Open Access
Add to BibSonomyAbstract
Thin films of GeAsS glass are prepared by e-beam evaporation technique. Photoinduced birefringence (PIB) is studied as function of the As content with concentrations ranging from 10% to 40%. Raman spectroscopy is used as additional tool to explain the corresponding changes undergone by the material system. The breakdown of homopolar bonds is suggested as a possible mechanism of photo induced structural changes leading to the creation of the PIB.
©2013 Optical Society of America
1. Introduction
Chalcogenide glasses (ChGs) have attracted great deal of attention thanks to their potential applications in integrated photo programmable infrared photonic circuits [1]. Very often, the research on ChGs was done on material systems, which were relatively easy to fabricate and had relatively large photosensitivity [2–7]. An example of such system is the arsenic sulphide (AsS), which is an excellent glass forming compound with good chemical stability and providing wide transmission range (from visible to IR), high index of refraction, photosensitivity and nonlinear optical properties [8]. However, this material also has important drawbacks, such as the relatively low glass transition temperature (Tg ≈180°C), which may compromise its practical applications. Thus, for the present study we have selected the glassy system Ge-As-S since the incorporation of Ge into AsS increases the glass transition temperature (up to Tg ≈350°C) and improves the mechanical and thermal properties of the glass [9].
The structure of those materials consists of interconnected trigonal (As2S3) and tetrahedral (GeS(1/2)4) units [10]. Optical studies were already reported on annealed films of such glasses fabricated by thermal vapor deposition. Thus, scalar photosensitivity properties of a glass composition (As20S60Ge20), containing relatively low content of As, were described [11]. The dependence of photo-induced bandgap changes upon the content of As was also reported [12]. Photo-induced birefringence (PIB) and photo-induced dichroism phenomena were studied in a glass composition containing low amount of As, e.g. (As8S60Ge32) [13]. We think that the understanding of corresponding mechanisms and structure formation as a function of As concentration is very interesting and useful task. In the present work, we present such detailed study of the PIB in thin films based on GeAsS. In addition, the e-beam technique used here for thin film deposition is different (allows better control of composition). Also, we start our investigations (in the current report) by using fresh (non annealed) films. Finally, Raman spectroscopy is used as parallel tool for their structural analysis.
2. Experimental
Glass samples were prepared by melting high purity starting elements (Ge, As and S) in a quartz ampoule evacuated up to 10−3 Pa. All procedures including synthesis, distillation and glass production were carried out in a closed system. After melting in a rocking furnace at 850°C during 8 h the ampoule was removed, quenched in water and annealed at the glass transition temperature, Tg, near 400°C. Glass rods about 60 mm in length and 15 mm in diameter were obtained after cutting the quartz ampoule. Pieces of around 2 mm in thickness were cut and polished. The obtained samples ranged in color from yellow to dark-red, corresponding to a shift of the visible absorption edge λvis from 450 nm to 630 nm.
Films of 1.5 μm thickness were deposited (by electron beam evaporation from the crushed ingots with an electron beam voltage of 4 k in a vacuum of 10−6 Pa) onto BK7 glass substrates held at room temperature. The deposition rate was 10 Å/s (measured continuously by a quartz-crystal monitor, Temescal FTM). It is known that such low deposition rate produces a chemical composition, which is very close to that of the bulk (starting) material. Indeed, the EDX analysis of our samples indicated that those compositions are close to within 1 at.% (see below). The study of those films was carried out in an electron microscope (FEI, Quanta 3D FEG).The corresponding analysis result is presented in the Table 1 for only one composition as example (see later for detailed analyses).
Table 1. Material Analyses of the Composition Ge25As30S45a
3. Results
The transmission spectra (obtained by means of Varian Cary 1500 spectrophotometer) are shown on Fig. 1 for different glass compositions of thin films with thicknesses of ≈1.5 µm. A shift of the bandgap position is observed as the As content increases.Namely, the corresponding numerical values of the bandgap (calculated from our experimental data by using the Tauc method [14]) are as follows: 3.26 eV for Ge25As10S65, 3.40 eV for Ge25As20S55, 2.42 eV for Ge25As30S45, 2.28 eV for Ge25As35S40, and 2.08 eV for Ge25As40S35. We note that there are two “groups” of compositions (with “low” and “high” contents of As), which have correspondingly regrouped bandgaps (see hereafter).
