potassium ion and were then freeze-dried below -30 "C.The concentration of metal ions in the freeze-dried gels was determined by energy dispersive X-ray analysis (EDXA, JEOL JED-2000). Simultaneous thermogravimetry-differential thermal analy-sis-mass spectroanalysis (Mac science TG-DTA-2000 with VG gas analysis systems) was carried out on the freeze-dried samples at a heating rate of 10 K min-' in an argon flow to determine the thermal decomposition behaviour of the gels. The phases formed after thermal analysis up to 1000°C were identified by X-ray diffraction (XRD, Rigaku RINT-2000). Results and Discussion Electrochemical synthesis of zirconia gels Fig.1 shows the current density change of the electrolysis at -1.8 V. When electrolysis started, the current density decreased sharply for several minutes, with vigorous gas evol- ution, and then gradually decreased with time. During this gradual current density decrease, the gel was observed to be deposited on the working electrode. After it had stabilized, the current density was maintained at a constant value of 2-3 mA cm-2 except for the Mg-containing solution, in which case the current density was greater than the others. During electrolysis at these constant densities, the gels continued to be deposited on the electrode. Transparent gels were deposited on the electrode, after electrolysis of pure zirconyl chloride solution without addition of metal chlorides for 24 h.Bubbles were seen to be included in the gels. During electrolysis the gels containing such bubbles were observed to partly detach from the electrode surface, with some portions settling on the bottom of the cell under the action of gravity and the gas evolution from the electrode surface. The solutions containing Ca, Y and Ce produced bubble-containing gels with similar appearances. The pure, Ca- and Y-containing gels were white and those containing Ce were yellow. Electrolysis of the solution containing Mg did not produce gels but a white solid with cracks on the working electrode. J. Muter. Chem., 1996,6( ll), 1795-1797 1795 Ob-I I 1 I 0 5 10 15 20 t/h Fig.1 Change of current density with time during potentiostatic electrolysis at -1.8 V: a, pure; b, Mg; c, Ca; d, Y; and e, Ce It has been reported that zirconia gel formation by cathodic electrolysis occurs as f01lows:~ (i) dissolution of zirconyl chloride in water, ZrOC1, -+Zr02 + 2CI-(1)+ (ii) electrolysis of water, 2H20+2e- +H2+20H- (2) (iii) formation of zirconium hydroxide, following reactions (1) and (a, Zr02+ +H20+Zr(OH)22+ (3) Zr(OH)22+ +20H- +Zr(OH), (4)(iv) formation of hydrous zirconia from zirconium hydroxide with an indefinite amount of hydration water: Zr(OH), + nH20-+Zr02-(n+ 2)H20 (5) It is thought that the hydration of zirconyl ions also occurs to produce hydrated oxide, ZrO(OH), or Zr02.H20." (v) In a similar way to the formation of zirconium hydroxide [steps (i)-(iii)], metal hydroxides were precipitated by the reaction of electrochemically generated hydroxide ion and metal ions, M"' +nOH- +M(OH), (8) The gas evolved during electrolysis must be H2 [reaction (2)].The formation of an insulating hydrous zirconia layer on the surface of electrode is expected to occur immediately after commencement of electrolysis, which is why the current density decreased rapidly. The constant current density after the current density decrease probably results from the repeated separation of the gels in the pure, Ca-, Y-and Ce-containing samples from the electrode. The electrolysis of the solution containing Mg continued in spite of formation of the solid on 1796 J.Muter. Chem., 1996, 6(11), 1795-1797 the electrode, probably because the layer on the electrode cracked, leading to facile migration of solution to the electrode. Thermal decomposition and phase transformation The gels were collected from the cell and electrode and freeze- dried to powders. The amount of metal ions in the powders was determined to be 4-10 atom% according to EDXA (Table 1). Chlorine was also found to be present in the samples. The XRD patterns of all the freeze-dried samples showed a small hump at 28=30" (Cu-Ka), indicating that they were amorphous. The TG-DTA-MS results for the pure zirconia gels are shown in Fig. 2. The mass change occurs in three steps in the temperature ranges 25-270, 320-350 and 350-520 "C, and the total mass loss reaches ca.70 mass%. The first loss (12%) proceeds relatively slowly, the second loss (56%) is very rapid, and the final loss of 2% again is slow. A large endo- thermic peak occurs at 20-270°C in the DTA curve, accompanying the mass loss in the first step followed by a small exothermic peak at around 350°C. The MS curve indicates the evolution of H20 occurred in the first step. In the second and final steps, an overlapping evolution of H20 and HC1 was observed. It is considered that the exothermic peak at ca. 350 "C results from the crystallization of amorphous zirconia to cubic zirconia, as confirmed by XRD measurement of the samples heated to 320 and 400°C. Fig. 3 shows the TG-DTA results for freeze-dried zirconia samples containing Ca, Mg, Y or Ce ions (Fig.3a-d). The mass losses of all the samples occur in the same three steps as the pure zirconia gels, except the Ce-containing samples. The DTA curves show a large endothermic peak around 150°C; The MS curves indicated a sequence of H20 and HC1 loss similar to that of the pure sample. The Ce-containing sample showed four mass loss steps (curve d). The DTA curve (d') showed a large endothermic peak around 150°C in addition to a small exothermic peak around 400°C; the cause of the exothermic peak cannot be explained at present. XRD con- firmed the formation of the cubic or tetragonal phases of Table 1 Metal contents of freeze-dried samples sample contents (atom%) Ca 612 Mg 4+2 Y 10+4 Ce 6+1 TI'C Fig.2 TG-DTA-MS curves of the pure, freeze-dried sample: (a) TG-DTA and (b)MS 200 400 600 800 1000 Tl"C Fig.3 TG-DTA curves of freeze-dried samples containing various metal ions: a, Ca; b, Mg; c, Y; and d, Ce. Curves a-d, TG; curves a'-d, DTA zirconia in all samples heated to 600 "C; all the samples heated to 400°C were shown to be amorphous by XRD. Thus, the three steps in the thermal decomposition of pure, Ca-, Mg- and Y-containing samples are (i) dehydration around 150"C, (ii) and (iii) the release of H20 and HCl around 330 and 400 "C, respectively. In the Ce-containing samples, the decomposition occurs in a similar way to that of the pure sample, but with an additional step between the first and second.Fig. 4 shows the XRD patterns of the samples heated to 1000°C. All the peaks were identified as monoclinic and/or tetragonal zirconia. For the pure sample, only the monoclinic phase is observed, whereas a mixture of the monoclinic and tetragonal phases is seen in the Ca-, Mg- and Y-containing samples (b-d); the tetragonal phase was the major one in the Y-containing samples (d). Only the tetragonal phase is seen in the Ce-containing sample (e). The mass fraction of tetragonal phase estimated from the integrated intensity of monoclinic (111) and tetragonal (111) reflections was 30% for the Ca-, 21% for the Mg- and 96% for the Y-containing samples.12 The lattice constants of the tetragonal phase measured in the metal-containing samples were smaller than those of the zir- conia solid solutions with the metal oxide, suggesting that solid solutions are not formed in the samples heated to 1000 "C.The crystallite sizes calculated from the tetragonal zirconia (111) peak using Scherrer's equation were in the range 20-30 nm. Garvie suggested that there is a critical crystallite size, ca. 30 nm, above which the metastable tetragonal phase cannot exist at room temperature.13 The crystallite sizes of 20 to 30 nm were sufficient to preserve the tetragonal zirconia at room temperature. Conclusion Zirconia gels containing various ions such as Ca2+, Mg2+, Y3+ and Ce3+ were continuously obtained by cathodic elec- I I I I I e CO 0 II A0 lo aH 25 30 35 40 45 50 2Bldegrees(Cu-Ka) Fig.4 XRD patterns of the samples heated up to 1000°C:a, pure; b, Ca; c, Mg; d, Y; and e, Ce. a, Monoclinic zirconia; 0, tetragonal zirconia trolysis of a zirconyl chloride solution containing the respective metal chloride. The gels were transparent except the Mg- containing samples which formed a white solid. The thermal decomposition of the gels was followed up to 1000°C and showed that they gave off water and hydrogen chloride, and changed to cubic or tetragonal phases at 600 "C, further transforming to monoclinic or mixed tetragonal and monoclinic phases at 1000 "C. The Ce-containing samples maintained a tetragonal phase even at 1000 "C. References 1 J. A. Switzer, Am. Ceram. SOC. Bull., 1987,66, 1521. 2 M. Yoshimura, S-E.Yoo, M. Hayashi and N. Ishizawa, J. Appl. Phys. Jpn., 1989,28, L2007. 3 P. Bendale, S. Venigalla, J. R. Ambrose, E. D. Verink Jr. and J. H. Adair, J. Am. Ceram. SOC., 1993,76,2619. 4 K. Kajiyoshi, K. Tomono, Y. Hamaji, T. Kasanami and M. Yoshimura, J. Am. Ceram. SOC.,1995,78, 1521. 5 L. Gal-Or, I. Silberman and R. Chaim, J. Electrochem. SOC., 1991, 138,1939. 6 Y.Matsumoto, H. Adachi and J. Hombo, J. Am. Ceram. SOC., 1993, 76,769. 7 R. Chaim, G. Stark and L. Gal-Or, J. Muter. Sci., 1994,29,6241. 8 K. Nakajima, S. Shimada and M. Inagaki, J. Ceram. SOC. Jpn., 1995,103,309. 9 S. Somiya, in Zirconia Ceramics I, ed. S. Somiya, Uchida Roukakuho, 1983, p. 1. 10 A. Fujisawa, M. Aizawa and T. Inoue, in Denki Kagaku Sokuteiho, Gihodo, Tokyo, 1984, p. 98. 11 M. Pourbaix, in Atlas of Electrochemical Equilibria in Aqueous Solution, Pergamon, London, 1966, p. 223. 12 H. K. Schmid, J. Am. Ceram. SOC., 1987,70,367. 13 R. C. Garvie, J. Phys. Chem., 1965,69,1238. Paper 6/051661; Received 23rd July, 1996 J. Muter. Chem., 1996, 6( ll), 1795-1797 1797
ISSN:0959-9428
DOI:10.1039/JM9960601795
出版商:RSC
年代:1996
数据来源: RSC
|
16. |
Synthesis and characterization of a shock-synthesized cubic B–C–N solid solution of composition BC2.5N |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1799-1803
Tamikuni Komatsu,
Preview
|
PDF (1069KB)
|
|
摘要:
Synthesis and characterization of a shock-synthesized cubic B-C-N solid solution of composition B&N Tamikuni Komatsu,*' Masayuki Nomura: Youzou Kakudateb and Shuzou Fujiwarab "Analytical Research Center, Asahi Chemical Industry Co. Ltd., 2-1 Samejima, Fuji, Sizuoka 416, Japan bNational Institute of Materials and Chemical Research, Tsukuba, Ibaraki 305, Japan A cubic BC,.,N solid solution (c-BC,.,N) was synthesized in 18% yield by shock-compression of a hexagonal BC2.5N compound. c-BC2.,N was confirmed to have a diamond structure without a long range order of atomic arrangement. The material is a polycrystal composed of microcrystals of 5-20 nm in size. All constituent atoms are teterahedrally coordinated, giving C-C, C-By C-N, and B-N bonds, and are homogeneously distributed on the lattice planes rather than showing a full short range order of atomic arrangement.Diamond and cubic BN (c-BN) are the hardest materials known; they are indispensable for polishing and cutting tools and have recently been noted for their potential as high- temperature semiconductors. Diamond is the hardest of all materials, but it is exceptionally weak for steel cutting and is burned to carbon dioxide at 800-900°C in air, whereas c-BN is strong for steel cutting and is inert up to 1600°C in the air but it is only half the Vickers hardness of diamond. It was hoped that a solid solution of diamond and c-BN would have both the hardness of diamond and the inertness of c-BN and that it would also be a useful electronic material.Wedlake and Penny prepared cubic BCN by a static high- pressure and high-temperature (HP/HT) treatment ( 15 GPa, >3OOO"C) of graphitic BCN, and they asserted that the prepared material has long-range ordering of the B-N pair aligning parallel to the ( 110) direction.' After their invention, Badzian prepared cubic BN-C mixed crystals by the same HP/HT technique (14 GPa, 3600 "C) and found that the mate- rial exhibited weak (200) and (420) reflections in the XRD pattern.' He believed the material to have a diamond-like structure and suggested the possibility of B-N pairs in the sublattices. Sasaki et al. attempted the phase transformation of graphitic BC,N into a cubic phase under relatively mild HP/HT conditions (5.5 GPa, 1400-1600 "C) using a Co cata- lyst, but they unexpectedly obtained the phase-separated cubic phases, i.e.a mixture of diamond and c-BN.~ Anticipating the catalytic promotion of the phase separation in the B-C-N system, Nakano et al. attempted the direct transformation without additives under higher HP/HT conditions (7.7 GPa, 2000-2400°C) and found ternary phases composed of dia- mond, c-BN and a cubic B-C-N compound to exist as an intermediate state before perfect separation into diamond and c-BN.~The result clearly means that the B-C-N system under the static HP/HT conditions is thermodynamically more likely to phase separate than to remain in the single phase, and the heating and cooling cause such phase separation. Knittle et al.recently prepared a cubic C,(BN)' -x (x=0.3-0.6) solid solu- tion by laser heating at static high pressure (30-50 GPa, 2000-2500K) and found the material to be similar to c-BN rather than to diamond in their bulk Young's modulus and bonding properties., Considering the solubility in the B-C-N system, we predicted that a cubic B-C-N solid solution would be a non-equilibrium material and therefore rapid heating- quenching on the hypothetical phase diagram (the phase diagram of the cubic B-C-N phase is presumed to be between those of diamond and c-BN) would be most suitable for synthesis. Based on this idea, we obtained cubic (BN)xCz(l-x) (x=0.42-0.50) by means of the hexagonal-to-cubic transform- ation induced by shock compression. We established that the material is a single-phase solid solution having a diamond-like structure without long-range ordering of the atomic arrange- ment, which possesses lattice dimensions larger than those predicted from Vegard's law.6 The material we obtained is similar to, but not the same as, the above-mentioned cubic B-C-N material: (1)the materials prepared by Wedlake et al.and Badzian exhibited very weak (200) and (420) reflections, whereas our material exhibits a very weak (200) reflection [the intensity relative to the (111)is 0.50/,] but no (420) reflection within an error of 0.1% of the (111) intensity; (2) cubic C0.33(BN)0,66prepared by Knittle et al. has a bulk Young's modulus of 355 f19 GPa, which is much smaller than the predicted value for an ideal solid solution of diamond and c- BN, whereas that of our c-BC,N is 401 f15 GPa is close to the predicted value.These subtle but essential differences are related not only to the chemical composition and bonding properties but also to the atomic distribution and arrangement on the lattice planes. Here we elucidate the crystal structure and the chemical structure of the cubic (BN)xCz(l-x) solid solution by X-ray diffraction, nuclear magnetic resonance and analytical electron , microscopy. Experimental Sample preparation A hexagonal BC2.,N compound (h-BC,.,N), the starting mate- rial, was obtained by chemical vapour deposition from a heated mixture of