2.1

Process curves and sample deposition

The generation of process curves and the layer deposition was carried out in an inline sputtering system with a top-down arrangement. The target geometry of this system is 750 mm × 100 mm. During the process, the substrate scans vertical to the narrow side of the target. The layer thickness is controlled by determining the deposition rate in a preliminary test and adjusting the number of scans and scan-velocity. Al₂O₃, SiO₂ and Si₃N₄ were sputtered by reactive dual-magnetron sputtering with MF-voltage from an Al- or a Si-target, respectively. The sputtering of these three materials was done with a power of 4 kW. Ag was sputtered from an Ag target by DC-sputtering with a power of 5 kW. The background pressure was below 3 × 10^-6 mbar. Ar was used as working gas and O₂ or N₂ were used as reactive gases. During the layer-deposition, Ar and N₂ were delivered into the coating chamber with a constant flow, while the oxygen partial pressure in the chamber was controlled by a feedback control system. The actual value of the oxygen partial pressure was not measured directly. With a ZIROX XS22.3H vacuum probe a lambda probe value “L” was determined. This value is directly inverse in proportion to the oxygen partial pressure in the coating chamber. The pumping power of the inline sputtering system was kept constant during the experiments.

In the case of oxides (Al₂O₃, SiO₂), S-shaped process curves have been generated by using the oxygen partial pressure as the control parameter. The lambda probe value L varied while the supply of the associated oxygen (gas flow into the coating-chamber) was recorded. The theoretical background of this kind of process curves is described in detail in the literature [21, 25, 26]. Three different values of Ar supply (gas flow into the chamber: 20 sccm, 90 sccm or 160 sccm) were used, whereby three different process curves could be generated. These curves are shown in figure 1. For each Ar-supply and thus for each process curve, an operation area has been determined (see figure 1). The sample deposition was done for every determined operation area. Thus, Al₂O₃ and SiO₂ where deposited under three different conditions each.

Figure 1:

S-shaped process curves for the reactive sputtering of Al₂O₃ (a) and SiO₂ (b). The curves differ with respect to the supply of working gas (Ar). The areas show the location of the operating points for the different Ar-supplies.

Si₃N₄ has been deposited with four different sets of coating parameters. The deposition of the sample was done with an Ar-supply of 20 sccm in combination with a low supply of N₂ (33 sccm) and a high supply of N₂ (120 sccm). In addition, the sample deposition was done with an Ar supply of 80 sccm in combination with a low supply of N₂ (47 sccm) and a high supply of N₂ (120 sccm).

In each deposition run (3 x Al₂O₃, 3 x SiO₂, 4 x Si₃N₄), a 3-inch Si-wafer and a smaller Si-substrate were placed, for the investigation of the film stress and the solubility behavior respectively. The layer thickness of the prepared samples is in the range of 280 nm ± 40 nm. The investigated samples and the dynamic deposition rates from the sample deposition are summarized in table 1.

Table 1.

Investigated layers (deposited on substrates), their sputtering conditions and the resulting deposition rate.

2.2

Solubility behavior of reactive sputtered layers

To determine the properties of the reactive sputtered layers (Al₂O₃, SiO₂ and Si₃N₄) regarding their solubility, these layers were deposited on polished Si-substrates. These samples were exposed to a solution, which consisted of distilled water with a certain amount of Na₂S. The exposure was interrupted frequently to determine the remaining layer thickness of the layers. With the help of this approach, it was possible to determine the removal rate of the layers in aqueous solution.

In a first test, all samples were exposed to a solution which consist of distilled water and 0.13 m% Na₂S. This solution has a pH-value between 11 and 11.5. The frequent determination of layer thickness was carried out by IR absorption spectroscopy with a Varian 3100 FTIR. The characteristic peaks resulting from absorption due to Al₂O₃, SiO₂ or Si₃N₄ have been analyzed. In this first test, only the Al₂O₃ layers and the layer “SiO_{2_}160Ar” reveal a significant decrease of layer thickness. This decrease is shown in Figure 2. The negative slope represents the removal rate. The slope and the uncertainty of the measurement were determined by linear regression.

