Crystallization Induced Changes

The crystallization of layers in a-MLs by annealing or laser exposure can cause a number of structural and mechanical changes. Among them are cracking [103], some increase in the magnitude of the internal strains [123], deterioration of the interface smoothness, ML periodicity [105, 140], ML surface [110], etc.

Laser power between 0.25 W and 3.0 W from an Ar+ cw laser were used for crystallization of A-Ge:H/a-SiNx:H MLs [103]. No visible changes or measurable signs of crystallization have been detected for MLs on crystalline silicon substrates. For multilayers on glass, however, changes have been observed. Even for powers below 1.0 W, some regions appeared "milky," caused by microscopic bubles that increased in density with power. For powers greater than 1 W, some rough and shiny areas appeared, corresponding to severe cracking and pitting when viewed through a microscope. The appearance of such mechanical defects can be avoided using slow thermal annealing [103].

As previously mentioned, in order to determine the average size of nanocrystals, X-ray diffraction and HRTEM studies are usually used. The Scherrer's equation dNC = X/820 ■ cos 0 (9)

is employed to a distinct band in the X-ray diffraction pattern of the sample in order to calculate the average nanocrystallite diameter dNC (Fig. 5). Here, 0 and 520 are the position and full width at half the maximum of the band. However, a discrepancy has been reported [119] between the average size of CdSe nanocrystals in nc-CdSe/SiOx MLs determined with X-ray diffraction and HRTEM studies. It is known that the full width at the half maximum of X-ray diffraction peaks depends not only on the NC size but also on existing microstrains and deformations in the NC network. The observed discrepancy has been related to the existence of microstrains in the nanocrystals. Using relation [141]

and the average nanocrystallite diameter obtained from HRTEM, the approximate levels of the microstrains have been estimated assuming the CdSe NC size and strain (e) distributions to be Gaussian. Values of around 20 ■ 10-3 and 28 ■ 10-3 have been obtained for nanocrystals having dNC = 10.0 and 5.0 nm, respectively. They are close to the values of 20-22 ■ 10-3 calculated [142] for the CdSe layers of a free-standing CdS/CdSe superlattices with layers of equal thickness. A value of e ~ 49 ■ 10-3 has been determined for nanocrystals having dNC = 3.0 nm, which indicates that, as it might be expected, the level of microstrains rises with decreasing NC size. Several reasons can cause this strain. The surface-to-volume ratio increases with decreasing layer thickness and nanocrystallite size, which is correlated with a higher intrinsic stain and caused by bond-angle deviation. Some interface strain can originate from the difference in the thermal expansion coefficient of the two materials. There may be also an intrinsic component resulting from the volume shrinkage with the transformation of an amorphous region into a more dense crystalline material. Keeping in mind that both the particle size and the lattice strain have a different angle dependent influence on the Bragg diffraction linewidth, the particle size and strain can be separated by two-line Scherrer analysis as described in [143].

The internal strain distribution has also been studied in nc-Si:H/a-Si:H MLs [144] by analyzing the intensity and line widths of X-ray diffraction peaks in the low- and high-angle ranges and the obtained results have been compared with those from Raman scattering. The elastic tensile strain e has been calculated approximately by using the X-ray diffraction angle of the (120) plane for Si single crystal (20 = 28.47°), which was reduced to 20 ~ 27° in MLs [145]. The relation

Diffraction angle, 20 (degrees)

Figure 5. X-ray diffraction spectrum of a SiOx (5 nm)/CdSe(5 nm) ML. The (100), (002), and (101) bands of wurtzite CdSe are not resolved while the (110) one is well resolved and can be used for determination of CdSe nanocrystallite size by applying Eq. (1); S& indicates full width at half maximum of the (110) band.

Diffraction angle, 20 (degrees)

Figure 5. X-ray diffraction spectrum of a SiOx (5 nm)/CdSe(5 nm) ML. The (100), (002), and (101) bands of wurtzite CdSe are not resolved while the (110) one is well resolved and can be used for determination of CdSe nanocrystallite size by applying Eq. (1); S& indicates full width at half maximum of the (110) band.

has been applied, where a0 is the lattice constant of single crystal Si and a is the measured lattice constant of MLs due to the internal strain. The elastic tensile strain has also been estimated from the Raman scattering data using the relation

where P = 1.43«C and Q = (-1.89)^2 are the phonon deformation potentials of Si, wc is the frequency of the crystalline peak, and ws is the vibrational frequency of the shell region between the crystalline core and outer space in nanocrystallites [123, 146]. The effective strains obtained by both approached are about 5% for MLs with nc-Si:H layer thicknesses of 2.2 and 3.0 nm and a-Si:H thickness of 8 nm. The strain can partially be released by using high annealing temperatures [118]. A quasi-equilibrium annealing (with a rate of temperature increase of ~10 K/min) from 870 K to 1320 K has been proposed [63] to reduce internal strains and interface defect density upon rapid thermal crystallization of Si/SiO2 a-MLs.

Small- and high-angle X-ray diffraction studies on a-Se/ a-CdSe MLs, annealed at various temperatures for a long time (18 hr) have shown [105, 140] that the ML periodic structure vanishes upon annealing at 333 K (the glass transition temperature of a-Se is Tg ~ 320 K). Upon annealing at 363 K, a crystalline phase was evident by its main diffraction peak situated at 20 = 29.6°. Also, upon annealing at 720 K in a vacuum used for preparation of a-Si:H/nc-Si MLs [147], it has been seen that the layered structure is partially destroyed. This observation has been explained by the effusion and redistribution of hydrogen over the volume of the films. On the other hand, the SAXRD spectra of a-Ge:H/ a-GeNx MLs annealed at temperatures higher than the crystallization temperature of a-Ge (723 K) have proved that both the ML period and the interface quality did not change appreciably after the crystallization of a-Ge:H layers [101]. The HRTEM results on those MLs, as well as on other Ge-based a-MLs [67, 103], have displayed that a-Ge layers crystallized without disrupting the multilayer structure, and that after crystallization the interfaces were atomically smooth and uniform. It has also been established that in a-Si:H/ SiO2, a-Si:H/SiCx, and a-Si:H/SiNx [63, 121], as well as in SiOx/GeOy [148] MLs, both thermal and laser-beam crystallization of the a-Si:H and GeOy layers do not destroy the regular periodic structure of multilayers. Finally, SAXRD studies of CdSe/a-SiOx MLs, annealed at 673 K for 60 min [93], have shown that these MLs also exhibited a very good periodicity. The above short review of the results on the periodicity of various nanocrystalline/amorphous multilayers shows that the periodic structure of amorphous multilayers is resistant to annealing or laser beam illumination when the constituent material with the higher crystallization temperature has a rigid structure. In such MLs, the growth of semiconductor microcrystals in the "low-temperature" sublayers is restricted along the ML axes and, thus, crystallization does not destroy the ML periodic structure.

Was this article helpful?

0 0

Post a comment