Solid state 77Se NMR investigations on arsenic-selenium glasses and crystals

Solid State Sciences 5 (2003) 219–224 www.elsevier.com/locate/ssscie

Solid state 77 Se NMR investigations on arsenic-selenium glasses and crystals Bruno Bureau a,∗ , Johann Troles a , Marie LeFloch a , Frédéric Smektala a , Gilles Silly b , Jacques Lucas a a Laboratoire des verres et céramiques, UMR-CNRS 6512, université de Rennes 1, campus de Beaulieu, 35042 Rennes cedex, France b Laboratoire de physique de l’état condensé, UMR-CNRS 6087, université du Maine, 72017 Le Mans cedex, France

Received 28 May 2002; accepted 24 June 2002 Dedicated to Sten Andersson for his scientific contribution to Solid State and Structural Chemistry

Abstract Some resolved solid state 77 Se NMR spectra are presented in the Asx Se1−x glass family at ambient temperature. They exhibit three different kinds of Se environments. A comparison with the parent crystalline phases permits to assign the lines to Se-Se-Se, Se-Se-As and As-Se-As Se atom neighborhoods. The measurements of the relative intensities of the lines prove the validity of the intermediate range order structural model known as the “chains crossing model” which is based on AsSe3 pyramids homogeneously distributed among the divalent Se atoms network. In particular, any scenario involving a selenium clustering process is refuted.  2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. Keywords: Chalcogenide glasses; 77 Se; Solid State NMR; Structural investigations

1. Introduction The arsenide chalcogenide Asx Se1−x glasses (g-Asx Se1−x ) possess important potential applications as semiconductors [1], non-linear and infra-red optical compounds [2–5]. Nevertheless, their structures have been quite rarely investigated during the past decades. Previous studies [6– 11] have permitted to describe precisely the short range order which is controlled by a chemical ordering process which privileges the heteropolar bonds. For the Se-enriched glasses, x
0.4, on the one hand As and Se atoms tend to form Se triangular base pyramids with an As atom at the summit [6]. On the other hand, the divalent Se atoms in excess tend to form chains or rings of variable length. Consequently, depending on their composition, the network of these glasses is expected to be built up with AsSe3 pyramids more or less connected to each other by chains of Se atoms. Two theories directly related to the way the pyramids are spread over the network were proposed to describe the medium range order for the Se-enriched chalcogenide glasses. The first model is the more simple and intuitive one, * Corresponding author.

E-mail address: [email protected] (B. Bureau).

known as the “chains crossing model”. It consists in considering that the As atoms are homogeneously distributed among the network. Then, the Se chains are cross-linking the AsSe3 pyramids and the mean number of Se between two pyramids depends on the initial composition: the richer the composition in Se, the longer are the chains. It is in favour of the largest distances between As atoms. The second model, so-called the “outrigger raft model” proposed by Griffiths et al. [12] would reckon the existence of clusters of corner or edge shared pyramids diluted in a chalcogenide matrix constituted by Se chains or rings [13,14]. Then, it appears that the Se atoms play a crucial role in the connectivity of the AsSe3 pyramids in the glass network. A similar phenomena was encountered in fluoride glasses [15] were the network is built of fluorine octahedra centred on transition metal ions connected by corners. In this latter case 19 F (I = 1/2) high resolution solid state NMR has been proved to be a powerful tool to investigate the connectivity of the glass by probing the anion site. Three ranges of isotropic chemical shift values were evidenced according to the F− ion is shared between two octahedra, belongs to only one octahedron or is in an interstitial position. 77 Se nuclei possesses also a nuclear spin I = 1/2. The isotropic component of the chemical shift interaction tensor

1293-2558/03/$ – see front matter  2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. doi:10.1016/S1293-2558(02)00102-4

220

B. Bureau et al. / Solid State Sciences 5 (2003) 219–224

δiso is very sensitive to the neighbourhood of the Se atoms since values lying from −900 ppm to +2430 ppm can be measured [16]. This must permit to differentiate the Se atoms involved in one or two pyramids, or belonging to a chain. Very recently, a 77 Se liquid state NMR study was carried out on molten g-Asx Se1−x sample [17]. Thanks to a discussion based on the dynamic exchange process versus temperature, the authors favour the “crossing chains model” since no signature of any clustering process was evidenced in these liquid phases. These reasons led us to perform solid state 77 Se NMR measurements to obtain a deeper insight of the arsenide chalcogenide glasses network at room temperature. The 77 Se natural abundance (7.58%), its low relative sensitivity (6.93 × 10−3 ) together with its long longitudinal relaxation time, make this nucleus very rarely studied in solid state materials [16–18]. Up to now it is a challenge to obtain valuable spectra in such circumstances. Nevertheless, it will be shown that these experiments allow us to find evidence for three Se environments. Thanks to the known structures of the parent crystalline phases (c-As2 Se3 and pure c-Se), three types of selenium neighborhoods will be reliably characterized and quantified leading to a precise description of the glassy network.

