Measurement results.

a. Silicon L, nitrogen and oxygen K edges

In Fig. 4.56 the solid line shows the transmission of the thin SiO₂

  

Figure 4.56: Transmission of the thin SiO₂ film as a function of energy (solid line). Dotted line represents standard Henke data.

film at energies below 800 eV. Very strong near-edge oscillations can be seen at energies above the silicon L and oxygen K absorption edges. In the areas where near-edge structure does not play a significant role, a fit to the standard Henke data was made with a thickness of the film as a free parameter. The best fit thickness was found to be 0.157 microns, and the corresponding transmission curve is shown on Fig. 4.56 as a dotted line. The deviation of the measured transmission from the Henke data is very strong above the absorption edges, and the edges themselves are shifted. The structure above 100 eV corresponds to silicon L₁, L₂, and L₃ edges at 158, 107, and 105 eV respectively. At the oxygen edge a very strong resonant absorption peak is found at 538 eV.

A transmission curve for the SiO₂+Si₃N₄+SiO₂ sandwich (which is an exact copy of the CCD gate isolator) is shown on the Fig.  4.57. In addition to the silicon and oxygen edges this plot shows a prominent nitrogen K edge, also shifted from the tabulated atomic value. The transmission of this sandwich was modeled as a combination of the Henke-derived Si₃N₄ transmission and the experimentally measured transmission of SiO₂. The best fit (thicknesses of both materials being free parameters) is shown in Fig. 4.57 as the dotted line.

  

Figure 4.57: Transmission of the sandwich (solid line).

b. Silicon K edge

Fig. 4.58 contains absorption curves (not transmission!) of polysilicon, SiO₂, and the SiO₂+Si₃N₄+SiO₂ sandwich in the close vicinity of the silicon K edge. Each sample shows a sharp resonant peak right above the edge. Due to chemical shifts each of the three materials exhibits the silicon K-edge at slightly different energy. For polysilicon the peak is at 1841 eV; for SiO₂ it is at 1847.3 eV. The SiO₂-Si₃N₄-SiO₂ sandwich shows two distinct peaks which, although are not well-resolved, can be determined to be at 1847.3 and 1844.7 eV. The first one can obviously be attributed to SiO₂, while the second one originates from Si₃N₄. Polysilicon absorption shows a lot of structure due to its crystalline and ordered nature, whereas silicon dioxide has fewer peaks because it is amorphous and uncorrelated interference from remote atoms smears out the features.

  

Figure 4.58: Absorption of the thin films of polysilicon (solid line), silicon dioxide (dotted line) , and SiO₂ - Si₃N₄ (dashed line) sandwich.

The difference between the edge and resonant peak energies of polysilicon and SiO₂ can be seen very nicely in the oxidized polysilicon sample transmission in Fig. 4.59. A dashed line in this plot represents the result of fitting to this data the product of the transmissions of the separate films of polysilicon and SiO₂. The quality of the fit is so good that the dashed line can hardly be seen under the solid line.

  

Figure 4.59: Transmission of the oxidized polysilicon film (solid line) and result of the best fit to it of the product of the polysilicon and SiO₂ transmissions (dotted line).

In order to fill a gap in the data (no measurements were made in the range from 900 to 1300 eV), and also to extend the results to higher energies, we used Henke data (which should be adequate at energies far enough from the edges).

For each of the materials standard Henke data were used to fit the transmission at energies below the edge and far above the edge, where the near edge oscillations become weak. This procedure allows us to define the thickness of each film. The results are slightly different from the nominal values and are reflected in the labels in the above Figures. The densities of the materials used for those calculations were 2.19, 2.33 and 3.44 g/cm³ for SiO₂, polysilicon and Si₃N₄, respectively. For both polysilicon and SiO₂ near-edge measurements were combined with Henke atomic scattering factors for silicon and oxygen, and mass absorption coefficient were produced in the range from 60 eV to 15 keV.

We encountered some difficulties in our attempt to separate optical constants of the Si₃N₄ film from the other components of the sandwich, most likely due to deviations from stoichiometry in one or more constituent layers. Because of that, for further applications we used a transmission of the sandwich as a whole unit, instead of deriving a mass absorption coefficient for each component.