Figure 1a shows the variation of total reflectance with the amount of carbon precursor (SUC) of C–SiO₂samples. The investigation was done with 6, 9, 11 and 12 g of SUC for a fixed TEOS: H₂O ratio of 12:9. The low to high reflectance transition appears to move towards an apparent limit of 2 μm with increasing amount of SUC.

It is striking that for wavelengths less than 0.8 μm, the reflectance increased with increasing amount of SUC. It was expected that an increase in SUC would lead to higher absorption, and hence lower reflectance, in the UV–VIS interval. This behaviour suggests that when the amount of SUC is low the resultant carbon chains in the silica matrix have on average a small size and more uniform distribution, as suggested by X-HRTEM image in Fig. 2a, and therefore would have a higher absorption cross section. When the amount of SUC was increased the resultant carbon chains agglomerated to form larger aggregates, as shown in Fig. 2b, with a large scattering cross section that resulted in higher reflectance. The gain in absorptance in the UV–VIS for the low SUC samples was accompanied by a loss in absorptance in the 0.8–2.0 μm interval due to resonance absorption by water molecules.

In the near infrared (NIR) and infrared (IR) regions, the reflectance of the samples with low SUC content is higher than that of samples with higher SUC content. The reason for this is that samples of low SUC content were thinner than those of higher SUC content. This is thought to be purely a viscosity effect of the sol during the spin-coating process. The net result of this behaviour is a lower thermal emittance by samples with low SUC content. It was thus clear that an optimum of SUC content had to be sought. The optimum for samples with high absorptance and low emittance was observed to be 11 g SUC based on the graphs in Fig. 1a.

Characteristic chemical bonds in the C–SiO₂ samples were identified from the FTIR reflectance spectrum presented in Fig. 1b. Three distinct absorption bands are observed. The major band at approximately 1,050 cm⁻¹ is assigned to stretching vibrations of Si–O–Si or Si–O–X, where X represents ethoxy groups bonded to silicon 8 9. The shoulder at about 1,200 cm⁻¹ is assigned to either the transverse optical mode of the out of phase mode of the asymmetric vibration or to the longitudinal optical mode of the high frequency vibration of SiO₂ 8 . On the other side of the major absorption band, at 900 cm⁻¹, is an absorption band that can be assigned to the stretching vibration of Si–OH or Si–O⁻ groups. A broad absorption band, situated between 3,000 and 3,600 cm⁻¹, and another one around 1,600 cm⁻¹ are assigned to O–H stretching and O–H bending vibrations, respectively 8 9. The latter absorption bands appear to indicate the hydrophilic nature of the sol–gel synthesized silica.

The reflectance spectra for C–ZnO and C–NiO samples are presented in Fig. 3a. An attempt has been made to check reproducibility of the samples of both types by maintaining the same deposition parameters. In Fig. 3a, the dips in the spectra between 6 and 7 μm are due to water absorption (O–H bending vibrations at 1,600 cm⁻¹) 6 9. The O–H bending vibrations are much weaker than the case for previously studied C–SiO₂ samples 6 ; this implies a lower emittance for the C–ZnO samples. The O–H stretching vibrations around 2.7 and 3.3 μm (3,000–3,600 cm⁻¹) of the C–SiO₂ samples (see Fig. 1a) are clearly absent in the C–ZnO samples resulting in an even lower emittance (NB: a high reflectance from the sample in the IR wavelength range means a low emittance by the sample). The absorption due to the Zn–O vibrations is expected between 20 and 25 μm 10 11, which is just beyond the measurement range for the present work. A major difference between these spectra and those of previous experiments on C-SiO₂ selective absorbers 6 is the absence of strong absorption between 2.5 and 20 μm, which signifies a great improvement in emittance characteristics of the C–ZnO absorber coatings compared with those of the previous study of C–SiO₂ selective absorbers. Another added advantage is that the UV–Vis absorption of the C–ZnO based samples seems somewhat better than that of C–SiO₂. However, a drawback of the C–ZnO based absorber surfaces is that the transition step from low to high reflectance is not steep enough for all samples investigated.

