Appl. Phys

Appl. Phys. A 73, 631–636 (2001) / Digital Object Identifier (DOI) 10.1007/s003390100851

Applied Physics A Materials Science & Processing

Structure and morphology of chemical bath-deposited CdS films and clusters A.E. Rakhshani∗ , A.S. Al-Azab Physics Department, Kuwait University, P.O. Box 5969, Safat 13060, Kuwait Received: 29 January 2001/Accepted: 30 January 2001/Published online: 3 May 2001 –  Springer-Verlag 2001

Abstract. Thin films of cadmium sulfide have been deposited on glass substrates and the structural properties of films have been investigated using scanning electron microscopy and X-ray diffraction techniques. The films consist of domains (groups of grains) and weakly bound grain clusters. The structural parameters of grains, domains and clusters and the effect of film thickness on these parameters are reported. From the measurement of lattice constants in CdS films and in free CdS clusters, it has become evident that the films on glass substrates have a tensile strain along their planes. The effect of thermal annealing on the partial relaxation of the strain is discussed. PACS: 68.55; 73.61.J; 81.05.Dz Cadmium sulfide (CdS) is a II–VI compound semiconductor which has been studied extensively for different applications, including photovoltaic applications [1–3]. It is known that CdS is an excellent heterojunction partner in CuInSe2 and CdTe-based solar cells. Commonly used techniques for the deposition of CdS films include vacuum evaporation, chemical vapor deposition, chemical spraying, sputtering and chemical bath deposition (CBD). The latter, which is a lowtemperature, low-cost and scaleable technique, has proven to yield very thin but pinhole-free films of CdS which have been used in the fabrication of CdTe/CdS solar cells having 15.8% efficiency [3]. In the CBD technique, CdS deposition is carried out in an alkaline solution (pH > 9) containing thiourea [SC(NH2 )2 ] and a cadmium complex. When ammonia is used as the complexing agent, the overall reaction for CdS deposition is likely to be: − Cd(NH3 )2− 4 +SC(NH2 )2 + 2OH → CdS + H2 NCN + 4NH3 + 2H2O .

(1)

The details of the deposition reaction and its optimization have been discussed in [4–8]. ∗ Corresponding

author. (Fax: +965/481-9374, E-mail: [email protected])

The formation of CdS can take place heterogeneously on the substrate surface, depositing a CdS film, and homogeneously in the bath solution, forming CdS colloids. The undesirable homogenous process that yields powdery overlayers on heterogeneously grown CdS films is normally minimized by slowing down the rate of reaction. This can be achieved by adding a buffer solution such as an ammonium salt. The heterogeneous deposition takes place more readily on crystalline substrates. In the case of amorphous substrates, the process can be promoted by chemical etching of the substrate, which creates nucleation sites [5]. Despite extensive studies that have been performed on CdS, the microstructure and surface morphology of these films and the influence of growth conditions and post growth treatments on these properties are not well understood. For instance, the attribution of X-ray diffraction results to the hexagonal or the cubic form of CdS is controversial [9]. Films with either of these structures, or with a mixture of them, are reported as a function of preparation conditions [10–12]. Films deposited from solutions containing iodine have a hexagonal structure with a band gap of 2.62 eV, whereas chlorine-containing solutions yield films with cubic structure and a band gap 2.45 eV [12]. The lattice constants of both types are smaller than the bulk values by about 1%. This difference diminishes by annealing the films at 400 ◦ C. Here we report the results of our study of the microstructure and surface texture of CdS films and colloids and the effect of annealing on these properties. 1 Experimental procedure Cadmium sulfide has been grown from solution using a wide range of bath compositions [5, 12–16]. The composition of the solution used for growing CdS samples is compared with other recipes in Table 1. We examined some other compositions based on CdSO4 and CdAc2 but the mechanical quality of the films obtained from the CdCl2 bath listed in Table 1 was superior. Films were grown on commercial soda lime glass substrates (and in some cases on stainless steel foils) in the temperature range 75 to 90 ◦ C. The substrate was

632 Table 1. The composition and molar concentration of chemical bath solutions used for deposition of CdS

[Cd2+ ] (mM)

[CS(NH2 )2 ] (mM)

[NH4 OH] (M)

[NH4 Cl] (mM)

