ARTICLE IN PRESS Journal of Physics and Chemistry of Solids 68 (2007) 866–872 www.elsevier.com/locate/jpcs Ag-photodoping in Ge-chalcogenide amorphous thin films—Reaction products and their characterization M. Mitkovaa,, M.N. Kozickib a Department of Electrical & Computer Engineering, Boise State University, Boise, ID 83725-2075, USA b Center for Solid State Electronics Research, Arizona State University Tempe, AZ 85287-6206, USA Abstract We make a brief review on the effect of silver photodiffusion in Ge-chalcogenide glasses and report some of our recent results in this aspect. Using Raman spectroscopy and X-ray diffraction analysis we demonstrate that the hosting backbone undergoes depletion in chalcogen due to the specific conditions of photodiffusion and the diffusion products are silver chalcogenides. While in the Ge–Se system preliminary binary Ag—chalcogenides are forming, in the Ge–S system formation of Ag2 GeS3 is evidenced. This effect is related to the ability of the Ge–S glasses to form ethane—like structure at much lower Ge concentration than the Ge–Se glasses. For this type of structures is known that Ag replaces Ge to form homogeneous material, hence formation of Ag-containing ternary occurs. r 2007 Elsevier Ltd. All rights reserved. Keywords: A. Chalcogenides; A. Nanostructures; C. Raman spectroscopy; C. X-ray diffraction 1. Introduction 2. Nature of the photodoping process Silver photodiffusion—accelerated silver diffusion under the action of light is a unique feature of chalcogenide glasses. Ever since this effect was first encountered [1] it has been profoundly investigated because it can be used in many applications of chalcogenide glasses. We are using this process to form the active material—the solid electrolyte based on silver doped chalcogenide glass—for the programmable metallization cell (PMC) two terminal memory devices which have an electrode made from Ag and the other being electrochemically inert (W, Ni Pt, etc.) [2]. At application of a bias with positive voltage towards the Ag electrode formation of a robust but reversible conducting pathway by way of electrodeposition at low voltage and current reduces the resistance of the electrolyte by several orders of magnitude. In this manner non-volatile memory is realized in elements that are highly electrically and dimensionally scalable. Our understanding is that the process of photodiffusion is driven by the formation of charged defects in the chalcogenide glass, which form by illumination with light and create an electrical potential. The light that is critical for metal photodissolution is absorbed at or near the interface between the reacted and unreacted (doped and undoped) chalcogenide layers [3,4]. In this process, electrons are trapped by silver ions [5,6], while holes move further into the chalcogenide film and are trapped there. The electric field formed by the negatively charged chalcogen atoms and positively charged silver ions can be sufficient for the silver ions to overcome the energy barrier at the interface. Therefore the penetration of the metal into the chalcogenide during photodoping is due to the difference in electrochemical potentials and the process was considered to be similar to that occurring in a galvanic cell, where the more electropositive metal is dissolved into the electrolyte [7]. Kluge [8] considered the process of photodiffusion of metals in chalcogenides as an intercalation reaction. The main reason this can be realized in chalcogenide glasses is the fact that they possess relatively rigid covalent bonds Corresponding author. Tel.: +1 208 426 3395; fax: +1 208 426 2479. E-mail address: MariaMitkova@boisestate.edu (M. Mitkova). 