CH3CN and BC1, according to ref. 7. The h-B&N powder was mixed with small copper balls in a h- BC2.,N :copper =6 :94 mass ratio, placed into a steel capsule, and pressed to form a disk 5 mm in thickness and 30 mm in diameter.The bulk density of the disk was set at 70% of the theoretical value. The disks were shock compressed using the cylindrical shock-compression apparatus described in Fig. 1. The apparatus was constructed with an inner copper capsule of a cylinder filled with the disks and an outer steel flyer surrounded by an explosive on the outside. The explosive used was a slurry explosive composed of cyclo( tetramethylene tetra- nitramine) and aq. NaC104. A shock wave was generated by impacting the explosively accelerated steel flyer onto the capsule. The incident shock pressure on the sample was estimated to be 35 GPa.The recovered sample was machined, immersed in aqua regia to remove the copper matrix, heated in conc. HC104 to remove unchanged h-BC2.,N, treated with molten Na,C03 to remove trace amounts of metallic contami- J. Muter. Chem., 1996,6(11), 1799-1803 1799 Fig. 1 Section of a cylindrical shock-compression apparatus: 1, disk sample; 2, cylindrical inner capsule; 3, flyer; 4, detonator; 5, explosive sheet; 6, explosive nants, washed with distilled water, and dried under vacuum at 200°C. The yield of the obtained material was ca. 18%. To characterize the obtained material, the reference standards, diamond (Du Pont, Ltd., shock-synthesized diamond 1/8 U QG) and c-BN [Showa Denco Co., Ltd., c-BN SBN-M no.325/400 (B46)], were used. Elemental analysis The chemical compositions of the materials were determined by the following quantitative analysis: (1) for the C, H and N contents, the sample (0.1 g) was mixed with Pb304 (0.02 g), placed into an Sn boat, heated at 1000°C in a stream of oxygen, and the evolved CO,, H,O and NO, gases were determined by gas chromatography; (2) for the boron content, the sample (0.1 g) was mixed with K2C03(1 g) and Na,C03 (1g), placed into a Pt crucible, completely melted using a burner, cooled to room temperature, dissolved in conc. HCl, diluted with ultrapure water, and the B(OH), present was determined by inductively coupled plasma mass analysis; (3) trace amounts of metallic contaminants were determined by an electron probe X-ray microanalyser.The analysis was carried out at least three times and the results were averaged. Characterization The structures of the materials were investigated using a Philips PW-1800 X-ray powder diffractometer (XRD) equipped with a position-sensitive proportional counter and a graphite monochromator on the detector, a Bruker MSL-400P solid-state NMR spectrometer equipped with 13C NMR and "B NMR probes, and a JEOL JEM-4000FX high-resolution trans- mission electron microscope (HRTEM) equipped with an electron energy-loss spectroscope (EELS) and an electron diffractometer (ED). The XRD measurement was conducted using Ni-filtered Cu-Ka radiation through a pinhole collimator as the incident X-ray source.The data were corrected using an Si internal standard and the lattice constant was obtained by least-squares refinements of the data. The NMR studies were performed in a magnetic field of 9.4T. The 13C NMR spectra were measured under the conditions of 100.614 MHz resonance frequency, 4500 Hz magic angle spinning (MAS) using a 7 mm MAS probe, 83 kHz spectral width, 90" pulse with a 4.5 ps pulse-width, 12.288 ms acquisition time and 10 ps delay time, 30 s pulse repetition time and 8 K data points. The observed chemical shift was corrected on the basis of the carbonyl peak (6 176.46) of the external standard, glycine. The 1800 J. Mater. Chem., 1996, 6(11), 1799-1803 "B NMR spectra were measured under the conditions of 128.33 MHz resonance frequency, 6000 Hz MAS using a 4 mm MAS probe, 26 kHz spectral width, pulse <90" with a 1.0 ps pulse width, 38.912 ms acquisition time, 30 ps delay time, 4 s pulse repetition time and 8 K data points.The observed chemical shift was calibrated using the peak at 6 2.0 associated with the tetrahedral BO, unit in the external standard borax [Na2B4O5(OH),43H,O]. The TEM samples were prepared as follows: the sample powders were diffused supersonically in a methanol-distilled water (1 : 1) solution for 5 min; a drop of the solution was then placed on a microgrid coated with a carbon-collodion membrane and air-dried. The TEM obser- vation was carried out as nearly as possible on this thin region at 400 keV accelerating voltage. The EELS spectra and ED patterns were obtained from selected areas of the observed TEM image. The EELS spectra were measured with a reso- lution of 1 eV using a Gatan model 666 parallel spectrometer.Results and Discussion Chemical analysis The elemental analysis of the starting material gave the follow- ing atomic percentages: B :C :N = 18.9: 51.8 :25.0, with ca. 0.1% hydrogen and several tens of ppm of Sn, Si, Al, Fe, Cr and Ni. The atomic ratio was thus approximated as B : C :N = 1.0:2.5 : 1.0. The chemical composition of the material, pre- pared by shock compression of h-BC2.5N, was approximately B :C :N = 1.0: 2.5 : 1.0, and the material included negligibly small amounts of hydrogen and metallic elements. Crystal structure Fig. 2(u) indicates that the starting material possesses a hexag- onal XRD pattern.The formula was then expressed as h- BC2.5N. Fig. 2(b) shows the shock-compressed material to have a cubic pattern, in which the (loo), (222) and (420) reflections are absent and the (200) reflection cannot be observed at the noise level. The d values, intensities and the lattice constant are given in Table 1. These were all the same as those obtained -20 40 60 80 100 120 140 2e/degrees Fig. 2 X-Ray diffraction patterns characterizing the crystal structures before and after shock-compression of the B-C-N compounds: (a) the starting compound (h-B(&N); (b)the purified compound after shock- compression (c-BC,.,N) Table 1 Interplanar spacings/nm for the c-BC,.,N solid solution cubic observed calculated" hkl d values d values difference Iobs 111 0.2083 0.208 1 -0.0002 100 200" -0.1803 220 0.1274 0.1275 -0.0001 24 311 0.1087 0.1087 -0.001 13 222" -0.1041 400 0.09015 0.0901 3 -0.0002 5 331 0.08272 0.8273 0.0001 12 420" -0.07359 --"These lines are inhibited for the diamond structure due to the extinction rule."Calculation is based on the cubic parameter a, = 0.3605k0.0001 nm. previously, expect for the absence of the (200) reflection6 The crystal structure is therefore assigned to a diamond structure without long-range ordering of the atomic arrangement, a class of a face-centred cubic form. The material is formulated as c- BC2.5N, with a regular tetrahedral structure where each atom in the unit cell is four-coordinated, in other words, a heterodia- mond.The atomic distribution and arrangement in the lattice planes have been discussed for c-BN and the cubic BN-C mixed crystals with the help of the theoretical structure ampli- tude, Fhkl, for adequate structure models.2,8 However, as mentioned by Badzian, the problem for the cubic B-C-N solid solutions was very difficult because of the close similarity of the atomic scattering factors of the neighbouring light elements.' In the present study, we attempted to discuss this problem. The scattering intensity, Ihkl, on the (hkl)plane was calculated in the following way:8 Ihkl x Phkl [( 1+COS' 28/sin2 e cos e)] I Fhkl I Fhkl=Cfnexp(2ni(hU,+kT/,+IWfl)) where Phkl is the multiplicity factor [the values are 8 for the (111) plane and 6 for the (200) plane], 8 is the Bragg angle, Fhkl is the structure factor,f, is the atomic scattering factor of the nth atomYg and (U,,V,, W,) is the fraction coordinate.Assuming a model in which the B-N pair lies parallel to the (110) direction, the same model as that for c-BCN claimed by Wedlake, the result was as follows: I Fiii I 2=32 i(fc+fB)2+(fc+fN)2) I F200 I 2=24 (fB-fN)2 the relative intensity 12,,/11,, was approximately 1.5%. In the case of full short-range order in which both B and N atoms occupy exclusively fcc sublattices, the 1200/1111ratio for c- (BN)xC(l-x) is given by x2/16.5 (where x is the composition) as reported by Badzian;2 therefore, the I,,,/I,,, ratio for c- BC2,,N is 0.5%.Clearly, these models are not compatible with the experimental result for C-BC~.~N. The only case permitting a negligible Izoo/llllratio is the model in which the atoms are randomly distributed to have an equal scattering factor on average. We believe, therefore, that the constituent atoms of c- BC2.5N are distributed homogeneously. We are also attempting a perfect solution of the problem using high-radiant XRD and neutron diffraction, and the results will be discussed elsewhere, Chemical structure The chemical structures of the cubic B-C-N solid solutions have never been investigated, possibly because of their small quantities and low purity. We were able to obtain a highly pure material in a sufficiently large amount to observe the NMR spectra.Fig. 3 shows the 13C NMR spectrum of c-BC2.,N. The spectrum apparently consists of three components and was therefore deconvoluted by assuming three components with the help of a curve-fitting technique. The result gave three 150 120 90 60 30 6 Fig. 3 Solid-state I3C MAS NMR spectra of cBC,.,N: (a)the observed spectrum; (b-1), (b-2) and (b-3) show deconvolution assuming three components, the curve fitting was carried out using a Lorentzian curve (b-1) and (b-2)and a Gaussian curve (b-3); (c) the difference between the observed spectrum and the sum of the separated curves separated peaks, and the sum of the three curves successfully fitted the original curve. The chemical shifts of the three peaks were over the range 6 37-55.The chemical shift, 6, of the tetrahedral bonding carbon of diamond is 6 35.2, and the reported three-coordinate bonding carbons of h-BC,N exhibit a broad peak with a maximum at 6 135.'' The three peaks of c-BC,.,N were, therefore, attributed to tetrahedral bonding carbons. The peaks were assigned as follows on the basis of the reference standards of diamond: the peak at 6 37.0 is due to the C-C bonding carbon; either the peak at 6 44.3 or the broad peak at around 6 55.2 is the C-N bonding carbon or the C-B bonding carbon. The assignment of the first peak is based on the chemical shift (6 35.2) of the C-C bonding carbon in diamond. Although the assignment of the remaining two peaks is uncertain, the third peak is possibly due to the C-B bonding carbon, because the C-B-C bonding carbon in B,C shows a peculiar, very broad linewidth." In addition, the chemical bonding for boron was investigated by "B NMR measurement.As shown in Fig. 4, a single broad peak was observed at 6 1.3. Regarding the reference standards, the peak of the four-coordinate bonding boron of c-BN was observed at 6 1.8 and those of the three-coordinate bonding borons of h-BN were at 6 12.2-23.9. The "B peak of c-BG,.,N was therefore assigned to the four-coordinate bonding boron. The atomic distribution of c-BC,.,N in the unit cell can be presumed to differ from those of diamond and c-BN, and this might reflect the slight differences in the chemical shifts of carbon and boron between c-BC,.,N and the cubic standards.Po Fig. 4 Solid-state "B MAS NMR spectrum of cBC2.,N J. Muter. Chem., 1996, 6(11), 1799-1803 1801 Electronic structure Fig. 5 and 6 show the TEM image and the selected area electron diffraction (SAED) pattern of c-BC2.,N, respectively. The observed lattice image and the SAED sharp rings con- Fig. 5 TEh image o the c-BC,., I powder sample obtained in this work: note the size of microcrystals is in the range of 5-20 nm Fig. 6 Selected area electron diffraction (SAED) pattern of c-BC2,,N the sharp rings consisting of a series of spots show the material to be a polycrystal 1802 J. Muter. Chew., 1996, 6(11), 1799-1803 0 100 200 300 400 energy loss/eV Fig.7 EELS spectrum at the K-edges for the constituent atoms of c-BC2.5N Table 2 EELS peak positions at the K-edges for the B-C-N compound and several reference standard compounds ls-m* (K-edge) 1s +r~* (K-edge) bonding compound orbitals B/eV C/eV N/eV B/eV C/eV N/eV c-BC2.,N sp3 ---194.0 287.0 403.0 diamond sp3 ----286.0 -C-BN SP3 ---196.0 -404.0 h-BC,.,N sp2+p, 189.0 282.0 398.0 197.0 290.0 404.0 h-BC,N12 sp2+p, 190.5 284.0 398.0 197.5 290.5 404.0 graphite sp2+p, -281.0 --290.0 -h-BN sp2+p, 188.0 -399.0 196.0 -406.0 sisting of a series of spots show the material to be a polycrystal in which microcrystals of size 5-20 nm are stacked at the grain boundaries.Fig. 7 shows the EELS spectrum of c-BC,.,N, and the data are given in Table 2 with additional data for h- BC2,,N, h-BC2N, diamond and c-BN.The core-loss region in the spectrum is useful for discussing the bonding properties of the constituent atoms. h-BC2.5N had both ls-m* and ls-+a* transitions at the K-edges (the spectrum was abbreviated). This is the same as the reported assignment of the EELS peaks for h-BC2N.I2 In contrast c-BC,.,N exhibited no ~s-+Tc* trans-ition, but 1s+c~* transitions very similar to those for diamond and c-BN were observed. That is, the constituent atoms of c- BC2.,N are combined with only sp3 a-bonds. Based on the EELS spectra, each atom of c-BC2.,N was found to be in an environment akin to that of each atom in diamond and c-BN, but the EELS fine structure of c-BC,.,N is slightly different from those of diamond and c-BN: (1) the EELS spectrum of c-BC2.,N was not a perfect superposition of those of diamond and c-BN, although the a* transition peaks of c-BC2.,N are similar to those of diamond and c-BN within the resolution power of EELS; (2) diamond and c-BN exhibit sharp pattern^,^ whereas c-BC,.,N exhibits blunt patterns.The similarity of the peak positions denies the possibility of the differences in the local atomic environment, because the differences in the peak positions are also very small for a series of hexagonal com- pounds (see Table2). The slight difference in the spectrum form might rather positively support the existence of a homo- geneous atomic environment, because, if B is bonded to C in addition to N, the superposition of the B K c~*transitions for B-C and B-N bondings makes a peak less sharp than that for only B-N bonding.Conclusions A cubic B-C-N compound was synthesized in satisfactory yields by the hexagonal-to-cubic transformation using a cylin- drical shock-compression apparatus and was investigated in terms of the chemical composition, crystal structure and bond- ing properties. The material obtained was confirmed to be a single cubic phase of the BC2.,N solid solution (c-BC,.,N) having a diamond-like structure without long-range ordering of the atomic arrangement. c-BC,.,N existed as a polycrystal composed of microcrystals of 5-20 nm in size. The constituent atoms of c-BC2.,N were all found to be tetrahedrally coordi- nated to give C-C, C-B, C-N and B-N bonds, consistent with sp3 o-bonds in each case.The NMR chemical shifts and the fine structure of the EELS spectrum of the c-BC2.,N suggested that the constituent atoms of c-BC,.,N are homogeneously distributed on the lattice planes. This work was conducted in the program ‘Advanced Chemical Processing Technology’ consigned to ACTA from NEDO, which is carried out under the Industrial Science and Technology Frontier Program enforced by the Agency of Industrial Science and Technology. References De Beers Industrial Diamond Division Ltd., Ger. Pat., 2806070, 1979; R. J. Wedlake and A. L. Penny, Chem. Abstr., 1979, 90, 428652. 2 A. R. Badzian, Muter. Res. Bull., 1981, 16, 1385. 3 T. Sasaki, M. Akaishi, S. Yamaoka, Y. Fujiki and T. Oikawa, Chem.Muter., 1993,5,695. S. Nakano, M. Akaishi, T. Sasaki and S. Yamaoka, Chem. Muter., 1994,6,2246. E. Knittle, R. B. Kaner, R. Jeanloz and M. L. Cohen, Phys. Rev. B, 1995,51,12 149. Y. Kakudate, M. Yoshida, S. Usuba, H. Yokoi, S. Fujiwara, M. Kawaguchi and T. Sawai, in Advanced Materials ’93, ed. M. Wakatsuki, J. C. Angus, A. T. Collins, N. Fujimori, H. Kawarada and H. Komiyama, Trans. Muter. Res. Soc. Jpn., Elsevier, Amsterdam, 1994, vol. 14B, pp. 1447-1450. 7 T. Sasaki and N. Bartlett, Proc. 197th ACS National Meeting (Inorganic), ACS, Washington, DC, 1990, p. 46. 8 F. P. Bundy and R. H. Wentorf, Jr., J. Chem. Phys., 1963,38,1144. 9 I. Nitta, X-Ray Crystallography, Maruzen, Tokyo, 1959. The value of an atomic scattering factor of the nth atom, fn, is given in Appendix 8 on p. 748. 10 R. Riedel, J. Bill and G. Passing, Adv. Muter., 1991,3, 551. 11 J. Conard, M. Bouchacourt, F. Thevenot and G. Hermann, J.Less-Common Met., 1986,117, 51. 12 J. Kouvetakis, T. Sasaki, C. Shen, R. Hagiwara, M. Lerner, K. M. Krishnan and N. Bartlett, Synth. Met., 1989,34, 1. Paper 6105034D; Received 18th July, 1996 J. Muter. Chern., 1996, 6(11), 1799-1803 1803
