XRD and EXAFS study of the local structure in some non-crystalline SbS compounds

Journal of Non-Crystalline Solids 107 (1989) 261-270 North-Holland, Amsterdam

261

XRD A N D EXAFS S T U D Y OF T H E LOCAL S T R U C T U R E IN S O M E N O N - C R Y S T A L L I N E Sb-S COMPOUNDS

G. DALBA, P. FORNASINI and G. G I U N T A Dipartimento di Fisica dell'Universit~ degli Studi di Trento e Centro CNR - ITC, 1-38050 Povo (TN), Italy

E. B U R A T T I N I Laboratori Nazionali 1NFN, Frascati, Italy

Received 23 May 1988 Revised 13 September 1988

The local structure of three non-crystalline Sb-S compounds - two vacuum-evaporated thin films f-Sb40S60 and f-Sb28S72 and a rapidly quenched glass g-Sb38S62 - has been studied by XRD and EXAFS. The local coordination around Sb atoms is not affected by the preparation procedures or the composition of the investigated samples: Sb atoms have 3 S as nearest neighbours, S atoms have 2 Sb; the mean distance between Sb atoms and their S nearest neighbours is almost equal for glass and thin films. Trigonal Sb-S 3 pyramids constitute the basic structural units which give rise to the disordered network through S bonds. On the contrary, static disorder of the basic units Sb-S 3 and medium range correlations are affected both by the investigated preparation methods and compositions.

1. Introduction

Sb-S compounds and their alloys with other chalcogenide systems have awakened a remarkable technological and theoretical interest in their photoconductive properties [1-3]. The connection between physical properties and local structure is a problem of great interest in non-crystalline materials where preparation techniques and composition can affect the short- and medium-range order [4]. Amorphous Sb-S non-crystalline compounds can be obtained in different compositions and forms: glasses, powders, thin films. The local structure of a-Sb2S 3 has been studied by several authors [5,6 and refs. therein[. From the analysis of the XRD radial distribution functions it has been established that the shortest interatomic Sb-S distance, the number N(S) of S atoms nearest neighbours of Sb atoms, and the number N(Sb) of Sb atoms nearest neighbours of S atoms can be considered equal within the experimental errors for the studied a-Sb2S 3 glasses and powders. On the contrary, the N(S) and N(Sb) values for the amorphous Sb2S3 thin films measured by vari0022-3093/89/$03.50/~:~ Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

ous authors do not agree and, in general, are different from those of bulk samples (table 1 in ref. [6]). As a consequence, these results gave rise to different structural models of the short-range ordering in thin films. Since the causes of the differences between a-Sb2S 3 bulk samples and thin films are not yet clear, this work proposed a comparative study of the local structure of the thin films f-Sb40S60 with the glass g-Sb3sS62. The study has been extended to a film with a non-stoichiometric composition: f-Sb28S72.

The structural analysis was done by two different and somewhat complementary techniques: Xray diffraction (XRD) and extended X-ray absorption fine structure (EXAFS). XRD measurements, whose interpretation does not require model compounds, were utilized to obtain an absolute determination of Sb-S distances, N(Sb) and N(S) values and some qualitative suggestions about medium range order (MRO). EXAFS analysis requires the knowledge of photoelectron phase shifts and backscattering am-

262

G. Dalba et al. / Local structure of non-crystalline S b - S compounds

plitudes to obtain an absolute determination of the structural parameters. Since an accurate study revealed that also the most suitable model compound c-Sb2S 3 was not reliable, theoretical phase shifts were used to obtain Sb-S distance. No use was attempted of theoretical amplitudes because of their unreliability, particularly at low K values. While the absolute values of structural parameters presented in this work rest mainly on XRD results, EXAFS was utilized for a more accurate relative comparison between the three samples and to monitor different degrees of distortion of the first coordination shell of Sb.

