J. CHEM. SOC. FARADAY TRANS., 1994, 90(21), 3261-3265 -EPR Study of NaCl :CO,' and NaCl :SO, Peter D. Moens,' Sabine E. Van Doorslaer,t Freddy J. Callens,$ Filiep R. Maes and Paul F. Matthys Laboratory of Crystallography and Study of the Solid State, Krijgslaan 28lSl,8-9000Gent, Belgium Johan M. D'heer Laboratory for General and Inorganic Chemistry, Krijgslaan 28 1-S3,8-9000Gent, Belgium Two new triatomic molecular ions with C,, symmetry i.e.C0,-and SO,-have been detected in NaCl by means of EPR. By analysing the 13C hyperfine tensor of the C0,-radical the spin densities in the 2s and 2p, carbon atomic orbitals can be calculated. Since the 14N hyperfine tensor of the isoelectronic NO, molecule in the same host lattice is also known, a comparison between the spin densities in both defects can be made.In addition, the spin densities for both defects are calculated using a6 initio methods and are compared with the experimental results. As is the case for the smaller C0,-ion, the SO,-radical is assumed to substitute for only one halide ion. The monovacancy model for the SO,-ion is discussed. Such a model is in accordance with that for the smaller ozonide 0,-ion. On the other hand, the larger S3-and Se,- ions occupy a trivacancy site. Owing to their inherent high degree of symmetry, alkali- metal halide single crystals are very interesting host matrices for the study of atomic and molecular impurities. Once the species are trapped in the alkali-metal halide lattice, the elec- tronic structure and the orientation of the impurity can be studied with EPR and/or electron nuclear double resonance (ENDOR) to yield a wealth of information for both theo- reticians and experimentalists.This paper reports the EPR spectra of two related triatomic molecules, C0,- and SO,-, with C,,symmetry, trapped in NaCl. The C0,- ion has already been studied extensively in a wide variety of host lattices.'-* However, it has only very recently been detected in alkali-metal halide single crystals, specifically in KC19 and KBr." In these matrices, the defect exhibits axial symmetry around a (111) axis and hence is denoted as a C0,- { 11 l} species. It was assumed that the C0,-ion substitutes for a single halide ion. The spin den- sities in the carbon atomic orbitals of the CO, -radical could be obtained by analysing the 13C hyperfine tensor. A com-parison with the isoelectronic NO, molecule in the same host lattices and in the same configuration (both (111) species) could be made, yielding results in contradiction with theoreti- cal deductions made by Atkins et ul." The present experi- ments were carried out in order to check whether the C0,- ion can be incorporated in the NaCl lattice and also to deter- mine whether the comparison with the NO, molecule in NaCl yields results that contradict the theory.Theoretical calculations were performed to test the hypothesis of Atkins et ul." regarding the spin distribution in NO, and CO,-. Recently, there has been interest in the C0,- ion among EPR spectroscopists because it serves as a possible candidate for a low-dose (doses < 10 Gy) EPR do~imeter.~~*'~ Both the C0,- and the SO,-ions are triatomic molecules possessing C,, symmetry.In the alkali-metal halides, the SO,-ion has also already been detected in single crystals of KCl and KBr.14*15 The monovacancy model for the larger SO,-ion, proposed by these authors, will be tested using the data presented in this paper. As the 33S hyperfine tensor could not be determined, the spin density of the SO,-ion cannot be discussed. Research Assistant of the N.F.S.R. (Belgium).3 Senior Research Associate of the N.F.S.R. (Belgium). Experimental Materials NaCl single crystals were grown using the Bridgman tech- nique. The quartz growth capsule contained a graphite cylindro-conical crucible, filled with, typically 15 g NaCl (Merck Suprapur Powder).The NaCl powder was vacuum dried at 200 "C for one week using a two-stage rotation pump (ALCATEL 2012 AC, chemical series). Approximately 0.7 wt.