Cd Source

Reference

2 1 10 50 1.1 0.5–5 5–15

3 5 75 50 70 10–150 50–100

0.64 0.4 pH = 11.5 1.75 1 1.34–2 0.1–0.3

15 20 (NH4 Ac) 50 470 (TEA) – – 10–30

CdCl2 CdAc2 CdCl2 ,CdI2 CdAc2 CdAc2 CdSO4 CdCl2

This work [5] [12] [13] [14] [15] [16]

placed vertically in the magnetically stirred solution after being cleaned and etched with 2% HF for about 5 min. The overall rate of deposition was in the range 1–7 nm/min. The pH of the solution was about 11. Several deposition runs, each from a fresh solution, were used for the preparation of films thicker than about 200 nm. The film thickness was measured from the mass deposited, assuming a density of 4.82 g cm−3 . The accuracy of this method was found to be adequate by comparing the results with those obtained from a Tencor (Alfa step 2000) surface profiler. Film structure was studied by X-ray diffraction (XRD) using a Siemens D500 (Cu K α : 0.15406 nm) and scanning electron microscopy (SEM), using a Jeol JSM-6300. The average size of crystallites (grains) was obtained from the Scherrer formula [17] D = kλ/(∆θ cos θ)

of clusters can be avoided by removing the sample from solution before the dominance of homogeneous deposition. The clusters are weakly bound to the surface and can often be removed by ultrasonic cleaning. In the 58-nm-thick sample (Fig. 1a), where some pinholes can also be detected, the average size of the observable features is about 50 nm. This is almost the same as the grain size (57 nm) obtained from the breadth of the sample XRD peak. This implies

(2)

in which λ = 0.15406 nm and ∆θ is the full-width at halfmaximum (FWHM) of the XRD peak appearing at the diffraction angle θ. The shape factor, k, the value of which depends on the crystallite shape, has a value close to unity. We have taken k = 1, in line with widespread practice. Considering the uncertainty in the value of the shape factor and the effect of any residual microstrain on ∆θ, the grain size cannot be determined with an accuracy better than 25% [17]. 2 Results and discussion 2.1 Film morphology Growth rate is one of the deposition parameters which influences the surface morphology and the grain geometry in thin polycrystalline films. In the CBD technique the growth rate increases with increase in temperature and reactant concentration, and with decrease in ammonia concentration [7], at the expense of grain size [9]. Unlike other deposition methods, the growth rate in the CBD technique can not be kept constant during film deposition. The growth rate decays with time as the concentrations of reactants in solution decrease. Consequently, the growth rate cannot be treated as a controllable parameter in systematic studies of the effect of various deposition parameters on the film’s properties. In this work the overall growth rate is used only as a gross parameter. In the following discussion, a group of grains deposited heterogeneously is referred to as a domain and a cluster is a much larger group of grains which is both weakly bound to the film surface and deposited homogeneously. The surface texture of three samples with different thickness is shown in Fig. 1. This figure shows that the films consist of domains and much larger clusters (brighter in contrast). Deposition

Fig. 1a–c. SEM micrographs of the surface of different CdS films deposited at 85 ◦ C. The film thickness and the overall deposition rate is a 58 nm, 2.1 nm/min b 170 nm, 3.8 nm/min, and c 1000 nm, 4.2 nm/min

633

that in Fig. 1a, each feature seen is a single crystallite grain. Grains group together to form domains in thicker films as shown in Fig. 1b and c. In Fig. 1c the group-structure of domains can be observed. Domains are as large as 600 nm in this sample though the grain size is only 76 nm, as obtained from XRD studies. Comparison of Fig. 1b and c indicates that the domain size increases with the film thickness. These samples have nearly the same overall growth rates. The result of studies on several samples is illustrated in Fig. 2. The domain size increases linearly with film thickness. The overall growth rate for samples thicker than 200 nm was in the range 3.8 to 5.2 nm/min and did not show any correlation to the domain size. These samples were prepared using more than one deposition run. For samples thinner than 200 nm, which were prepared in one deposition run, the overall growth rate is higher for thinner films. Since in this thickness range the growth rate varies appreciably, a clear conclusion about the dependence of the domain size on thickness is not possible. When the overall growth rate was varied by varying the pH whilst keeping the thickness constant, the domain size showed an inverse relationship to the growth rate. Figure 3 shows the SEM micrograph of a sample which has the same thickness as that of the sample of Fig. 1b, but which was deposited at a lower ammonia concentration (lower pH), re-

sulting in a higher growth rate of 7 nm/min. The smoother surface of this film (very small domains) is due to the higher deposition rate used. 2.2 Cluster structure The homogeneously grown clusters sticking to the film surface are much larger than the domains. These clusters are structurally the same as the colloids precipitated in the solution. Some of the clusters show a hexagonal symmetry, as may be observed in Fig. 1. A regular hexagonal shape can appear only when the crystallites have a hexagonal structure with their c-axis, or a cubic structure with their [111] axis, parallel to the electron beam. Figure 4 shows the XRD patterns for two CdS films with different thickness and also for the homogeneously formed CdS clusters. For the latter, a powder sample was prepared from the precipitated CdS in an identical bath. The precipitated material was filtered, washed and dried at 100 ◦ C at atmospheric pressure. For reference, an XRD measurement was also performed on a commercial CdS powder (Alfa 20108). The results indicate that the interplanar distances (d) for the commercial powder are the same as the standard values given in Table 2. Those peak intensities which