3. X-ray diffraction: experimental details and data reduction XRD measurements were carried out at room temperature in parafocussing geometry utilizing Mo K~ radiation and a Zr Kp filter. Thin-film measurements were performed both on the supported film and on the bare glass support; the scattering intensity of the film 1 ( 2 0 ) was obtained by subtracting the scattering intensity I~ of the glass support, corrected for the film absorption, from the total intensity I m scattered by the supported film: 1 ( 2 0 ) = I m ( 2 0 ) - I~(20) e x p ( - 2 # t / s i n O),

2. Sample preparation Thin films f-Sb40S60 and f-Sb28S72 were prepared by thermal evaporation of commercial powders Sb2S 3 and Sb2S5 respectively and were deposited on both mylar and glass substrates maintained at 300°C [6]; the films on glass substrates were used for XRD measurements, those on mylar for EXAFS measurements in transmission mode. The densities of f-Sb40S60 and f-Sb28872 were 4.22 g / c m 3 and 3.81 g / c m 3 respectively. Glassy Sb3sS62 samples in massive forms were obtained for the first time [7] by heating to 650 o C commercial non-stoichiometric Sb2S5 powder (instead of stoichiometric c-Sb2S3) and then quenching the melt at liquid-nitrogen temperature. Sb2S5 is a prevailingly amorphous compound containing crystalline phases of S and SbzS 3 [8-10]. The excess of S in Sb2S5 is lost during the glass preparation process and the final yield, g-Sb38S62, has a quasi-stoichiometric composition, g-Sb38S62 is dark red coloured and has a density of 4.354 g / c m 3. Differential thermal analysis of g-Sb3sS62 showed a glass transition temperature of 277 o C. The composition of the samples was measured by dispersive X-ray fluorescence analysis (XRF) and electron probe micro-analysis (EPMA). X R D diffraction patterns did not contain any peaks due to crystallinity; also the morphological surface analysis of the films did not present any image contrast confirming the lack of crystallinity zones in the samples.

where /z is the absorption coefficient of the film and t its thickness. The glassy sample g-Sb38S62 was finely handpowdered and compressed in a disk. X-ray intensity data were collected in the angular range 2 0 = 5-35 o by steps of 0.25 ° using 0.25 o slits and in the range 30-120 ° by steps of 0.50 ° and 1 ° slits. The experimental curves I(k), where k = 47r sin O/X, were corrected for polarization and absorption using the absorption factors given by Milberg [11]. The normalization in electronic units (eu) was made according to the high-angle method [12] by fitting I(k) to the independent scattering curve. The interference function i(k) was obtained using the coherent and incoherent scattering coefficients reported by Hajdu [13] and Vedene and Chapman [14], respectively. The mean electronic scattering factor fe calculated over the unit of composition SbzS 3 was used as sharpening coefficient [15]. Further details on data analysis and the procedure for extracting the radial electron density distribution functions (REDD) are reported in ref. [6].

4. XRD results Figure 1 shows the experimental spectra of scattering intensity for g-Sb38S62 and f-Sb40S60; the dashed curves represent the relative independent scattering intensities. Both spectra in fig. 1 are characterised, as for other amorphous chalcogenide compounds [16], by a first sharp peak at 1.2

G. Dalba et al. / Local structure of non-crystalline S b - S compounds

1200

263

functions obtained from the interference functions

ki(k) of fig. 2 are shown in fig. 3; the positions of 960 ,~.

720

>-

480

p-

x\~\ \

g - Sb

S

the main structures of the spectra are reported in table 1. The quantitative analysis of the first peak of the R E D D functions (fig. 3(a, b)) was performed by assuming the existence of trigonal structural units, Sb-S 3 bonded by S atoms as proposed in ref. [5]. In this picture each Sb is surrounded by 3 S and each S by 2 Sb. The areas under the peaks at 2.5 A of the R E D D functions (fig. 3(a, b), shaded areas) are proportional to the coordination

-GLASS"

N

z 240 LU Iz

o

240 r

0

0

2

4

6

K

8

(A-1)

10

12

14

r"

1

T

0

-

Table 1 XRD results drawn from the REDD functions in fig. 3 (the theoretical value of the first peak is 1920 e 2 according to eq. (1))

T

600

Fig. 1. XRD scattering intensity for g-Sb38S62 and f-Sb40S60 (solid lines) corrected for polarization and absorption and normalized to electron units; dashed lines represent independent scattering intensities.

~, which is more pronounced for the thin film than for the glass. The interference functions ki(k) obtained from the intensity data of fig. l(a, b) are reported in fig. 2(a, b) respectively. For comparison, in fig. 2(c) the interference function obtained by Cervinka et al. [5] for the glass g-Sb40S6o prepared by melt spinning is presented. The greater differences in the spectra are in the region 6-14 ,~ 1. R E D D

"1

1200

600

1200

IE .2.

