Application of molecular dynamics techniques and luminescent probes to the study of glass structure: the SiO2–GeO2 case

(1)

Application of molecular dynamics techniques and luminescent probes to the study of glass structure: the SiO

2

±GeO

2

case

C. Bernard

^a,*

, S. Chaussedent

^a

, A. Monteil

^a

, N. Balu

^a

, J. Obriot

^a

, C. Duverger

^b

, M. Ferrari

^c

, M. Bouazaoui

^d

, C. Kinowski

^e

, S. Turrell

^e

aLaboratoire POMA, Universited'Angers, UMR CNRS 6136, 2 bd Lavoisier, 49045Angers cedex, France

bINFM ± Dip. Fisica, Universitadi Trento, Via Sommarive 14, 38050 Trento, Italy

cCNR ± CeFSA, Centro Fisica Stati Aggregati, Via Sommarive 14, 38050 Trento, Italy

dPhLAM, Laboratoire de Physique des Lasers, Atomes et Molecules, UMR CNRS 8523, CERLA ± Centre d'Etudes et de Recherches Lasers et Applications USTL, 59655 Villeneuve d'Ascq cedex, France

eLASIR, Laboratoire de Spectrochimie Infrarouge et Raman, UPR CNRS A2631L, CERLA ± Centre d'Etudes et de Recherches Lasers et Applications USTL, 59655 Villeneuve d'Ascq cedex, France

Abstract

In this paper, we report on the results obtained from molecular dynamic simulation of a Eu³-doped germanosilicate glass. This simulation provides further information on the structure. In particular it reveals a homogeneous distribution of SiO4 and GeO4 units, a decrease of defects compared to SiO2 and GeO2 glasses, and a trend to clustering of the doping ions. Using the modi®ed crystal-®eld theory, the luminescence spectroscopic properties have been computed and comparison with experimental data has allowed a correlation of the spectral features with two main types of local environment depending on the coordination number and on the medium-range arrangement around the doping ions. Ó 2001 Elsevier Science B.V. All rights reserved.

PACS:61.43; 68.55; 78.20; 78.55; 78.66

1. Introduction

In addition to numerous experimental studies, the modeling of doped glasses using the molecular dynamics (MD) method can provide further information on the structural and spectroscopic properties of disordered materials. Particularly, doping with the rare-earth ions leads to various technological applications such as optically active

devices [1]. On the other hand, doping ions can be used as structural probes [2] when their spectroscopic features are simple enough.

In this paper, we present the results that we have obtained on the structural and luminescence properties of the SiO2±GeO2:Eu³ binary system.

The experimental sample of reference is a waveguide obtained by dip-coating using the sol±gel technique [2]. The detailed experimental analysis has been reported elsewhere [2]. While its properties render this waveguide a promising device for integrated optics, the structure is not yet well understood which makes more dicult its experimental optimization. For this reason, MD simulations of a similar system have been performed.

www.elsevier.com/locate/jnoncrysol

*Corresponding author. Tel.: +33-2 41 73 50 04; fax: +33-2 41 73 53 52.

E-mail address:christophe.bernard@univ-angers.fr (C. Ber- nard).

PII: S 0 0 2 2 - 3 0 9 3 ( 0 1 ) 0 0 3 8 1 - 7

(2)

The structure has been analyzed by means of distribution functions, ring statistics, and systematic studies of the local arrangement around the doping ions. The luminescence properties have been computed from the MD con®gurations. All results of our simulations are presented and discussed in the light of conclusions drawn from the experimental analysis.

2. Computational procedure

The numerical sample has been composed with respect to the experimental composition (Ge/Si1 with 1.17 mol% of Eu³). The MD simulations have been performed using an eective 2 + 3-body interatomic potential [3]. Among the dierent sets of parameters describing each interaction, those involving Si and O atoms were kept to the values established by Peres et al. [4], while those required for Ge-interactions have been determined as the result of a preliminary study on the GeO2 glass structure. As for silica, the 3-body term has been included to have the best control of the GeO₂ angular distributions. To determine this new set of parameters a comparison has been made between the total correlation functions, Tr, obtained from experimental measurements and from the simulated con®gurations (Fig. 1). The simulated Tr has been reconstructed according to the

method described by Wright [5] and using the experimental processing parameters of Desa et al. [6]

in their neutron diraction experiment. The agreement factor that has been reached Rv5:7% is one of the best obtained for a simulated glass, thus validating our potential parameters for Ge-interactions. The potential parameters involving Eu³ have been determined by successive adjustments that allowed the Eu2O3

crystal structure at 300 K to be maintained.