2. Experimental 2.1. Materials Raw materials with 99.999% elemental abundance were used for glass preparation. Arsenic and selenium were further purified of remaining oxygen by the volatilization technique. This method uses the greater vapor pressure of oxides over these of the metals to remove the oxide species. It consists in heating selenium at 250 ◦ C and arsenic at 290 ◦ C under vacuum for several hours. After this treatment, the required amounts of Se and As are sealed in a silica tube under vacuum. Then the mixtures are maintained at 650 ◦ C for 12 hours in a rocking furnace to ensure a good mixing and homogenization of the liquid. To condense a maximum of vapor in the liquid the temperature is reduced to 500 ◦ C for 1 hour. Then the ampoules are quenched in air and annealed near glass transition temperatures (Tg ) to reduce the mechanical stresses occur during the cooling. The glassy nature was confirmed by X-ray diffraction and differential scanning calorimetry (DSC). The measured Tg are 40 ◦ C (pure g-Se, x = 0), 70 ◦ C (AsSe9 , x = 0.10), 100 ◦ C (AsSe4.5 , x = 0.18), 112 ◦ C (AsSe3.3 , x = 0.23) and 185 ◦ C (As2 Se3 , x = 0.40). These glass compositions are otherwise candidates for non linear optical investigations in this binary system [4,5]. The crystalline compounds, c-As2 Se3 and c-Se (hexagonal c-Se phase) were synthesized using the same method but instead of quenching, a very slow cooling, about 1 ◦ C per hour, was applied to the melt. Their structures were controled by X-ray diffraction led on powders.

2.2. NMR measurements The 77 Se (S = 1/2) NMR spectra were recorded at room temperature on an ASX 300 Bruker spectrometer operating at 57.28 MHz with a 4 mm MAS probe spinning at 15 kHz. The spectra were recorded with the maximum spinning speed reachable (15 kHz) that simultaneously averages the chemical shift and reduces the homonuclear dipolar interaction between Se atoms. Note that this interaction is expected to be weak in view of the low abundance of the 77 Se (7.58%). So, the remaining broadness of the lines are due to the isotropic chemical shift distributions characteristic of the vitreous states. Moreover, due to this strong residual broadness of the lines, a Hahn spin echo sequence was applied to refocus the whole magnetization and thus to avoid any distortion of the baseline. For the glasses, the Fourier Transformation were done from the whole echos (so-called full shifted echos) in order to increase the signal to noise ratio and to directly obtain absorption mode lineshapes [19, 20]. The processing and acquisition parameters were: 3.5 µs π/2 pulse duration, 10 s recycle time, spectral width 1 MHz, time domain 1 K. To ensure that MAS together with spin echo acquisition do not occur any distortion of the lineshapes we have to choose a time delay between the pulses equal to 1/vR where vR is the spinning speed. So the 15 kHz MAS spectra were recorded with a delay of 67 µs. Owing to the weak sensibility, the scan numbers were contained between 2000 and 10000 versus the Se atom rate in the glass. The external reference used for the chemical shifts is a saturated solution of Me2 Se in CDCl3 [16]. The simulations of the experimental spectra were performed using a modified version of the Winfit Bruker Software [21].

3. Results and discussion 3.1. NMR lines assignment The obtained 77 Se 15 kHz MAS NMR spectra for five vitreous material are compared in Fig. 1. Three different line positions appear for this glass family at about 850 ppm (labeled a), 550 ppm (labeled b) and 380 ppm (labeled c) on the chemical shift scale. In order to determine the neighborhood of the three corresponding Se sites, the c-Se and c-As2 Se3 crystalline phases were investigated. The corresponding 15 kHz MAS spectra are given together with the one of the related glasses in Figs. 2 and 3. The isotropic chemical shift values associated to the lines of c-Se and g-Se (Fig. 2) are respectively 792 ppm and 864 ppm in very good agreement with the published data from [16] who gives respectively 794 ppm and 865 ppm for both phases. As expected the g-Se spectrum is much broader (17 kHz half-height width) in accordance with the distribution of site characterizing a vitreous structure. The network is built up with chains of divalent Se atoms, well aligned to each other in the crystalline phase while interlaced in glassy

B. Bureau et al. / Solid State Sciences 5 (2003) 219–224

221

Fig. 1. Comparison of the 77 Se NMR spectra for Asx Se1−x glasses exhibiting three types of line labeled a, b and c.