,v,

600 0

- 600

1200

600 Positions of the main peaks in REDD (]~)

g-Sb38S62 present work f-Sb~S60 present work g-Sb4oS6o glass-bulk [5] a-SbzS 3 amorphous powder [5]

R1

R2

R3

2.49

3.75

5.85

First peak area (e 2 )

N (S)

N (Sb) 0

- 600 g - Sb4o S6o 1985

3

2 -1200

2.45

3.75

5.72

2020

3

2

2.50

3.85

5.95

2000

3

2

2.50

3.90

5.85

0

2

4

6

8

10

12

14

[K (~-1)] Fig. 2. Interference functions k i ( k ) multiplied by the convergence factor M ( k ) = e x p ( - a 2 k 2 ) , where a = 0.07 ~,, for (a) g-Sb38S62: (b) f-Sb4oS6o; and (c) the melt spinning g-Sb4oS6o glass studied by (~ervinka et al. [5].

264

G. Dalba et al. / Local structure of non-crystalline S b - S compounds

30000

,

~

24000 18000 12000 6000

a

g - Sb38S62 a

0

s

~/~

/

--~

ILia

f-Sb4oSeo b

.//~

~ ~ .

11

~' 8 0 o 0 0

/ ,.//~1/.,. ,,

-'

n

g - Sb40 S60

8000 0

~" " i

2

4

8

8

R CA) Fig. 3. REDD functions: (a) g-Sb38S62; (b) f-Sb4oS60; (c) g-Sb40S60 glass studied by Cervinka et al. [5].

numbers N(S) and N(Sb). The measured areas have been compared with the calculated ones. The theoretical area is given by

A t = U(Sb)csP(S, S b ) + U(S)CsbP(Sb , S),

(1)

where P(S, Sb) and P(Sb, S) are the integrals of the pair correlation functions and c~ and CSb indicate the relative concentrations [5]. The agreement, within the experimental uncertainty, between measured and calculated areas (table 1) supports the assumption that the trigonal Sb-S 3 pyramids constitute the basic structural units in thin films and bulk samples.

SbL 1 edges (4.1 and 4.7 keV respectively). We had then to face the typical limitations of the L edges. The SbL 3 EXAFS is limited to a useful range of 240 eV by the onset of the edge L2, the SbL~ EXAFS has a worse S / N ratio than L 3 and is affected by the EXAFS of the preceding edges. In this paper we will show only results drawn from SbL 3 EXAFS. The results obtained from the analysis of the SbL~ edges were in agreement with those at the L 3 edges, though generally less accurate. EXAFS measurements were carried out in transmission mode at room temperature at the Synchrotron Radiation Facility PWA of the National Laboratories in Frascati [18]. The energy resolution of the experimental apparatus was 0.5 eV at the photon energy E = 4 . 1 keV (SbL3). EXAFS signals were extracted from experimental absorption spectra according to a standard procedure: the continuum atomic-like background /%(E) above the absorption edge was obtained by fitting a 4th degree polynomial to the experimental absorption coefficient /~(E) relative to the L 3 edge [19,20]. This procedure then allowed the calculation of the normalised EXAFS function: x ( k ) = ( ~ - ~0)/~'0, where k = [ 2 m / h Z ( E - E0)] 1/2 is the photoelectron wavevector, E the incident photon energy. The absorption threshold E 0 was assumed equal to the absorption limit of the spectrum (maximum of the first derivative). EXAFS analyses were based on the Fourier filtering of the data expressed in momentum space and on the single electron, single scattering plane wave approximation, which gives: x ( k ) = ~_,A,.(k) sin(2kRj + cbi/(k)) J

5. EXAFS: experimental details and data reduction

with

It is well known that the information obtainable from EXAFS decreases progressively when increasing the photon energy beyond about 25 keV, because of decreasing excited core state lifetime and instrumental resolving power [17]. For this reason we could not utilize the Sb K edge (30.1 keV). We measured EXAFS at the SbL 3 and

A / ( k ) = [ N / ~ ( k )/kR~] e x p ( - 20i2k z ) × exp( - 2 R J X j ) ,

(2)

where ~ ( k ) is the backscattering amplitude for each of the ~ neighbouring atoms of the j-type at a distance from the absorbing atoms i. The Debye-Waller factor o/ accounts for thermal vibra-