The MD box has been built to contain 854 Si, 854 Ge, 3446 O and 20 Eu ions, respecting the composition of the experimental glass. The volume of the box has been adjusted to obtain correct experimental density value. After a complete mixing at 12 000 K, a quenching procedure has been applied, the system being cooled to 300 K through nine successive relaxation stages.

3. Results

The structure of the simulated glass has been analyzed using the typical radial, cumulative and angular distribution functions (RDF, CDF and ADF) and various ad hoc computational procedures. Si±O bond length has been found to be 1.62 A with a full width at half maximum (FWHM) of 0.08 A and Ge±O bond length is 1.70 A with a FWHMof 0.1 A, giving coordination number of 4.00 for both cations.

In Fig. 2, the Ge±Si CDF indicates that Ge atoms are, in average, surrounded by nearly three Si atoms and one Ge atom. We have also computed the bridging oxygens statistics and found that 67% of them are bonded to one Si and one Ge at the same time.

A more detailed statistical analysis has shown that the small amount of non-bridging oxygens (5%) are mainly found in germanium local environment. A subsequent step in understanding the glass structure consists in determining rings statistics. Since Si and Ge are both network former, and due to the homogeneous distribution of Ge and Si atoms in the glass, we have called n-fold ring a ring involving a numbernof Si or Ge atoms.

Comparing results obtained from SiO2;GeO2, and

Fig. 1. Experimental [6] (±±) and simulated (- - -) total correlation functionsTrfor a pure GeO2glass.

(3)

SiO2±GeO2:Eu³, we have found that 3- and 4-fold rings are less numerous in the binary system.

Despite poor statistics due to the concentration in Eu³, the RDFs and CDFs of Fig. 3 can give an insight of the arrangement of this local probe in the germanosilicate glass. The very well de®ned

®rst coordination shell displayed by the Eu±O RDF has permitted us to obtain a reliable mean coordination number of 6.4 which has been statistically resolved into two kinds of site, namely 6- and 7-coordinated sites. Fig. 3 presents Eu±Si and Eu±Ge RDFs, and shows that the medium- range environment of Eu³ is richer in Ge than in Si.

To study the clustering of Eu³, we have based our analysis on the fact that the Eu±O distances are very well de®ned and are assumed to be bonds.

In such a way, within a cluster, Eu³ ions are in- terconnected through Eu±O±Eu linkages. By means of ad hoc computational procedures, we have been able to count the number of clusters, the size of a cluster being de®ned as the number of europium ions involved. It has been found that more than 60% of the doping ions are isolated. We also mention that the Eu³ average coordination number increases with clustering. Indeed, while the average coordination number is 6.40, this mean number is decreased to 6.12 when only isolated dopants are taken into account.

To link the simulated structure and the luminescent properties of the sample, we have used the modi®ed crystal-®eld theory. From the statistical set of simulated environments, crystal-®eld parameters,Bnm, have been computed and according to a numerical procedure described elsewhere [7] a spectral envelope has been generated to be directly compared with the experimental emission spectrum.

The comparison is shown in Fig. 4. It points out agreement to within 8.8% which is indicative of the simulated local structure reliability. The inset in Fig. 4 shows the spectral contributions of the 6- and 7-fold coordinated sites: the splitting of the

7F1level is less large for 7-coordinated sites. Fur- thermore the second-order crystal-®eld strength S2, which is mainly responsible for the spectral features, has also been calculated: S2was found to be less for 7-coordinated sites than for 6-coordinated ones.

Fig. 3. RDF and CDF involving the doping ion.

Fig. 4. Experimental and simulated Eu³ emission spectra.

Inset: Contribution of each type of environment to the total

5D0!⁷F0,⁷F1transitions.

Fig. 2. Si±Si, Ge±Si and Ge±Ge radial and cumulative distribution functions in the SiO2±GeO2:Eu³numerical glass.

