Raman spectroscopic study of SbxSe100−x phase-separated bulk glasses

Raman spectroscopic study of SbxSe100−x phase-separated bulk glasses

Journal of Non-Crystalline Solids 355 (2009) 2040–2044 Contents lists available at ScienceDirect Journal of Non-Crystalline Solids journal homepage:...

524KB Sizes 0 Downloads 33 Views

Journal of Non-Crystalline Solids 355 (2009) 2040–2044

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids journal homepage: www.elsevier.com/locate/jnoncrysol

Raman spectroscopic study of SbxSe100x phase-separated bulk glasses O. Kostadinova, S.N. Yannopoulos * Foundation for Research and Technology, Hellas-Institute of Chemical Engineering and High Temperature Chemical Processes (FORTH/ICE-HT), P.O. Box 1414, GR-26504, Rio, Patras, Greece

a r t i c l e

i n f o

Article history: Available online 22 July 2009 PACS: 63.50.-x 63.50.Lm 64.75.Qr 65.60.+a Keywords: Raman scattering Chalcogenides Raman spectroscopy Medium-range order Short-range order Calorimetry Glass transition

a b s t r a c t The structure of SbxSe100x bulk glasses is investigated with the aid of Raman scattering over a wide composition range. The Raman spectra of the glasses exhibit unusual features when compared with other structurally similar binary glasses, owing to the phase separation of the present glasses in a certain composition range. The evolution of the Raman spectra and the depolarization ratio of the polarized and depolarized Raman intensities are consistent with the phase separation effect. The present findings are discussed alongside with calorimetric data from the literature that have been used up to now to extract structural information in an indirect way. The capability of Raman scattering as a tool for investigating phase separation in homogeneous media is demonstrated. Ó 2009 Elsevier B.V. All rights reserved.

1. Introduction Binary selenide glasses, mainly those of As and Ge, have intensively been investigated over the last decades owing to the ease of glass-formation and the plethora of applications that these materials find [1–3]. On the contrary, considerably fewer efforts have been dedicated on Sb–Se binary glasses. The lack of systematic studies has a twofold origin. First, Sb–Se materials are not good glass-formers exhibiting high crystallization tendency. Second, even in the case of glass-formation under rapid cooling of the melt, the obtained glasses are phase-separated on a microscopic scale. Phase separation is generally considered as drawback to applications and further complicates the study of the SbxSe100x binary system. Phase separation is avoided in thermally evaporated amorphous films [4] where the non-crystalline range can be extended beyond the range of the bulk glasses, i.e. the stoichiometric Sb2Se3 is the limiting composition that can be prepared in bulk form. However, using splat-cooling techniques, foils of glasses were prepared for Sb content up to 50 at.% [5]. Despite the above-mentioned shortcomings the non-crystalline phases of antimony selenides meet a number of applications in view of their interest* Corresponding author. Tel.: +30 2610 965 252; fax: +30 2610 965 223. E-mail address: [email protected] (S.N. Yannopoulos). 0022-3093/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2008.07.046

ing optical properties [6] and could be potentially used as phase change non-volatile memory materials [7]. The phase diagram of the SbxSe100x binary system – reproduced in Fig. 1 for convenience – reveals an incipient (or metastable sub-liquidus) immiscibility over the concentration range 4 < xSb < 35 [4]. In particular, while the binary alloy forms a homogeneous liquid above the liquidus curve, the decrease of temperature towards the glass transition temperature Tg causes decomposition of the mixture via the spinodal decomposition mechanism. As a result, the alloy separates in two different glassy phases with compositions similar to those defined by the immiscibility dome limits. Despite the interesting physics in understanding the formation of heterogeneous glassy materials and the fact that antimony selenides could meet a number of applications, a limited number of studies on bulk glasses have been undertaken, devoted mainly to their thermal properties [4,8–11]. Structural studies on bulk Sb–Se glasses are practically absent. In this work, we undertake a structural study of the glassy binary system SbxSe100x (0 < xSb < 30) employing off-resonant Fourier transform Raman (FT-Raman) scattering, which is used to reveal the local bonding arrangement of Sb and Se atoms. The Raman results are consistent with the existence of phase separation in Sb–Se glasses in two phases whose compositions fall within the immiscibility dome shown in Fig. 1.