Fig. 2. The variation of domain size, D, with film thickness. Samples were deposited in the temperature range 75–85 ◦ C. The overall growth rate for samples thicker than 180 nm was in the range 3.5–5.2 nm/min

Fig. 3. SEM micrograph of a 174-nm-thick CdS film deposited at 85 ◦ C at a rate of 7 nm/min

Fig. 4a–c. X-ray diffraction patterns for a the homogeneously formed CdS clusters, b a thin CdS film (sample of Fig. 1a), and c a thicker sample (sample of Fig. 1c). H and C, respectively, refer to hexagonal and cubic structure

634 Table 2. Diffraction data for powder CdS using CuKα radiation (ASTM Card Nos. 41-1049 and 10-0454). Our results for the commercial CdS powder are given in brackets

Hexagonal Phase d (nm) 0.358 0.336 0.316 0.290 0.245 0.207 0.206 0.190 0.179 0.176 0.175 (0.173) 0.168 0.158 0.145 0.134 0.119

were found to differ from the standard values are listed in brackets. Figure 4a shows that the XRD spectrum for the colloidal powder is different from the standard pattern. Peaks are generally broader and overlap. This is the case for the group of H(100), H(002)/C(111) and H(101) peaks and also for the H(200), H(112) and C(311) peaks. The H(103) peak was faint and the H(102) peak could not be detected. Using the breadths of the peaks appearing at values of 2θ about 27◦ , 44◦ and 52◦ and using the Scherrer formula, the sizes of the grains in the clusters were determined to be 20 nm, 17 nm and 20 nm, respectively. From the comparison of the peaks’ intensities with the standard data given in Table 2, a mixture of cubic and hexagonal structures can be assigned to the clusters. The lattice constant, a, for both structures was obtained from the d-spacing of different peaks. The result is plotted in Fig. 5 against F(θ) = cos2 θ/ sin θ + cos2 θ/θ. This is a method by which the effect of random and systematic errors on the measurement of θ is minimized and the precise value of the lattice constant is obtainable from the vertical intercept of the straight-line fit [17, 18]. The cubic lattice constant, a, obtained from the equation dhkl = a/(h 2 + k 2 + l 2 )1/2 is plotted in Fig. 5a. The calculated value is 0.5818 nm, which is the same as the reported value for the cubic structure [19]. If a hexagonal structure is assigned to the XRD pattern in Fig. 4a, the lattice constant c (along the c-axis) and the lattice constant, a, can be determined from the interplanar spacing of different (hkl) planes using [20] −2 dhkl = 4(h 2 + hk + k 2)/3a2 + (l/c)2 .

Cubic Phase

I

hkl

2θ(◦ )

I

hkl

2θ(◦ )

75(54) 60(37) 100(100)

100 002 101

24.8 26.5 28.2

100

111

26.5

40

200

30.8

25(37) 55(74)

102 110

36.6 43.7 80

220

43.9

40(74) 18(10) 45(59)

103 200 112

47.9 50.9 51.9

4(4) 8(10)

004 202

54.6 58.3

60 10

311 222

52.1 54.6

20 30 30

400 331 422

64.0 70.3 80.7

The experimental value for (c/a) = 1.611 is 1.4% smaller than (c/a) = 2(2/3)1/2 for ideally close-packed hexagonal structures. Despite the excellent agreement of experimental results with the cubic and hexagonal lattice constants, the relative intensities of XRD peaks do not fit either structure. Since the sample was in powder form, the intensity effects are not due to the preferential orientation. Therefore the struc-

(3)

The lattice constant c = 0.6662 nm was obtained from the d-value of the H(002) peak. Also c = 0.6694 nm was obtained from the d-values of H(103) and H(101) peaks accord−2 −2 ing to c−2 = (d103 − (d101 )/8, which was deduced from (3). The average of these two values, c = 0.6678 nm, is slightly smaller (0.6%) than the standard value 0.6722 nm (Table 2). Taking the c-value obtained from the H(002) peak and the d-values of other diffraction peaks in Fig. 4a, and using (3), several values were obtained for the lattice constant a. These data are plotted against F in Fig. 5b. The vertical intercept of the fitted line yields a = 0.4134 nm which is in excellent agreement with a = 0.4137 nm reported in the literature [1].