0022-3697/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jpcs.2007.01.004 ARTICLE IN PRESS M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 mixed with soft van der Waals interconnections. This type of structure ensures formation of voids and channels where the diffusing ions can migrate and can be hosted. The reaction can be efficient when the reversible transport of ions and electrons can be achieved, accompanied by formation of bonds with the host matrix, according to the reaction: suggest that the kinetics of the diffusion process is highly influenced by free space availability in the chalcogenide matrix and the chemical reaction between S and Ag in the case of sulfur-rich glasses. When Ge rich glasses are considered, the photodiffusion rate could be related to the formation of the number of chemically active photo excited localized centers [20]. (1) This reaction describes the transition of an initially twofold covalently bonded chalcogenide atom ðC02 Þ into a C 1 charged unit possessing only a single covalent bond and an excess electron that establishes an ionic bond with Agþ ðMþ Þ. Eq. (1) shows the importance of the potential þ in forming the new C bonds of the intercalation 1M product. The possible number of these bond-units is fairly high as the chalcogenide glasses are capable of forming a number of single C 1 centers under the influence of light illumination. Once silver is introduced into the chalcogenide glass, its further migration into the chalcogenide glass continues. The photodiffusion kinetics depends on a number of factors such as light intensity [9], light wavelength [10], temperature [11], pressure [12], external electric field [13], composition of the hosting glass [14], and the atmosphere in which the diffusion process is performed [15]. Many details in this respect are given in the work of Kolobov and Elliott [16]. In this review, we will specifically discuss the data concerning Ge-chalcogenide systems. 3. Silver photodiffusion in the Ge–S system 3.1. Basic data The most profound investigation of silver diffusion in sulfur-rich Ge glasses has been made by Oldale and Elliott [17]. They have found that the Ag photo-dissolution rate in a-Ge29 S71 has no induction period and the process has 2 stages—phase 1, which is an acceleratory stage leading to a maximum in the photodissolution rate, and phase 2, a final deceleratory stage. As the time to develop the acceleratory stage shows a spectral dependency, it is obvious that the absorption of actinic radiation in the photodoped film is responsible for this stage of the photodissolution kinetic profile. Maruno and Ban [18] reported that when the silver film is deposited on previously illuminated Ge30 S70 film, the diffusion process proceeds very slowly. Speculations are made upon the changes in the structure of the chalcogenide film due to the light illumination. However, we assume that oxidation of the chalcogenide film during the initial illumination could also be responsible for the occurrence of this effect. The defects that can be created by illumination are indeed the driving force for the oxidation process. Kawaguchi and Maruno [19] obtained very important data about the compositional dependences on the initial photodoping rate and photodoping kinetics. Their data 3.2. XRD data generated in our research We contributed with some research related to GeS2 photodiffused with Ag [21] towards further understanding of what the diffusion products are and how diffusion affects the hosting backbone, combined with data about the influence of annealing at 150, 300 and 430 C since in the real world of application, the glasses are usually processed at similar temperatures. The X-ray diffraction (XRD) spectra of photodiffused films are shown on Fig. 1 (a)–(d). As shown on the figure diffusion products are Ag2 S and Ag2 GeS3 and only during annealing at 430 C Ag8 GeS6 forms. The most impressive result of this experiment is the fact of a fast growth of the diffusion products during annealing. The studies of the density of Ge–S glasses show that the stoichiometric composition GeS2 is expected to have the highest density [22]. However, the real composition of the backbone after introduction of Ag is much more Ge rich because of the reaction of Ag with the negatively charged defects and it is expected to have much lower density. This easily allows formation of channels, which, because of the low polymerization, can offer substantial space where the Ag containing phase is located. Numerical simulations of the structure also suggest their existence [23]. Therefore, we believe this enables introduction of high amount of Ag and the rapid growth ^ + + Intensity (Arb. units) þ C02 þ e þ Mþ ! C 1M . 