2041

O. Kostadinova, S.N. Yannopoulos / Journal of Non-Crystalline Solids 355 (2009) 2040–2044

Sb2Se3

Tm

600

liquidus

o

Temperature [ C]

500

binodal

400 300 200

spinodal

100 0 0

10

20

30

40

Sb content [at %]

Se

Sb2Se3

Fig. 1. Se-rich part of the phase diagram of the binary SbxSe100x system, reproduced form Ref. [4]. The liquidus of this system exhibits an inflection point, which is indicative of a hidden miscibility gap. The binodal and spinodal curves are shown by solid and dashed lines, respectively. The vertical arrows indicate the glass compositions studied in this work.

lower glass-forming ability of these mixtures. Melting and quenching cycles were performed repeatedly in order to minimize the concentration gradient between the top and the bottom part of the tube. The amorphous nature of the glasses was tested by X-ray diffraction measurements using a Bruker D8 Advance diffractometer. The radiation source was a Cu Ka line at 1.5405 Å and the signal was measured by LynxEye, PSD detector. Fig. 2 contains representative diffraction patterns for three of the glasses for xSb = 2, 10 and 20. Traces of crystallinity were observed only for the two highest concentrations studied in this work. Raman spectra were recorded with the aid of a Fourier transform (FT) Raman spectrometer (model FRA 106/S, Bruker) using the 1064 nm laser line as the excitation source. The signal was detected and analyzed by a liquid-nitrogen – cooled charge-coupled device (Ge-type CCD detector). The backscattered light was analyzed employing two scattering geometries: the polarized (VV, vertical polarization of the incident beam, vertical analysis of scattered light); and the depolarized one (VH, vertical polarization of incident beam, horizontal analysis of scattered light). Polarization calibration was checked by liquid CCl4. In order to get maximum spectral accuracy a resolution 1 cm1 was used for all spectra. The laser power was set at 70 mW using a rather defocused spot on the sample in order to avoid heat- and photo-induced effects. 3. Results

Reduced Intensity [arb. units]

Intensity [arb. units]

Glasses with compositions xSb = 0, 2, 5, 10, 15, 20, 25, and 30 were prepared from high purity (99.9999%) Sb and Se. Appropriate amounts of the elements, of a total mass of 1.5 g, were loaded in silica tubes with 6 mm outside diameter – 4 mm inside diameter. Silica tubes were thoroughly cleaned by dilute hydrofluoric acid and rinsed several times with triple distilled water. Finally, the tubes were baked by an oxygen–butane flame at 1100 °C in order to enhance their optical quality and to remove impurity inclusions. The Raman cells with the appropriate material amounts were heated under dynamic vacuum at 200 °C for few hours in order to remove traces of humidity and then were sealed under vacuum. The cells were placed in a furnace heated at 800 °C and kept therein for almost 20 h with periodic shaking to ensure homogenization before being quenched in water. For compositions with xSb = 25 and 30, quenching took place from 1000 °C in silica tubes with 4 mm outside diameter – 2 mm inside diameter due to the

Stokes-side Raman spectra of the Sb–Se glasses studied in this work are presented in Fig. 3 in the reduced representation. For comparison, the left-hand inset contains the experimental Raman spectra of these glasses. The spectra have been normalized at the peak maximum of the 250 cm1 band characteristic of the Se chain vibrational mode. The enhanced baselines of the two highest glass compositions at high wavenumbers originate from the second order scattering mechanism of the Sb–Se band, which peaks at 380 cm1 (not shown in the figure). Because the intensities of the various Raman bands depends on the sample temperature and the wavenumber of each band it is important to remove this effect so as to isolate the vibrational density of states (weighted of course by the Raman coupling coefficient or cross section of each vibrational mode). To achieve this,