The reflectance spectra of the C–NiO samples are presented in Fig. 3b. It is clear that both modes of the O–H vibrations of the C–ZnO samples are absent in the C–NiO samples; this yields even better emittance characteristics. The step transition for all the samples is between 2 and 3 μm and is steeper than that of the C–ZnO samples. This gives the C–NiO samples the closest to a step transition at 2.5 μm expected for an ideal selective solar absorber surface for domestic water heating.

The selected area electron diffraction (SAED) result shown in Fig. 4a also demonstrated that in most regions, the coating consisted of nanocrystalline NiO. The electron diffraction reveals that there exists a small amount of Ni grains with diameter of about 30 nm in some regions of the film (see Fig. 4b).

The Raman spectroscopy data of the samples were first analysed using the Tuinstra–Koenig (TK) equation. The background to this equation, which is important for Raman spectroscopy characterization of carbon allotropes, is given in a review article by Gouadec and Colomban 12 . This equation relates the ratio of phonon intensities of the disorder (defect), I _D, and the prefect graphite peak, I _G, to the carbon allotrope grain size. The empirically calibrated TK equation is given as:

In our samples, the D peaks are weaker than the G peaks and this suggests that the grain sizes could be larger than 44 nm 12 . From the peak positions ω_D and ω_G and intensities I _D and I _G of the D and G-peaks, respectively, we calculated the shifts Δω_D and Δω_G, ratio I _D/I _G, and hence the grain sizes L _grain and these are presented in Table 1. (It must be noted that grain size is different from the size of the crystallites that agglomerate to form the grain.)

From the calculations, it was found that in all samples, the D and G-peaks shift in position to higher Raman shift. The reason for this “blue shift” is due to compressive stress on the carbon allotrope caused by the matrix. Tensile stress leads to red-shift as shown in the following relation between the observed peak position ω_vib and the strain ε_lb of the carbon–carbon bond of length l _b given by 12 :

Here ω₀if the peak position for the strain-free bulk sample. The symbolsaandrare, respectively, attractive and repulsive exponents in the Mie-Grunsen potentials in solids that govern the bond energies as a function of interatomic distances. (The values foraandrare respectively 6 and 12 for van der Waal’s forces in the 6–12 Lennard-Jones potential, 1 and 9 for ionic bonding anda + r = 3 for covalent bonding.) It can be clearly seen that a positive strain (tensile strain) reduces the observed peak ω_vibleading to a red shift whereas a negative strain (compressive strain) leads to the opposite effect in peak position shift—a blue shift. This means that in all samples, carbon clusters are compressively strained by their respective host materials as expected.

Phonon confinement models have been used to fit the asymmetrical broadening of the Raman peaks. We assume that the carbon clusters and nanocrystallites are perfect spheres. In this case, we can fit to the Raman spectral data the following Richter et al. (1981) 13 equation for asymmetrical Raman line-shapes due to phonon confinement in nanomaterials given as:

In this equation, A ₀ is a pre-factor to be determined from the fitting session, d is the carbon cluster size, q is the wave vector of the exciting light source in the Raman spectroscopy set-up, α is the scaling factor, ω(q) is the phonon dispersion relation for the material under study and Γ₀ is the full width at half maximum of the Raman line-shape for bulk form of the same material. For materials of different shapes, a modification of the d ³ q is required. For instance, d ³ q becomes 2πqdq for nanorods 14 and proportional to q ² dq for quantum dots 15 .

The typical Raman spectroscopy plots are presented in Figs. 5–7for vibrational properties of carbon in SiO₂, ZnO and NiO, respectively. No significant asymmetry was observed in the D peak in all samples. This means that the size of the defect in these carbon clusters is large. However, asymmetry in the G peak was observed in all samples indicating that the nano-carbon is graphitic in structure.

In Figs. 5–7, we have re-drawn the G peak in a separate graph in order to demonstrate this asymmetry. The Richter equation was fitted to the experimental data and the relevant parameters in this equation were extracted. The results for C–NiO and C–ZnO samples are similar except that intensity is higher in ZnO than NiO. It is interesting to note that the C–SiO₂sample shows completely different results: extremely high scattering intensity and an increasing background intensity as the Raman shift increases. Due to the increasing background intensity, the phonon confinement model could not be used for the C–SiO₂samples. Typical parameters for the best C–NiO sample are tabulated in Table 2. It can be seen that the crystallite size that is responsible for the observed asymmetry in the Raman spectral line-shape for carbon in NiO is 6 nm.