ISSN:0959-9428
DOI:10.1039/JM9960601799
出版商:RSC
年代:1996
数据来源: RSC
|
17. |
Low-temperature stabilisation of tetragonal zirconia by antimony |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1805-1809
Antonino Gulino,
Preview
|
PDF (1150KB)
|
|
摘要:
Low-temperature stabilisation of tetragonal zirconia by antimony Antonino Gulino," Russell G. Egdellb and Ignazio Fragala" "Dipartimento di Scienze Chimiche Universith di Catania, Kle A. Doria 6, 95125, Catania, Italy bInorganic Chemistry Laboratory, South Parks Road, Oxford, UK OX1 3QR Low-temperature (400-800 "C) stabilisation of tetragonal zirconia, prepared by a hydroxide gel route and a metal oxalate method, both followed by low-temperature annealing, has been achieved by partial substitution of Zr4+ with Sb3+. XRD and Raman spectroscopy have been used for characterisation of the doped zirconia powders as well as for identification of the crystal symmetry. A transition from the tetragonal to the monoclinic phase takes place above 800 "C. XPS results provide evidence of pronounced Sb segregation upon increasing the firing temperatures. The grain size dependence upon the annealing temperature has been evaluated using the Scherrer equation.Particle morphology and size have been imaged directly by scanning electron microscopy. Zirconia polymorphs represent an intriguing area of active interest. At room temperature, ZrO, adopts a thermo-dynamically stable,lP6 monoclinic structure based on seven- coordinate Zr4+ architecture. However, it transforms to a tetragonal polymorph around 1100 "C and, in turn, to a cubic phase around 2300 OC.' Although there have been reports of the formation and stability of tetragona17-' and cubic" pure zirconia thin films as well as on the stability of the cubic phase in nanocrystalline pure ZrO,," much effort has been devoted to the stabilisation of zirconia in tetragonal and cubic phases by using various aliovalent cation dopants.The dopants are believed to substi- tute for the Zr4+ ion in the cation framework, thus creating oxygen vacancies due to charge compensation.'2-16 Partially stabilised tetragonal and fully stabilised cubic zirconia have attracted considerable attention because these solid solutions possess remarkable physico-chemical properties which, in turn, have stimulated their applications as oxygen sensors, fuel cells, catalysts and other area^.'^-,^ Moreover, partially stabilised zirconia is a tough material with excellent mechanical properties? The factors governing the stability of doped stabilised zir- conia were first investigated as early as 1924.27 Nevertheless, several aspects of the prerequisites of the dopant needed to stabilise the high-temperature polymorphs remain unclear.A model has been proposed setting out criteria for suitable dopant cations.28 In particular, it is claimed that stabilising cations must have larger ionic sizes, a lower formal charge state and a higher ionicity than Zr.,' Although there is a body of experimental data to support this several excep- tions to its predictions have been found. Thus, there is evidence that (i) undersized dopants,', (ii) some tetravalent cation^'^-'^ which do not introduce oxygen vacancies, and (iii) cations which are less ionic than Zr,2-5v12,32 may stabilise ZrO,.In a previous investigation the ability of group VA metals to stabilise ZrO, was studied and it was found that Bi3+ stabilised tetragonal ZrO, at low temperature^.^^ In the present paper we report the novel low-temperature preparation of tetragonal Sb3 +-doped ZrO, using coprecipitation methods. The present, low-temperature, synthetic procedure is of rel-evance in the context of the fabrication of doped ZrO, for several catalytic applications since high-area surfaces are gener- ally associated with low annealing temperatures.32P34 Experimenta1 Antimony-doped (3, 5, 10 and 15 mol%) ZrO, samples were synthesised by a coprecipitation method using high-purity zirconyl chloride hydrate (99.99'30, Aldrich) and antimony(II1) oxide (99.99%, Aldrich).Appropriate quantities of the starting materials were dissolved separately in water (ZrOC1,) and in concentrated HC1 (Sb,O,). The solutions were mixed and, under continuous stirring, neutralised with concentrated NH, solution to pH 10 (gel procedure). The hydroxide gels thus formed were filtered off, washed with distilled water and dried at 90°C. Effects due to both the synthetic procedure adopted and the nature of the precursors on the zirconia phase composition were investigated by synthesising 10% Sb-doped ZrOz samples from zirconyl chloride hydrate and antimony oxide by a metal oxalate method in ethanol at pH 0.1 (oxalate pr~cedure).~' Reference undoped ZrO, samples were prepared by similar procedures. The resulting powders were stepwise fired (24 h step-') at progressively higher temperatures (AT=100"C), in the interval 200-1000 "C, in recrystallised alumina crucibles.X-Ray powder diffraction (XRD) data were recorded on a computer-interfaced Philips PW 1130 diffractorneter.,, Raman spectra were measured with a JOBIN YVON UlOOO spectrometer at 300 K in the 90" geometry (496.5 nm Ar' laser, 100 mW). The instrumental resolution was 4 cm-l. The Sb-doping levels of all samples were determined by wavelength-dispersive X-ray fluorescence using the Zr KP and Sb Ka emissions. A Philips PW 1410/00/60 spectrometer, equipped with a tungsten anode and an LiF (220) crystal as a secondary radiation dispersive element, was ~sed.~~,~~ Samples for X-ray fluorescence analysis were prepared as described el~ewhere.,~ XPS measurements on powders held on indium foil were made with a PHI 5600 Multi Technique System (operating in the lo-" Torr pressure range).Resolution, correction for satellite contributions, background removal38 and a procedure to account for steady-state charging effects,' have been described elsewhere.,, Morphological analyses of sample surfaces were carried out in a 340 Cambridge scanning electron microscope (SEM) with a 20 keV electron beam energy and a resolution >0.2 pm. Results and Discussion X-Ray fluorescence (XRF) data show that all the samples prepared via the gel route have final compositions almost identical to the nominal doping levels (Table 1)after firing up to 800 "C.Higher firing temperatures (1000"C) resulted in pronounced Sb loss (ca. 50%),possibly as Sb406 vapo~rs~~,~' J. Muter. Chem., 1996, 6(11), 1805-1809 1805 Table 1 X-Ray fluorescence analysis" doping level from X-ray nominal doping level firing temperature fluorescence analyses (%I /"c (%I 3 500 2.80 5 500 4.50 10 not fired 9.40 10 400 9.40 10 500 9.38 10 800 9.35 10 1000 4.75 15 600 13.70 10 (pH 0.1) not fired 6.85 10 (pH 0.1) 600 6.70 " Estimated uncertainties are k5%. (Table 1). Note that 10% Sb-doped samples heated directly at 500 "C and fired for 24 h suffer a loss of 33% of the Sb. Therefore, sudden heating procedures cause Sb406 volatilis- ation prior to stabilisation of the ZrO, phase. By contrast, a stepwise heating procedure in the range 200-800 "C allows immobilisation of Sb cations in the crystal architecture.Samples prepared by the oxalate procedure do not have the nominal compositions. Low Sb (ca. 30% of nominal value) contents are always observed due to the non-quantitative precipitation of the antimony oxalate at pH 0.1 (Table 1). XRD data of undoped ZrO,, investigated as a reference, show that the material is amorphous after firing below 500 "C. Evidence of crystalline phases, provided by monoclinic and less intense tetragonal XRD reflections, is found after firing at 500 "C. Above this temperature the tetragonal lines become weaker and at 800°C the samples are completely monoclinic (Fig. l).33 The XRD patterns of all samples annealed up to 400 "Care uninformative and typical of amorphous samples, although the broad peak profiles do show some dependence on the synthetic procedure adopted.35 Above 400 "C crystalline phases emerge, and at 500°C the XRD patterns of the 5% and 10% doped samples show peaks indicative of tetragonal and/or cubic phases (Fig.2). The formation of crystalline phases occurs at higher temperatures (600°C) in the case of 15% doped samples. Thus, almost identical XRD patterns, free of monochic features, are obtained for 5% and 10% doped 20 40 60 80 28/degrees Fig. 1 X-Ray powder diffraction pattern for undoped ZrO, stepwise fired at 800°C 1806 J. Mater. Chem., 1996, 6(11), 1805-1809 20 30 40 50 60 70 80 2 Bldegrees Fig.2 X-Ray powder diffraction patterns, over a 20" <28 c80" angular range, for (a) 3% Sb-doped ZrO, stepwise fired at 500°C; (b)5% Sb-doped ZrOz stepwise fired at 500 "C; (c) 10% Sb-doped ZrO, stepwise fired at 500 "C and (d) 15% Sb-doped ZrO, stepwise fired at 600 "C. Peaks at 28=38.47, 44.73, 65.13 and 78.22" are due to the A1 sample holder (*). (A)Tetragonal phase; (a)monoclinic phase. samples (Fig. 2, 3), while only negligible quantities of a mono- clinic phase (28=24.35) are evident in the 15% doped sample [Fig. 2(d)].42 Complete stabilisation is never achieved with 3% Sb doping [Fig. 2(a)] and the XRD patterns show distinct features of both monoclinic and stabilised ZrO, phases. Note, in this context, that previous studies have shown that full tetragonal stabilisation occurs in 3 % Bi3 +-doped ~irconia.~~ It has been reported that the difference between the c and (a, b) lattice constants in tetragonal ZrO, gives rise to characteristic line splittings which are absent in the cubic 20 40 60 80 2Bldeg rees Fig.3 X-Ray powder diffraction patterns, over a 20" <28 <80" angular range, for 10% Sb-doped ZrO, samples: (a) obtained by the gel procedure and fired directly at 500°C; (b) obtained by the oxalate method and fired stepwise (200, 300,400, 500 "C). Peaks at 28 =38.47, 44.73, 65.13 and 78.22" are due to the A1 sample holder. 0 200 400 600 800 frequency/cm-' Fig. 4 Representative Raman spectrum of 10% Sb-doped ZrO, powder fired at 500 "C and measured in the 200-700 cm-' frequency range structure.43t These splittings are most obvious in higher-order, low intensity, XRD reflection^.^^ The splittings evident in Fig.2(c), (d) are, therefore, characteristic of the tetragonal nature of the Sb-doped zirconia prepared in the present work. Interestingly, identical line splitting has been observed already in tetragonal Y,03-Ce0,-ZrO, ceramic pellets pressed at 190 MPa.44 Of course, an unambiguous phase assignment requires additional supporting evidence, and Raman spectroscopy pro- vides another method for the identification of crystal symmetry in zirconia polymorph~.~~9~~-~~ The number of Raman-active bands in ZrO, crystals can be deduced, for the various polymorphs, from symmetry consider- ation~.~~Raman spectra of the present zirconia samples, in the Sb doping range between 5 and 15% (Fig.4), closely resemble those reported previously for Y-doped tetragonal zirconia and point to dominance of the tetragonal pha~e,~'.~~ although small amounts of cubic Sb-doped zirconia cannot be ruled out. Upon increasing the temperature, Sb-doped samples remain tetragonal up to 8OO"C, whilst at 1000°C they become mono- clinic (Fig. 5).42 The solid solution range of Sb203 in ZrO, is expected to increase with increasing temperature. On the other hand, Sb406 is a volatile oxide and the vapour pressure of Sb406, in equilibrium with the Sb-doped ZrO, sample, will increase with increasing temperature: the rate of mass transport of Sb to the surface will also show strong temperature dependence.Thus, on firing in an open system at lOOO"C, transport of Sb to the surface and loss as Sb406 becomes significant on the timescale of the annealing cycle. At the same time, activation barriers to phase transformations can be overcome allowing the monoclinic phase to form [Fig. 5 (d)]. Similar behaviour has been observed previously in Fe203 and Ga203-doped zirconia.', A further effect of firing at progressively higher temperatures is that the XRD lines become somewhat sharper (Fig. 5), an indication of some crystal growth. Particle sizes of the 10% Sb-doped sample at various temperatures were determined from XRD data using the Scherrer and Warren equation:47 s=0.9 n/(Bcos 6,) (1) t The splittings of the XRD lines observed in the present samples are: (002)-(200) at 28~35"; (202)-(220) at 28%50"; (113)-( 131) at 28~60".I"'I"'I"'I 20 40 60 80 28/degrees Fig. 5 X-Ray powder diffraction patterns, over a 20" <28 <80" angular range, for a representative Sb-doped ZrO, (loo/,) solid solutions stepwise fired at 400°C (a), 600°C (b), 800°C (c) and 1000°C (d). Peaks at 28=38.47, 44.73, 65.13 and 78.22" are due to the A1 sample holder. where S is the crystallite grain size, Iz is the X-ray wavelength and 6B is the Bragg angle of the considered XRD peaks. B represents the FWHM XRD broadening, obtained as follows: B2=Bm2-B,2 where Bm2 and B: represent the FWHM line width of the material itself and of a crystalline a-A1203 standard, respectively. The resulting values of crystallite sizes are shown in Table 2.Note that the largest value (27nm) is found in the samples fired at 1000"Cwhere the transition to the thermodynamically more stable monoclinic phase takes place. Similar behaviour has been noted previously in Bi-doped Zr02,33 and Chatterjee et al. have observed a similar temperature dependence for nanocrystalline pure cubic ZrO,." Fig. 6(a)shows the XP spectra of the 500°C annealed 10% Sb-doped ZrO, in the Zr 3d energy region. Several interesting features are apparent. The Zr 3d features consist of the main 3d5/2 and 3d3/, spin-orbit components at 181.8 and 184.1 eV, respe~tively.~~-'~Moreover, a satellite feature is found at about 15 eV from the main Zr 3d,,, peak.In analogy to previously reported XPS results on pure and doped zirconia, this feature can be interpreted in terms of shake-up proces~es.