CdS semiconductor nanoparticles in silica sonogel matrices

Journal of Sol-Gel Science and Technology, 2, 689-694 (1994) © 1994 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.

CdS Semiconductor Nanoparticles in Silica Sonogel Matrices Code: F12 M. PIlqERO, R. LITRAN, C. FERNANDEZ-LORENZO, E. BLANCO, M. RAMIREZ-DEL-SOLAR, N. DE LA ROSA-FOX AND L. ESQUIVIAS Dpto. Estructura y Propiedades de los Materiales, Universidad de Cddiz, Apdo 40, l]510-Puerto Real (Cddiz), Spain

A. CRAIEVICH Labaratorio Nacional de Luz Sincrotron/CNPq, CP 6192-13081, Campinas, Sao Paulo, Brazil

J. ZARZYCKI Laboratory of Science of Vitreous Materials (LA 1119)-C.069, University of Montpellier II, 34095-Montepellier, France

Abstract. SiO2 gels obtained by sonocatalytic method combined with DCCA were used as host-matrices for extremely fine dispersions of CdS semiconductor particles. Small crystallites were produced "in situ" by H2S gas diffusion method. The particles were characterized by TEM and HRTEM, EXAFS, UV-Vis and Raman spectroscopies. The size of crystallites ranged from 5 to 10 nm. The optical transmission spectra showed the characteristic blue shift as a function of the particles size, as predicted by the theory. The optical and mechanical qualities of the samples were substantially improved by an infiltration method using a sono-sol which sealed the superficial pores thus ensuring greater longevity and the possibility of obtaining transparent gels by polishing. Keywords: II-VI semiconductor, ultrasounds, quantum dots, nanoparticles, DCCA 1. Introduction

of which can be modified by aging or heat-treating

[3]. Materials containing semiconductor nanocrystallites immersed in a dielectric matrix have recently received attention for their non-linear optical third order susceptibility and drastic changes in the optical absorption spectra regarding the bulk semiconductor [1]. To obtain these materials it is indispensable to control the size distribution of the semiconductor particles as well as their crystallinity. An important advantage of the sol-gel route for new materials' processing is that it allows the molecular manipulation in the liquid state [2]. However, the control of the chemical parameters governing the chemical ordering is not easy in most cases. Another characteristic that has attracted much interest to processing materials scientists is the specific nature of gels, i.e., they are porous solids, the textural features

A II-VI semiconductor such as CdS is often chosen because it easily precipitates in silica gel matrix with a good optical transmission. Several authors have recently reported interesting results on the sol-gel processing of CdS~ Sel_~ semiconductor doped glasses [4, 5, 6] which differ in the manner that semiconductor is immersed in the matrix or in the matrix composition. We have developed this work with the aim of including some innovative variations in the processing. Thus, on the basis of a previous work, sonocatalysis and a Drying Control Chemical Additive, DCCA were combined to obtain monolithic and transparent matrices

[71. Here we present a first approach to the particles' structure and size distribution of nanocrystals. The photon absorption energy ~as recorded to check the validity of the t~fros and t~fros mode [8] of "quantum dots" systems.

690

2.

Piaero,et al. Experimental

2.1. CompositePreparation Silica sono-xerogels were prepared by ultrasonically assisted hydrolysis of a mixture of TMOS: Water (pH < 1): Formamide, in the molar ratio 1:10:3, using an ultrasonic dose of 64 Jcm -3 [7]. Different amounts of Cd(NO3)2 were added under mechanical agitation to the sono-solutions before gelation. Once the resulting gels had been left to age at room temperature, H2S gas was diffused by thermal decomposition of thioacetamide (TAA). Small CdS crystals are then produced by precipitation into the silica matrix. The samples are referred to as follows: Sx(x = 1, 5, 10), x being the particles percentage (in weight) that would be in the sample if the whole of the Cd added had reacted to form CdS. Also, a non-diffused sample was monitored and it is referred to in the text as "white". The CdS/SiO2 composite xerogels were stabilized by an impregnation process consisting of the total immersion of the gel sample in a sonosol with the same composition as the host gel at room temperature. The samples were left submerged in the solution under varying pressure. The impregnated samples were then removed from the solution and introduced into hermetically closed containers and left in an oven at 40°C for 24 hours. After this time the cover was removed and the impregnated gels were left to dry at 40°C for 24 hours. The samples were then ready for optical polishing.