20

Sb20S80 10

Sb10S90

10

200

0 10

20

30

40

50

60

70

80

220

240

260 -1

280

Raman shift [cm ]

5

50

100

150

200

250

300

Sb30 Sb25 Sb20 Sb15 Sb10 Sb5 Sb2 Sb0

350

-1

Raman shift [cm ]

0

Sb2S98

As30 As25 As20 As15 As10 As5 As0

Intensity [arb. units]

2. Experimental

100

150

200

250

300

350

-1

Raman shift [cm ] 90

2θ Fig. 2. Representative diffraction patterns for the glasses with x = 2, 10 and 20. Scan rate: 25 s/degree.

Fig. 3. Reduced Raman spectra of SbxSe100x glassy system for various Sb contents as shown in the legend. The spectra have been normalized with respect to the Se–Se symmetric stretching vibrational mode. The left-hand side inset shows the same data without treatment as obtained from the experiment. The right-hand side inset shows for comparison the corresponding spectra of the binary As–Se glasses [13].

2042

O. Kostadinova, S.N. Yannopoulos / Journal of Non-Crystalline Solids 355 (2009) 2040–2044

of the glass. In the course of the present study we have measured correspondingly sharp Sb–Se Raman peaks in cases where the quenching rate was not enough to ensure the absolute absence of crystallinity.

Sb15 0.5

ρ

4. Discussion

250

300

-1

Fig. 4. Frequency dependence of the Raman spectra depolarization ratios for three glasses.

one has to employ the so-called reduced representation. Due to the Boson-like statistical description obeyed by phonons, their mean number at any temperature is given by ~=kB TÞ  11 where  ~; TÞ ¼ ½expð hm h and kB are the Planck and nðm Boltzmann constants, respectively. Therefore, the Stokes-side reduced Raman intensity (Ired ) is related to the experimentally measured one (Iexp ) via the equation [12]:

~ ½nðm ~ Þ ¼ ðm ~0  m ~Þ4 m ~; TÞ þ 11 Iexp ðm ~Þ; Ired ðm

ð1Þ

where the term in the fourth power is the usual correction for the ~ is the Raman wavelength dependence of the scattered intensity; m ~0 denotes the wavenumber of the incident shift in cm1, and m radiation. It is obvious from Fig. 3 that the intensity of the vibrational mode at 195 cm1, characteristic of the Sb–Se bond in SbSe3/2 pyramidal units, increases systematically with respect to the intensity of the Se chain peak, upon increasing xSb. Two important observations emerge from Fig. 3: (a) the energy of the vibrational mode of the Se matrix remains practically constant over the very wide range of Sb content. (b) The broad band of SbSe3/2 pyramids exhibits an unusual self-similarity, i.e. its shape and energy is essentially the same for all glasses studied except for xSb = 2. Both observations are unique for this binary system since the Raman spectra of other structurally similar arsenic–chalcogen glasses (i.e. As–S, As–Se, etc.) do not share common features with antimony selenide glasses. For example, the right-hand inset in Fig. 3 contains the reduced Raman spectra of As–Se glasses at various compositions [13]. This distinction reflects structural heterogeneity in these phase-separated glasses as will become clear in the next section. Fig. 4 illustrates the frequency dependence of the depolarization ratio, calculated via q = IVH/IVV. As is evident, the depolarization ratio of elemental Se exhibits fine structure as expected for a homogeneous medium whose structural units obey different symmetry operations in the course of vibrational motion. The fine structure of q is gradually lost when entering the phase separation regime, lending support to the heterogeneity of glass structure. Raman spectra of Sb–Se bulk glasses have previously been reported for the composition range 0 6 xSb 6 8 [11]. However, in that work the Sb–Se band is rather sharp indicating partial crystallinity