4. Photosensitivity study in thin films
To study the photosensitivity of thin ChG films, a linearly polarized CW Argon ion laser (operating at 514.5 nm) was used for excitation at normal incidence (from the ChG side). Samples were exposed during 60 min with different intensities, varying from 2 W/cm2 to 15 W/cm2. We first notice, from the Table 1, the significant increase of the amount of oxygen after photo exposition indicating a photoinduced oxidation of the film. In addition the absorption coefficients were determined from transmittance measurements using the Lambert-Beer-Bouguer law and those spectral measurements show a clear shift of the bandgap towards the shorter wavelengths (the so-called photo bleaching process), as presented in the Fig. 2 for one glass composition (see also [15]).Note that, in the As2S3 composition, usually the opposite process is observed (photo darkening).
Fig. 2 Absorption coefficients of the Ge25As30S45 as function of probe’s energy obtained for photoexposition intensity of 8W/cm2 for 60 min. The solid curve corresponds to the unexposed case; the dashed curve corresponds to the photoexposed case.
The experimental setup, used to study the PIB, is schematically shown in Fig. 3.The photo excitation was performed by using the s-polarized beam of the CW Argon ion laser with an angle of incidence of 10°. The power of this excitation beam was controlled by a filter (F1). Glan prism polarizer (P) was used to obtain linear polarization. A CW He–Ne laser beam (operating at 632.8 nm) was used as probe (at normal incidence on the sample) to analyze the PIB. Its polarization was oriented (by means of a half-wave plate, λ/2) at 45° with respect to the polarization of the excitation beam. The two laser beams were aligned to overlap on the sample (S) with the help of a dielectric mirror (M). After passing through the sample, the He–Ne beam was filtered by an analyzer (A) that was crossed with respect to the original polarization of the probe beam (defined by the λ/2 plate). The intensity of the He–Ne beam was set as low as possible (0.1 mW/cm2). An interferential filter (F2) was used to transmit wavelengths above 600 nm (to reduce the noise from the excitation beam). A diaphragm (d) was also used, in front of the photo detector (D), to further reduce the noise.
Fig. 3 The experimental setup used for the study of PIB: P-polarizer, M–mirror, λ/2- half wave plate, S-sample; A-analyzer; F1 and F2-filters, d-diaphragm, D-detector.
The as-deposited ChG thin films were optically isotropic, and thus there was no initial transmission through the crossed polarizer-analyzer system (Fig. 4). When the excitation beam was switched on, we observed an increase of probe transmission followed by stabilization after t > 6000 sec. We observed a partial relaxation of the PIB when the excitation beam was switched off leading to a significant remnant PIB.The theoretical fit of obtained experimental curves was made by a bi-exponential function I = I0 + I1 exp(-t/τ1) + I2 exp(-t/τ2) and the corresponding coefficients for two excitation contributions were found to be comparable (of the same order of magnitude) in terms of their amplitudes both during excitation (I1/ I2 = 1.75) and relaxation (I1/ I2 = 3.06). However, their characteristic times were drastically different (at least by an order of magnitude) during the excitation (τe1 = 106 s and τe2 = 1818 s) and during the relaxation (τr1 = 11 s and τr2 = 115 s). Based on those data one cannot conclude about the exact nature of the excitation microscopic mechanisms. Yet, the partial (approximately 30%) relaxation of the PIB (after the removal of the excitation laser beam) demonstrates one very fast (τr1 = 11 s) and important channel of relaxation and another, relatively smaller (by a factor of 3) and significantly (τr2 = 115 s) slower, channel. This information is used hereafter in our further analyses.
Fig. 4 Typical cycle of excitation and partial relaxation of the PIB in the Ge25As30S45 film. The solid curve shows the experimental result and the dashed one (behind the experimental curve) represents the fitted curve. The thickness of the film was 1.5 µm and the excitation intensity was 8W/cm2.
We have studied the behavior of PIB for various As contents. We have chosen the 4 W/cm2 exposition intensity to study the PIB in GeAsS films of the same thickness (≈1.5 µm), but containing different levels of As. Figure 5 shows the obtained dependence of the (stabilized under excitation) PIB upon the content of As for concentrations up to 35% (the probe beam was strongly absorbed for glass compositions with higher As contents).We can see, from Fig. 5, that the PIB shows a change in trend (an “inflection point”), that is, it increases with the increase of the content of As up to 30% (the composition Ge25As30S45), but then decreases for higher As contents. This also is important information to be used in our discussion (see hereafter).