G. Dalba et al. / Local structure of non-co'stalline Sb-S compounds"

tions and static disorder, ~ i j ( k ) is the total phase shift experienced by the photoelectron and Xj indicates the electron mean free path [21]. An EXAFS spectrum can be affected by different errors connected with sample preparation: thickness effects and inhomogeneities in the samples can cause a reduction of the signal amplitude [22]. In order to avoid the thickness effect we made absorption measurements for various sample thicknesses and found that the EXAFS amplitude did not significantly change for edge jumps txx varying from 0.2 to 1.5. Holes or thickness inhomogeneities can cause amplitude reduction as well. To avoid these types of errors we prepared c-Sb2S s and g-Sb3sS62 samples by slow deposition of fine powders on polyacetate filters; the samples were about 20 /am thick. A more subtle cause of inhomogeneities in powder samples is due to the mean size of the powder particles; if it is much larger than the sample thickness the EXAFS amplitude suffers a notable reduction [22]. Our c-Sb2S 3 and g-Sb38S62 powder particles had a mean diameter less than 20 /tm.

6. EXAFS results

EXAFS measurements have been carried out on the crystalline compound c-Sb2S 3 to check its usefulness as a model compound and to establish the local structural relation between crystalline and amorphous compounds. The c-Sb2S 3 lattice belongs to the orthorhombic system (a = 11.310 ,~, b = 3.836 A, c = 11.228 ,~) [23]. The structure is constituted by ribbons of (Sb486) n units forming sheets infinitely extended parallel to the b axis (fig. 4(a)) [24]. Two different coordinations having equal weights are possible for Sb atoms (fig. 4(b)). In first (trigonal pyramid site) Sb~ atoms are located at the edges of the ribbons and have 3 S neighbours (2 Sit atoms at 2.54 ,~ and 1 S m at 2.52 A); S atoms are connected to Sb I by simple covalent bonds and form a trigonal pyramid with Sb~ at the vertex and 3 S atoms at the corners of the base. In the second coordination (square pyramid site) the Sbi~ atoms are bound to 5 S

265

a

b S

S

Sbl

S,u

Sb.

Su

S

Fig. 4. Sb2S~ crystalline structure: (a) projection of ribbon configuration along the a-axis: (b) trigonal and square base pyramid sites.

atoms located at the corners of a slightly distorted square base pyramid (fig. 4(b)), the Sb atom being slightly displaced outwards of the pyramid base. The S b n - S bonding distances vary considerably: 2.46 A for 1° S x at the vertex of the pyramid, 2.68 and 2.85 A for 2 Snl and 2 S~ respectively at the corners of the pyramid base [25,26]. In conclusion, in c-Sb2S 3 Sb atoms are surrounded by an average of more than 3 S first neighbours. The experimental EXAFS signals x(k) of cSb2S3 (solid line) and f-Sb40S6o (dashed line) are compared in fig. 5(a). Whereas the x ( k ) signal of the crystalline compound is a sum of several sinusoidal-like contributions, the x(k) of the amorphous thin film appears to be made up by only one sinusoidal-like component. The amplitude of the EXAFS signal of c-SbzS 3 is lower than that of f-Sb40S60. This difference is reflected in the Fourier transforms (FT) of the spectra, fig. 5(b): the peak at 2 A of c-Sb2S 3, which corresponds to the first S b - S coordination shell, is lower than the analogous peak of the amorphous film. The peak at 3.5 ~, corresponds to a second shell of Sb atoms at 3.97 A and 4.02 A [24]. The backtransforms x l ( k ) of the main peaks (1.2 to 2.8 A) in fig. 5(b) are compared in fig. 5(c); they represent the contribution to EXAFS of the first sulphur shell around Sb absorbers. The amplitude of x~(k) for c-Sb2S 3 is lower than for

G. Dalba et aL / Local structure of non-crystalfine Sb- S compounds

266

sample preparation we exclude that the amplitudes are significantly influenced by thickness effects or inhomogeneities of the samples• An explanation can be found in terms of negative interference amongst the sinusoidal-like contributions which give rise t o c - S b 2 S 3 x](k)• As a matter of fact, in c - S b 2 S 3 there are 13 Sb-S distances in the range 2•57-3•33 ,~, hence the main peak in the FT of c - S b 2 S 3 c o m e s from the contribution of a large number of different Sb-S distances. We simulated xl(k) for c - S b 2 S 3 by considering only the distances of the 3 S belonging to the trigonal pyramid site and the 5 S of the square pyramid site. Choosing a different edge jump position, E 0, for each site and utilizing theoretical phases shifts and amplitudes [27] we succeeded in fitting the experimental xl(k) with AE o = 6 eV between the two sites• The different E 0 values can be justified by different chemical bonds of Sb