% Na, l3Co3 was added with a I3C enrichment of 99% (MSD Isotopes, Miinchen, Germany) together with Na metal. The capsule was re-evacuated and sealed off. The samples thus grown were X-irradiated at room temperature for 30 min with a tungsten anticathode Philips X-ray tube, operated at 60 kV and 40 mA, corresponding to a dose of CQ. 40 kGy. Crystals were cut for rotation around a (1 10) axis. There was no intentional doping with a sulfur-containing species. The sulfur may originate from the carbon crucible where it is present as contamination.Methods The EPR spectra were recorded using a Bruker ESP300 X-band spectrometer. The maximum power of this spectrom- eter is 200 mW. The magnetic field was modulated at 100 kHz with a peak-to-peak amplitude of 0.5 x T. All spectra were normalised to the same frequency uiz. 9.47 GHz and hence can be compared directly. The magnetic field was measured using a Bruker NMR035M Gaussmeter. With this equipment it is possible to measure accurately the relative line positions of the EPR signals present. Small shifts of the magnetic field position down to 0.1 x lop4 T can be detected. Using an Oxford ESRlO flow cryostat, temperatures down to 4 K can readily be obtained. For absolute g value determination, a cali-bration using the g standard DPPH at 0.1 mW (g = 2.0036) was performed.Results No resonances could be detected in unirradiated crystals. After X-irradiation, resonances due to two different paramag- netic species were visible. The first defect (defect I) is visible J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 tk 3485e 3285' I <110> Cool>aldegrees Fig. 1 Angular variation in a (1 lo} plane of the resonances of the C0,-ion in NaCl. The height and the width of the rectangles is proportional to the line width and the signal height, respectively. The solid lines represent the theoretical angular variation, assuming zero line width. from 4 K to 180 K. The optimum detection temperature and microwave power for this defect are 4 K and 1 mW, respec- tively.The second defect (defect 11) is visible from 80 K up to 150 K and is preferentially measured at 100 K and 10 mW. Both defects will be discussed in more detail separately. Defect I The angular variation of the resonances of defect I in a { 110) plane is shown in Fig. 1. The observed spectra consist of three major groups of lines i.e. one group due to the unen- riched species (I = 0, "C) and two groups of lines (doublet) caused by the interaction of the electron spin with a 13C nucleus; this is the only nucleus with I = 1/2 in view of the doping procedure described above. The central line around g = 2 can be explained by considering the use of the graphite The central line has twice the intensity of the 13C hyperfine lines.The paramagnetic centre exhibits trigonal symmetry around a (1 11) axis. Owing to the large line width of the EPR resonances (compared with the g and A tensor anisotropy), the individ- ual resonances cannot be resolved in most orientations. Only when the applied magnetic field is oriented approximately along a (1 11) or a (1 10) direction, a small additional reson- ance is visible on the low-field hyperfine satellite (see Fig. 1). No additional substructure is detected on the central and high-field hyperfine lines. As was argued by Moens et al.," one must be very careful in determining the spin Hamiltonian parameters from such an angular variation by the 'classical methods' (ix.by fitting the resonance positions).As a result of the strong line overlap of the individual resonances, the observed line positions can be shifted from the real positions. Therefore, a program was developed for fitting the line profiles of all the observed spectra instead of only fitting the line positions of the EPR resonances. The resulting spin Hamiltonian parameters thus obtained are (hyperfine param- eters in MHz): g1 = 2.0022 (2), A, = 413 (3); gll = 1.9951 (3), All = 351 (3). The line width was optimized to be 11 (1) x T. The numbers in parentheses indicate the error on the last digit. The gI1axis is oriented parallel to a (111) axis. The theoretical angular variation, calculated with the above-mentioned spin Hamiltonian parameters is also shown in Fig.1. As can be seen, the small additional resonances of the low-field hyperfine line lie somewhat above the theoreti- cally predicted positions. That this is an effect of the large line 3350:: <loo> <110> aldegrees Fig. 2 Angular variation in a (100) plane of the resonances of the SO,-ion in NaCl. Same remarks as for Fig. 1. width and not of the fitting procedure, can be demonstrated by computer simulations. lo Defect I1 The angular variation of defect I1 exhibits orthorhombic symmetry and is shown in Fig. 2 in a (100) plane. The reson- ance positions can be adequately reproduced with the follow- ing g tensor: gx = 2.0017 (2), g,, = 2.0113 (2), gz = 2.006312). Here, the x and y directions are along the (110) and (110) axes, respectively, the z direction is along a (001) axis.As the line width is relatively small compared to the g tensor aniso- tropy, the individual resonances can be isolated in most orientations and hence the 'classical' optimization routines can be applied. The theoretical angular variation using the g values listed above, is also shown in Fig. 2. Discussion It will be demonstrated that both defects have to be ascribed to two different molecular ions. Hence defect I and I1 are discussed separately. Defect I As defect I exhibits hyperfine interaction with an I = 1/2 nucleus and as 13C is the only nucleus having nuclear spin I = 1/2, this defect must contain one 13C nucleus. Comparing the spin Hamiltonian parameters for defect I with data from the literature,'-'' the radical giving rise to the resonances of defect I is undoubtedly a C0,- ion.Table 1 lists the spin Hamiltonian parameters for the C0,- ion in KCl, KBr and NaCl. As can be seen, there is a strong resemblance between the C0,-ion in NaC1, KCl and KBr i.e. the spin Hamilto- nian parameters for the C0,- ion do not vary much with the host matrix. This can be readily understood by considering the electronic configuration of the radical. The CO, -molec-ular ion exhibits CZVsymmetry with a 2Al ground state.'v2 Thus, as the ground state is orbitally non-degenerate, the crystal field will merely cause a shift of the energy levels, having little effect on the g and A values. Table 1 Spin Hamiltonian parameters for the C0,-ion in several alkali-metal halide single crystals 91 II A, A,, AB/10-4T ref.~ ~~ KCl 2.0026 (1) 1.9962 (3) 388 (1) 328 (1) - 9 KBr NaCl 2.0026 (2) 2.0022 (3) 1.9948 (3) 1.9951 (3) 383 (3) 413 (3) 341 (3) 351 (3) 13 (1) 11 (1) 10 this work Hyperhe parameters in MHz. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Most probably, the C0,-radical substitutes for one halogen ion in the alkali-metal halide host lattice. This model is depicted in Fig. 3. As a result of the 'A, ground state of the C0,- ion, the one-electron molecular orbital containing the unpaired elec- tron can be written as: Here, the z axis is oriented along the main symmetry axis of the molecule (i.e. the C2 axis).The y axis is parallel to the direction connecting the two oxygen atoms, whereas the x direction is perpendicular to the molecular plane. The super- script in the notation for the atomic orbitals denotes the atom on which the orbital is centred. Normally, the C0,-species exhibits orthorhombic g and A Typical values for g and A, measured along the distinct molecular axes, are: gx = 2.0030, g, = 1.9970, gz = 2.0015 and A, = 513 MHz, A,= 506 MHz, A, = 630 MHz. The axial g value found for defect I indicates that the C0,-radical is rapidly rotating around the y axis (i.e. the axis connecting two oxygens). That such a rapid tumbling can occur at such low temperatures, has already been established for the isoelec- tronic NO, m01ecule.'