Various excitation intensities were used to study the PIB for the “optimal” glass composition Ge25As30S45. For each intensity, we have studied the PIB behavior at different parts of the same ChG sample to avoid the errors due to sample to sample variation. As expected, the growth of the PIB was faster for higher intensities (not shown here) and the growth process stabilized at different levels (see Fig. 6); higher levels of PIB were achieved for higher excitation intensities. The obtained curves also were fitted by bi-exponential functions to characterize the dynamics of excitation, the established values of the PIB as well as their partial relaxation. It is worth to note that the use of higher intensities (>15 W/cm2) in our samples was limited by the damage threshold of the ChG films (in fact by their photo oxidation and evaporation of As).As one can see, from the Fig. 6, the PIB first increases almost linearly with the excitation intensity and then saturates for intensities at the order of 8 W/cm2. It seems that the PIB is created by a limited number of pre-existing photosensitive units, but not newly photo created ones. We note that very large values of PIB were achieved (Δn ≈0.03 ± 0.003) at excitation intensities of the order of 8 W/cm2. Compared to typical ChG compositions, those are rather high values of PIB, particularly for the ChG composition with such high value of Tg.
Fig. 6 The dependence of the established (saturated) value of PIB upon the excitation intensity for the composition Ge25As30S45.
To further study the possible mechanisms involved in the process of described above PIB, Raman scattering measurements were performed (by using a LABRAM 800HR Raman spectrometer from Horiba Jobin Yvon) with normal back scattering configuration in the wavenumber region, ranging from 150 cm−1 to 600 cm−1. The 632.8 nm line of a CW He-Ne laser (30 mW/cm2) was used as light source. Duration of acquisition was kept very short (<10 sec) to ensure the absence of material modification caused by the irradiation of the He-Ne laser.
Figure 7 shows the measured Raman spectra for the compositions of Ge-As-S films (versus the ratio As/S) used in the study of PIB (see Fig. 5). In the Raman spectra of GeAsS films, four bands and one shoulder may be clearly seen at 215, 242, 345, 490 and 430 cm−1, respectively. To facilitate their analyses, the recorded spectra have been first reduced by subtracting a polynomial baseline linking the first minimum (before the first band observed near to 200-250 cm−1) and the minimum at the end of the spectrum (at 600 cm−1). Then the spectra were normalized at 242 cm−1 (which corresponds to the vibrational mode of Ge-Ge bond) in order to observe the dynamical changes of homopolar As-As bonds (corresponding to 215 cm−1) and heteropolar Ge-S, As-S bonds (corresponding to 345 cm−1).
Fig. 7 Normalized Raman spectra of thin Ge-As-S films for different compositions: Ge25As10S65 (dotted black line), Ge25As20S55 (short dash dotted red line), Ge25As30S45 (dashed green line), Ge25As35S40 (short dotted blue line), Ge25As40S35 (solid cyan line).
The structure of GeAsS glasses already was studied by several authors using Raman spectroscopy [9,10,16,17] and other techniques such as Extended X-Ray Absorption Fine Structure [18] or Fourier Transform Infrared Spectroscopy [10]. It consists mainly of a three-dimensional network of Ge (S1/2)4 tetrahedra and As (S1/2)3 pyramids. The main Raman signature of those structural units is centered at 345 cm−1 and is attributed to overlapped symmetric stretching Ge-S and As-S bonds. In our results, we note (in Fig. 7) that the relative intensity of this band is larger for the Ge25As10S65 sample (which is stoichiometric), than for other glass compositions, all being S-deficient. The lack of sulfur atoms in the network will indeed limit the formation of Ge(S1/2)4 and AsS3 structural units, which are predominant in stoichiometric compositions and even in sulfur-excess ones [18]. The attribution of the band at 430 cm−1 is more ambiguous but was discussed by Tanaka ([17] and references therein) and was associated either to vibrational modes of edge-sharing bitetrahedra Ge2S2+4/2 (a structural unit known to exist in both crystalline and glassy GeS2) or to vibrational modes of S-S dimers, which can be encountered into the network through small sulfur chain and/or sulfur bridging 2 tetrahedra.