2.2. Characterization Electron Microscopy1 micrographs were obtained with a standard transmission microscope with a side entry operating at 120 kV and a High Resolution (2.1 /~) microscope with a top entry operating at 200 kV. The sample UV/Vis transmission 2 was measured in the wavelength range from 350 nm to 800 nm. All the specimens were 2.5 4- 0.1 mm thick. Transmission detection was used for XANES and EXAFS experiments 3 as described elsewhere [9]. The estimation of energy resolution (2 eV), calibration, convolution from the low energy region and normalization were carried out according to the reported procedure [9]. EXAFS spectra were recorded over a range of 1000 eV above the Cd K-edge (26625 eV). The data were treated with the available software [ 10]. The photoelec-

tron energy (E-E0), E0 being the threshold energy, is converted into wave vector, k. E0 is approximately estimated (AE0 = ±15 eV) and then improved during data analysis. To evaluate the neighbours' positions around the absorber atom, the well-known EXAFS equation was used [11]. This equation describes the EXAFS oscillations for a Gaussian distribution of neighbours around the central atom in the single scattering and the planewave approximations, the amplitude of which is a function of Nj (average coordination number for the Gaussian distribution of distances centered at the Rj value), crj (Debye-Waller contribution) and Fj (related to the photoelectron mean free path). The Fourier Transforms (FT) of the kZx(k) weighted signals are the Pseudo Radial Distribution Functions (PRDF) around the Cd atoms. These functions are affected by a shift caused by the backscattering phase that may be evaluated [12]. To avoid spurious oscillations due to the lack of high backscattering data and signal noise, FT were calculated with a Hanning window from 3 to 13 ]k-1 above the edge. A FT filtering was used to extract the EXAFS contribution of the first coordination sphere which is supposed to be bi-layered (S and O atoms around Cd). The structural parameters were obtained fitting the EXAFS equation [11] to the filtered data. This was performed by an interactive process in the k and R spaces [ 10] in which the backscattering amplitude and phase function were experimentally evaluated by means of the EXAFS equation from X-ray absorption data of pure CdO and CdS, the structural parameters of which are known: No = 6, dcdo = 2.35 /~ and Ns = 4, dcds = 2.54/~, according to their respective spatial group unit cells, dimensions and atomic coordinates [13]. The output phases of Raman 4 signals were analyzed with a photo counting electronics. An Ar + laser with a power of 0.4 w at 514.5 nm wavelength was used as an exciting radiation source.

3.

Results

On TEM micrographs of the sample S10, CdS particles can be distinguished from the amorphous silica matrix (Figure 1). T h e average size of the CdS particles was estimated from measurements carried out on diverse micrographs. HRTEM provided another display of the particles' crystalline structures. Figure 2 shows a hexagonal CdS

CdS Semiconductor Nanoparticles

691

50 . . . . .

Fig. 1. TEM micrograph of CdS/SiO2 composite corresponding to the S10 sample. Upper inset is a particle size histogram.

particle embedded in the amorphous silica matrix (sample $5). Identification of the particles' specific crystalline phase was carried out by means of selected area electron diffraction (SAED) technique. The experimental patterns show diffuse rings, each ring matching either to the known d spacing for hexagonal CdS or cubic face-centered CdO type. The recorded spectra (Figure 3) present an absorption band assigned to the semiconductor particles. Regarding the position of the relative minimum of the bands, a blue shift from the bulk CdS absorption band position at room temperature (512 nm) is exhibited in all cases. The absorption increases with the cadmium content. Normalized near-edge absorption spectra are shown in Figure 4. As XANES have an atomic origin, spectra of the same compound in a different state are very much alike. A table of theoretical spectra of mixed "white"-CdS compounds--linearly interpolated from the experimental absorption spectra of pure CdO and CdS--has therefore been included for comparison. XANES of low Cd 2+ content samples are quite similar to the "white" and CdO ones. The S10 sample has a structure very near to the CdS sample. Figure 5, corresponding to the PRDF, also presents similar results. The fitting of the PRDF is also shown in Figure 5. The resulting structural parameters appear in Table 1. We do not take into consideration of Cd atoms shared by O and S inside the particles as the HRTEM micrographs show homogeneous CdS hexagonal particles and the electron diffraction presents separately CdO and CdS crystals.

Fig. 2. HRTEM micrograph of $5 sample. Left upper inset is the electron diffraction patterm corresponding to the CdS greennockite phase. Right upper inset is the electron diffraction pattern of CdO crystalline phase. 0.8

0.6 ~

'~ 0.4

0.0

350

i

i

i

i

400

450

500

550

600

;~ (nm) Fig. 3. RoomtemperatureUV/VistransmissionspectraofCdSdoped gels with different cadmium contents, top to bottom. Also is included a non-diffused sample.

Figure 6 shows the Raman scattering from 200 to 700 cm -1. In this region modes attributed to CdS vibrations [14] around 610 cm -1 (2LO) and 300 c m - 1(1LO) can be seen in samples with different Cd 2+ content. The LO modes shift towards lower energies.