867 430 C 300 oC * 20 d + * * + * 150oC RT * * ^ * o + * ** 30 + + + + + 40 2 Theta (Deg.) + c + b ** ** a 50 Fig. 1. XRD data for: (a) photodiffused Ge–S film; (b) photodiffused Ge–S film annealed at 150 C; (c) photodiffused Ge–S film annealed at 300 C; (d) photodiffused Ge–S film annealed at 430 C; * denotes appearance of Ag2 GeS3; þ denotes appearance of Ag2 S; ˆ denotes appearance of Ag8 GeS6 . Figure taken from Balakrishnan et al. [21]. ARTICLE IN PRESS M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 Ge-Ge bond vibration (ethane-like structure) 1.0 d 0.5 0.0 1.0 c 0.5 Intensity (Arb. units) Intensity (Arb. units) symmetric stretch of Ge(S1/2)4 tetrahedra clu mo ster de ed g 1.0 e Sclu S str ste etc r e h fr dg om ed im ers 868 0.0 200 400 300 Rel. Wavenumber (cm-1) 0.5 0.0 1.0 b 0.5 500 Fig. 2. Raman spectra for pure Ge–S films at room temperature. Figure taken from Balakrishnan et al. [21]. 0.0 1.0 a 0.5 of the diffusion products through agglomeration of the nanoclusters as established by the XRD data—Fig. 1(a)–(d). 3.3. Raman data generated in our research The Raman results for pure Ge–S films Fig. 2 taken from [21] show appearance of relatively high intensive mode at 343 cm1 of the A1 symmetric stretch of GeðS1=2 Þ4 [24] combined with scattering at 370 and 427 cm1 from the edge sharing structures. There is a well resolved peak at 252 cm1 corresponding to the vibrations coming from the ethane like structures available in glasses containing more than 33 at% Ge [25]. The deconvolution of the Raman modes between 340 and 440 cm1 manifests formation of a peak at 427 cm1 . We believe that this signal occurs from the vibrations of the S chains available due to formation of wrong bonds in these glasses. The variety of building blocks forming this film is indication for the specific structure that develops in the Ge–S system. Boolchand et al. [26] have demonstrated that the formation of ethanelike structural units containing Ge–Ge bonds starts at the stoichiometric composition GeS2 . On grounds of stoichiometry an equivalent number of S–S bonds are available. This is a very rare case in chalcogenide glasses in which almost all possible building blocks emerge in one composition. The implication of this effect is that upon illumination metastable states [27] occur on sulfur atoms with different surroundings. When Ag is photodiffused in the Ge–S glass, well expressed changes occur in the Raman activity of the newly formed material, (Fig. 3(a)). One observes intensity growth of the mode at 250 cm1 , indicating formation of a higher number of ethane like structural units. Meanwhile, the relative intensity of the mode at 343 cm1 is reduced, and the vibrations at 370 and 400 cm1 strengthen. We attribute the latter modes to development of thiogermanate 0.0 200 300 400 500 Rel. Wavenumber (cm-1) Fig. 3. Raman spectra for: (a) photodiffused Ge–S film at room temperature; (b) photodiffused Ge–S film annealed at 150 C; (c) photodiffused Ge–S film annealed at 300 C; (d) photodiffused Ge–S film annealed at 430 C. Solid lines are fitted results. Figure taken from Balakrishnan et al. [21]. bonds ðGe2S Þ forming metathiogermanate tetrahedra ðGeS2 3 Þ and dithiogermanate tetrahedra ðGeS2:5 Þ as suggested by Kamitsos et al. [28]. Illumination with light causes formation of defects not only on S atoms forming S chains but also on S that is part of other structural units. It is for this reason that we observe formation of both Ag2 S and Ag2 GeS3 after Ag diffusion by which the Ag ions react with the photoinduced defects in the chalcogenide glass as suggested by Eq. (1). The growth of the mode at 250 cm1 indicates significant sulfur depletion of the initial composition (Fig. 3) of the hosting backbone after Ag is photodiffused in the Ge–S film (Fig. 3) and a large number of ethane-like structures are formed. After annealing of Ag diffused films, we realize that their structure keeps the initial character and the intensity of the ethane like structures increases with annealing temperature (Fig. 3(b)–(d)). The decreasing ratio between the intensity of the mode at 252 cm1 