~~ Fig. 6(b)shows the XP spectra of the same sample in the 0 Is and Sb 3d binding energy region. As noted previo~sly~~.~~ the Sb 3d5,, peak overlaps with the 0 Is peak and, therefore, only the Sb 3d,,, (539.4 eV) peak can be used for quantification of the surface Sb content. Note that the binding energy (EB) of the Sb 3d3/, peak is identical to the value observed in pure Sb203 .51 Table 2 Particle sizes of 10% Sb-doped ZrO, at different temperatures firing temperaturePC particle size/nm 500 16 600 22 700 23 800 25 1OCQ 27 J. Mater. Chem., 1996, 6(ll), 1805-1809 1807 I'I.I',.I -1 * 1.1 170 175 180 185 190 195 200 205 525 530 535 540 545 binding energy/eV Fig.6 Al-Ka excited XPS of 10% Sb-doped ZrOz solid solution fired at 500°C and measured in the Zr 3d binding energy region (a)and in the 0 Is-Sb 3d binding energy region (b). Structure due to satellite radiation has been subtracted from the spectra. Fig.7 SEM image of a representative 10% Sb-doped ZrO, solid solution fired at 1000"C The XPS Sb 3d3,,/Zr 3d intensity ratio taken at 45" emission relative to the surface plane, has been used to measure the effective surface Sb/Zr atomic ratios once due account is taken of the relevant atomic sensitivity factors.', Thus, the Sb/Zr atomic ratio is close to the nominal value in the sample fired at 500°C, whilst the value of 0.37 measured in the 1000°C- fired sample indicates that there is some Sb segregation to the ~urface.~'.~~9 '3 Finally, SEM images allow us to investigate the morphology of the present samples even though, of course, they are not diagnostic of particular crystalline phases.Evidence of a pro- nounced dependence of the sample morphology on the annealing temperature has been obtained. A higher crystallinity is observed at 1000°C where the monoclinic phase becomes predominant (Fig. 7).33 Conclusions The coprecipitation of Sb (doping level>5%) and Zr as a hydroxide gel followed by thermal annealing in air, in the range 400-800 "C, stabilises ZrO, in the tetragonal phase. The oxalate procedure does not lead to a quantitative precipitation of the Sb.Grain growth has been observed upon increasing the firing temperature. Both XRD and Raman measurements point to the tetragonal symmetry of the present 5-15% Sb-containing samples. Progressive heating treatments cause surface Sb segregation. The observed stabilisation by Sb3+ ions is in agreement with other st~dies,~~~.'~~~~,~~~'~-~~even though it is at variance with the original model which suggested that only cations more ionic than Zr should be capable of stabilising effects in zirconia.28 Finally, we must mention that the interesting surface com- position of the present systems provides motivation for further surface structural investigations. I. F. and A. G. thank the MURST and the CNR (Rome) for financial support.Dr. G. Compagnini (Co.Ri.M.Me., Catania) is gratefully acknowledged for Raman measurements. References 1 A. F. Wells, Structural Inorganic Chemistry 4, Clarendon Press, Oxford, 1975. 2 P. Li, I-W. Chen and J. E. Penner-Hahn, Phys. Rev. B., 1993, 48,10063. 3 P. Li, I-W. Chen and J. E. Penner-Hahn, Phys. Rev. B., 1993, 48,10074. 4 P. Li, I-W. Chen and J. E. Penner-Hahn, Phys. Rev. B., 1993, 48,10082. 5 V. Sergo, C. Schmid, S. Meriani and A. G. Evans, J. Am. Ceram. SOC.,1994,77,2971. 6 K. Ishida, K. Hirota, 0. Yamaguchi, H. Kume, S. Inamura and H. Miyamoto, J. Am. Ceram. SOC., 1994,77,1391. 7 B. J. Gould, I. M. Povey, M. E. Pemble and W. R. Flavell, J. Muter. Chem., 1994,4,1815 and references therein. 8 C.S. Hwang and H. J. Kim, J. Muter. Res., 1993,8, 1361. 9 C. M. Scanlan, M. Gajdardziska-Josifovska and C. R. Aita, Appl. Phys. Lett., 1994,64,3548 and references therein. 10 Z. Xue, B. A. Vaartstra, K. G. Caulton and M. H. Chisholm, Eur. J. Solid State Inorg. Chem., 1992,29,213 and references therein. 11 A. Chatterjee, S. K. Pradhan, A. Datta, M. De and D. Chakravorty, J. Muter. Res., 1994,9,263 and references therein. 12 P. Li, I-W. Chen and J. E. Penner-Hahn, J. Am. Ceram. SOC., 1994, 77, 118. 13 J. Xue and R. Dieckmann, Solid State Ionics, 1994,73,273. 14 M. Fukuya, K. Hirota, 0. Yamaguchi, H. Kume, S. Inamura, H. Miyamoto, N. Shiokawa and R. Shikata, Muter. Res. Bull., 1994,29, 619. 15 S. Chen. W. Deng and P. Shen, Muter. Sci. Ezg.B, 1994,22,247. 16 M. M. R. Boutz, A. J. A. Winnubst, F. Hartgers and A. J. Burggraaf, J. Muter. Sci., 1994,29,5374. 17 T. H. Etsell and S. N. Flengas, Chem. Rev., 1970,70, 1970. 18 K. Sasaki, H. P. Seifert and L. J. Gauckler, J. Electrochem. SOC., 1994,141,2759. 19 0. Yamamoto, Y. Arati, Y. Takeda, N. Imanishi, Y. Mizutani, M. Kawai and Y. Nakamura, Solid State Ionics, 1995,79, 137. 20 J. Van Herle, A. J. McEvoy and K. Ravindranathan Thampi, J. Muter. Sci., 1994,29, 3691. 21 H. Kurosawa, Y. Yan, N. Miura and N. Yamazoe, Chem. Lett., 1994,1733. 22 G. Z. Cao, J. Appl. Electrochem., 1994,24,1222. 23 K. Otsuka, T. Ando, S. Suprapto, Y. Wang, K. Ebitani and I. Yamanaka, Catal. Today, 1994,24, 315. 24 R. S. Garvie, R. H. Hannink and R.T. Pascoe, Nature (London), 1975,258,703. 25 G. Fischer, Ceram. Bull., 1986,65, 1355. 26 T. Yokoyama, T. Setoyama, N. Fujita, M. Nakajima, T. Maki and K. Fuji, Appl. Catal., 1992,88, 149. 27 E. A. Van Arkel, Physica, 1924,4,286. 28 S. M. Ho, Muter. Sci. Eng., 1982,54,23. 29 M. Yoshimura, Am. Ceram. SOC.Bull., 1988,67,1950. 30 J. Dexpert-Ghys, M. Faucher and P. Caro, J. Solid State Chem., 1984,54, 179. 31 J. G. Allpress and H. J. Rossell, J. Solid State Chem., 1975,15,68. 32 A. Keshavaraja and A. V. Ramaswamy, J. Muter. Res., 1994,9,837 and references therein. 1808 J. Muter. Chem., 1996, 6(ll), 1805-1809 33 A. Gulino, S. La Delfa, I. Fragala and R. G. Egdell, Chem. Muter., 48 D. Majumdar and D. Chatterjee, J. Appl. Phys., 1991,70,988. 34 1996,8, 1287.P. Afanasiev, C. Geantet and M. Breysse, J. Catal., 1995,153, 17. 49 D. D. Sarma and C. N. R. Rao, J. Electron. Spectrosc. Relat. Phenom., 1980,20,25. 35 G. Gongyi and C. Yuli, J. Am. Ceram. SOC., 1992,75,1294. 50 G. Marletta and S. Pignataro, Chem. Phys. Lett., 1986,124,414. 36 G. G. Condorelli, G. Malandrino and I. Fragala, Chem. Muter., 51 W. E. Morgan, W. J. Stec and J. R. Van Wazer, Inorg. Chem., 1973, 1994,6, 1861. 12,953. 37 C. S. Hutchison, Laboratory Handbook of Petrographic Techniques, 52 D. Briggs and M. P. Seah, Practical Surface Analysis, 2nd edn., 38 Wiley, New York, 1974. M. Repoux, Surf. Interface Anal., 1992,18,567. 53 Wiley, Chichester, 1994. A. E. Hughes, J. Am. Ceram. SOC., 1995,78,369.39 40 G. M. Ingo, Appl. Surf. Sci., 1993,70171,235. A. Gulino, A. E. Taverner, S. Warren, P. Harris and R. G. Egdell, Surf. Sci., 1994,315, 351. 54 55 D. J. Kim, H. J. Jung and D. H. Cho, Solid State Ionics, 1995,80,67. D. J. Kim, J. W. Jang, H. J. Jung, J. W. Hug and I. S. Yang,J. Muter. Sci. Lett., 1995, 14, 1007. 41 A. Gulino, G. G. Condorelli, I. Fragala and R. G. Egdell, Appl. 56 M. Hartmanova, I. Travenec, A. A. Urusovskaya, K. Putyera and Surf. Sci., 1995,90,289. D. Tunega, Solid State Ionics, 1995,76,207. 42 American Society for Testing and Material, Powder Diffraction Files (Joint Committee on Powder Diffraction Standards, USA, 57 V. P. Dravid, D. Ravikumar, M. R. Notis, C. E. Lyman, G. Dhalenne and A. Revcolevschi, J. Am. Ceram. SOC., 1994, 77, 43 1974),set 1-5 (revised), Inorganic. R. Srnivasan, S. F. Simpson, J. M. Harris and B. H. Davis, J. Muter. Sci. Lett., 1991,10,352. 58 2758. M. Hartmanova, F. W. Poulsen, F. Hanic, K. Putyera, D. Tunega, A. A. Urusovskaya and T. V. Oreshnikova, J. Muter. Sci., 1994, 44 J. U. Wan and J. G.Duh, J. Muter. Sci. Lett., 1993, 12, 575. 29,2152. 45 D. W. Liu, C. H. Perry and R. P. Ingel, J. Appl. Phys., 1988, 64, 1413. 59 M. Fukuya, K. Hirota, 0.Yamaguchi, H. Kume, S. Inamura and H. Miyamoto, Muter. Res. Bull., 1994,29,619. 46 J. Cai, Y. S. Raptis and E. Anastassakis, Appl. Phys. Lett., 1993, 62, 2781. 60 61 M. Feng and J. B. Goodenough, J. Am. Ceram. SOC., 1994,77,1954. H. Naito and H. Arashi, Solid State Ionics, 1994,67, 197. 47 B. E. Warren, X-Ray DifSraction, Addison-Wesley, Reading, MA, 1969. Paper 6/03418G; Received 16th May, 1996 J. Muter. Chem., 1996, 6(11), 1805-1809 1809
ISSN:0959-9428
DOI:10.1039/JM9960601805
出版商:RSC
年代:1996
数据来源: RSC
|
18. |
Preparation and characterisation of high refractive index PbO–TiO2–TeO2glass systems |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1811-1814
Raul F. Cuevas,
Preview
|
PDF (506KB)
|
|
摘要:
Preparation and characterisation of high refractive index Pb0-Ti0,-Te02 glass systems Raul F. Cuevas,' Ana M. de Paula,' Oswaldo L. Alves,b Norbert0 Aranha," JosC A. Sanjurjo; Carlos L. Cesar' and Luiz C. Barbosa*." 'lnstituto de Fisica, Universidade Estadual de Campinas, Caixa Postal 61 65, 13083-970 Campinas-SP, Brazil bInstituto de Quimica, Universidade Estadual de Campinas, Caixa Postal 61 54, 13083-970 Campinas-SP, Brazil We describe the preparation and characterisation of high refractive index Pb0-Ti02-Te02 glass systems. Highly homogeneous glasses were obtained by agitating the oxide mixture during the melting process in an alumina crucible. The characterisation was performed by X-ray diffraction, density, dilatometry, Raman scattering, light absorption and linear refractive index measurements.The results show a change in the glass structure as the PbO content increases: the TeO, trigonal bipyramids characteristic of TeO, glasses transform into TeO, trigonal pyramids. However, the measured refractive indices are almost independent of the glass composition. We show that third-order non-linear optical susceptibilities calculated from the measured refractive indices using Lines' theoretical model are also independent of the glass composition. Tellurium oxide-based glasses are potential materials for optical device applications due to their non-linear optical properties. They show high refractive indices, and consequently high third-order non-linear susceptibilities x(,), and high rela- tive permittivities.''2 Also, they are transparent over the visible and near IR regions. Kim et ul., have measured for a pure TeO, glass a value of x(~)=1.41 x lo-', esu; however, it is not easy to obtain this pure glass.Some recent works have shown that the addition of TiO,, PbO and Nb205 increases the glass stability and refractive In particular, the PbO-Ti0,-TeO, glass systems have presented the highest f3) values, in the range 2 x lo-', to 4 x lo-', esu.'y6 Nonetheless, the dependence of the non-linear properties on the concentration of the three components is not yet clearly established. The discrepancies in the measured x(,) values are still quite high, probably due to inhomogeneities in the gla~ses.~?~ In this work we present a study of the properties of this glass system as a function of the PbO composition.We have synthesised xPbO : 10TiO,:( 100 -x)Te02 glasses with x =5, 10, 15, 20 and 25 mol%. We obtained highly homogeneous glasses by agitating the mixture during the melting process. The glasses were characterised by means of X-ray diffraction (XRD), density, dilatometry, Raman scattering, optical absorp- tion and linear refractive index measurements. Most of the measured properties show the previously observed6v8 change in the glass structure as the PbO content increases: the TeO, trigonal bipyramids characteristic of TeOz glasses break into TeO, trigonal pyramids. Nonetheless, the measured refractive indices are almost independent of the glass composition. Calculated f3) values using Lines' theoretical modelg*'* are also independent of the glass composition. Experimental We studied a series of five xPbO : 10Ti0, :( 100-x)TeO, glass samples with x =5, 10, 15, 20 and 25 mol%, which we denote as PTT1, PTT2, PTT3, PTT4, and PTTS, respectively (see Table 1).The materials used were reagent-grade TiO, (Merck), TeO, (Merck) and PbO (Riedel). A batch of 25 g of the mixture for each composition was placed directly in an alumina crucible in an electric furnace. The mixture was agitated during the melting process in order to obtain homogeneous concen-trations. Melting was achieved at 900 "C for 20 min. The melt was quenched between a pair of stainless-steel plates and later annealed for 2 h at 300°C.The samples were prepared in two forms: powder for the XRD, and approximately 1 mm thick slabs for the other measurements. The glassy state was confirmed by X-ray diffraction analysis using Cu-Ka radiation in a Shimadzu difractometer. The densities were determined by the pycnometer method using helium as the displacement gas. The thermal properties were measured with a Harrop conventional horizontal dilatometer. The Raman spectra were measured using the argon ion laser line at 514.5 nm at backscattering geometry and a triple Jobin- Yvon spectrometer with multichannel detection. The absorp- tion measurements were obtained using a Perkin-Elmer Lambda 9 spectrophotometer. For the refractive index measurements we used a manual Rudolph null ellipsometer (model 436) with a 150 W tungsten lamp as a light source.The desired wavelengths were selected by interference filters. The output power of the ellipsometer optics was measured with a commercial liquid-nitrogen-cooled InAs detector, whose signal was integrated by a lock-in ampli- fier. The use of a less noisy detector instead of the cooled PbS detector supplied by the manufacturer proved to be essential to achieve experimental errors as low as 0.1%. Results and Discussion Glass formation The preparation procedure described in the Experimental section permitted us to obtain bubble-free transparent glasses with high homogeneity and a yellowish colour. All the composi- tions showed the XRD patterns typical for a glass phase presenting a halo near 28 =22.5".Density Fig. 1 shows the glass density dependence on the PbO concen- tration. The density increases almost linearly with increasing PbO content. This is an indication that the PbO enters the glass structure as a network modifier. The linear behaviour may be explained by the fact that TeO, was substituted by the heavier PbO. Dilatometry The glass-transition temperatures (q),the softening-point tem- peratures (%) and the thermal expansion coefficients (K) are J. Muter. Chew., 1996, 6(11), 1811-1814 1811 Table 1 Parameters used in the x(3)calculations and the obtained values considering f= 1.72 composition glass (mol%) PTTl 5Pb0 :lOTi0, :85Te02 1.871 PTT2 lOPbO :10Ti0, :8OTeO, 1.850 PTT3 15Pb0 :10Ti0, :75Te0, 1.854 PTT4 20Pb0 :10Ti0, :7oTeo2 1.855 PTTS 25Pb0 :10Ti02:65Te02 1.854 6.0 5.8 m I6 5.6 s? .