Fig. 5. The plots of lattice constant, a, for cubic and hexagonal phases of CdS clusters against F = cos2 θ/ sin θ + cos2 θ/θ. The lattice constant was calculated from different XRD peaks. Here, θ is the diffraction angle of the peak

635

ture of clusters is most likely polytypic, consisting of closedpacked hexagonal planes stacked in a sequence corresponding neither to the face-centered cubic nor to the closed-packed hexagonal system [21]. 2.3 Film structure The XRD patterns for the thin film samples shown in Fig. 4 are typical for samples used in this study. All films showed a strong preferential orientation of either H(002) or C(111) planes parallel to the substrate surface. The XRD peaks appearing at values of 2θ about 40◦ , 47◦ and 68◦ belong to the Pt/Rh sample holder. For some samples, as in Fig. 4c, two minor peaks corresponding to values of 2θ between 55◦ and 56◦ , and 22◦ and 23◦ , could be detected. The latter is close to the position of the sulfur H(101) peak as reported for CdS prepared by sputtering [22]. This peak disappeared after the sample of Fig. 4c was annealed in air at 350 ◦ C. The very small but well-defined peak at 2θ close to 24.8◦ is the H(100) peak of CdS. The hexagonal lattice constant, a, for the films could be determined only from the position of this peak. Figure 6 shows the variation of average grain size with film thickness. The average grain size in samples with different thickness was obtained from the breadth of the H(002)/C(111) XRD peak. The number of deposition runs used for the preparation of samples in Fig. 6 was 1, 1, 2, 4, 4 and 5, in increasing order of sample thickness. The corresponding overall growth rates increased successively in the range 3.8–5.2 nm/min. The overall growth rate in multi-run samples could be controlled, to some extent, by adjusting the number and the period for each deposition run. This is because the instantaneous rate in each run decays with the growth time. Figure 6 shows that the grain size is smaller in thinner films, as is expected, and approaches the size of grains in clusters at the limit of zero thickness. The plot in Fig. 6 also shows an upper limit of about 80 nm for grains. This is due to the fact that the grain size increases appreciably with the film thickness, as in the case for domains, but also decreases with increase in growth rate. Due to these two opposite trends, a peak value is obtained in Fig. 6.

Fig. 7. The d-spacing of the preferential planes parallel to the substrate surface as a function of grain size

The interplanar distance, d002 , of the preferential planes parallel to the film surface (d-spacing) is less than that for the grains in clusters or for the powder CdS. Figure 7 shows the d-spacing obtained from the position of the dominant H(002)/C(111) peak for many samples as a function of the film grain size. The grain size was calculated from the breadth of this peak. For thick films (≈ 1000 nm) with large grains the d-spacing approaches 0.3264 nm, which is 2% smaller than that for the cluster grains and 3% smaller than the standard value for CdS powder. For columnar grains, the a-spacing of grains in the vicinity of grain boundaries is expected to be larger than that in the grain interior. This is because the grain boundaries are defective regions with a lower atomic density than the grain interior. Therefore the grain boundaries set up an elastic tensile strain in the grains close to the boundaries. The average value of strain assessed from the measurement of a-spacing or d-spacing depends on the grain size, which determines the extent of penetration of the strain into the grain interior. The plot in Fig. 7 shows that the compression of the preferential planes is more for films with smaller grains (larger tensile strain). The a-spacing value could be obtained only for four samples whose H(100) peaks were detectable. The range in grain size for these samples was 65–77 nm. The result is a = 0.416 ± 0.001 nm. This value is 3% greater than that obtained from the H(100) peak for cluster grains (0.404 nm) (Fig. 5b). 2.4 Effect of annealing

Fig. 6. The average grain size as a function of film thickness. The data point at d = 0 corresponds to the grains of clusters. Samples were deposited in the temperature range 75–85 ◦ C. The overall growth rate was in the range 3.5–5.2 nm/min

The effect of annealing on the film structure was studied. Films with different thickness were annealed in air at 350 ◦ C for 1 h. The annealing process did not make a noticeable change in film morphology. However, annealing broadened the preferential XRD peak and shifted it to a slightly lower angular position. The relative broadening of the peak, which is defined as the ratio of the increment in FWHM value to its value before annealing, is a measure of disorder occurring in the grains as a result of annealing. Figure 8 shows that the disorder is relatively high for films with low thickness, possibly due to the influence of substrate, but drops by a factor of 5 and then vanishes as the film thickness reaches about 1000 nm. Figure 8 shows also the effect of annealing on the relative increase in the interplanar spacing of H(002) and C(111) planes as a function of film thickness. Annealing