~*'~ From the principal values of the 13C hyperfine tensor of the C0,-ion, one can easily calculate the spin densities in the 2s and 2p, carbon atomic orbitals [see eqn.(l)]. The spin density in the 2s carbon atomic orbital can be estimated by comparing Ai, with 2p0 g, gN 1,pN IYzs(o)12/3. The latter being calculated to be 3108.66 MHz.18 On the other hand, the spin density in the 2p, carbon atomic orbital is estimated by comparing A, -with 0.29, gN 1,pN(r2p-')//I, the latter being calculated as 45.36 MHz.'* Because the "O hyperfine tensor is unknown, only the spin density in all the oxygen atomic orbitals can be deduced from the normalisation to unity of the total spin density. The results are given in Table 2. As the 14N hyperfine tensor of the isoelectronic NO, mol- ecule in the same configuration and in the same host lattices is also known (i.e.an NO,{lll) rnolec~le),'~ we are in an excellent position to compare the spin densities of both the C0,-and NO, ions in KCl, KBr and NaCl. The results are also shown in Table 2. <001> Fig. 3 Model for the C02-ion in NaCl. The different ions are drawn to scale, taking into account their ionic radii. 3263 Table 2 Experimental spin densities on the individual carbon and nitrogen orbitals: S&N) and P&%N) for co2-and NO, in KC1, KBr and NaC1;" ck is the spin density on the oxygen atoms, derived from the normalisation to unity of the total spin density co,-NO, ','C CL 'i CZN ';,N '6 KCl 0.118 (1) 0.43 (2) 0.45 (2) 0.099 (2) 0.47 (2) 0.43 (2) KBr 0.119 (2) 0.32 (4) 0.56 (4) 0.099 (2) 0.40 (4) 0.50 (4) NaCl 0.127 (2) 0.45 (4) 0.42 (4) 0.102 (1) 0.49 (3) 0.41 (3) The values in parentheses denote the error on the last digit.From these results, it can be noticed that the spin density on the oxygen atoms is always larger for the C0,-ion than for the NO, molecule. These results are at variance with the theoretical deductions about C0,-and NO, ,''9l9 that predict that the spin density on the oxygen atoms is larger for NO, than for CO,-. In order to check this contradiction between theory and experiment, theoretical calculations of the spin densities of both free molecular ions using the GAMESS9O2O program were performed. The program was run on an IBM RS/6000, model 320 workstation.The spin densities of C0,-and NO, were calculated using five different basis sets. These were thus chosen to include minimal (ST0-3G),,' split valence (6-31G,,' 6-31 lG2') and polarized (6-31G*, 6-311G*)24 basis sets. The geometry of the molecules was optimized for each basis set. The spin densities calculated using the Lowdin spin population analysis are summarized in Table 3. For the oxygen atoms, only the total spin density is given. As can be seen from Table 3, the spin density on the oxygen atoms is always larger for NO, than for CO,-, in agreement with the theoretical deductions made by Atkins et al.' The discrepancy with the experimental results most '9" probably has two origins. First, the theoretical calculations are performed on free ions whereas the experiments are carried out on molecules embedded in a crystal matrix and secondly, the calculations do not account for the rapid rota- tion of the C0,- ion around the oxygen-oxygen axis.Such a rotation introduces a spin-rotational coupling term in the Hamiltonian which may have an effect on the magnetic res- onance spectrum. This was considered in some detail by Bojko and Silsbee16 for the NO, molecule. Ab initio calcu-lations, performed on radicals embedded in a host matrix can possibly give an indication of the importance of the effect of the crystal field. Calculations incorporating the effect of the crystal field through the Madelung potential will be carried out in our laboratory in the near future.Table 3 Comparison of the theoretical spin densities in the differ- ent atomic orbitals of carbon and nitrogen in C0,- and NO,, and the total spin density on the oxygen atoms ion basis 1s 2s 2p, 