0.5 t

.o

~ C

I

P~

I

•

~

/

Xa

-0.5

a

o

I

I

I

I

I

I

I

2

3

4

5

6

7

8

K (A-~/ .8 I

~f

.6

I

I,

g¢ it

.....

f - Sb4o S6 0

~

C- Sb 2 S 3

I

.4

1

2

0

4

6

i

,

,

8

I

.5

a'

R (I) !

!

I

i

i ISb_S b I

o

i

0.4 i/~'1

0.2

t

0.0

'~

~

m

B

~tf

f - Sb4oSeo"

~

~

f - Sb,mS ~

@-

-.5

b

""

.0.... _

~

,'; "

.

-0.2 '~"

-

v

C

x

oF

0

0.4

2

I

I

I

I

I

3

4

5

6

7

8

-.05 f - Sb2e S72

K (~-1) Fig. 5. (a) c-Sb2S3 and f-Sb40S60 experimental EXAFS spectra; (b) FTs of the spectra in the range 3-7.3 ,~-~; (c) Back transforms of FT main peak (the vertical bars at 1.2 and 2.8 indicate the limits of the inverse FTs).

.5

C

f

-.5 ~

,,

- Sb28S7:

¢'

.0~ 0

"

0

-.05

f-Sb4oS6o. The explanation of this difference is not trivial, since the EXAFS amplitude is proportional to the coordination number N, and XRD measurements give a value N = 3 for f-Sb4oS60 while one half of the Sb atoms in c-Sb2S 3 (Sbll) are surrounded by 5 S atoms and the other half by 3 S. Because of the care taken in the procedure of

b'

g - Sb38 S62 I

I

r

2

4

6

K (K ~)

-,5

8

0

I

I

I

2

4

6

R (A)

Fig. 6. (a-c) f-Sb40Sro, f-Sb28572 and g-Sb3sS62 experimental EXAFS spectra; ( a ' - c ' ) modulus (dashed lines) and imaginary part (full lines) of FTs of the spectra in (a-c) respectively, in the range 3-7.5 A - a (the vertical bars indicate the limits of the inverse FTs).

G. Dalba et al. / Local structure of non-co,stalline S b - S cornpound~

atoms in the two sites [28]. The preceding analysis demonstrates that c-Sb2S 3 is not a good model compound for obtaining experimental phaseshifts q)ij(k) and backscattering amplitudes ~ ( k ). The higher xt(k) amplitude in f-Sb40S60 (fig. 5(a)) indicates a lower degree of local disorder around Sb atoms in the amorphous sample with respect to c-Sb2S 3. In this sense EXAFS supports the XRD result of a unique coordination Sb-S 3 in the film. The experimental EXAFS spectra of f-Sb40S60, f-SbzsS72 and g-Sb38S62 are compared in fig. 6(a-c). The FTs of the spectra (fig. 6(a'-c') respectively) are characterized by a strong peak due to Sb-S first-shell bonds. The structures beyond the first-shell peak contain artifacts of the FT algorithm. The peak at about 3.5 A is probably due to the Sb-Sb correlation. As a matter of fact, taking into account the effect of phase shifts, it corresponds to the well defined peak at 3.8 A in the REDD of fig. 3. The value of Sb-S interatomic distance in f-Sb4oS60, f-SbzaS72 and gSb38S62 has been calculated from x~(k) by the R-constant method [29]. Since c-Sb2S 3 could not be used as a standard for extracting experimental amplitudes and phase shifts, we resorted to the calculated phaseshifts [27]. A good linearization of the interatomic Sb-S distance R~ as a function of k was obtained for all the samples: in fig. 7 the Rl(k ) function of f-Sb40S60 is presented. The R] values shown in table 2 are essentially the same for all the samples. R 1 values obtained by EXAFS (table 2) and XRD (table 1) are the same for f-Sb40S60 and slightly different for the glass g-SbssS62. We did not attempt to determine absolute values of coordinai

i

i

Table 2 First shell Sb-S distance obtained from EXAFS measurements by the R-constant method with theoretical phase shifts