Sb-Se bandwidth [cm-1]

200

Raman Shift [cm ]

45

Sb10Se90

40 35

Sen

30

(Se-Se)

150

4 2

/I

0.0 100

(Sb-Se)

Se 0.5

I

0.0

Before proceeding to the discussion of the obtained Raman data in terms of structural features of Sb–Se glasses it is important to briefly refer to the vibrational modes assignment of elemental Se. An extensive analysis of this issue has been presented elsewhere [14,15]. The Raman spectrum of Se can be divided in two spectral regions: in the bond-bending region [70–150 cm1] and the bond stretching one [200–300 cm1]; the latter is the main focus of the present study. Detailed analysis [15] revealed, in contrast to previous studies, that the bond stretching vibrational region contains three peaks. The peak located at 234 cm1 manifests the Se–Se bond stretching vibrations in closely packed Se chains in a configuration resembling that of trigonal Se (t-Se). The peak at 250 cm1 is composite. The stronger component at 250 cm1 was assigned to Se–Se bond stretching vibrations in ‘amorphous’ chains and the 260 cm1 high frequency shoulder to Se8 rings, in direct analogy with liquid sulfur [16] which also contains rings and chains in a delicate and temperature sensitive equilibrium. In order to present quantitative results, which will be useful for the subsequent discussion, we analyzed the reduced data shown in Fig. 3 by fitting the broad bands with Gaussian distributions. A representative fit is shown in Fig. 5 for the Sb10Se90 glass. The dependence of the intensity ratio of Sb–Se and Se–Se (ISb–Se/ISe–Se) modes, the position of this peak and its bandwidth on Sb content is shown in the inset. The following experimental facts suggest that phase separation in Sb–Se glasses for xSb > 5 occurs which is in agreement with the phase diagram of this binary system.

Sb-Se peak position [cm-1]

Sb5

0.5

Reduced Intensity [arb. units]

0.0

0 220 210 200 190 0

5

10

15

20

Sb content [at. %]

SbSe3

150

200

Se8

25

t-Se

250

300

-1

Raman shift [cm ] Fig. 5. Representative fitting example for Sb10Se90. Only 20% of the experimental data (open circles) are shown for clarity. The inset shows the obtained results for the full width of the Sb–Se peak, the ratio of the intensity of the Sb–Se mode (blue line) to the intensity of the Se–Se modes (green lines) and the peak position of the Sb–Se peak as a function of xSb.

2043

O. Kostadinova, S.N. Yannopoulos / Journal of Non-Crystalline Solids 355 (2009) 2040–2044

The aforementioned observations provide compelling evidence for the decoupling of two glassy phases. Regions of Sb2Se3 (or slightly sub-stoichiometric concentration) are immersed into a Se-rich (or slightly Sb-doped) phase. Considering this and taking into account the details of the Sb–Se phase diagram the following picture can be envisaged for the structure of the SbxSe100x binary glasses. For xSb P 5 the glass structure is organized in two phases, the Se-rich and the stoichiometric-like one whose compositions are determined by the limits of the immiscibility dome shown in Fig. 1. The two phases are characterized by such spatial extent, presumably some tens of nanometers, so that their vibrational modes are decoupled. Up to now no direct evidence has been provided for the shape and size of the two phases. However we know that in the present case the growth of phase separation is the spinodal-type mechanism. Nucleation and growth is characterized by distinct spherical droplets of the nucleated phase in a continuous matrix of a second phase. Preliminary scanning electron microscopy measurements in our lab showed no evidence for such droplets, implying a composition fluctuation spatial profile much smaller than one micron. On the other hand, the non-preservation of the depolarization ratio indicates that the spatial extent of composition fluctuation cannot be smaller than few tens of nanometers because otherwise the structure should be ‘seen’ as homogeneous by the laser light. The addition of Sb reduces the population or the volume of the Se-rich phase in a way proportional to xSb. Thus the system behaves as a ‘mechanical’ mixture of two phases that exchange population, while maintaining their local structure unchanged. The experimental finding shown in the inset of Fig. 5 supports this conclusion. Indeed the ratio of the intensity of the Sb–Se mode to the intensity of the Se–Se modes grows almost linearly with xSb. Given the above context, the Raman spectrum of any glass in the phase separation regime can be written as a linear combination of the Raman spectra that correspond to the two phase, i.e. the Sb-poor one and the Sb2Se3-like one. The present study also suggests that the depolarization ratio is indeed a sensitive indicator of monitoring phase separation at the nanoscale. Loss of fine structure in the depolarization ratio might originate from ‘optical stresses’ or gradients in the refractive index trapped in the glass structure as a result of fast quenching. However, after sufficient annealing these optical inhomogeneities are