Fig. 2. Comparison of the 77 Se spectra of the glassy and crystalline pure Se phases. The stars labeled the spinning side bands.

phase [22]. In any case, each Se atom remains identically connected to two other Se atoms. Then the a lines (Fig. 1) are attributed to such Se neighborhood noted Se-Se-Se. At the opposite side of the chemical shift scale, the unique broad line of the g-As2 Se3 spectrum appears at 380 ppm (c line in Fig. 1). For c-As2 Se3 , the line narrowing at 15 kHz is sufficient to evidence three different 77 Se NMR lines at 210, 309 and 327 ppm with equal relative intensities (Fig. 3). This layered structure is built up with AsSe3 pyramids di-

Fig. 3. Comparison of the 77 Se spectra of the glassy and crystalline As2 Se3 phases. The stars labeled the spinning side bands.

rectly connected to each other by the Se atoms. Nevertheless, whatever the authors [23–26], the structure is expected to exhibit three slightly different crystalline sites for the divalent Se atoms. The As–Se distances provided by [24–26] are regularly spaced (about 2.40 Å, 2.42 Å and 2.44 Å for the three Se crystallographic sites) and can not justify the fact that two NMR lines are very closed to each other. On the other hand, the NMR spectrum is in better agreement

222

B. Bureau et al. / Solid State Sciences 5 (2003) 219–224

with the Se–As distances given in [23]: the Se atoms are connected to two As atoms with similar Se–As bond length of 2.38 Å and 2.40 Å for the crystallographic sites Se1 (33%) and Se2 (33%) respectively and with a larger bond length of 2.51 Å for Se3 (33%). The integrated NMR line intensities are also in agreement with identical site multiplicities. Expecting that the shorter the distances the higher the chemical shift, we propose the line attributions given in Fig. 3 for Se1, Se2 and Se3. The broad g-As2 Se3 Gaussian line (20.5 kHz half-height width) recovers the whole range of the c-As2 Se3 chemical shift values showing that the corresponding Se neighborhoods are very similar in both phases (Fig. 3). Then, the g-As2 Se3 structure is constituted by a floppy network of pyramids directly connected to each other by bridging Se confirming the X-ray results from [6]. The gAs2 Se3 line position being closer to the Se1 one, we suggest that the mean Se–As bond length is not far from 2.38 Å in the glass. In any case, the chemical shifts belonging to this range (c lines) are assigned to Se atoms connected with two As first neighbors noted As-Se-As. Between both a and c lines, appears clearly on Fig. 1 a third type of line with intermediary chemical shift values around 550 ppm labeled b. In particular this type of line is in large majority in g-AsSe3.3 . According to the above discussion, these lines are attributed to Se atoms connected with one Se and one As as first neighbors noted As-Se-Se. Note that these three assignments are in agreement with the results from [17] on the molten g-Asx Se1−x phases.

Fig. 4. Reconstruction of the 77 Se MAS (15 kHz) spectrum of AsSe4.5 glass.

Fig. 5. Schematic representation of As2 Se3 (a), AsSe3 (b), AsSe4.5 (c) and AsSe9 (d) glass structures according to the “chains crossing model” that predicts a homogeneous distribution of the AsSe3 pyramids.

3.2. Description of the glassy network glass composition to the other. The following discussion is based on the measured intensities that enable to directly quantify the percentage of Se atoms of each type. Firstly, the relative intensity evolution confirms the assignment done: the more there are Se atoms in the glass composition, the higher the intensity of the a lines to the detriment of the b and overall c lines. Moreover, from the initial composition of each mixture it is possible to calculate the numbers of Se atoms of each type expected in the glassy structure for both intermediate range order models. The “chains crossing model” foresees that the Asx Se1−x glassy network is built up with Se chains cross-linking