A weak band can be observed at 490 cm−1 (Fig. 7) on the Ge25As10S65 spectrum whereas it is strongly reduced (invisible) for other compositions. It can be attributed to the S-S stretching vibration in S8 rings or short Sn chains [9]. Such sulfur chains or rings are observed in S-rich glasses, more seldom in stoichiometric composition, and inexistent in S-deficient glasses, which is consistent with the spectra presented in Fig. 7.
The band, centered at 242 cm−1, that exists not only for the stoichiometric composition, Ge25As10S65, but also for all S-deficient compositions, can be attributed to the vibration of ethane-like structure S3Ge-GeS3 characterized by homopolar Ge-Ge bonds [19]. Finally, a shoulder at 215 cm−1 is observed in the spectrum of the Ge25As20S55 film. The relative intensity of the band at 215 cm−1 is increased with the increasing As content (As/S ratio) in the film, or in other words, with increasing the lack of S in the network. This band was attributed to the As-As homopolar bond [20]. The relative intensity of the As-As band becomes higher than the Ge-Ge band in the spectra of films where As content exceeds the Ge one. This result is in accordance with previous studies reported by Aitken et al [18] where it was evidenced that As-As homopolar bonds are formed prior to Ge-Ge ones in S-deficient GeAsS and no Ge-As bond was observed.We can also notice (in Fig. 7) that the bands corresponding to both homopolar As-As and Ge-Ge bonds are relatively more intense (for all S-deficient films) than the band centered at 340 cm−1 that is attributed to the heteropolar bonds Ge-S and As-S. Although it is difficult to make some quantitative conclusions from these observations, it is clear that S-deficient films contain much larger amount of Ge-Ge and As-As homopolar bonds than the stoichiometric one.
5. Discussions
The increase of the PIB with increasing As content, observed in Fig. 5, can be thus correlated to the increasing content of homopolar bonds (pre-existing before the photo exposition) within the glass network evidenced trough the Raman spectra. Indeed the PIB intensity of all S-deficient films is higher than the one obtained for stoichiometric composition. For the moment, it is difficult to conclude on which of the As-As and Ge-Ge bonds plays more important role in the PIB. On one hand, the highest value of PIB was measured for the Ge25As30S45 thin film (see Fig. 5), i.e. a composition whose content of As-As bonds is expected to be larger than Ge-Ge ones, Fig. 7. On the other hand, a lower value of PIB was measured in both Ge25As20S55 and Ge25As35S40 samples. It is worth noting that the relative intensities of the As-As Raman band of the latter samples are respectively lower and higher than that of the Ge25As30S45 sample, as shown in Fig. 7. This thus suggests the existence of an optimal ratio of homopolar bonds As-As/Ge-Ge to obtain strong PIB. By considering the PIB results and the Raman spectra discussed above, this optimal ratio is here close to 1. In other words, a larger PIB is observed when the amounts of As-As and Ge-Ge homopolar bonds are similar in the material. Nevertheless, Raman spectroscopy does not bring us to absolute quantitative results. Indeed, despite the fact that Aitken et al. [18] have shown that As-As homopolar bonds are formed prior to Ge-Ge ones in S-deficient GeAsS bulk glasses, our study was performed on thin films. The method of preparation of thin films through e-beam evaporation (used here), is different from the melt-quenching technique used for bulk samples, and is known to promote formation of homopolar bonds to the detriment of heteropolar ones [21]. The Raman spectra of bulk and thin film of same composition, presented in Fig. 8, corroborate this feature: the thin film spectrum exhibits a higher relative intensity of the 215-240 cm−1 band, assigned to the As-As and Ge-Ge homopolar vibrational modes, than the spectrum recorded on the bulk glass of same composition. It is worth to remind that the thin films have been prepared from the bulk glassy samples. The origin of higher degree of structural disorder, i.e. formation of homopolar bonds, observed in thin films vs bulk glass can be explained by the higher cooling rate obtained during evaporation process [21].
Fig. 8 Normalized Raman spectra of Ge25As30S45 bulk glass (dotted black line) and thin films unexposed (short dash dotted red line) and exposed at 2.14 W/cm2 (dashed green line), 4.24 W/cm2 (short dotted blue line) and 7.87 W/cm2 (solid cyan line) for 60 min.