and the mode at 334 cm1 could be related to some continuing reaction between the three elements at annealing. Note that the peak at 252 cm1 is very stable due to a higher rigidity of the structure and better filling of the intercluster space with introduction of Ag in the film. This prevents incidence of intercluster changes. In fact, due to the low dimensional nature of the clusters, the Ge–S host is expected to be less strained ARTICLE IN PRESS M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 compared to its isoionic counterpart, Ge–Se glass. This, along with the bond strength of the covalent bonding in the studied glass accounts for the lower polymerization relative to the Ge–Se system. As a result a more relaxed structure is formed where the Ag-containing products experience much lower pressure from the surrounding backbone and only the low-temperature forms of the respective compositions occur that are known to have larger volume than the hightemperature forms. The dramatic decline in the scattering intensity of the edge and corner sharing tetrahedra during annealing at 430 C (Fig. 3(d)) can be related to formation of a new ternary composition—Ag8 GeS6 (Fig. 1(d)). For this composition the structure is formed by isolated GeS4 tetrahedra as well as S atoms that are not bonded to the Ge atoms [29]. The anion parts GeS4 and S are connected by Ag atoms to form 3D structure. The Ag atoms are bi-, three- and fourfold coordinated with S [30]. In other words, formation of Ag8 GeS6 brings about a serious depolymerization of the structure and hence decreases the intensity of the modes related to particular structural coordination of the Ge–S tetrahedra. This effect can be also accompanied with high concentration of Ag on the surface at the highest annealing temperature that would be the most natural effect considering the highest number of wrong bonds that are related to the surface defects. The scattering coming from this Ag rich medium with a narrow band gap will significantly reduce the scattering from the Ge–S host. However, note that the second order Si substrate mode at 303 cm1 yields almost consistent intensity in all the fits, indicating that the laser beam reaches with ample energy the studied material and the sample penetration depth through the films is not influenced by Ag clustering near the surface. 4. Silver photodiffusion in the Ge–Se system 4.1. Basic data Most extensive data about Ag diffusion in Ge–Se glasses have been reported by Kluge et al. [31]. They have found that the diffusion kinetics and the total amount of diffused Ag are very closely related to the composition of the hosting backbone. While for a backbone containing 75290 at% Se there is almost no induction period for the diffusion process, at lower Se concentration this period grows with decreasing Se content. This is somewhat correlated with the depth profile of the Ag diffusion. At high Se concentration, for GeSe5:5 , a step-like profile is found [32,33] by which a Ag-depleted layer lies over the Ag-enriched layer situated just above the substrate. Leung et al. [34] also confirm that surface diffusion is much smaller than the bulk diffusion. Indeed, some irregularities and discontinuity of the silver doped film with formations of islands have been found also by Rennie et al. [35]. For GeSe2 glasses, the process is slower and follows the classical distribution [36]. Wagner et al. [37] established a significant difference in the diffusion profiles of laterally 869 diffused Ag in Ge20 Se80 and Ge40 Se60 glasses, with sharp edge diffusion for the Se-rich glass and a classical diffusion profile for the Ge-rich glass. Their interpretation of this effect is related to the existence of two glass-forming regions in the Ge–Se–Ag system. While for the Ge-rich glasses the diffusion process goes through compositions characteristic only for one of the sub regions, in the case of Se-rich glasses the composition of the diffused product resembles those of the two regions. This requires some structural rearrangements that affect the diffusion profile. Structural investigations on the photodiffused material reveal formation of a heterogeneous structure