RI (£) g-Sb3~S62 f-Sb4oS6o f-Sb2gS72

2.45 + 0.03 2.45 + 0.03 2.46 ± 0.03

tion numbers N(S) because of the low reliability of theoretical amplitudes at low k values and the difficulty in accurately estimating the contribution to EXAFS of inelastic effects, such as the electron mean free path. We made a relative quantitative comparison among the xl(k) of the three EXAFS spectra applying the amplitude ratio method [30]. Using the amplitude functions A(k) of the EXAFS signal for a single shell of N identical atoms (eq. (2)) and denoting the reference compound by the subscript m, the logarithm of the amplitude ratio gives

ln[ A( k )/Am( k )] = ln( UR2m/UmR2 ) -2k2(o2-

Ore).

The plots of ln[A(k)/Am(k)] vs. k 2 should ideally be straight lines with slope 2(o 2 - o,~,) and intercepts ln(NR2,jNmR 2). These plots are shown in fig. 8 for each couple of the three amorphous

~

'

f _ Sb28S72 / f - Sb4oS~o

~

r_....~ " . 5

E v

I

267

b

f - Sb2eSr2 / g

-

Sb3eS6~ I

0

f - Sb4o Sso

c

2.50 "

n,

J

2.45

~

o

2.40

/

,b

Sb38S62

f _ Sb40S60 / g -

2'0

s'o

4'0

80

K 2 ( K 2) 410'510

'

6.0 I

'

710

'

8.0

K (~")

Fig. 7. f-Sb4oS6o first-shell radius: R-constant method with theoretical phase shifts.

Fig. 8. Comparison between the EXAFS amplitudes A ( k ) according to the amplitude ratio method: comparison between (a) f-Sb2sS72 and f-Sb40S60; (b) f-Sb28S72 and g-Sb3sS62; (c) f-Sb~S60 and g-Sb38S62. The dashed lines are linear fits from 3 to7A.

268

G. Dalba et al. / Local structure of non-crystalline S b - S compounds

Table 3 Comparison of the first-shell coordination numbers and Debye-Waller factors among the three amorphous samples

f-Sb28572/(f-Sb40Sr0) m f-Sb28Sv2/(g-Sba8S62)m f-Sb4o$6o/(g-Sb38 $62) m

N/Nm

o2 _ O~m ( x I O -4)

1.09 + 0.2 1.21+0.2 1.13+0.2

68 + 20 1 + 0.2 --70--+21

compounds. From the linear fits from 3 to 7 (dashed lines in fig. 8) we obtained N / N m and Ao 2 = o 2 _Omz values reported in table 3. While the coordination numbers can be considered the same for the three samples, within the typical experimental uncertainty of EXAFS, a remarkable difference is noted for the Debye-Waller factor a 2 which is lower for f-Sb40S60 than for f-Sb28S72 and g-Sb38S62.

7. Discussion The values R I , N(S), and N(Sb) obtained by the analysis of the first peak in the R E D D (fig. 3) for g-Sb38S62 prepared by rapid quenching agree with the analogous results obtained for amorphous powder and g-Sb40S60 glass prepared by melt spinning by other authors [5] (table 1). Discordant results about the short range order (SRO) in thin films f-Sb4oS60, which have given rise to several different local ordering models, are reported in the literature [31,32]. Our comparative analysis (table 1) between films and glass however demonstrates that the local coordination for films and bulk samples is the same and independent of the preparation techniques and concentrations which were used in our experiment. This local coordination is based on Sb-S 3 pyramidal units similar to the trigonal pyramid site in c-Sb2S 3 (fig. 4(b)). Therefore our studies enable us to exclude the presence of more complicated configurations, such as the pentahedral one in c-Sb2S 3 (fig. 4(b)). The absence of long-range order in our samples can explain the reduction of the mean Sb-S interatomic distance with respect to c-Sb2S3, where the shortest Sb-S distance is 2.52 ,~.