-3 v)s 5.4 U 5.2 5.0 0 5 10 15 20 25 30 PbO content (mol%) Fig.1 The PTT glass density dependence on the glass composition shown in Fig. 2 as a function of the PbO content. The temperatures and Td decrease linearly with increasing PbO concentration, whereas the thermal expansion coefficient shows a linear increase with the PbO content. A decrease in Tgand usually indicates a more open glass network, whereas a increase in the thermal expansion coefficient indicates weaker bonds, as weaker bonds increases the anharmonic contri- butions of the inter-ionic potentials to the thermal expansion. When the PbO enters the glass structure as a network modifier, some of the Te-0 bonds are broken with TeO, trigonal bipyramids transforming into TeO, trigonal pyramids with non-bridging This transformation weakens the Te-0 bonds and opens the glass network, and may explain the observed dependence on PbO concentration.380 360 9'340 320 .... i2.0 t -1.9 I 7 I Y 1.8 -I Lo 1.7 I OI\. Li -1.6 I 1.5 1.834 4.87 1.923 3.49 1.808 4.73 1.932 3.48 1.813 4.75 1.942 3.54 1.814 4.78 1.952 3.54 1.813 4.73 1.962 3.63 Raman spectra The Raman spectra from the PTT glasses are presented in Fig. 3. We observed the following features in the spectra: a band at 135 cm-' which is assigned to the PbO vibrations,8 a broad band at ca.478 cm-' which is assigned to Te-0-Te bending modes," whose intensity decreases as the PbO content increases. The two bands observed at ca. 682 and 773 cm-' have been assigned to the stretching vibrations of Te04 trigonal bipyramids, characteristics of TeOz glasses, and stretching vibrations of the Te03 trigonal pyramids, Note that as the PbO content increases the intensity of the Te03 band increases relative to that of the TeO, band, and the Te03 band is the most intense band for the samples PTT4 and PTTS. Also these bands shift slightly to lower Raman energy as the PbO concentration increases. There is also a weak band at ca. 300 cm-l for PbO contents greater than 15 mol%, which has been assigned to the bending vibrations of TeO, trigonal pyramids with non-bridging oxygen.These results are consist- ent with the structural change observed for PbO-Te02 glass systems; as the PbO content increases the TeO, trigonal bipyramids break into Te03 trigonal pyramids. Nonetheless, this change is slower when compared to Li,O-Ti0,-Te02 gla~ses.~Yamamoto et aL5 presented quantitative measure-ments of the Te coordination number in these two glass systems that showed this same trend. Optical absorption We have measured the glass transmission in the wavelength range 300-1500 nm. The transmittance data are shown in the inset of Fig. 4 for samples PTTl and PTT2, the curves are almost flat from 800 to 1500nm. The curves for the other samples (PTT3 to PTTS) are indistinguishable from the PTT2 curve.The UV-VIS cut-off wavelength is at ca. 420 nm for all compositions. From the transmission curve we calculated the optical absorption coefficient, a. In order to obtain the optical energy gap we plotted (aE), as a function of energy E. Fig. 4 shows the curves for samples PTTl and PTT2. The values of the optical energy gap were obtained from the extrapolation of the linear regions of the plots to (uE)~=O.The optical gap is at ca. 2.94 eV for all the PbO compositions. 0 5 10 15 20 25 30 IPbO content (mot%) 0 200 400 600 800 1000 Fig.2 Thermal properties of the PTT glass system as a function of Raman shift/cm-lthe PbO content. The transition (q;a) and softening (Td; 0)glass temperatures are shown in (a) and the thermal expansion coefficient Fig.3 Raman spectra of the PTT glass system. The spectra have been (K) in (b). arbitrarily shifted vertically for clarity. 1812 J. Muter. Chem., 1996,6(11), 1811-1814 50 -40 -N 5 (v 30 -2 70-c 20 -400 600 800 cu-Y% 10 2.5 2.6 2.7 2.8 2.9 3.0 3.1 EIeV Fig. 4 Plot of (aE)' as a function of energy for the PTT glass system, the curves for the PTT3, PTT4 and PTTS glasses are indistinguishable from the curve for the PTT2 glass. The solid and dashed lines are the linear fits to obtain the optical gap for PTTl and PTT2 samples, respectively. The inset shows the transmission curves. Linear refractive index We have measured the linear refractive index, n, of the glasses for wavelengths centred at 546, 633, 800, 1300, 1750 and 2000nm.The bandwidth of the used filters was 70nm. The results are listed in Table 2, the experimental errors are ca. 0.1%. Fig. 5 shows the linear refractive index dispersion, the solid lines are extrapolation curves. In Fig. 6 the refractive index is plotted as a function of PbO content for all the measured wavelengths. Note that there is just a small change with the PbO content: the values for the PTTl sample are Table 2 Measured refractive indices as function of wavelength for the PTT glasses refractive index, n wavelength/nm PTTl PTT2 PTT3 PTT4 PTTS 546 2.204 2.209 2.208 2.207 2.208 633 2.150 2.152 2.153 2.151 2.152 800 2.088 2.085 2.085 2.086 2.084 1300 1.990 1.981 1.983 1.981 1.983 1750 1.907 1.891 1.892 1.893 1.892 2000 1.852 1.835 1.835 1.836 1.834 2.3 2.2 XQ) Ut.-a) 2.1>._4-0E c 2 2.0 za C.--1.9 500 1000 1500 2000 Alnm Fig.5 Linear refractive index as a function of wavelength for the PTT glass system: W, PTT1; 0, PTT2. The results for the PTT3, PTT4 and PTTS are indistinguishable from those for the PTT2 glass. The lines are extrapolation curves. x 2.12 - a) nc 43 A 800. .-$ 2.04 - E c.0 c - V v 1300 1.96P t.--1.88 - 0 o 1750 #f #I #f #I 2000' I Fig. 6 Linea refractive index of the PTT glass system as fu ction of the PbO content. The wavelengths in nm are indicated at the right side of the plot.The crosses are the data of Yamamoto et a1.6 slightly larger than for the other samples at the high wavelength region. Also plotted in Fig. 6 are the linear refractive index data at 632.8 nm (crosses) from Yamamoto et aL6 The values are comparable; however, their results show an increasing trend with PbO content, which we do not observe. We have calculated Ed and E, using the expression for n as a function of the energy E proposed by Wemple:12 l/(n2 -1)= E,/Ed -E2/(E,Ed), where E, is the average excitation energy for electronic transition and Ed is the electronic oscillator strength related to dispersion. The E, values (listed in Table 1) are used to calculate the non-linear refractive index. Non-linear refractive index To calculate the non-linear refractive index we used Lines' bond-orbital the~ry,~.'~ which considers the influence of cat- ionic empty d-orbital on the glass non-linear optical response. The contribution of the d-orbital to the non-linear response is proportional to the decrease of the bond length between cation and anion, ld, and to the increase of the orbitals overlap (dlp).He studied a number of transparent transition-metal oxides and fouFd that this contribution is negligible for bond lengths &>2.3A, but increases rapidly as tbe bond length decreases and becomes dominant for Zd <2.0 A. The non-linear optical response was obtained using a variational method to describe the effect of an applied electric field j7Zx along the bond direction x in the perturbed molecular orbital.The factor f takes into account any local field enhancement or shielding effects. The frequency-dependent non-linear refractive index, n2,is expressed by the empirical formula:" 25f3fL31d2(n2-1)E,6n2(av.)/10-13 esu= n(Es2-h202)4 (1) where fL =(n2+ 2)/3 is the Lorentzian local-field enhancement fa~tor,~n is the long-wavelength limit value of the refractive index, E, in eV is the effective Sellmeier energy gap which takes in cccount the contributions of the sp and d orbitals, and Id inA is the bond length for the ternary glass. The Sellmeier gap is in practice equal to E, in Wemple's equation;12 values are listed in Table 1. The bond lengths for the ternary glasses were estimated using the method proposed by with the P,b-0, Ti-0 and Te-0 bond lengths as 2.24, 1.96 and 1.91 A, respectively." Also, we used the following equationg to convert n2 into x(~): J.Mater. Chem., 1996, 6(11), 1811-1814 1813 0 5 10 15 20 25 30 PbO content (mol%) Fig. 7 Third-order susceptibility x(3)as a function of the PbO content for the PTT glass system. The triangles represent the data from Yamamoto et aL6 and the circles are our calculated values. The estimated x(,) at the wavelength of 1.907 pm (the wavelength for the measured data of Yamamoto et aL6) are given in Table 1 and plotted in Fig. 7 (circles), for the PTT glasses. Note that the results are almost independent of the PbO content. To compare our results with the data of Yamamoto et uL6 (triangles) we have used f=1.72, we chose to make the values of x(,) equal to the experimental value for the PbO content of 25mol%.This f value is close to the value of 1.9 kO.1 obtained by Lines'' for transition-metal oxides dominated by d-band responses. This is an indication that Lines' model, which assumes that the high non-linear optical response is due to virtual excitations from the cationic sp levels to virtual d levels, is valid for Te0,-based glasses. Kim et uL3 have already shown that this model can be used for a TeO, pure glass, where the most probable optical transition is the electron transfer from 2p, orbitals of the oxygen to the empty non-bonding 5d orbitals of tellurium. We have also shown that this model holds for the Li20-Ti02-Te02 glass ~ystem.~ It should be pointed out that for the Li,O-Ti0,-Te02 glass system the refractive index and f3) decrease as the Te02 is substituted by the Li,O, owing to the change of TeO, trigonal bibyramids into TeO, trigonal pyramids.One would then expect that slower structure changes in the Pb0-Ti02-Te02 glasses would result in a slow decrease in the refractive index. Nonetheless, the measured values are almost independent of the PbO content. Thus, the presence of the PbO should increase the refractive index to compensate for the decrease due to the glass structure change. Conclusions We present highly homogeneous Pb0-Ti02-Te02 glasses with high refractive indices. The density, softening-point tempera- ture, glass-transition temperature, thermal expansion coefficient and Raman peaks were found to depend on the PbO content.In contrast, the UV-VIS cut-off, optical gap and the linear refractive index were found to be almost independent of the PbO content. These trends are consistent with a glass structure change, with TeO, trigonal bipyramids breaking into TeO, trigonal pyramids as the PbO content increases. The almost constant refractive indices may be explained by two opposing contributions. The PbO itself increases the refractive index; however, the increase in PbO content also increases the breaking of TeO, trigonal bipyramids into TeO, trigonal pyramids, which decreases the refractive index. Nonetheless, the x(3)values are larger than those of the pure TeO, glasses.The authors acknowledge the Conselho Nacional de Desenvolvimento Cientifico e Tecnologico (CNPq), Fundaqgo de Amparo a Pesquisa do Estado de S5o Paulo (FAPESP), Programa de Apoio a0 Desenvolvimento Cientifico e Tecnologico (PADCT) and the Telecomunicaqaes Brasileiras S/A (Telebras) for financial support. References 1 S. Inoue and A. Nukui, in Proc. Int. Conf. on Science and Technology of New Glasses, ed. S. Sakka and N. Soga, The Ceramic Society of Japan, Tokyo, 1991, p. 77. 2 H. Nasu, Y. Ibara and K. Kubodera, J. Non-Cryst. Solids, 1989, 110,229. 3 S-H. Kim, T. Yoko and S. Sakka, J. Am. Ceram. SOC., 1993, 76, 2486. 4 H. Nasu, 0. Matsushita, H. Kamiya, H. Kobayashi and K. Kubodera, J. Non-Cryst. Solids, 1990,124,275. 5 H. Nasu, 0. Matsushita, H. Kamiya, H. Kobayashi and K. Kubodera, Jpn. J. Appl. Phys., 1992,31, 3899. 6 H. Yamamoto, H. Nasu, J. Matsuoka and K. Kamiya, J. Non-Cryst. Solids, 1994, 170, 87. 7 R. F. Cuevas, L. C. Barbosa, A. M. de Paula, Y. Liu, V. C. S. Reynoso, 0.L. Alves, N. Aranha and C. L. Cesar, J. Non-Cryst. Solids, 1995,191, 107. 8 S. Khatir, F. Romain, J. Portier, S. Rossignol, B. Tanguy, J. J. Videau and S. J. Turrel, J. Mol. Strut., 1993,298, 13. 9 M. E. Lines, Phys. Rev. B, 1990,41,3372; 3383. 10 M. E. Lines, Phys. Rev. B, 1991,43,11978. 11 T. Sekiya, N. Mochida, A. Ohtsuka and M. Tonokawa, J. Non-Cryst. Solids, 1992, 144, 128. 12 S. H. Wemple, J. Chem. Phys., 1977,67,2151. 13 K. Nassau, Electron Lett., 1981, 17, 769; Bell. Syst. Tech. J., 1981, 60, 327. Paper 6/02829B; Received 23rd April, 1996 1814 J. Muter. Chem., 1996,6(11), 1811-1814
ISSN:0959-9428
DOI:10.1039/JM9960601811
出版商:RSC
年代:1996
数据来源: RSC
|
19. |
Comparison of X-ray patterns and Raman spectra ofn= 2 and 3 Aurivillius phases by principal component analysis |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1815-1821
Peter J. Klar,
Preview
|
PDF (993KB)
|
|
摘要:
Comparison of X-ray patterns and Raman spectra of n =2 and 3 Aurivillius phases by principal component analysis Peter J. Klar,*“ Limei Chen“ and Thomas Rentschlerb “School of Physics, University of East Anglia, Norwich, UK NR4 7TJ bInstitute of Inorganic and Applied Chemistry, University of Hamburg, Martin-Luther-King-Platz6, 20146 Hamburg, Germany Eight series of n=2 and 3 Aurivillius phases with structures derived from the original Bi,A, -‘BflO3,+ structure by fractional substitution in the (Bi,O,)’+ sheets as well as in the perovskite-type blocks (A,-1Bn03n+1)2-have been prepared by solid-state reaction. Raman spectra of all 45 samples are shown and the origin of the vibrational modes discussed. Plots of the first two principal components of principal component analysis (PCA) of the X-ray data and Raman data for all the samples as well as within the n =2 and 3 subsets are presented.The structural and compositional information contained in the Raman spectra and in the X-ray patterns is preserved in the PCA. Differences between the PCA results of the X-ray data and the Raman data are caused by the different sensitivities of the two techniques towards the various properties changed by the fractional substitutions. The availability of fast data acquisition and the quality of the PCA results suggest that a Raman microscope system in connection with a multivariate data analysis technique based on PCA can be a powerful tool for in situ quality control of devices based on these materials. X-Ray diffractometry and Raman spectroscopy are standard tools for material characterization. The origins of X-ray and Raman spectra are different, although both reveal the lattice symmetry of crystalline materials.X-Rays probe the atoms around their equilibrium positions and are therefore a measure of the static lattice. In contrast, Raman-scattered light is a probe of the lattice dynamics because of its sensitivity towards a subset of the lattice vibrations. X-Ray patterns of crystalline materials consist of a series of very sharp, well separated diffraction peaks, whereas in Raman spectra, signals of the different vibrational modes may extend over a wide wave- number range with adjacent features overlapping. The clearness of the X-ray patterns together with advanced theoretical modelling makes X-ray diffractometry the major tool for the determination of crystal structures.Raman spectra are usually not modelled in great detail; instead a group theoretical analysis based on the crystal structure (generally determined by X-ray diffractometry) predicts the number of Raman-active vibrational modes and their symmetry characters. If materials of different composition have the same symmetry, then the differences in the masses of the constituent compounds and the differences in the chemical bonding will still result in changes of the Raman shifts and variation of the signal intensities. Thus, a Raman spectrum is, as an X-ray pattern, a fingerprint of the crystalline structure of a material.The Aurivillius phases belong to the class of mixed bismuth oxide layered compounds and are of general composition Bi,A,-1Bn03n+3. They were first described in 1949.1-3 They consist of two structural elements, (Bi,O,)’+ sheets which are interleaved by perovskite-type blocks (A, -1Bn03n which+ increase in thickness as n increases. The research interest in these materials is quite considerable owing to their interesting physical properties such as their structural similarity to high- temperature superc~nductors,~~~ ion conductivity in oxygen- deficient compounds6-” and their outstanding ferroelectric propertie~.l’-’~Some of the compositions discussed here, e.g. SrBizTazOg and Bi4Ti3012, are used in commercially available ferroelectric memories.16-18 When a material is to be employed in commercial devices quality control becomes very important.It is desirable to have quick, non-destructive methods for material characterization which can be used even in situ during the production of the device. This requires fast data acquisition as well as fast data analysis. Optical characterization methods fulfil the first requirement and are widely applied in microelectronics manufacturing.” Raman spectroscopy is one of the most powerful optical methods for material characterization. Recent developments and improvements in basic spectroscopic instrumentation such as notch filters, charge-coupled device detectors and lasers allow the construction of low-cost Raman instruments with a high throughput and good sensitivity.” Raman microscopy makes Raman spectroscopy applicable to device structures. Similar developments have taken place in X-ray technology.In recent years area detectors which can be used in conjunction with laboratory X-ray sources have become commercially a~ailable.