after photodiffusion [38,39]. Chen and Tai [38] report formation of bcc Ag2 Se, Ag2 SO4 and small amounts of orthorhombic Ag2 Se when Ag is diffused into GeSe2 glass and bcc Ag2 Se and traces of free Ag when the diffusion process is conducted in Ge0:1 Se0:9 glass. We assume that these differences in the diffusion processes are closely related to the availability of space and channels for the diffusion of Ag in the particular hosting glasses. Formation of Ag2 Se has been submitted also by Zembutsu [39]. Considering the existing results, Kawaguchi et al. [40] proposed a schematic model for the evolution of the structure of the chalcogenide glasses during Ag diffusion that depict the formation of the two phases—Fig. 4. 4.2. XRD data generated in our research We studied [41] the diffusion products in the Ge–Se system and their growth during annealing at 85, 110, 125 and 150 C in hosting materials with composition Ge20 Se80 , Ge30 Se70 , Ge33 Se67 and Ge40 Se60 . Fig. 5 gives representative curves of the XRD spectra of the photodiffused glasses for an initial composition of Ge33 Se67 . In all cases, the hosting Ge–Se glass remained amorphous during the annealing while the silver containing species formed nanocrystals. We found orthorhombic, bAg2 Se, as well as cubic, aAg2 Se with Ag8 GeSe6 appearing only when Ag is introduced in a host containing 33 at% and higher concentration of Ge. The crystals forming after diffusion a b c Ag-poor phase Ag-rich phase d e f Ag particles Fig. 4. Schematic illustration of the change of structure of ðGex SeðSÞ1x Þ1y Agy deposited films with increasing Ag content: (a) shows the initial glass in which the amount of Ag gradually increases (b)–(e) until Ag particles phase separate (f). Figure taken from Kawaguchi et al. [40]. ARTICLE IN PRESS M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 870 * + + + ^ + * * d Intensity (Arb. units) * * * + + ^ * + * + + + + ^+ + + ^ + + * * * c * b * * 30 50 40 2 Theta, Deg. 60 b Ge30Se70 c Ge33Se67 d Ge40Se60 e after Ag diffusion Film resulting after photodiffusion a 100 20 Ge20Se80 a Intensity, Arb. Units * 70 Fig. 5. Representative XRD plots of Ge33 Se67 glass photodiffused with Ag annealed (a) at 85 C for 15 min, (b) at 85 C for 120 min, (c) at 150 C for 15 min and (d) at 150 C for 120 min; * peaks characteristic for Ag8 GeSe6 ; ˆ peaks characteristic for aAg2 Se; þ peaks characteristic for bAg2 Se. Some peaks were reduced to fit on a single graph. Figure taken from Mitkova et al. [41]. are relatively small because they can only form in the free interspaces available in the matrix of the hosting glass. Although in the case of Ge20 Se80 glass the initial structure is floppy, following the initial silver inclusion and formation of Ag2 Se, the glass structure becomes depleted in Se and stiffer. The internal space limitation produces the same effects as elevated pressure, stabilizing some clusters in the high temperature form which has the closest packing. With Ge-enrichment of the hosting backbone, the intensity of the peaks of aAg2 Se becomes higher, suggesting reflectance from a larger number of planes. At the same time, Ag8 GeSe6 clusters are formed and we assume that these occur at terminal defects on the Ge–Se tetrahedra in the case of the Ge33 Se67 host or develop within the volume of the films when Ag is diffused in a Ge40 Se60 host. Indeed, Mössbauer spectroscopy definitely shows that replacement of Ge by Ag occurs in Ge-rich glasses [42] so a combined effect could be the reason for the development of the ternary composition. 4.3. Raman data generated in our research Raman features of initial hosts closely match those of bulk materials with the same composition, as illustrated in Fig. 6(a)–(d). However, after diffusion, the spectra of all samples show a vibrational band at 180 cm1 and a higher frequency band at 200 cm1 independent of the composition (Fig. 6(e)), suggesting the formation of a structure containing ethane-like units with Ge–Ge bond as well as the Ge–Se tetrahedra. These spectra remained unchanged following the moderate annealing. 