Some comments on medium range order (MRO) can be made looking at the R E D D functions in fig. 3: the R E D D of f-Sb40S60 is more structured than the R E D D of the glass in the range 3-8 ,~. Hence f-Sb40S60 is more ordered than g-Sb38S62 also in the medium range. This result is confirmed by the first sharp peak at 1.2 ,~ in fig. 1 which is more pronounced for thin film than for glass. The first sharp peak has been associated, by some authors, to certain medium range order forms [5,331. EXAFS results of figs. 6 and 7 and table 2 confirm the existence of a similarity in the first coordination shell of Sb in films and glass. The differences in the R 1 values of tables 1 and 2 fall within the combined experimental uncertainties of the two techniques, EXAFS and XRD. EXAFS is more sensitive than X R D to the sharp features in the radial distribution function of the nearest neighbours but it is not able, in our case, to give reliable information on the MRO in amorphous Sb-S samples. In fig. 6 the second shell contribution (at 3.5 ~,) is scarcely distinguishable from the F T background. It is nevertheless possible, from a deeper knowledge of the first shell distribution, to obtain some hints about MRO. From the data in table 3 one can see that f-Sb40S60, i.e. the film with stoichiometric composition, has the lowest Debye-Waller factor. Supposing the same thermal contribution to the Debye-Waller factor for the 3 samples, we deduce that f-Sb40S60 is characterized by the highest degree of static order in the first coordination shell of Sb atoms. Concerning the MRO (fig. 3, R E D D structures beyond 5 A) the f-Sb40S60 appears more ordered than the glass. This could mean that the relatively slow process of film formation and the stoichiometric composition allow the growth of a structure well-defined in the short range and with a high degree of correlation in the medium range. In the case of the thin films the change of composition from the stoichiometric one to the sulphur-richer one, f-Sb28Sv2, does not affect the nearestneighbour mean distance Sb-S (table 2), however it is reflected in the increase of coordination number and Debye-Waller factor (table 3). While the increase of N, by a factor N / N m = 1.09 in passing

G. Dalba et al. / Local structure of non-crystalline Sb- S compounds

from f-Sb40S60 to f-Sb28S72, can be considered within the uncertainty of the EXAFS technique, the variation of Ao 2 is worth considering. If we suppose the coordination of Sb with the three nearest S atoms as perfectly trigonal (say with three equal Sb-S distances) in f-Sb40Sro, then the value of Ao 2 obtained for f-Sb28S72, when attributed only to static disorder, indicates a remarkable distortion of the trigonal Sb-S 3 pyramid. The distortion of the trigonal coordination (and possibly the slight increase of N) can be correlated with the necessity to accomodate a number of S atoms larger than the stoichiometric number. This number would result in the presence of nonbridging sulphurs breaking the regularity of trigohal motifs. The glass g-Sb38S62, with its quasi-stoichiometric composition, has a slightly lower coordination number and a remarkably higher Debye-Waller factor than f-Sb40Sro. The higher Debye-Waller factor, and the lower degree of MRO (figs. 2 and 3), of the glass with respect to the stoichiometric film can be correlated to the preparation techniques. The higher quenching rate of the glass causes a loose arrangement of the chemical bonds among the nearest neighbours and creates a lower degree of MRO order.

8. Conclusions

In this work the local structures of three noncrystalline Sb-S compounds: g-Sb38562, f-Sb40S60 and f-Sb28S72, differing in composition and in preparation technique, have been studied by XRD and EXAFS. The different and complementary peculiarities of the two techniques have been exploited. From our measurements it follows that for both glass g-Sb38S62 and film f-Sb40S60 as well as for the non-stoichiometric film f-Sb28S72 the basic structural unit of the random network is the trigonal S b - S 3 pyramid. Thus, the more complex local configurations around Sb atoms suggested in the literature for f-SbzS 3 are not consistent with our results.

269

EXAFS measurements show different degrees of static disorder in the first coordination shell of Sb atoms depending on the composition and preparation techniques. These differences reflect different degrees of distortion of the basic unit Sb-S 3. Beyond the first shell (MRO) there exist only minor differences in the coordination among the three samples. The best geometrical regularity in the Sb-S 3 units and the highest degree of MRO was found in the thin film with stoichiometric composition f-Sb40S60. We are grateful to prof. L. (~ervinka (Institute of Physics, Czechoslovak Academy of Sciences), Dr. M. Mobilio and Dr. A. Balerna (Laboratori Nazionali INFN di Frascati) for helpful discussions and Dr. A. Borgese (ITI A. Castelli, Brescia) for collaboration in sample preparation.