In order to obtain further results related to the way the AsSe3 pyramids are connected in this glass family, all the spectra were deconvoluted using an extended version of Bruker Winfit software [21]. Since the spectra evidence two maximums, each line shape is reconstructed with two contributions as shown in Fig. 4 for AsSe4.5 . Each contribution has a Gaussian shape due to the disorder surrounding the 77 Se sites in such glasses. The results of the deconvolution are gathered in Table 1. Of course, the three types of line positions (a, b and c) are retrieved with chemical shifts remaining at the quite same values from a

Table 1 Parameters used for the reconstruction of the 77 Se NMR static spectra. The lines are pure Gaussians Line

x = 0.000 vitreous Se

x = 0.100 AsSe9

x = 0.182 AsSe4.5

x = 0.233 AsSe3.3

x = 0.400 As2 Se3

Chemical shift (±10 ppm)

a b c

865 – –

865 570 –

850 550 –

850 525 –

– – 380

Integrated intensities (±5%)

a b c

100 0 0

62 38 0

35 65 0

15 85 0

0 0 100

Widths (±0.5 kHz)

a b c

17.0 – –

13.2 17.7 –

12.6 20.0 –

11.5 22.9 –

– – 20.5

B. Bureau et al. / Solid State Sciences 5 (2003) 219–224

223

Table 2 Expected percentage of each type of Se atoms with the “chains crossing model” Se neighborhoods Se-Se-Se Se-Se-As As-Se-As

x = 0.000 vitreous Se

x = 0.100 AsSe9

x = 0.182 AsSe4.5

x = 0.233 AsSe3.3

x = 0.400 As2 Se3

100 0 0

66 33 0

33 66 0

10 90 0

0 0 100

Fig. 6. Comparison between the a, b and c line normalized intensities and the expected Se-Se-Se, As-Se-Se and As-Se-As ratios with the “chains crossing model”. The x values are given in Table 2.

AsSe3 pyramids. As examples, we have schematized in Fig. 5 the network of the typical following compositions: As2 Se3 , AsSe3 , AsSe4.5 and AsSe9 . The richer the composition with Se, the longer are the Se chains, leading to the percentages of each Se type given in Table 2 in the case of a perfect distribution of the pyramids among the network. For example, in g-AsSe4.5 , there are three Se atoms between two As atoms leading to the proportions: 33% of Se-Se-Se and to 66% As-Se-Se. Note that with this model the three types of Se, Se-Se-Se (a lines), Se-Se-As (b lines) and AsSe-As (c lines) can not be present at the same time in a given glass. Finally, in Fig. 6 the percentages of Se of each type are compared with the relative intensities obtained from the deconvolution of the spectra (see Table 1). It appears that both data fit very well to each other confirming the validity of the “chains crossing model” for this chalcogenide glass family. With the second model based on selenium phase separation forming independent clusters, the expected percentages of Se atoms of each type would be very different. Firstly, the As-Se-As neighborhood would remain very numerous even in the Se atom enriched glasses since the AsSe3 pyra-

mids would agglomerate. The 77 Se NMR spectra refute this assumption since the corresponding c lines vanish as soon as x < 0.3. Secondly, the percentage of Se-Se-As neighborhood should be very low because, with a phase separation model, the Se chains or ring would be rarely linked to the pyramids. These expectation is also in contradiction with the NMR spectra where the b type lines are clearly evidenced with high intensities for AsSe3 or AsSe4.5 . Finally, The intermediate range order description based on a clustering process is incompatible with the positions and the relative intensities of the 77 Se solid state NMR spectrum lines. Note that the vitreous transition temperature and the densities evolution of the glasses are coherent with the “chains crossing model”. Firstly, very logically, the longer the floppy chains of Se, the lower are the Tg . The density values [27] increase regularly with the percentage of As in the composition from 4.28 g·cm−3 (g-Se) to 4.49 g·cm−3 (AsSe3 ) showing that the shorter the chains, the higher is the rigidity of the network. Then, for As richer compositions, the densities remain quite stable (4.48 g·cm−3 for As2 Se3 ).