Therefore, the mere role played by As-As homopolar bonds in the PIB effect cannot be argued from the results described here and the existence of both As-As and Ge-Ge homopolar bonds has to be considered in the origin of such effect. In addition, our photo excitation dynamic study has shown that more than one excitation mechanisms are involved in the PIB process.
In fact, the role played by homopolar bonds in photoinduced phenomena in ChG glasses has been well established by several authors in various material systems [22–24]. Interestingly, it can be noticed that the inflection point of PIB in our studies (see Fig. 5) happens for the composition Ge25As30S45, which is placed on the Ge – As2S3 tie line in the Ge-As-S ternary diagram [25]. Further studies, based on composition belonging to this tie line, are ongoing in order to provide more information about the origin of the PIB effect.
Finally, to study the evolution of heteropolar vs homopolar bonds before and after the laser irradiation, we have recorded the Raman spectra on the Ge25As30S45 thin films exposed with the same Ar ion laser beam at different intensities (the exposition time was 60 minutes). As previously, the obtained Raman spectra were corrected (baseline subtraction) and normalized at 242 cm−1, presented in Fig. 8. It can be seen that, by increasing the laser exposition, the band at 340 cm−1 increases monotonously. It can be therefore assumed that the laser exposition indeed promotes the breaking of homopolar bonds and the formation of heteropolar bonds.Such effect may be seen as a structural relaxation of the network with the replacement of homopolar bonds by heteropolar bonds, tending toward the polymerized structure encountered in bulk glass. Similar results already were reported in the literature [21]. We believe that the inflection phenomenon (when the PIB starts to decrease while the As content continues to increase above the 30%) is related to the phase transition between the two zones of glass formation, described in Ref [6].
6. Summary and conclusion
Thin films, produced by e-beam evaporation deposition, were produced with good control of stoichiometry in GeAsS glassy system. PIB study was performed in such films with different As contents, ranging from 10% to 40%. High PIB values were observed which increase as the content of As increases up to 30%. Raman spectroscopy was performed for different glass compositions before and after photo exposition for better understanding of this phenomenon. It was shown that increasing the concentration of the arsenic is favoring the formation of the As-As and Ge-Ge homopolar bonds (because of the decrease of the concentration of sulfide anions in the network). It was also shown that the exposition with visible light, promotes the breaking of homopolar bonds, increases the mobility of atoms and favors the formation of heteropolar bonds, accompanied by structural polymerization.
Acknowledgments
We acknowledge the financial support of Canadian Institute for Photonic Innovations (CIPI), Fonds Québécois de la Recherche sur la Nature et les Technologies (FQRNT) and Natural Sciences and Engineering Research Council of Canada (NSERC). We thank Y. Ledemi for valuable comments on Raman spectra. We also thank TLCL Optical Research Inc. for the material help and advice.
References and links
1. B. J. Eggleton, B. Luther-Davies, and K. Richardson, “Chalcogenide photonics,” Nat. Photonics 5, 141–148 (2011).
2. K. Tanaka, A. Saitoh, and N. Terakado, “Giant photo-expansion in chalcogenide glass,” J. Optoelectron. Adv. Mater. 8, 2058–2065 (2006).
3. Y. Ledemi, S. H. Messaddeq, I. Skhripachev, S. J. L. Ribeiro, and Y. Messaddeq, “Influence of Ga incorporation on photoinduced phenomena in Ge-S based glasses,” J. Non-Cryst. Solids 355(37-42), 1884–1889 (2009). [CrossRef]
4. K. Shimakawa, A. Kolobov, and S. R. Elliott, “Photoinduced effects and metastability in amorphous semiconductors and insulators,” J. Non-Cryst. Solids 44, 475–588 (1995).