55

50

o

Tg [ C]

140 120

o

Tg [ C]

(i) The composite band at 250 cm1 exhibits neither a shift in energy nor a change in bandwidth with increasing Sb concentration for glasses in the phase separation regime. A contrasting behavior occurs for similar concentrations in As–Se binary glasses (see inset in Fig. 3) where As atoms are homogeneously dispersed within the Se matrix preventing phase separation in the As–Se binary system. The Se–Se Raman band exhibits an appreciable blue-shift with increasing As content as a result of the effect of chains’ interconnection by means of As atoms. On the contrary, the existence of Se-rich phase, in Sb–Se glasses, with a negligible amount of Sb atoms linking Se chains results in an insignificant influence of Sb atoms on Se chains’ vibrational properties and hence justifies the constant energy of the symmetric bond stretching Se–Se mode at 250 cm1. (ii) The broad band at 195 cm1 is characteristic of the Sb–Se stretching mode of the SbSe3/2 pyramidal units. This band exhibits two noticeable features. Its width is practically constant for all xSb P 5. The mode frequency exhibits a slight red-shift for 0 < xSb < 10 while it remains constant for higher xSb. The effect of self-similarity in shape (constant width) can be accounted for by considering the existence of only one kind of species for all antimony concentrations. This is a stoichiometric-like environment which builds in the glass structure for xSb > 5.

45

100

0

5

10

15

20

25

Sb content [at. %]

Ref. [8] bulk Ref. [9] bulk Ref. [11] bulk Ref. [4] films Ref. [4] annealed films Ref. [4] bulk

80 60 40 0

5

10

15

20

25

30

35

40

Sb content [at. %] Fig. 6. Dependence of Tg of Sb–Se glasses on the Sb content compiled from different sources as shown in the legend. The inset shows an enlarged part of the phaseseparated glasses.

eliminated and the depolarization ratio attains the correct form. In the present case where the annealed glass structure is phaseseparated, fluctuations in the refractive index originate from the different compositions of the two phases. Therefore, light will be depolarized due to these inhomogeneities and mainly due to the scattering that light undergoes at the interfaces of the two phases. As mentioned above, thermal studies [4,8–11] have provided the only available data on Sb–Se bulk glasses. In most of the studies the Tg values exhibit unusual temperature dependence. Specifically, the expected increase up to xSb  4 is followed by a decrease and practically constant trend for higher xSb values. Similar behavior was exhibited by other thermodynamic parameters, such as thermal stability, crystallization temperature, etc. To account for these observations Mehta et al. [10] considered the highly unlikely scenario that in glasses with xSb > 4 homopolar Sb–Sb bonds start to form at the expense of Sb–Se bonds. Further, they considered that at this critical composition (xSb = 4) a dimensionality change (from 1D to 2D) occurs. Evidently, Raman spectra show evidence against this suggestion. In another thermal study [9] the maximum in the Tg’ vs. xSb curve was observed at somewhat low composition, i.e. xSb  1.5. The authors suggested that initially up to this limit Sb is bonded to terminal Se atoms and that up to this limit all Se chains ends have been saturated by Sb atoms. For high compositions Sb atoms are incorporated into Se chains. An illuminating thermal study has been presented by Myers et al. [4] where both bulk glasses and evaporated films were examined. As can be seen from Fig. 6, in the case of as-prepared films Tg values increase almost linearly with xSb as expected for a homogeneous, non-phase-separated binary mixture. This is because Sb atoms are uniformly distributed in the Se structure coordinated by three Se atoms in a pyramidal configuration. Indeed, producing films through thermal evaporation one avoids the incipient immiscibility, which is a genuine feature of the melt-quenched glasses. However, when the homogeneous films are annealed at T > Tg phase separation takes place and the thermal features of these films are practically similar to those of the bulk glasses [4]. 5. Conclusions Structural studies of bulk antimony selenide glasses are largely absent. Therefore, a detailed Raman spectroscopic study on the structural details of SbxSe100x bulk glasses over a wide composi-