200 300 400 500 Raman shift, cm-1 Fig. 6. Raman spectra of the undoped Ge–Se glasses and spectrum of the photodiffused material. Compositions are noted in the figure. Figure taken from Mitkova et al. [41]. We assume that as in the case of Ge–Se glasses, the illumination with light causes formation of charge defects that can react with Ag and form the diffusion product. This fact has important consequence since some Se is extracted from the initial Ge–Se backbone to react with the diffused Ag. So the remaining chalcogenide glass backbone becomes Se deficient, as demonstrated by the appearance of a Raman signature that is characteristic of a Ge-rich glass, independent of the initial composition of the host. In this composition the underlying molecular phase consists of face-sharing quasi one-dimensional ethane-like Ge2 ðSe1=2 Þ6 chain fragments whose presence is manifested on the Raman spectra by the appearance of the mode at 180 cm1 [43] depicted in Fig. 6(e). The Raman spectrum of the resulting material shows a lower intensity ratio between the modes at 180 cm1 and the mode of the Ge-tetrahedral units at 200 cm1 when compared to the intensity ratio of these modes for a Ge40 Se60 initial glass film indicating that the number of ethane like units is lower than in Ge40 Se60 glass. However this structure still contains Ge–Ge bonds. They are the result of the spontaneous reaction of Ag with charged metastable states on the chalcogen initiated by light illumination and with charged defects occurring at bond conversion [44]. This reaction will be preferred since the energy that it requires is less than the energy for the Ge–Se bonding (48.4 vs. 113 kcal/mol). We suggest that this, together with space organization in the material is the reason for the extraction of some Se from the Ge–Se backbone for the formation of Ag2 Se in addition to the reaction of Ag with the initially available free Se chains. The structure of Ge–Se backbone formed after photodoping is depolymerized to some extent due to the extraction of Se and formation of crystalline products. It is for this reason that the organization of the photodiffused hosting glass does not change with the moderate annealing applied, ARTICLE IN PRESS M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 as happens with pure Ge–Se films [45] where the local stressed configurations with a high free energy relax through breaking of the Ge–Ge bonds and formation of Ge–Se corner-sharing units due to reaction with Se–Se wrong bonds. 5. Conclusions In this work we gave a brief review with extended references of the published results about the Ag photodiffusion in Ge–S(Se) chalcogenide glasses and combined them with our recent results. We demonstrate that the photodiffusion effect can be well characterized by XRD method which gives direct evidences about the diffusion products that are crystalline and by Raman spectroscopy which supplies data about the structure of the hosting Ge–S(Se) backbone. We found out differences in the photoinduced effects in the investigated systems which can be summarized as follows: For the Ge–S system: The diffusion products are nanocrystals of Ag2 S and Ag2 GeS3 which grow via agglomeration with increasing the annealing temperature. At 430 C Ag8 GeS6 forms which is product of reaction of agglomerated Ag2 S with the hosting backbone. The Raman data about the hosting Ge–S backbone show that it becomes more rigid and Ge-rich after the act of Ag photodiffusion in it. The intensity of the Raman mode characterizing the formation of Ge–Ge bond after introduction of Ag grows with the annealing temperature up to 430 C. At this temperature the intensity characterizing the Ge–S tetrahedra decreases drastically because of structural rearrangement and formation of Ag8 GeS6 ternary which is build up by isolated GeS4 tetrahedra and this essentially affects the structure of the host. For the Ge–Se system: Regardless of the initial composition of the hosting glass, the photodiffused material shows Raman features characteristic for Ge-rich material. The glassy component becomes Se-deficient due to consumption of Se in the formation of the diffusion products. The diffusion products are nanocrystalline regions dispersed into the glassy matrix and their composition is dependent upon the hosting glass composition and develops from Ag2 Se to a combination of Ag2 Se and Ag8 GeSe6 with enrichment of the host in Ge. The cluster size of the crystalline products depends on the molar volume of the host in close relation to its rigidity. Isothermal annealing at moderate temperatures results in diffusion limited slow growth of the Ag2 Se clusters and homogeneous growth of the Ag8 GeSe6 clusters. 