5. T. V. Galstyan, J. F. Viens, A. Villeneuve, K. Richardson, and M. A. Duguay, “Photoinduced self-developing relief gratings in thin film chalcogenide As2S3 glasses,” J. Lightwave Technol. 15(8), 1343–1347 (1997). [CrossRef]
6. G. Pfeiffer, M. A. Paesler, and S. C. Agarwal, “Reversible photodarkening of amourphous arsenic chalcogens,” J. Non-Cryst. Solids 130(2), 111–143 (1991). [CrossRef]
7. K. Tanaka and K. Ishida, “Photoinduced anisotropic structure in chalcogenide glasses,” J. Non-Cryst. Solids 227-230, 673–676 (1998). [CrossRef]
8. A. Zakery and S. R. Elliott, “Optical properties and applications of chalcogenide glasses: a review,” J. Non-Cryst. Solids 330(1-3), 1–12 (2003). [CrossRef]
9. B. G. Aitken and C. W. Ponader, “Property extrema in GeAs sulphide glasses,” J. Non-Cryst. Solids 274(1-3), 124–130 (2000). [CrossRef]
10. G. Lucovsky, F. L. Galeener, R. C. Keezer, R. H. Geils, and H. A. Six, “Structural interpretation of infrared and Raman-spectra of glasses in alloy system GE1-xSx,” Phys. Rev. B 10(12), 5134–5146 (1974). [CrossRef]
11. Y. Ohmachi and T. Igo, “Laser-induced refractive-index change in As-S-Ge glasses,” Appl. Phys. Lett. 20(12), 506–508 (1972). [CrossRef]
12. K. Petkov and B. Dinev, “Photo-induced changes in the optical properties of amorphous As-Ge-S thin films,” J. Mater. Sci. 29(2), 468–472 (1994). [CrossRef]
13. M. Klebanov, V. Lyubin, D. Arsova, E. Vateva, and V. Pamukchieva, “Photoinduced anisotropy in photobleached Ge-As-S films,” Physica B 301(3-4), 399–404 (2001). [CrossRef]
14. J. Tauc, R. Grigorovici, and A. Vancu, “Optical properties and electronic structure of amorphous germanium,” Phys. Status Solidi, B Basic Res. 15(2), 627–637 (1966). [CrossRef]
15. J. P. De Neufville, S. C. Moss, and S. R. Ovshinsky, “Photostructural transformations in amourphous As2Se3 and As2S3 films,” J. Non-Cryst. Solids 13(2), 191–223 (1974). [CrossRef]
16. P. Boolchand, J. Grothaus, M. Tenhover, M. A. Hazle, and R. K. Grasselli, “Structure of GeS2 glass: spectroscopic evidence for broken chemical order,” Phys. Rev. B 33(8), 5421–5434 (1986). [CrossRef]
17. K. Tanaka and M. Yamaguchi, “Resonant Raman scattering in GeS2,” J. Non-Cryst. Solids 227-230, 757–760 (1998). [CrossRef]
18. S. Sen, C. W. Ponader, and B. G. Aitken, “Ge and As x-ray absorption fine structure spectroscopic study of homopolar bonding, chemical order, and topology in Ge-As-S chalcogenide glasses,” Phys. Rev. B 64(10), 104202 (2001). [CrossRef]
19. J. Heo, J. Min Yoon, and S.-Y. Ryou, “Raman spectroscopic analysis on the solubility mechanism of La3+ in GeS2-Ca2S3 glasses,” J. Non-Cryst. Solids 238(1-2), 115–123 (1998). [CrossRef]
20. P. J. S. Ewen and A. E. Owen, “Resonance Raman-scattering in As-S glasses,” J. Non-Cryst. Solids 35-36, 1191–1196 (1980). [CrossRef]
21. R. Ston, M. Vlcek, and H. Jain, “Structure and photoinduced changes in bulk and films of As-Ge-S system,” J. Non-Cryst. Solids 326-327, 220–225 (2003). [CrossRef]
22. M. Popescu, F. Sava, A. Lorinczi, E. Skordeva, P. J. Koch, and H. Bradaczek, “Photo-induced softening and hardening in Ge-As-S amorphous films,” J. Non-Cryst. Solids 227-230, 719–722 (1998). [CrossRef]
23. L. F. Gladden, S. R. Elliott, and G. N. Greaves, “Photostructural changes in bulk chalcogenide glasses—an EXAFS study,” J. Non-Cryst. Solids 106(1-3), 189–192 (1988). [CrossRef]
24. I. P. Kotsalas, D. Papadimitriou, C. Raptis, M. Vlcek, and M. Frumar, “Raman study of photostructural changes in amorphous GexSb0.4S0.6,” J. Non-Cryst. Solids 226(1-2), 85–91 (1998). [CrossRef]
25. A. Feltz, Amorphous Inorganic Materials and Glasses (Wiley-VCH, 1993).