References [1] I. Watanabe, S. Noguchi and T. Shimizu, J. Non-Cryst. Solids 58 (1983) 35. [2] S. Manabe, K. Gonpei, T. Kuwahata and M. Nakabayashi, Toshiba Rev. (Int. Ed.) 151 (1985) 34. [3] L. (~ervinka, O. Smotlacha and L. Tichy. J. Non-Cryst. Solids 97&98 (1987) 183. [4] R.J. Nemanich, G.A.N. Connel, T.M. Hayes and R.A. Street, Phys. Rev. B 18 (1978) 6900. [5] L. Cervinka and A. Hruby, J. Non-Cryst. Solids 48 (1982) 231. [6] G. Dalba, P. Fornasini and G. Giunta, Sol. St. Commun. 62 (1987) 773. [7] G. Dalba, P. Fornasini, G. Giunta, E. Burattini and A. Tomasi, J. Non-Cryst. Solids 97&98 (1987) 411. [8] G.G. Long, J.C. Stevens, L.H. Bowen and S.L. Ruby, Inorg. Nucl. Chem. Lett. 5 (1969) 21. 19] T. Birchall and B. Della Valle, Chem. Commun. 12 (1970) 675. [10] G. Dalba, P. Fornasini and E. Burattini, Nuovo Cim. 7 (1986) 293. [11] J. Milberg, J. Appl. Phys. 29 (1958) 64. [12] F. Hajdu, J. Appl. Cryst. 5 (1972) 392. [13] F. Hajdu, Acta Crystallogr. A 28 (1972) 250. [14] H. Vedene and D.C. Chapman, Acta Crystallogr. A 31 (1975) 391. [15] B.E. Warren, X-ray Diffraction (Addison-Wesley, New York, 1969). [16] L. (~ervinka, J. Non-Cryst. Solids 90 (1987) 371. [17] P. Eisenberger, in: EXAFS for Inorganic Systems, Proc. Study Weekend (Daresbury Lab., England, 1981) p. 1. [18] E. Buranini. A. Reale, E. Bernieri, N. Cavallo, A. Morone,

270

[19] [20] [21] [22] [23] [24] [25]

[26]

G. Dalba et al. / Local structure of non-crystalline Sb-S compounds

M.R. Masullo, R. Rinzivillo, G. Dalba, P. Fornasini and C. Mencuccini, Nucl. Instr. and Meth. 208 (1983) 91. B.K. Teo, EXAFS: Basic Principles and Data Analysis (Springer, Berlin, 1986). I.B. Boroviskii, R.V. Vedrinskii, V.L. Kraizman and V.P. Sachenko, Sov. Phys. Usp. 29 (1987) 539. F.W. Lytle, D.E. Sayers and E.A. Stern, Phys. Rev. B 11 (1975) 4825. E.A. Stern, in: EXAFS for Inorganic Systems, Proc. Study Weekend (Daresbury Lab., England, 1981) p. 40. P. Bayliss and W. Nowacki, Z. Kristallogr. 135 (1972) 308. S. Scavnikar, Z. Kristallogr. 114 (1960) 85. I.P. Grigas and A.S. Karpus, in: Chemical Bonds in Solids, Vol. 2, ed. N.N. Sirota (Consultants Bureau, New York, London, 1972) p. 95. K.R. Rao, S.L. Chaplot, V.M. Padmanabhan and P.R. Vijayaraghavan, Pramana 19 (1982) 593.

[27] B.K. Teo and P.A. Lee, J. Am. Chem. Soc. 101 (1979) 2815. [28] B.K. Teo and D.C. Joy, EXAFS Spectroscopy: Techniques and Applications (Plenum, New York, 1981) p. 25. [29] G. Martens, P. Rabe, N. Schwentner and A. Werner, J. Phys. C: Sol.. St. Phys. 11 (1978) 3125. [30] E.A. Stern and S.M. Heald, in: Handbook on Synchrotron Radiation, eds. E.E. Koch, D.E. Eastman and Y. Farge Vol. lb (North-Holland, Amsterdam, 1983) p. 955. [31] V.P. Zacharov and V.S. Gerasimenko, in: Structural Properties of Semiconductors in the Amorphous State (Izd. Naukova, Kiev, 1972) p. 124. [32] L.I. Tatarinova, Electronography of Amorphous Materials (Izd. Nauka, Moskow, 1972) p. 61. [33] J.C. Phillips, J. Non-Cryst. Solids 43 (1981) 37.