871 References [1] M.T. Kostyshin, E.V. Mikhailovskaya, P.F. Romanenko, Fiz. Tverd. Tela 8 (1966) 571 (Sov. Phys. Solid State (1966) 451). [2] M.N. Kozicki, M. Park, M. Mitkova, IEEE Trans. Nanotechnol. 4 (2005) 331. [3] T. Wagner, M. Frumar, V. Suskova, J. Non-Cryst. Sol. 128 (1991) 197. [4] J.H.S. Rennie, S.R. Elliott, J. Non-Cryst. Sol. 97&98 (1987) 1239. [5] A.V. Kolobov, S.R. Elliott, M.A. Taguirdzhanov, Philos. Mag. B 61 (1990) 859. [6] I.Z. Indutni, V.A. Danko, A.A. Kudryavtsev, E.V. Michailovskaya, V.I. Minko, J. Non-Cryst. Sol. 185 (1995) 176. [7] A.V. Kolobov, G.E. Bedel’baeva, Philos. Mag. B 64 (1991) 21. [8] G. Kluge, Phys. Stat. Sol. (A) 101 (1987) 105. [9] A. Urena, M. Fontana, B. Arcondo, M.T. Clavaguera-Mora, J. NonCryst. Sol. 320 (2003) 151. [10] S.A. Lis, J.M. Lavine, Appl. Phys. Lett. 42 (1983) 675. [11] M.T. Kostyshin, V.I. Minko, Ukr. Fiz. Zh. 29 (1984) 1560. [12] Ke. Tanaka, Phys. Rev. Lett. 65 (1990) 871. [13] G.E. Bedel’baeva, A.V. Kolobov, V.M. Lyubin, Fiz. Tech. Polupr. 25 (1991) 197. [14] P.J. Ewen, A. Zakery, A.P. Firth, A.E. Owen, J. Non-Cryst. Sol. 97–98 (1987) 1127. [15] A.V. Kolobov, V.M. Lyubin, J. Troltzsch, Phys. Stat. Sol. (A) 115 (1989) K139. [16] A.V. Kolobov, S.R. Elliott, Adv. Phys. 40 (1991) 625. [17] J.M. Oldale, S.R. Elliott, J. Non-Cryst. Sol. 128 (1991) 255. [18] S. Maruno, S. Ban, Jpn. J. Appl. Phys. 19 (1980) 97. [19] T. Kawaguchi, S. Maruno, J. Appl. Phys. 71 (1992) 2195. [20] R. Ishikawa, Sol. State Comm. 30 (1979) 99. [21] M. Balakrishnan, M.N. Kozicki, C.D. Poweleit, S. Bhagat, T.L. Alford M. Mitkova, J. Non-Cryst. Sol. (2007), to be published in spring. [22] H. Takebe, H. Maeda, K. Morinaga, J. Non-Cryst. Sol. 291 (2001) 14. [23] M.F. Thorpe, private communication. [24] G. Lucovsky, F.L. Galeener, R.C. Keezer, R.H. Geils, H.A. Six, Phys. Rev. B 10 (1974) 5134. [25] K. Jackson, A. Briley, S. Grossman, D.V. Poresag, M.R. Pederson, Phys. Rev. B 60 (1999) R14 985. [26] P. Boolchand, J. Grothaus, M. Tenhover, M.A. Hazle, R.K. Grasselli, Phys. Rev. B 33 (1986) 5421. [27] K. Shimakawa, A. Kolobov, S.R. Elliott, Adv. Phys. 44 (1995) 475. [28] E.I. Kamitsos, J.A. Kapoutsis, G.D. Chryssikos, G. Taillades, A. Pradel, M. Ribes, J. Sol. State. Chem. 112 (1994) 255. [29] P. Armand, A. Ibanez, J.-M. Tonnerre, B. Bouchedt-Fabre, E. Philippot, Phys. Rev. B 56 (1997) 19852. [30] D. Carre, R. Ollitrault-Fichet, J. Flahaut, Acta Cryst. B 36 (1980) 245. [31] G. Kluge, A. Thomas, R. Klabes, R. Grötzschel, P. Süptitz, J. NonCryst. Sol. 124 (1990) 186. [32] R. El Ghrandi, J. Calas, G. Galibert, Phys. Stat. Sol. (A) 123 (1991) 451. [33] J. Calas, R. El Ghrandi, G. Galibert, A. Traverse, Nucl. Instrum. Meth. Phys. Res. B 63 (1992) 462. [34] W. Leung, N. Chung, A.R. Neureuther, Appl. Phys. Lett. 46 (1985) 543. [35] J. Rennie, S.R. Elliott, C. Jeynes, Appl. Phys. Lett. 48 (1986) 1430. [36] J.H.S. Rennie, S.R. Elliott, J. Non-Cryst. Sol. 77&78 (1985) 1161. [37] T. Wagner, R. Jilkova, M. Frumar, M. Vlcek, Int. J. Electr. 77 (1994) 185. [38] C.H. Chen, K.L. Tai, Appl. Phys. Lett. 37 (1980) 605. [39] S. Zembutsu, Appl. Phys. Lett. 39 (1981) 969. [40] T. Kawaguchi, S. Maruno, S.R. Elliott, J. Appl. Phys. 79 (1996) 9096. [41] M. Mitkova, M.N. Kozicki, H.C. Kim, T.L. Alford, J. Non-Cryst. Sol. 352 (2006) 1986. ARTICLE IN PRESS 872 M. Mitkova, M.N. Kozicki / Journal of Physics and Chemistry of Solids 68 (2007) 866–872 [42] M. Mitkova, Yu. Wang, P. Boolchand, Phys. Rev. Lett. 83 (1999) 3848. [43] P. Boolchand, in: P. Boolchand (Ed.), Insulating and Semiconducting Glasses, World Scientific, Singapore, 2000, p. 214. [44] N. Bondar, N. Davydova, V. Tishchenko, M. Vlcek, J. Mol. Struct. 555 (2000) 175. [45] Y. Wang, K. Tanaka, T. Nakaoka, K. Murase, J. Non-Cryst. Sol. 299&302 (2002) 963.