4.

Discussion

According to l~fros and l~fros model [8], the absorption threshold energy for quantum dots is given by: E = E 9 + 2#_R---~

(1)

692

Pifiero, et al.

Table 1. Structuralparameters extracted from EXAFS equationfittings.

Sample S10 $5 S1

Atom

R (/~)

a (/~-1)

P (A-z)

N (atoms)

AEo (eV)

O S O S O S

2.23 2.55 2.31 2.62 2.31 2.61

0.056 0.056 0.027 0.027 0.042 0.042

3.55 3.55 2.77 2.77 3.09 3.09

0.50 3.55 5.70 0.30 5.70 0.30

13.48 12.33 10.11 10.11 15.10 15.10

Table 2. Spectrophotometricexperimentaldata.

Sample S1 $5 S10

I

2oo

Minimum (nm)

Size (/~)

Size width (•)

436 425 410

42 22 20

20-30 20-28 18-27

":'.150 100 i:z:; 12.,

$1o

50 jf 1.0

,/,,/

1.5

st

2.5

2.0

3.0

Fig. 5. PT magnitude of the k 3 weighted EXAFS signal(fullline) and thc fittingof the EXAFS equation (dashed line)for $I, $5 and $I0 samples.

a~

I

35

26600

26700

$1

E n e r g y (eV) Fig. 4. XANESspectra of CdS doped gels (S1, $5, S10) with different theoretical mixed "white"-CdS compounds.

where !~ = ___1 m* + ml-Tis the electron effective reduced / 1 ,

I

I

200

250

300

350

400

450

500

550

600

650

700

(~,~-~)

mass with m~ = 0.2 me and m~ = 0.7me, m~ being the mass of the electron, _Ris the particle radius and Eg is the band gap for the bulk CdS crystal. These energy values can be obtained from the position of absorption band. Figure 7 shows the absorption threshold energy values vs. the inverse square of particle mean radius, mea-

Fig. 6. Ramanscatteringof the differentCdS dopedxerogels. Dotted lines indicatethe 1LO (300 cm-1) and 2LO (610 cm-1) vibration models of the CdS crystal.

sured from TEM images. The data are well fitted by a linear function, indicative of quantum size effect occurring in CdS microcrystallites according to equation (1). The slope of the linear function gives a reduced

CdS Semiconductor Nanoparticles

3.2

,

,

,

3.1 -~- 3.0

o

2.9 2.8 2.7 2.6

' ' ' ' 0.00 0.03 0.06 0.09 0.12 0.15

i/R a

(rim -2)

Fig. 7. Band gap energy of CdS doped silica xerogels from transmission espectra vs. the reciprocal square of nanocrystals size with the best linear interpolation. Straight line slope is proportional to CdS electron and hole reduced mass, according to eq. (1).

mass of 0.17m~ which is a fairly good agreement to the theoretical one (0.16m~). This result allows the relation of the absorption band with the semiconductor size distribution in the samples. Thus, the band position will correspond to the mean microcrystallites' size and the band width is related to the dispersion of particle size. In Table 2 are given the band positions of the UV-Vis spectra, mean particle size and its distribution width for different cadmium contents. Therefore, the zero extrapolation value in Figure 7 would be in disagreement with the bulk CdS band gap, since for particles with/~ ---+ oc the linearity of equation (1), and then the quantum confinement, are lost. Values of Table 2 show that the mean particle size decreases when increasing Cd content, in agreement with our Raman spectroscopy results. As a matter of fact, according to Ferrari et al. [15], who indicate the LO modes shift toward lower energies when the size of nanocrystals decreases, the sample S 10 contains smaller CdS particles than the S 1 and $5 ones. This behaviour can be interpreted by the cation (Cd 2+) effect in the xerogel texture: low Cd 2+ contents lead to a fine and closed xerogel porosity where the H2S diffusion is not possible through small pores. Recent results show that higher Cd 2+ contents shift the pore size distribution toward higher values allowing a more complete sulphuration [16]. The percentage of Cd-S bonds with respect to the total amount of C d - X bonds can be estimated from the