2044

O. Kostadinova, S.N. Yannopoulos / Journal of Non-Crystalline Solids 355 (2009) 2040–2044

tion range has been presented in this paper. The Raman spectra of the present glasses exhibit unusual features when compared with other structurally similar binary glasses, such as As–S and As–Se. The experimental findings that the Sb–Se and Se–Se bond stretching frequencies and widths are independent from the glass composition are consistent with the phase separation that takes place in bulk, melt-quenched Sb–Se glasses over the composition range 4 < xSb < 35. Further, it was shown that the particular frequency dependence of the depolarization ratio is another quantity that can be used as a sensitive indicator for detecting nanoscale phase separation in condensed media, such as glasses. Acknowledgements The authors would like to acknowledge financial support from the General Secretariat for Research and Technology – Hellas in the framework of the program PENED 2003 (03ED887); the Hellenic Telecommunications Organization (OTE S.A.) is also thanked for support.

References [1] Z.U. Borisova, Glassy Semiconductors, Plenum, New York, 1981. [2] A. Feltz, Amorphous Inorganic Materials and Glasses, VCH, Weinheim, 1993. [3] A. Kolobov (Ed.), Photoinduced Metastability in Amorphous Semiconductors, Wiley-VCH, Weinheim, 2003. [4] M.B. Myers, J.S. Berkes, J. Non-Cryst. Solids 8&10 (1972) 804. [5] D. Brasen, J. Non-Cryst. Solids 15 (1974) 395. [6] E.R. Shaaban, M. Abdel-Rahman, El Sayed Yousef, M.T. Dessouky, Thin Solid Films 515 (2007) 3810. [7] M.J. Kang, S.Y. Choia, D. Wamwangi, K. Wang, C. Steimer, M. Wuttig, J. Appl. Phys. 98 (2005) 014904. [8] M.M. El-Zaidia, A. El-Shafi, A.A. Ammar, M. Abo-Ghazala, Thermochim. Acta 116 (1987) 35. [9] D. Tonchev, S.O. Kasap, J. Non-Cryst. Solids 248 (1999) 28. [10] N. Mehta, R.S. Tiwari, A. Kumar, Mater. Res. Bull. 41 (2006) 1664. ˇ ernošková, J. Optoelectron Adv. Mater. Rapid [11] J. Holubová, Z. Cˇernošek, E. C Commun. 1 (2007) 663. [12] F.L. Galeener, P.N. Sen, Phys. Rev. B 17 (1978) 1928. [13] O. Kostadinova, S.N. Yannopoulos, unpublished data. [14] S.N. Yannopoulos, K.S. Andrikopoulos, Phys. Rev. B 69 (2004) 144206. [15] S.N. Yannopoulos, K.S. Andrikopoulos, J. Chem. Phys. 121 (2004) 4747. [16] A.G. Kalampounias, K.S. Andrikopoulos, S.N. Yannopoulos, J. Chem. Phys. 118 (2003) 8460.