2py 2p, 0 C0,-STO-3G 0.0018 0.1301 0.0323 -0.0347 0.4060 0.4645 6-31G 0.0025 0.2205 0.0439 -0.0267 0.4394 0.3204 6-311G 0.0017 0.2298 0.0428 -0.0271 0.4294 0.3234 6-31*G 0.0015 0.1739 0.0347 -0.0242 0.4206 0.3171 6-311*G 0.0017 0.1806 0.0341 -0.0248 0.4148 0.3116 NO, STO-3G O.ooo6 0.0558 0.0102 -0.0450 0.3188 0.6597 6-31G 0.0013 0.1093 0.0389 -0.0274 0.3930 0.4845 6-311G 0.0010 0.1085 0.0406 -0.0266 0.3947 0.4819 6-31*G 0.0011 0.0846 0.0293 -0.0232 0.3718 0.4832 6-311*G 0.0011 0.0861 0.0315 -0.0258 0.3792 0.4826 Data obtained using the Lowdin population analysis calculated with the GAMESS90 program package.Defect I1 The paramagnetic radical giving rise to the resonances of defect I1 exhibits no hyperfine interaction. Hence it cannot contain a carbon nucleus. In view of the doping procedure, the most plausible candidate for this radical would be the 0,-ion. However, the ozonide ion has to be rejected on the basis of the g values, as can be seen from Table 4.However, comparing the g values of defect I1 with those of the SO2-ion in KCI and KBrl49I5 indicates that this paramagnetic species must be identified as an SO,-molecular ion (see Table 4). The sulfur most probably originates from the carbon crucible where it is present as contamination.Reac- tion of the sulfur with the dopant during the growth process can result in the formation of SO2and SO, molecules. The SO2-ion exhibits CZy symmetry and has a ,B, ground ~tate.'~ Hence, within the LCAO scheme, the unpaired electron resides in a molecular orbital of the form: where the same notations and conventions for the molecular axes as for eqn. (1) are used. As can be seen from eqn. (2), the electron resides mainly in the px atomic orbitals i.e. the orbitals perpendicular to the molecular plane. Unfortunately no conclusions can be drawn about the spin densities in the distinct atomic orbitals as no data are available for either the 33Sor the 170 hyperfine tensors. The SO,-ion most probably substitutes for one halide ion. The model in depicted in Fig.4.The incorporation of the SOz-ion will be governed by the electrostatic interactions between the paramagnetic px lobes of the defect (as these contain the unpaired electron) and the surrounding Na' and C1- ions of the host matrix. In the case of a monovacancy model, the electrostatic interaction will be mainly between the px lobes and the six nearest Na' ions. As the p, lobes are oriented perpendicular to the molecular plane, the only Na' ions which are of interest here are the two Na+ ions situated Table 4 g tensor values for the 0,-and SO,-ions in several alkali-metal halide single crystals lattice defect 9x gv 9, ref. KCI 03 -2.0032 2.0182 2.0118 22 K Br 03 -2.0027 2.0180 2.0113 23 KI 03 -2.0030 2.0185 2.0116 24 KCI SO,-2.0025 2.0110 2.0071 14 KBr SO,-2.0050 2.0100 2.0075 14 NaCl defect I1 2.0017 2.0113 2.0063 this work <110> i <OOl> Fig.4 Model for the SO,-ion in NaCI. Same remarks as for Fig. 3' J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 above and the two situated below the molecular plane. The two Na ions lying above the molecular plane are indicated + in Fig. 4 by 1 and 2. Thus, it can be seen that the SO,-ion has to be placed slightly off-centre, so that the px lobes avoid the electron clouds of the Na+(l) and Na'(2) ions as much as possible. However, the total C,, symmetry is preserved. The situation is in some respects analogous to the case of the smaller 0,-ion in KCl, KBr and KI.25-28 B y assuming that the incorporation of this ion is also governed by electro- static interactions, it was predicted that the ozonide ion could not be incorporated untilted in the KC1 lattice because of the too small dimensions of the C1- vacancy.On the other hand, the Br -and I -vacancies offer enough space for the 0,-ion to be untilted. All these aspects were indeed observed with EPR, giving experimental evidence for the proposed model for 0,-to be extended in this paper to SO,-. By the same arguments the larger S,-and Se,- ions should occupy a tri- vacancy site, as was confirmed e~perirnentally.'