~’-~~These new detectors permit X-ray patterns to be evaluated directly after exposure without the need for time- consuming scanning operations. To fulfil the second requirement the analysis of the highly correlated Raman spectra or X-ray patterns needs to be automated. Multivariate analysis techniques are statistical techniques dealing with a computational reduction and extrac- tion of information from highly correlated data sets. One commonly used technique is principal component analysis (PCA).For example, to search a set of similar spectra for an ‘outsider’ is very difficult and time consuming by conventional techniques such as lineshape fitting. In particular, subtle fea- tures (which comprise the differences in similar spectra) such as shoulders are hard to fit without introducing additional fit parameters. In contrast, for a PCA no parameters apart from the experimental data are used. A set of spectra can be presented as two-dimensional graphs depicting clusters of points, where each point represents a full spectrum. Outlying points (spectra) can be easily recognised. The central mathemat- ical idea of principal component analysis is to reduce the dimensionality of a data set consisting of a large number of interrelated variables or coordinates, while retaining as much as possible of the variation present in the data set.This reduction is achieved by transforming to a new set of variables or coordinates the principal components (PCs), which are uncorrelated, and which are ordered so that the first few retain most of the information present in all of the original variables (for a review, see ref. 24). PCA is often applied as a preliminary J. Muter. Chem., 1996,6(11), 1815-1821 1815 to, or in conjunction with, other statistical techniques such as regression, discriminant analysis, cluster analysis and canonical correlation analysis. Examples of successful applications of PCA, sometimes in conjunction with other techniques, to Raman spectroscopic as well as X-ray diffraction data27*28can be found in the literature. In the following we will first discuss the Raman spectra obtained for the eight series of Aurivillius phases from a group theoretical point of view to relate the spectroscopic features to the underlying crystal structure.Then we will carry out a PCA of the X-ray data and the Raman data for all the samples as well as within the structurally similar groups consisting of n=2 and 3 Aurivillius phases respectively. This analysis will show how structural and compositional differences between the samples, which can be revealed by interpreting the two techniques in the conventional way, are preserved in the few principal components of the PCA.This is a major test for the applicability of Raman microscopy in conjunction with PCA as a tool for in situ characterisation of devices containing these materials. Experimental We have synthesised four series of n=2 and four series of n= 3 Aurivillius phases. The n=2 series are the two charge- compensated systems Biz -,Pb,Sr, -,LaxM209 and the two single Pb-substituted systems Bi2-,Pb,SrM20g (M =Ta, Nb), over the whole compositional range 0 <x<1. The n =3 series are all charge compensated and given by Bi4-,Pb, Ti3-,NbxO12 and Bi,-,Pb,La,Ti3-,Nb,012 for O<x< 1, Bi2Sr2-xPb,TiNb,012 and Bi,-,Pb,Sr, -,La,TiNb,O,, for Obx<2. The compositions of the samples prepared and the abbreviations used in the following are given in Table 1.The bulk ceramic samples were prepared from La203 (Merck, LAB and pre-annealed at 1000 K for 3 h), Bi203 (Merck, reinst), PbO, Sr(N03),, Ta205, TiO, (all Merck, p.a.) and Nb205 (Fluka, puriss.) by solid-state reaction. After decomposition of nitrates the samples were heated in alumina crucibles (Degussit A123) starting from 1120 K up to a maximum temperature of 1320K. The samples were annealed in air for 2 to 12 days, with several interruptions to regrind the powders and to perform X-ray diffractometry. The reactions were stopped after obtaining single phase materials or when no structural changes were observable by X-ray diffractometry. For the powder X-ray diffraction a Philips PW 1050 powder diffractometer with Cu-Ka radiation and SiO, standard was used.The spectra were taken in 28 ranges between 3 and 90" with an angular resolution of typically 0.02". The time to acquire an X-ray trace was of the order of 1h. The use of new area detectors should allow a much faster acquisition time for an X-ray pattern. The Raman spectra were acquired at room temperature using a modified Nikon Metallurgical Microscope with a Nikon MPlan Achromat ELWD40X objective. The excitation light was provided by the 514.5 nm line of a Spectra Physics Ar ion laser model 164. After filtering out the plasma lines with a premonochromator the laser light was coupled into the microscope and focussed onto the powder sample. The scat- tered light was directed via a beamsplitter and collecting optics into the spectrometer.The spectrometer was a modified Spex 1401 equipped with a nitrogen cooled ISA charge-coupled device detector. The Rayleigh-scattered light was suppressed by using a Kaiser Optical Systems holographic supernotch filter HSNF-514.5-1.5 with a spectral bandwidth of <350 cm-l. All the spectra were taken with the spectrometer centred at a Raman shift of 550cm-l, the detection window was about 1000cm-I wide and consisted of 578 data pairs. The spectral resolution was better than 2 cm-l. The laser power on the sample was less than 100mW and typical acquisition times were of the order of 30 s. The PCA was carried out using the software package Win- Discrim 3.1.,' The X-ray spectra were numerically reduced to 600 data points in the range 10-70" by averaging all data points over 0.1" intervals to obtain the same 28 coordinates for all spectra in order to use roughly the same number of data pairs as in the Raman spectra for the PCA.The area under each X-ray trace and under each Raman spectrum was normalised to unity before the PCA was carried out. X-Ray patterns and Raman spectra The X-ray results on the Aurivillius phases studied here are discussed in great detail in other publication^^'*^^ where Table 1 Compositions of the 21 n =2 Aurivillius phases and the 24 n n=2 Bi, -,Pb,SrTa20g Biz- ,Pb,Sr, -,La,Ta,O, Bi, -,Pb,SrNb20g Bi,-,Pb,Sr, -,La,Nb,O, X name" 0.0 A-1 0.2 A-2 0.4 A-3 0.6 A-4 0.8 A-5 1.o A-6 0.2 B-1 0.4 B-2 0.6 B-3 0.8 B-4 1.0 B-5 0.0 c-1 0.25 c-2 0.5 c-3 0.75 c-4 1.o C-5 0.2 D-1 0.4 D-2 0.6 D-3 0.8 D-4 1.o D-5 n=3 Bi, -,Pb,Ti, -xNb,O,, Bi, -,Pb,La,Ti, -,Nb,Ol2 Bi,Sr, -,Pb,Nb2TiO12 Bi,-,Pb,Sr, -,La,Nb2TiO12 =3 Aurivillius phases X name" 0.0 E-1 0.2 E-2 0.4 E-3 0.6 E-4 1.o E-5 0.0 F-1 0.2 F-2 0.4 F-3 0.6 F-4 0.8 F-5 1.o F-6 0.0 G-1 0.25 G-2 0.5 G-3 0.75 G-4 1.o G-5 1.25 G-6 1.5 G-7 1.75 G-8 0.25 H-1 0.5 H-2 0.75 H-3 1.0 H-4 1.5 H-5 "Abbreviations used in the figures and in the text.The samples are divided into eight groups defined by the fractional substitution series.The assignments of samples A-1, C-1 and G-1 to series A, C and G are arbitrary, they are also the x=O members of series B, D and H respectively. 1816 J. Muter. Chem., 1996, 6(11), 1815-1821 examples of the spectra are given. In summary, both charge- compensated n =2 material systems Bi,-,Pb,Sr, -xLaxM209 (M =Ta, Nb) are single phase, while for the Pb-substituted n =2 systems Bi,-,Pb,SrM20g (M =Ta, Nb) traces of a second, unidentified phase appeared with increasing fraction of substi- tution. All four n =3 Aurivillius phases Bi,-,Pb,Ti, -xNb,012, Bi, -,Pb,La,Ti, -,Nb,O,, and Bi,Sr, -,Pb,TiNb,Ol, are basi- cally single phase over the whole range of fractional substi- tution.Only in the sample G-8 Bi,Sro.2sPbl~,sTiNb2012 and in the system Bi2-,Pb,Sr2-,La,TiNb2012for x 2 0.75 was a small admixture of a second unidentified phase detected. The X-ray patterns of the powder samples can be indexed using an orthorhombic symmetry Fmmm for the Bi4 -,Pb,Ti, -,NbxO12 series or pseudo-tetragonal unit cells for all the other samples. The Raman spectra of the n =2 and 3 Aurivillius phases are .~shown in Fig. 1 and 2 respectively. Graves et ~ 1 calculated~ the Raman modes of Aurivillius phases for n= 1-3 assuming tetragonal I4/rnrnm symmetry (which corresponds to the sym- metry above the ferroelectric Curie transition temperature) 100 500 900100 500 900 Raman shiWcm-' Fig. 1 Raman spectra of the four series of n=2 Aurivillius phases.The spectra were taken at room temperature using 514.5 nm excitation. The abbreviations are explained in Table 1. '100 . 500 900100 ' 500 ' 900 Raman shiftkm-' Fig. 2 Raman spectra of the four series of n= 3 Aurivillius phases. The spectra were taken at room temperature using 514.5 nm excitation. The abbreviations are explained in Table 1. and no distortion of the M06 octahedra. This calculation leads to 12 Raman-active modes for n=2 and 16 Raman-active modes for n =3: 4A1, +2B,, +6E, for n =2 (1) 6A1, +2B1, +8E, for n =3 (2) Reducing the symmetry from tetragonal I4/mmm to ortho- rhombic Fmmm the doubly degenerate E, modes split into B2,+B3, modes, which are both Raman active, thus resulting in 18 Raman-active modes for n=2 and 24 Raman-active modes for n=3.Looking at the Raman spectra displayed in Fig. 1 and 2 the number of resolved features is far less than 18 or 24 even taking into account the existence of modes below 100 cm-1.33-35 The observed spectral features originate from overlapping Raman bands. It is difficult to resolve these features at room temperature, which is the most convenient operating temperature for a Raman system used for in situ quality control. Better resolution of the different Raman bands can be obtained at 77 K.36When dealing with powder samples the use of optical selection rules is very restricted. Attempts to reveal the character of the different bands by investigating single crystals have been made by Graves et al.and the polarisation results, together with the fact that the intragroup binding energy of the octahedron is much larger than the crystal binding energy, lead to the conclusion that the Raman spectra are dominated by vibrational modes related to the three Raman-active modes of an octahedron M06. These are commonly referred to as v,, the symmetric M-0 stretching of Al, symmetry; v2 of E, symmetry, where the octahedron is compressed along one axis and stretched in the plane perpen- dicular to it; and vs, an interbond angle bending vibration of symmetry F2,. These modes split owing to the lower site symmetry of the octahedron: oh D4h D2h The FZg-BZg-Blg mode splitting path is shown as a dashed line in our simple site-symmetry analysis because it is sup- pressed as a lattice vibrational mode as shown by the results of the factor-group analysis of Graves et al.The strong Raman bands in the range 150-350cm-' can be related to modes originating from vs. The weaker bands around 600 cm-I originate from v2, and the band between 800 and 900 cm-I is essentially the v, mode of the octahedron. These assignments are in good agreement with the results for other perovskite- type structures containing Ta and Nb.37 It is interesting that in the spectra of some samples, including series A and C for increasing x,the samples D-4 and D-5 in Fig. 1 and series H for x 30.75 in Fig. 2, a second peak emerges from the v, related peak. This splitting is not a lifting of mode degeneracy because the v1 related peak is of symmetry A,,.It is more likely to be related to the second unidentified phase in the material, which was detected by X-ray diffractometry for all these samples apart from samples D-4 and D-5. Results of the principal component analysis PCA is basically a coordinate transformation to a coordinate system which displays most of the differences in a set of data (here, a set of spectra) in the first few coordinates. A spectrum consisting of m intensity values (y, y,, ...)ym) taken at different positions can be regarded as a point in an m-dimensional J. Muter. Chern., 1996, 6(11), 1815-1821 1817 space, where the rn positions represent the rn orthonormal coordinate axes and where the rn intensity values yi (i= 1, 2, ..., rn) are the coordinates of the spectrum in this coordinate system.The origin of the coordinate system can be moved to the centre of the cloud of points given by all the spectra, defining a new orthonormal coordinate system with coordi- nates xi=yi-jji. The covariance matrix of the coordinates xi is defined as 1" 1" (4)k=l k=l where n is the number of spectra and k denotes the kth spectrum. The covariance matrix C, is a symmetric rn-dimen- sional matrix. In general the covariance matrix has rn positive eigenvalues, which can be labelled that A1 2A2 2...2Am for which orthonormal eigenvectors pl,p2, ..., pm can be derived. These eigenvectors define a new coordinate system. The rn-dimensional matrix with the x-coordinates of the eigenvectors of the covariance matrix C, as column vectors is the transform- ation matrix between the x-coordinate system and the p-coordinate system.The covariance matrix in the p-coordinate system Cpis diagonal with (Cp)ii=;li and the total variance of the data points is conserved. In general the first I coordinates in the p-coordinate system, I<<rn, contain most of the variance and, in other words, best represent the differences within the data set. To introduce some conventional terms, the axes of the p-coordinate system are called the principal components, the loading of the ith principal component is defined as the coordinates of the vector pi in the x-coordinate system multi- plied by the square root of its variance (Cp)ii,the projections of a spectrum onto the loadings are called the scores.38 It is worth mentioning here that definition of groups, which we use in the following to represent the