(4)

4. Discussion

4.1. Structure of the SiO₂±GeO₂ doped glass The Si±O and Ge±O distribution functions have revealed that the binary system is composed of SiO4and GeO4tetrahedral units. Furthermore, the Ge±Si CDF (Fig. 2) indicates that the SiO4 and GeO4tetrahedral units are not randomly scattered but homogeneously distributed in the matrix since a random distribution should have given, in average, two Ge and two Si atoms in the ®rst neighborhood of each tetrahedral unit. The homogeneous distribution of the tetrahedral units has also been con®rmed by the large number of bridging oxygens yielding a large fraction of Si±O±Ge linkages. The simulated glass is therefore orga- nized as a very ecient Si±Ge melt ruling out any phase segregation possibility. This structure agrees with the Raman spectra interpretations [2] which have shown that ``1 xSiO2±xGeO2binary glass structure is homogeneous with Si±O±Si, Si±O±Ge and Ge±O±Ge bonds and forms mixed network rather than phase segregation''. On the other hand, the small amount of non-bridging oxygens are mainly found in germanium local environment, and the weaker Ge±O bond energy led us to assert that the network structure is more ¯exible around germanium ions.

Concerning rings statistics, 3- and 4-fold rings are of particular interest because they can be related to the D₁and D₂default bands in the Raman spectra of amorphous SiO2. The decreases of the number of these 3-fold and 4-fold rings in the binary system compared to SiO2 and GeO2 glasses are also in agreement with the interpretations of the experimental Raman measurements which have shown a decrease of the intensity of the related bands [2].

4.2.Eu³ local environment and luminescence properties

The presence of germanium seems to be a fa- vorable factor for the higher coordinated sites: the above-mentioned gain in the network ¯exibility allows the doping ion to satisfy its natural trend to capture a maximum number of oxygens. By way of

proof, the comparison between Eu±Si and Eu±Ge RDFs (Fig. 3) shows that the medium-range environment of Eu³ is richer in Ge than in Si. In addition, a study of the linkage statistics has shown that, whether bridging or non-bridging, the oxygens of the Eu³ ®rst coordination shell are in most cases linked to germanium.

On the other hand, the local structure analysis has also shown that about 40% of the Eu³ ions are arranged in clusters: in this case, Eu³ ions are linked to each other by means of two or three oxygens giving rise to the double peak observed in the Eu±Eu RDF (Fig. 3).

The Eu³ luminescence spectra in Fig. 4 show that the inhomogeneous broadening due to site-to- site variations is well reproduced by the simulation. We have paid particular attention to the

5D0!⁷F1 transition because its features can be interpreted in terms of the second-order crystal-

®eld strength S2 which is strongly sensitive to the local structure, and consequently to the dierent site arrangements encountered in the host matrix.

The simulated spectrum has revealed that 6- and 7- coordinated sites yield dierent contributions to the overall features of this transition. The associ- ated S₂was found to be less for 7-coordinated sites than for 6-coordinated ones. Since the larger coordination number sites have been preferentially related to clustered con®gurations, the following assumptions can be proposed: the clustering trend could be explained by the tendency of doping ions to occupy low-energy sites which are mainly found in clustered con®gurations.

5. Conclusion

The structure analysis of the Eu³-doped germanosilicate glass shows that SiO4and GeO4units are homogeneously distributed in the glass. It is shown from rings statistics that the introduction of germanium in the silica glass decreases the number of defects such as 3- and 4-fold rings. Further- more, germanium is also responsible for a gain in the network ¯exibility which explains that Eu³

local environment contains more Ge ions than Si ions. The dopant clustering trend is quanti®ed and interpreted in terms of crystal-®eld strength

(5)

minimization. Finally, the agreement between experimental and simulated emission spectra allows the spectral features to be related to the dierent kinds of local structure encountered by the doping ions in this binary host matrix.

Acknowledgements

This research has been supported by a French±

Italian Program Galileo 99. A DLPOLY package from the CCLRC, Daresbury Laboratory, Nr Warrington, has been used at the IDRIS com- puting center (991007). The CERLA is supported by the Ministere Charge de la Recherche, the Region Nord/Pas de Calais and the Fond Eu-

ropeen de Developpement Economique des Regions.

References

[1] A. Polman, J. Appl. Phys. 82 (1997) 1.

[2] C. Duverger, S. Turrell, M. Bouazaoui, F. Tonelli, M. Montagna, M. Ferrari, Philos. Mag. B 77 (1998) 363.

[3] B.P. Feuston, S.H. Garofalini, J. Chem. Phys. 89 (1988) 5818.

[4] T. Peres, D.A. Litton, J.A. Capobianco, S.H. Garofalini, J. Non-Cryst. Solids 221 (1997) 34.

[5] A.C. Wright, J. Non-Cryst. Solids 159 (1993) 264.

[6] J.A. Desa, A.C. Wright, R.N. Sinclair, J. Non-Cryst. Solids 99 (1988) 276.

[7] S. Chaussedent, A. Monteil, M. Ferrari, L. Del Longo, Philos. Mag. B 77 (1998) 681.