~*' Conclusions The EPR spectra of two small triatomic molecules with C,, symmetry in NaCl uiz. C0,-and SO,-, are presented. The C0,-ion rotates rapidly around the axis connecting the two outer oxygen atoms, even at liquid-helium tem- peratures. The same peculiar feature was observed for the iso- electronic NO, molecule. By analysing the 13C hyperfine tensor, the spin densities in the 2s and 2p, carbon atomic orbitals could be calculated.In addition, the total spin density on the oxygen atoms could be deduced from the nor- malisation to unity of the total spin density. When these data are compared with those for the NO, molecule, results in contradiction with theoretical calculations using the GAMESS90 program are obtained. Most probably, this dis- crepancy between theory and experiment must be explained by the fact that the calculations were performed on free rad- icals, whereas the experiments concern radicals embedded in a crystal matrix.The monovacancy model for the SOz-ion, as proposed by other authors, is confirmed by our measurements. The authors wish to thank the Executieve van de Vlaamse Gemeenschap-Departement Onderwijs and the Inter-universitair Instituut voor Kern Wetenschappen (IIKW) for financial support. References 1 D. W. Ovenall and D. H. Whiffen, Mol. Phys., 1961,4, 135. 2 S. A. Marshall, A. R. Reinberg, R. A. Serway and J. A. Hodges, Mol. Phys., 1964, 8, 225. 3 R. J. Cook and D. H. Whiffen, J. Phys. Chem., 1967,71,93. 4 S. Schlick, B. L. Silver and 2. Luz, J. Chem. Phys., 1971,54,867. 5 P. Meriaudeau, J. C. Vedrine, Y.Ben Taarit and C. Naccache, J. Chem. SOC.,Faraday Trans. 2,1974,71,736. 6 G.Bacquet, V. Quang Truong, M. Vignoles, J. C. Trombe and G. Bonel, Calcif Tissue Znt., 1981,33, 105. 7 F. J. Callens, R. M. H. Verbeeck, P. F. A. Matthys, L. C. Martens and E. R. Boesman, Calcif. Tissue Int., 1987,41, 124. 8 R. Debuyst, M. Bidiamambu and F. Dejehet, Bull. SOC. Chim. Belg., 1990,99, 535. 9 F. Callens, P. Matthys and E. Boesman, J. Phys. Chem. Solids, 1989,50,377. 10 P. D. W. Moens, P. F. A. Matthys, F. J. Callens, F. R. C. Maes and E. R. Boesman, Phys. Status. Solidi,B, 1991,168,289. 11 P. W. Atkins, N. Keen and M. C. R. Symons, J. Chem. Soc., 1962, 3, 1873. 12 P. N. Keizer, J. R. Morton and K. F. Preston, J. Chem. SOC., Faraday Trans., 199 1,87, 3 147. 13 R. Gray and M. H. Stevenson, Bruker Report 91192,1992, p. 9. 14 J.Schneider, B. Dischler and A. Rauber, Phys. Status. Solidi, 1966, 13, 141. J. CHEM. SOC. FARADAY TRANS., 1994, VOL.90 3265 15 16 17 18 J. Suwalski and H. Seidel, Phys. Status. Solidi, 1966,13, 159. I. Bojko and R.H. Silsbee, J. Magn. Reson., 1971,5, 339. J. R. Brailsford and J. R. Morton, J. Magn. Reson., 1969,1,575. A. H.Koh and D. J. Miller, At. Data Nucl. Data Tables, 1985, 23 24 R. Krishnan, J. S. Binkley, R. Seeger and J. A. Pople, J. Chem. Phys., 1980,72,650. P. C. Chariharan and J. A. Pople, Theoret. Chim. Acta, 1973,28, 213. 19 20 21 33,235. P. W. Atkins and M. C. R. Symons, The Structure of Inorganic Radicals. An Application of Electron Spin Resonance to the Study of Molecular Structure, Elsevier, Amsterdam, 1967. M.W. Schmidt, K. K. Baldridge, J. A. Boatz, J. H.Jensen, S. Koseki, M.S. Gordon, K. A. Nguyen, T. L. Windus and S. T. Elbert, QCPE Bulletin, 1990,10,52.W.J. Hehre, R.F. Stewart and J. A. Pople, J. Chem.Phys., 1970, 25 26 27 28 F. Callens, P. Matthys and E. Boesman, J. Phys. C, 1989, 21, 3159. F. Maes, F. Callens, P. Matthys and E. Boesman, Phys. Status. Solidi, B, 1990,161, K1. P. Moens, F. Callens, S. Van Doorslaer, F. Maes and P. Matthys, Phys. Status. Solidi, B, 1994, 182, 21 1. F. Callens, P. Moens, S. Van Doorslaer, F. Maes and E. Boesman, J. Chem. Soc., Faraday Trans., 1993, W,3691. 22 52,2769. W. J. Hehre, R. DitcMeld and J. A. Pople, J. Chern. Phys., 1972, 56,2257. Paper 4/03002H;Received 20th May, 1994