results, does not affect the PCA results: in other words PCA is a non-discriminating technique.The main difference between the n=2 and 3 Aurivillius phases is the thickness of the perovskite layers: they contain two and three octahedra sheets respectively. This structural difference can easily be detected by analysing X-ray data in the conventional way. The spectra in Fig. 1 and 2 demonstrate that this can be difficult in the case of Raman results, although group theoretical considerations suggest that it is possible. In order to show that the PCA of X-ray results conserves these differences and to see how the PCA of the Raman results compares to the PCA of the X-ray data, we have carried out PCA in both cases, taking spectra of all 45 samples into account.In both analyses the distribution of the relative variance on the first few PCs was very similar, about 55% for PC 1 and 20% for PC 2, which indicates that the data are well presented by the plots of the scores of PC 1 us. the scores of PC 2 in Fig. 3 for the X-ray data and in Fig. 4 for the Raman data. In both cases we obtain a full separation of the two groups. The assumption that most of the separation between the two groups originates from the fact that for the n=3 Aurivillius phases a high percentage of the central ions of the MOB octahedra are M=Ti, in contrast to the n=2 Aurivillius phases where only M =Ta or Nb are incorporated, can be ruled out in the following way.If the assumption were right, the effect should be most significant in the PCA of the Raman spectra, because these are very sensitive to the nature of the central ion. One would expect the PCA data for the n=2 group to split into two subgroups given by the M=Ta samples and the M=Nb samples, but this is not the case in Fig. 4. Consequently the separation between the n=2 and 3 Aurivillius phases in the PCA of the X-ray data and the Raman data is due to the different crystal structures. This demonstrates that structural differences, which are detectable in the conven- tional analysis of the X-ray data and which are expected for the Raman data from group theoretical considerations, are 1818 J.Muter. Chern., 1996,6(ll), 1815-1821 0.0501 I -I 0 0.0Z5 loF I $1 r 03 Iw-0 I8 -0.025 v) 1-0.050 0 a a 0 0 0 n=3 n=2 I -0.05 0.00 0.05 0.10 score of PC1 Fig. 3 Plot of the first us. the second principal component (PC) of the principal component analysis of the X-ray data of all the Aurivillius phases (21 different n=2 samples, 24 different n=3 samples). The X-ray traces used in the analysis were normalised per unit area and consisted of 600 equidistant points in the 28 range 10-70". O.O2*i0.01 I 0 01 0 -0.01 Fig. 4 Plot of the first us. the second principal component (PC) of the principal component analysis of the Raman data of all the Aurivillius phases (21 different n=2 samples, 24 different n=3 samples).The Raman spectra used in the analysis were normalised per unit area and consisted of 578 points covering the Raman shifts between 100 and 1000 cm-'. preserved in the PCA. In both figures the n=3 group seems to spread wider than the n =2 groups, which can be accounted for by the fact that the compositional differences between the four series comprising the n=3 group are bigger than the compositional differences between the four n =2 series. The structures within the n =2 and 3 groups will be discussed in the following. For this purpose we have carried out a PCA for each of the four sets of spectra. Fig. 5-8 show the plots of the first us.the second principal component of the respective PCA of Raman and X-ray data for n =2 and 3. The formation of the n =2 and 3 clusters will differ from the respective clusters in Fig. 3 and 4, where n=2 and 3 were analysed together, because now n=2 data are not considered in the PCA of the n= 3 data and vice versa. Fig. 5 displays the results of the PCA of the X-ray patterns of the 21 n =2 Aurivillius phases investi- gated here. 50% of the total variance within the data set fell on the first PC and about 20% on the second PC. All four series A, By C and D show characteristic traces as a function of composition. Two subgroups reveal similar behaviour, the two Pb-substituted series A and C are stretched out, the two charge-compensated series describe a curve.The A- 1 sample Bi2SrTa09 is the x=O member of the A and the B series, the spacial closeness to both series indicates this. Similar behaviour is expected for the C-1 sample Bi2SrNb09, which is the x=O 8 C-1 D c-2 rn N 0.02 c-3 rn2 Y--0.02ii -0.04 -0.02 0.00 0.02 0.04 score of PC1 Fig.5 Plot of the first us. the second principal component of the principal component analysis of the X-ray data of the 21 n=2 Aurivillius phases. The X-ray traces used in the analysis were normalised per unit area and consisted of 600 equidistant points in the 26 range 10-70". The abbreviations are explained in Table 1. 0.01t aD-1 DC-1 0.06 I I I I I I 0.04{ 0.02+ I-. -G-3 2-K 1 -0*06 B G8t t -0.08' I I I I I -0.08 -0.04 0.00 0.04 0.08 score of PC1 Fig.7 Plot of the first us.the second principal component of the principal component analysis of the X-ray data of the 24 n=3 Aurivillius phases. The X-ray traces used in the analysis were normalised per unit area and consisted of 600 equidistant points in the 28 range 10-70". The abbreviations are explained in Table 1. OF-' F-2 -0 p! -0.005 H-3A 0F-5 E-10 v) OF-6 8 t H4A -0.010 H-511 -0.01 0.00 0.01 score of PC1 ,,-0.015 ( 2 -0.02 -0.01 0.00 0.00 0.02 score of PCl Fig.8 Plot of the first us. the second principal component of the principal component analysis of the Raman data of the 24 n=3 Aurivillius phases. The Raman spectra used in the analysis were normalised per unit area and consisted of 578 points covering the Raman shifts between 100 and lOOOcm-'.The abbreviations are explained in Table 1. clear traces as a function of composition along the first PC, but cannot be separated in a plot PC 1 us. PC 2. As in Fig. 5 sample C-1 is far away from the D and C series compared to the internal distances. A possible reason for this behaviour of the x=O members of the four series might be the sensitivity of the X-ray diffractometry and Raman spectroscopy to a symmetry reduction introduced by compositional changes, but to establish this unambiguously on the basis of PCA results would require more thorough analysis than is attempted here. Fig. 7 and 8 deal with the PCA of the X-ray patterns and the Raman spectra of the 24 n =3 Aurivillius phases respect-ively.In the case of the X-ray data 60% of the total variance accumulates on the first PC and 15% on the second PC. In the Raman case 75% of the variance is found in the first PC and about 10Y0in the second PC. Again, in the Raman case more of the total variance is accumulated on the first two PCs than in the X-ray analysis. In both Figures, the PCA data arrange in the same two subgroups, with the E and F series corresponding to the Ti3-xNb,012 samples and the G and H series corresponding to the TiNb2012samples. This result is in contrast to the result for the n=2 Aurivillius phases, where the data are grouped differently for Raman and X-ray, a fact J. Mater.Chem., 1996, 6(11), 1815-1821 1819 Fig.6 Plot of the first us. the second principal component of the principal component analysis of the Raman data of the 21 n=2 Aurivillius phases. The Raman spectra used in the analysis were normalised per unit area and consisted of 578 points covering the Raman shifts between 100 and 1000 cm-'. The abbreviations are explained in Table 1. member of the C and the D series, but lies much further out compared to the internal distances between the other members of the two groups. Fig. 6 shows the results of the PCA of the Raman spectra within the YE =2 Aurivilliusphases corresponding to the analysis shown in Fig. 5. Here 55% of the total variance was within the first PC and 25% within the second PC. In the Raman case more of the total variance is accumulated on the first two PCs than in the X-ray analysis.The PCA data of the four series group in a different way than in Fig. 5. Again we obtain two subgroups, but this time they consist on the one hand of the Ta-containing samples series A and B and on the other hand of the two Nb-containing sample series C and D. This can be understood by considering the dominance of the octahedron-related features in the Raman spectra compared to the X-ray data; the latter are more sensitive to the structural differences such as the second unidentified phase, which appears with increasing x composition in the A and C series. Series A and B can be well separated by the PCA, B shows a well defined trace as a function of composition, whereas series A forms more of a cloud.A-1, which is also the x=O member of the B series, is far away from the B series compared to the internal distances within this series. The C and D series show that has already been explained in terms of the stronger sensitivity of X-ray diffractometry to the structural changes introduced by the second unidentified phase than to the octahedra-related features. In the n =3 case none of the series of samples reveal a continuous development of a second phase, so that this aspect does not play a role in the subgroup arrangement of the X-ray PCA data. However, the second unidentified phase might influence the traces of the four series displayed in Fig. 7. Series E and F, which are structurally pure in the sense that they consist of only one phase, are aligned well showing an increase of the score of PC 1 with increasing degree of substitution, x,whereas series G and H show some peculiarities in the region where the second phase occurs.Series H bends sharply after H-3 and G-8 is a long distance away from G-7. The traces of the four series displayed in Fig. 8 for the PCA of the Raman data are well separated. The peculiarities in the traces discussed for Fig. 7 do not occur in Fig. 8. Here G-8 is close to the rest of the G series and there is no significant kink in the H-series after H-3. Remarkable is the long distance between E-1 Bi4Ti3012 and the rest of the E series, which, we can speculate, displays the strong change in the Raman spectra due to the reduction of the symmetry as the composition changes. This is not very conclusive because the same be- haviour does not happen for F-1 Bi,La,Ti,012 and G-1 Bi,Sr,Nb2TiO12.On the other hand these already consist of more different compounds than E-1, so that the compositional change might not be so significant. Conclusions We have prepared eight series of new Aurivillius phases by solid-state reaction. These new structures are obtained by fractional substitutions in the original Bi,A,- 1Bn03,+3 struc- ture leading to substitutions in the (Bi,02j2+ sheets as well as the perovskite-type blocks (An- 1Bn03n+ j2-. These are four n =2 series Bi,- ,Pb,Srl -,La,M209 and Bi, -,Pb,SrM,O, for M=Ta, Nb and O<x<l, and four n=3 series Bi, -,Pb,Ti3 -,NbxO12 and Biz-,Pb,La2Ti, -,Nb,012 for 0 <x<1, Bi2Sr2-,Pb,TiNb2012 and Bi, -,Pb,Sr, -,La, TiNb2012 for OGxG2.X-Ray and Raman experiments have been performed. The X-ray results are discussed in detail el~ewhere.~~-~~The conventional analysis gives the full struc- tural information about the samples: the analysis of the X-ray diffraction patterns revealed an orthorhombic distortion Fmmm of the system Bi,~,Pb,Ti3~xNbxO12 while all the other series can be indexed applying the pseudo-tetragonal symmetry I4/mmm. Based on this result we have discussed the Raman spectra by group theoretical methods. Taking chemical bond- ing into consideration, the origins of the dominant spectral features have been related to the three Raman-active modes of the MO, octahedra.The Raman microscope set-up in connection with multichan- nel detection technique allowed fast data acquisition at about 30 s per spectrum. This suggests that Raman spectroscopy can be applied in situ for quality control in the production of devices containing these materials (e.g. ferroelectric memories containing Bi,SrTa,Og or Bi4Ti3OI2). Fast data analysis as well as fast data acquisition is required in in situ quality control. A method commonly used for such tasks, alone or in conjunction with other multivariate techniques, is principal component analysis (PCA). We carried out PCA of the X-ray data as well as of the Raman data: PCA of all the spectra gave a full separation of the n=2 and 3 Aurivillius phases with perovskite layers consisting of two and three sheets of octahedra respectively.This shows that differences in the crystal structure, which can be easily detected in conventional analysis of X-ray data and are expected from the group theoretical discussion of the vibrational Raman modes, are preserved in the plot of the first principal component us. the 1820 J. Mater. Chem., 1996, 6(11j, 1815-1821 second principal component of the PCA of the data. To investigate the sensitivity of PCA to compositional changes we have carried out PCA within the n=3 and 2 Aurivillius phases for both Raman and X-ray results and have analysed the plots of the first principal component us. the second principal component.In particular, in the n =2 case there were differences between the results of the analysis of the X-ray data compared to the analysis of the Raman data. These occur because Raman spectroscopy and X-ray diffractometry probe different symmetry-influenced properties of the specimen (the former the polarizability changing lattice vibrations, the latter the electron density) and therefore exhibit varying sensitivity towards the various changes resulting from the fractional substitution. The grouping of the Raman data reveals sub- groups according to the different central ions in the MO, octahedra: these structural features dominate the Raman spec- tra. In contrast, the X-ray data group in such a way as to distinguish between structurally pure samples on the one hand and samples with traces of a second phase of unidentified structure on the other.In all analyses the series show distinct traces as a function of fractional substitution for X-ray and Raman data. We did not have enough different samples within the eight series to investigate the effects of fractional substi- tution on the PCA of one individual series, but, nevertheless, one can anticipate that some of the internal distances and some of the pecularities in the traces of the series display the sensitivity of Raman and X-ray towards some of the various changes introduced by increasing the fraction of substitution. For example, the x=O end members such as Bi4Ti3012 in the Bi,-,Pb,Ti, -xNbx012 series are of higher symmetry than the rest of the series with xfO, shown in the Raman PCA by the significant difference between the x =0 sample and the rest of the series when comparing internal distances.For the series Bi, -,Pb,Sr, -,La,TiNb201, conventional X-ray analysis reveals the occurrence of a second unidentified phase for x >0.75. At this substitution fraction, a rapid change of direc- tion of the trace of this series in the PCA of the X-ray data is observed. In conclusion, differences in the crystal structure and in the composition of the Aurivillius phases, which can be studied and explained very accurately by a conventional X-ray analysis, are preserved in the first few principal components in the PCA of X-ray data as well as of Raman data.The quality of the Raman results suggests that a Raman microscope system in connection with a multivariate data analysis based on PCA can be a powerful tool for in situ quality control of devices based on these materials. P. J. K. and L. C. thank the University of East Anglia and the Norwich Research Park respectively for the provision of their research studentships and financial support. T. R. is grateful to the 'Fonds der Chemischen Industrie' for financial sup- port. We also thank Prof. A. Reller, Prof. J. J. Davies, Dr. D. Wolverson and Dr. E. K. Kemsley for fruitful discussions and careful reading of the manuscript. References B. Aurivillius, Arkiv. Kemi., 1949, 1,463. B. Aurivillius, Arkiv. Kemi., 1949, 1,499. B.Aurivillius, Arkiv. Kemi., 1950, 2, 519. K. A, Yee, T. A, Albright, D. Jung and M.