224

B. Bureau et al. / Solid State Sciences 5 (2003) 219–224

4. Conclusion In this work, some well resolved 77 Se solid state NMR spectra are shown with high signal to noise ratio in vitreous and crystalline condensed materials. Two main messages arise from our results. The first one concerns c-As2 Se3 itself where we verified the existence of the three kinds of selenium site as expected from the X-ray crystalline structure. This demonstrates the power of the 77 Se solid state NMR as a local probe method in a complex selenide crystalline structure. The second message deals with the possibility to collect information on glassy materials through the sensibility of the 77 Se isotropic chemical shift to its own environment. Indeed, we had been able to discriminate between three types of Se neighborhoods in the Asx Se1−x glass family at room temperature. A comparison with the parent crystalline phases allows to attribute the lines to the three following situation: Se-Se-Se, Se-Se-As and As-SeAs. Concerning the network description, the measurements of the line relative intensities prove the validity of the intuitive model, so-called “chains crossing model” that predicts a homogeneous distribution of the AsSe3 pyramids. In particular, no signature of any clustering process was evidenced. Finally, this work confirms that solid state NMR is a tool of choice to investigate the vitreous network by probing the polyhedron connector atoms (19 F, 77 Se, 17 O, . . .) or the polyhedron coordinator atoms (27 Al, 31 P, 29 Si, . . .).

References [1] K.L. Bathia, N. Kishore, R.S. Kundu, M. Singh, P. Singh, J. NonCryst. Solids 180 (1995) 251. [2] T.I. Kosa, R. Rangel-Rojo, E. Hajto, P.J.S. Ewen, A.E. Owen, A.K. Kar, B.S. Wherrett, J. Non-Cryst. Solids 164/166 (1993) 1219.

[3] M. Asobe, T. Kanamori, K. Naganuma, H. Itoh, T. Kaino, J. Appl. Phys. 77 (1995) 5518. [4] C. Quemard, F. Smektala, V. Couderc, A. Barthélémy, J. Lucas, J. Phys. Chem. Solids 62 (2001) 1435–1440. [5] G. Boudebs, F. Sanchez, J. Troles, F. Smektala, Opt. Comm. 199 (2001) 425–433. [6] Y. Iwadate, T. Hattori, S. Nishiyama, K. Fukushima, Y. Mochizuki, M. Misawa, T. Fukunaga, J. Phys. Chem. Solids 60 (1999) 1447. [7] A.J. Leadbetter, A.J. Apling, J. Non-Cryst. Solids 15 (1974) 250. [8] V. Mastelaro, H. Dexpert, S. Benazeth, R. Ollitrault-Fichet, J. Solid State Chem. 96 (1992) 301. [9] A.L. Renninger, B.L. Averbach, Phys. Rev. B 8 (1973) 1507. [10] P. Boochland, W.J. Bresser, P. Suranyi, Hyperfine Interact. 27 (1986) 385. [11] J.C. Phillips, C.A. Beevers, S.E.B. Gould, Phys. Rev. B 21 (1980) 5724. [12] J.E. Griffiths, G.P. Espinosa, J.C. Phillips, J.P. Remeika, Phys. Rev. B 28 (1983) 4444. [13] J.C. Phillips, Phys. Rev. B 31 (1985) 8157. [14] M.F. Thorpe, Physics of Disordered Materials, Plenum, New York, 1985, p. 55. [15] B. Bureau, G. Silly, J.Y. Buzaré, J. Emery, C. Legein, C. Jacoboni, J. Phys.: Condens. Matter 9 (1997) 6719. [16] H. Duddeck, Progress in NMR spectroscopy 27 (1995) 1. [17] C. Rosenhahn, S.E. Hayes, B. Rosenhahn, H. Eckert, J. Non-Cryst. Solids 284 (2001) 1. [18] R. Maxwell, D. Lathrop, H. Eckert, J. Non-Cryst. Solids 180 (1995) 244. [19] D. Massiot, I. Farnan, N. Gautier, D. Trumeau, A. Trokiner, J.P. Coutures, Solid State Nucl. Magn. Reson. 4 (1995) 241. [20] P.J. Grandinetti, J.H. Baltisberger, A. Llor, Y.K. Lee, U. Werner, M.A. Eastmann, A. Pines, J. Magn. Reson. A 103 (1993) 72. [21] D. Massiot, H. Thiele, A. Germanus, Brucker Rep. 140 (1994) 43. [22] A.F. Wells, Structural Inorganic Chemistry, Oxford Press, 1962, p. 143. [23] A.L. Reninger, A.L. Averbach, Acta Crystallogr. B 29 (1973) 1583. [24] A.A. Vaipolin, Soviet. Phys. Crystallog. 10 (1966) 509. [25] A.C. Stergiou, P.J. Rentzeperis, Z. Kristallographie 173 (1985) 185. [26] A.S. Kamishcheva, Y.N. Mikhailov, E.G. Zhokov, T.G. Grevtseva, IVNMA 19 (1983) 1744. [27] C. Quémard, Thesis 2377, Université de Rennes 1 (2000).