636

20 nm. The lattice constants of the clustered grains match those for powder samples, indicating the presence of stressfree conditions in the formation of these clusters. This is likely due to the absence of substrate-related constrictions. (3) Films deposited on glass substrates are under tensile stress along the film plane. If the crystal structure of grains is taken to be hexagonal, the microstrain and its relaxation during annealing can be described by the role of grain boundaries. Acknowledgements. The support of the Research Administration of Kuwait University under Research Project No. SP057 is gratefully acknowledged.

References 350 ◦ C

Fig. 8. The effect of annealing at for 1 h on the relative broadening of the H(002) peak (∆θ/θ) and on the (002) interplanar distance (∆d002 /d002 ) in films with different thickness. Film deposition conditions are same as those for Fig. 6

relaxes the tensile strain induced by grain boundaries along the film plane, probably as the result of a slight grain growth, and lets the interplanar distance increase towards its stressfree powder value. Since the H(100) peak was not detected in the annealed samples, the decrease in the a-spacing as the result of annealing could not be confirmed. As Fig. 8 shows, strain relaxation is more effective in thicker films with larger grains where the bulk properties of grains dominate the grain boundary constrictions, and also where the bulk of the film is less influenced by the constrictions imposed by substrate.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

3 Conclusions The study of CBD-grown CdS films with different thickness ranging from 58 to 1037 nm reveals that: (1) Films are composed of domains whose size increases almost linearly with film thickness. Domains are formed from much smaller grains whose size varies in the range of 40 to 80 nm. (2) Grain clusters formed by homogeneous process are much bigger than domains and consist of grains as small as

17. 18. 19. 20. 21. 22.

T.L. Chu, S.S. Chu: Solid State Electron. 38, 533 (1994) T.L. Chu, S.S. Chu: Int. J. Solar Energy 12, 121 (1992) J. Britt, C. Ferekides: J. Appl. Phys. Lett. 62, 2851 (1993) C. Ferekides, D. Marinskiy, D.L. Morel: Proc. 26th Photov. Special. Conf. (IEEE, New York 1997) p. 339 T.L. Chu, S.S. Chu, N. Schultz, C. Wang, C.Q. Wu: J. Electrochem. Soc. 139, 2443 (1992) B.R. Lanning, J.H. Armstrong: Int. J. Solar Energy 12, 247 (1992) R. Ortega-Borges, D. Lincot: J. Electrochem. Soc. 140, 3464 (1993) I.O. Oladeji, L. Chow: J. Electrochem. Soc. 144, 2342 (1997) M.A. Martinez, C. Guillen, J. Herrero: Appl. Surf. Sci. 136, 8 (1998) J.M. Dona, J. Herrero: J. Electrochem. Soc. 144, 4081 (1997) D. Lincot, R. Ortega-Borges, M. Froment: Philos. Mag. B 68, 185 (1993) T. Nakanishi, K. Ito: Sol. Energy Mater. Solar Cells 35, 171 (1994) P.J. Sebastian, H. Hu: Adv. Mater. Opt. Electron. 4, 407 (1994) A. Kylner, J. Lindgren, L. Stolt: J. Electrochem. Soc. 143, 2662 (1996) C. Guillen, M.A. Martinez, J. Herrero: Thin Solid Films 335, 37 (1998) S. Abd-Lefdil, C. Messaoudi, M. Abd-Lefdil, D. Sayah: Phys. Status Solidi 168, 417 (1998) A. Ashour: J. Mater. Sci. Mater. Electron. 5, 47 (1994) B. Qi, D. Kim, D.L. Williamson, J.U. Trefny: J. Electrochem. Soc. 143, 517 (1996) F. Pfisterer, H.W. Schock: Advances in Solar Energy, ed. by K.W. Boer, Vol. 4 (Plenum, New York 1988) p. 210 W.R. Runyan, S.B. Watelski: Handbook of Materials and Processes for Electronics, ed. by C.A. Harper, Chapt. 7 (McGraw-Hill, New York 1970) P.N. Gibson, M.E. Ozsan, D. Lincot, P. Cowache, D. Summa: Thin Solid Films 361–362, 34 (2000) T. Hayashi, T. Nishikura, T. Suzuki, Y. Ema: J. Appl. Phys. 64, 3542 (1988)