-H. Whangbo, Angew. Chem., 1989,101,789. 5 W. Zhou, Adv. Mater., 1994,2,94. 6 K. V. R. Prasad and K. B. R. Varma, Mater. Chem. Phys., 1994, 38,406. 7 A. Q. Pham, M. Puri, J. F. Dicarlo and A. J. Jacobsen, Solid State Ionics, 1994,72, 309. 8 K. R. Kendall, J. K. Thomas and H.-C. zur Loye, Chem. Mater., 1995,7, 50. 9 J. K. Thomas, K. R. Kendall and H.-C.zur Loye, Solid State Ionics, 1994,70171,225. 10 S. Lazure, R. Vannier, G. Nowogrocki, G. Mairesse, C, Muller, 26 F. Cadet, D. Bertrand, P. Robert, J. Maillot, J. Dieudonne and 11 M. Anne and P. Strobel, J. Muter. Chem., 1995,5,1395. R. W. Wolfe and R. E. Newnham, J. Electrochem. SOC., 1969, 116, 27 C.Rouch, Appl. Spectrosc., 1991,45,166. M. Nofz, B. Himmel, Th. Gerber and F. Ehrentreich, J. Non-Cryst. 832. Solids, 1992,143, 191. 12 L. Korzunova, Ferroelectrics, 1992,134, 175. 28 M. E. E. Harju, P. Minkinnen and J. Valkonen, Chemometrics 13 14 15 E. C. Subbarao, J. Phys. Chem. Solids, 1962,23,665. S. E. Cummins and L. E. Cross, J. Appl. Phys., 1968,39,2268. E. G. Fresenko, A. T. Shuvaev, V. G. Smotrakov, G. A. Geguzina, V. D. Komarov, V. G. Gavrilyachenko and E. S. Gagarina, Inorg. 29 30 31 Intelligent Laboratory Systems, 1994,23, 341. E. K. Kemsley, Win-Discrim 3.1., Norwich, 1996. T. Rentschler, Muter. Res. Bull., submitted for publication. T. Rentschler, M. Karus, A. Wellm and A. Reller, Solid State Ionics, 16 Muter., 1977,30,977. J. F. Scott, Physics World, 1995,8,46. 32 in the press. P. R. Graves, G. Hua, S. Myhra and J. G. Thompson, J. Solid State 17 18 C. A. Paz de Araujo, J. D. Cuchiaro, I. D. McMillan, M. C. Scott and J. F. Scott, Nature (London), 1995,374,627. T. Nakamura, R. Muhammet, M. Shimizu and T. Shiosaki, 33 34 Chem., 1995,114,112. K. Hisano and K. Toda, Solid State Commun., 1976, 18, 585. J. Liu, G. Zou, H. Yang and Q. Cui, Solid State Commun., 1994, 19 Integrated Ferroelectrics, 1995,6, 35. S. Perkowitz, D. G. Seiler and W. M. Duncan, J. Res. Natl. Inst. 35 90, 365. S. Kojima,R. Imaizumi, S. Hamazaki and M. Takashige, Jpn. Stand. Technol., 1994,99,605. J. Appl. Phys., 1994,33, 5559. 20 21 22 23 24 B. Chase, Appl. Spectrosc., 1994,48, 14A. T. Gao, Y. J. Li, J. Rousseau, K. Linliu and B. Chu, Rev. Sci. Instrum., 1993,66,394. H. Goldner, R&D Magazine, 1994,36,63. J. C. Phillips, J. Appl. Crystallogr., 1994,27, 853. I. T. Joliffe, Principal Component Analysis, Springer-Verlag, New 36 37 38 H. Idink, V. Srikanth, W. B. White and E. C. Subbarao, J. Appl. Phys., 1994,76, 1819. V. V. Fomichev, Russ. Chem. Bull., 1994,43,1943. D. L. Massart, B. G. M. Vandeginste, S. N. Deming, Y. Michotte and L. Kaufman, Chemometrics: a textbook, Elsevier, Amsterdam, 1988. 25 York, 1986. C. A. Drumm and M. D. Morris, Appl. Spectrosc., 1995,49, 1331. Paper 6/03929D; Received 5th June, 1996 J. Muter. Chern., 1996,6(11), 1815-1821 1821
ISSN:0959-9428
DOI:10.1039/JM9960601815
出版商:RSC
年代:1996
数据来源: RSC
|
20. |
A neutron diffraction study of structural distortions in the Ruddlesden–Popper phase Na2La2Ti3O10 |
|
Journal of Materials Chemistry,
Volume 6,
Issue 11,
1996,
Page 1823-1825
A. J. Wright,
Preview
|
PDF (1144KB)
|
|
摘要:
A neutron diffraction study of structural distortions in the Ruddlesden-Popper phase Na2La2Ti,010 A. J. Wrightt and Colin Greaves" School of Chemistry, University of Birmingham, Birmingham, UK B15 2TT The structure of Na,La,Ti3010 (tetragonal, space group I4/mmm) has been re-examined using time-of-flight powder neutron diffraction in order to clarify uncertainties remaining from an X-ray diffraction study. One of the oxygen atoms is shown to be displaced significantly from its ideal position, the shift being consistent with the rotation of one of the TiO, octahedra by ca. 12.5" around the [OOl] direction. This rotation allows for a 2.5% increase in four of the Ti-0 bond lengths and a corresponding decrease in the bond valence sum from 4.9 to 4.1. The structural distortion is similar to that observed previously at overbonded octahedral sites in related structures.The Ruddlesden-Popper series of materials have the general formula M,(A,- 1Bn03n+1) and have layered structures related to perovskite. Their structures, first reported for the Srn+lTif103fl+lseries by Ruddlesden and Popper in the late 1950s,' contain perovskite blocks of corner-linked BO, octahedra (usually B =Ti or Nb) with A atoms in conventional twelve-coordinate sites. These blocks are separated by two adjacent MO layers with the rocksalt structure. The Dion-Jacobson phases, M(A,- 1Bn03n+ 1), have related structures,' which contain only half the M ions in the link between the perovskite slabs. Dion-Jacobson and Ruddlesden- Popper phases have been found to have a wide range of potential uses, such as ion-exchange materials3 and ionic conductor^.^ The wide and varied intercalation chemistry possible between the perovskite slabs in these materials has attracted significant attention and prompted extensive research5 into the structures and properties of these materials.Recently, the synthesis and structure of Na2La2Ti3010 have been reported by Toda et aL6 The structure, determined by Rietveld refinement using X-ray powder diffraction data, was reported to be of the Ruddlesden-Popper type with perovskite blocks containing three layers of corner-linked TiO, octahedra (see Fig. 1 and Table 1). However, certain features of the reported structure are' anomalous.In particular, a number of the Ti-0 bond lengths are unusual, since bond valence sum (BVS) calculations7 suggest Ti valences significantly greater than four for the two Ti sites within the structure: 4.87 for Ti(l),which forms the central layer of octahedra, and 4.21 for Ti(2) (see Fig. 1). In addition, the four equivalent equator- ial oxygens around Ti( 1) haveo an exceptionally high iso- tropic thermal parameter (B=8 A,). In a recent study of the related tripled perovskite phases NdBa2Cu,TaOs8 and (Nd,Ce)zSrzTaCu2010,9 rotations of the TaO, octahedra were observed and were assumed to occur in order to relieve 'overbonding' at the Ta sites which, in an undistorted structure would also have had high BVS values (ca. 6). It seemed plausible that similar effects may be present in Na,La,Ti3010 to optimise the Ti environments and a neutron diffraction study has therefore been undertaken in order to maximise the precision of the oxygen positions.It was anticipated that this would clarify the structure, especially with respect to possible rotations of the TiO, octahedra, and provide a guide as to whether such rotations may be ordered. Experimental High-purity Na,C03, La20, and Ti02 were used to prepare Na2La2Ti3010 by solid-state reaction following the method described by Toda et al., To compensate for the partial loss of sodium through volatilisation, a 30% molar excess was added to the stoichiometric mixture. The reactants were ground intimately and fired initially at 550°C for 12 h, with a sub- sequent heating at 1100°C for 6 h.The product was then washed with distilled water to remove any remaining traces of Na2C03, and then dried at 100 "C for 24 h. Initial characterisation was achieved using X-ray powder diffraction (Siemens D5000). Time-of-flight neutron powder diffraction data were collected using the POLARIS diffrac- tometer at the Rutherford Appleton Laboratory. The structural t Present address: Department of Chemistry, University of Cambridge, Fig. 1 Idealised unit cell of Na2La2Ti,010 (the smaller spheres indicate Lensfield Road, Cambridge, UK CB2 1EW. Na, the larger indicate La and the octahedra comprise TiOG units) J. Mater. Chem., 1996,6( ll), 1823-1825 1823 Table 1 Structural parameters for Na,La,Ti,O,, (Toda et aL6) unit cell atom position x y Z B,~~/A~occupancy 4e 4e 2a 0 0 0 0 0 0 0.2895( 9) 0.4246( 1) 0 0.5 0.06(11) 0.1 4e 4c 4e 0 o 0 0 + 0 0.1491 (4) 0 0.065(1) 0.1 0.58(2) 8g4e 0 0 3 0 0.137( 1) 0.210( 1) 0.4( 6) 0.8(8) u =3.83528( 7) A, c=28.5737( 7) A.refinements were achieved using the multi-phase program FORTY 1 based on the Cambridge Crystallographic Subroutine Library." Scattering lengths adopted for La, Ti, Na and 0 were 0.827, -0.3438, 0.363 and 0.5805 (all x cm). Results and Discussion The synthesis of Na2La2Ti,010 was achieved, yielding a prod- uct with purity >go%; the main impurity was identified as the Ti" perovskite La0.66Ti03 -x. The impurity is consistent with the loss of sodium and is difficult to avoid. Fortunately, the La:Ti ratio is identical in these two phases so that no additional impurities were observed.Structure refinement involved parameters relating to both the main phase and the impurity. La0.66Ti03 -x was found to be {escribed adequately by a simple perovskite unit cell (a =3.889 A) and refinement of La and 0 site occupancies revealed no significant deviation from the ideal formula La0.66Ti03. These site occupancies were subsequently fixed accordingly. As expected, the O(1) site [ideal position 4c (0,1/2,0)] provided the major problem in refining the structure of the primary phase. An acceptable temperature factor was obtained only by allowing relaxation off the (100) mirror plane to two half-occupied sites [8j, cu.(0.1,1/2,0)]. The apical O(2) and O(4) sites, and the Na and La sites, all with ideal position 4e (O,O,z), were also allowed to move off the (001) axis to 16n sites in order to reduce temperature factors, but the shifts were much smaller than those observed for O(1). Refined parameters for Na2La2Ti3010 are listed in Table 2, with the profile shown in Fig. 2. The structural interest in this phase centres on the orien- tation of the TiO, octahedra (see Fig. 3). Rotation of the octahedra centred on the Ti( 1)atom about the c axis (ca. 12.5') produces a 2.5% expansion of the Ti(1)-O(1) bond. This, coupled with a minor tilting of these octahedra, suggested by the displacements in O(2) and 0(4), produces bond lengths more in keeping with a Ti valence of four.This is borne out by BVS calculations which indicate a Ti( 1) valence of 4.13, as compared to 4.87 calculated from a structure with no rotations or tilting of the octahedra [reported from X-ray diffraction Table 2 Refined parameters for Na,La,Ti,O,, (this study) unit cell atom position x y z B~~~/A~occupancy Na 16n 0.047( 5) 0.2911(3) 0.65(25) La 16n 0.020( 5) 0.4255( 1) 0.48 (9) a-Ti(1) 2a 0 0 Ti(2) 4e 0 0.1496(2) -' O(1) 8j 0.111( 1) 0 0.53(8) O(2) 16n 0.027( 7) 0.0672(2) 1.07(20) O(3) 8g 0 0.1344(1) 0.70(4) O(4) 16n 0.028( 7) 0.2090(2) 1.01( 17) "Anistropic temperature factors (in A')): Ti( 1): B,, =O.31( 19); B,,= 0.31( 19); B,,= 3.34(55). 0Ti(2): B,, =0.02(!0); B,,=0.02( 10); B,,= 0.39(21).a= 3.84397( 3) A, c =28.5882(4) A; 14/mmm; Rpro =5.40%, R,,=5.67%, RE= 1.90%. 1824 J. Muter. Chem., 1996,6(11), 1823-1825 45 I ~~I~UIIIIIIIIIIIIII~~III~III~IIIIII II II 111 I I II I I 1111II1111IIIIIII I I I I I I I I I I 4 254 05 20-als 15-0& 10-.-v) 5-0--441-10 : I I 0.6 1.1 1.6 2.1 d spacing/A Fig. 2 Observed (+), calculated and difference neutron powder diffraction profiles of Na,La,Ti,O,o (upper ticks) and La,.,,TiO, (lower ticks) Fig. 3 Cry$al structure around the Ti atoms in Na,La,Ti,O,, (bond lengths in A). Ti( 1) is linked to four equatorial O(1) and two apical O(2); Ti(2) is linked to four equatorial O(3) and apical 0(2), bridging to Ti( l), and O(4). study by Toda et a/., (see Table 3)].This would then appear to provide the driving force behind these rotations, just as with the rotations of the TaO, in (Nd,Ce)2Sr2Cu2TaOlo.9 The neutron diffraction study provides no evidence for similar rotation of the Ti(2)06 octahedra about the c axis, but it does indicate that the O(3) atoms (equatorial positions in these octahedra) are positioned out of the (001) plane contain- ing the Ti. This may be a consequence of the local electric field along c experienced by the Ti(2) and O(3) sites which would lie in a plane perpendicular to c in the ideal structure. Either side of these planes are (Na0)- and (Lao)' layers which would result in an electrostatic attraction of O(3) Fig. 4 A comparison of the structures reported by Toda et a1.,6 structure a, and that of the present study, structure b towards the (Lao) layer and Ti(2) towards the (NaO) layer.This is exactly what is observed and it results in longer Ti(2)-O( 3) bond lengths, which are more consistent with Ti"; the bond distances yield a BVS of 4.21. Just as with Ti(l), the Ti(2) octahedra are probably slightly tilted, leading to a further marginal increase in bond lengths. It should also Table 3 A comparison of selected bond distances/A for Na,La,Ti,Olo Toda et aL6 present study Ti( 1)-O( 1) 1.918(0) Ti(1)-O( 1) 1.969(8) Ti(1)-0(2) 1.848(35) Ti(1)-0(2) 1.924(7) Ti( 2)-O( 2) 2.412( 37) Ti( 2)-O( 2) 2.358( 9) Ti(2)-O( 3) 1.951( 5) Ti(2)-O( 3) 1.971( 8) Ti(2)-0(4) 1.734(38) Ti(2)-0(4) 1.702(8) be noted that the La and Na ions are displaced slightly from their ideal positions, the shifts being consistent with the proposed tilting of the TiO, octahedra.No evidence was found for any oxygen vacancies in this material. No superstructure reflections which would be indicative of long-range ordering of these rotations/tilts were apparent in the neutron diffraction profile. However, this does not preclude the possible existence of short-range order, and this will be investigated using electron microscopy. We are grateful to the EPSRC for providing a studentship (to A.J.W.) and the use of neutron diffraction facilities. We are grateful to R. Smith for assistance in the collection of neutron diffraction data. References 1 S. N. Ruddlesden and P. Popper, Acta Crystallogr., 1958,11,54. 2 M. Dion, M. Ganne and M. Tournoux, Muter. Res. Bull., 1981, 16, 1429. 3 J. Gopalakrishnan and V. Bhat, Inorg. Chem., 1987,26,4299. 4 V. Thangadurai, A. K. Shukla and J. Gopalakrishnan, Solid State Ionics, 1994,73,9. 5 S. Hardin, D. Hay, M. Millikan, J. V. Sanders and J. W. Turney, Chem. Muter., 1991,3,977. 6 K. Toda, Y. Kameo, M. Fujimoto and M. Sato, J. Ceram.SOC.Jpn., 1994,102,737. 7 I. D. Brown and D. Altermatt, Acta Crystallogr., Sect. B, 1985, 41,244. 8 C. Greaves and P. R. Slater, Physica C, 1989,161,245. 9 A. J. Wright, R. A, Jennings and C. Greaves, Supercond. Sci. Technol., 1993,6, 514. 10 J. C. Matthewson, P. Thompson and P. J. Brown, J. Appl. Crystallogr., 1982, 15, 167. Paper 6/03395D; Received 15th May, 1996 J. Mater. Chem., 1996, 6(11), 1823-1825 1825
ISSN:0959-9428
DOI:10.1039/JM9960601823
出版商:RSC
年代:1996
数据来源: RSC
|
|