693

average number of atoms of the species j in the first coordination sphere, Nj resulting from the fitting of the PRDF (Table 1). According to these Nj values, the sample S 10 has a concentration of CdS particles much higher than the others, in agreement with the XANES, US-Vis and Raman spectroscopic analysis. An estimate of the order of magnitude of the particle concentration can be obtained from these values. Let us take, for instance, the sample $5 (1.3 g.cm -3 from Hg pycnometry), if all the Cd z+ ions were reacted with S = ions to form CdS, it would be of the order of 102o CdS molecules per cm 3. Taking into account the experimental values of N s (Table 1), the CdS molecules number would be roughly given by us(exp) m20 Ns(CdS) ~V . If particles of 2.2 nm average radius are considered (Table 2), there would have been 1016 particles per cm a in $5 composite. A concentration of 1015 and 1017 particles per cm a for the S1 and S10 sample results respectively. That is the increase of the CdS particle concentration is very sensitive to Cd 2+ content. The first neighbours' bond lengths differ from their corresponding standards by a maximum of 3%. This distortion is caused, in our opinion, by the existence of an important number of O - C d - S mixed bonds on the particle-matrix interface, caused by the small particle size involving a high surface/volume ratio. 15% of Cd atoms in the particles are at a distance from the particlematrix interface that they can be bonded with matrix atoms. This effect is enhanced in Cd-O bond length of the S 10 sample because their particles are smaller than in the other samples. In this sample it is worth to note that the Cd-S bond length is very close to the standard.

5. Conclusions The sonocatalytic method combined with DCCA is adequate to prepare silica matrices for fine and uniform dispersion of "quantum dots", the particle concentration of which increases exponentially with the amount of Cd included in the starting sol. The influence on the porous texture of the Cd 2+ ion concentration renders difficult the precipitation process for low Cd 2+ contents. All the characterization methods used indicate that the sample S10 contains a better dispersion of CdS nanocrystals than the others. Although in all cases the sulphuration is incomplete, it is enough to show quan-

694

Pi~ero, et al.

rum c o n f i n e m e n t behaviour. It is manifested by the characteristic blue shift in the UV-Vis spectrum which is consistent with the Efros and l~fros model.

Acknowledgments This work has b e e n supported by the "Programas de A c c i o n e s Integrada H i s p a n o - F r a n c e s a s " and by the project M A T 9 1 - 1 0 2 2 of the " C o m i s i 6 n Interministerial de C i e n c i a y Tecnologfa" ( C I C Y T ) of the Ministerio de E d u a c i 6 n Ciencia, Spain.

2. Bagnall, C.M. and Zarzycki, J., J. Non-Cryst. Solids 121 (1990) 221. 3. Nogami, M. and Nagasaka, K., J. Non-Cryst. Solids 122 (1990) 101. 4. Takada, T., Yano, T., Yasumori, A., Yamane, M., and Mackenzie, J.D., J. Non-Cryst. Solids 147 & 148 (1992) 631. 5. Tohge, N., Asuka, M., and Minami, T., J. Non-Cryst. Solids 147 & 148 (1992) 652. 6. Spanhel, L., Arpac, E., and Schmidt, H., J. Non-Cryst. Solids 147 & 148 (1992) 657. 7. Blanco, E., Ph.D., University of Cadiz (1993). 8. l~fros, AI.L. and l~fros, A.L., Sov. Phys. Semicond. 16(7) (1982) 772. 9. Gomez-Vela, D., Esquivias, L., and Prieto, C., Phys. Stat. Sol. (b) 169 (1992) 303.

Notes

10. Bonnin, D., Kaiser, E, and Desbarres, J., FORTRAN Software for EXAFS analysis, Lab. Phys. Quant. (CNRS) Paris (1988).

1. JEOL JEM- 1200 EX and a JEOL JEM-2000 EX of the Servicio Central de Cienciay Tecnologfa of the CfidizUniversity facilities. 2. Perkin-Elmer UV/Vis/NIR spectrometer (LAMBDA- 19).

11. Teo, K., EXAFS: Basic Principles and Data Analysis in Inorgarlic Chemistry Concepts 9 (Springer-Ver!ag,Berlin, 1986).

3. Beam line of EXAFS-III station at the DCI storage ring (LURE, Orsay, France). 4. Jobin-Yvon U- 1000 double monochromator, coupled to an RCA 31034 photomultiplier.

References I. Jain, R.K. and Lind, R.C., J. Opt. Soc. Am. 73 (1983) 647.

12. Mckale, A.G., Veal, B.W., Paulikas, A.E, Chan, S.K., and Knapp, G.S., J. Am. Chem. Soc. 110 (1986) 3763. 13. Am. Soc. Test., and Mat. (CDS/6-0314, CDO/5-0640). 14. Champagnon, B., Adrianasolo, B., and Duval, E., J. Chem. Phys. 94(7) (1991) 5237. 15. Ferrari, M., Champagnon, B., and Barland, M., J. Non-Cryst. Solids 151 (1992) 95. 16. Blanco, E., Limln, R., Ramfrez-del-Solar, M., de la Rosa-Fox, N., and Esquivias, L. (submitted to the J. Mat. Res.)