Figure 2:

Removal rates in nm / h of the Al₂O₃ layers (a) and the SiO₂ layer “160Ar_2” in a basic solution (pH-value between 11 and 11.5), expressed through the negative slope m.

Those samples which show no significant decrease in the first test were exposed to a solution consisting of distilled water and 0.63 m% Na₂S. This solution has a pH-value between 12,5 and 13. The determination of the remaining layer thickness was not based on absorption spectroscopy, like in the first test, but on optical characterization. For this determination, the measurement of reflectivity in a spectral range from 400 nm to 1100 nm by a spectrophotometer (Lambda 850/900/950 from Perkin Elmer) and a subsequent characterization of the measurement by the software OptiChar [27] was realized (a dispersion model according to the Cauchy equation was applied). The reason for the change of the determination of the layer thicknesses is a lower measurement uncertainty for the optical characterization. The results of this second test are shown in Figure 3. The negative slope represents the removal rate. The slope and the uncertainty of the measurement were determined by linear regression.

Figure 3:

Removal rate in nm / h of the SiO₂ layers “SiO₂_20Ar” and “SiO₂_35Ar” (a) and the Si₃N₄ layers (b) in a basic solution (pH-value between 12.5 and 13), expressed through the negative slope m.

2.3

Mechanical stress in reactive sputtered layers

A Tencor FLX-2320 was used to determine the deflections of polished Si-wafers (3″) before and after the coating. Taking the layer geometry, the wafer geometry, the intrinsic material properties of the substrate and the change of deflection into account, the stress of the coated layers can be determined by applying the Stoney equation [28]. With this method, the stress of the coated layers was determined frequently (over a period > 3500 h after coating). The repeatability of measurements was < 5 MPA and the measurement uncertainty for the absolute value is <30 MPa. During the period of frequent measurements, the samples were stored under ambient conditions of 45% ± 20% humidity and a temperature of 23 ° C ± 4 ° C. The determined stress is compressive, which is indicated by the minus sign.

3.1

Process curves and sample deposition for reactive sputtering

For Al₂O₃ and SiO₂, the S-shaped process curves for reactive sputtering with O₂ could be generated by using the partial pressure of oxygen as a control parameter (figure 1). As described in the literature [21, 25, 26], these curves reveal areas, where sputtering takes place in the metallic mode, in the transition mode and in the compound mode:

Sputtering in the metallic mode takes place if the operating point is in the part of the process curve that stresses from the lowest oxygen partial pressure to the first inflection point. In this mode, the target is predominantly not poisoned due to reactions with a reactive gas. Thus, mainly the pure target material is sputtered. Sputtering in this mode often results in a deposition of layers which are highly absorbing in the visual spectral range. Therefore, this mode is not suitable for the deposition of protective layers for Ag-coatings.

Sputtering in the transition mode takes place if the operating point is between the first and the second inflection point. In this mode, the surface of the target is partly occupied by the reaction products (Al₂O₃ or SiO₂).

Sputtering in the compound mode takes place, if the operating point is in the part of the process curve that stresses from the second inflection point in the direction of higher oxygen partial pressure. In this mode, the surface of the target is predominantly poisoned with Al₂O₃ or SiO₂ respectively. The deposition rate for sputtering in the compound mode is expected to be lower than for sputtering in the transition mode and in the metallic mode.

Like shown in figure 1 and table 1, the sample deposition has been done with operating points in the transition mode. Thereby the sputtering of transparent layers with a relatively high deposition rate (compared to the sputtering in the compound mode) was carried out. The only exception is the SiO₂ layer which was sputtered with an Ar supply of 20 sccm and an operating point in the compound mode. The deposited rate of this SiO₂ layer is lower than the deposition rate of the other SiO₂ layers (table 1).

The chosen parameters for the Si₃N₄ have an influence on the deposition rate. A higher supply of N₂ leads to a lower deposition rate. This is true for a supply of working gas (Ar) of 20 sccm and of 80 sccm (table 1).

3.2

Solubility behavior of reactively sputtered layers

The exposure of the samples to an aqueous (basic) solution reveals that the removal rates of the three different Al₂O₃ layers are higher than the removal rates of all other tested layers (three different SiO₂ layer and 4 different Si₃N₄ layer). Even in an aqueous solution with a relative low pH-value between 11.0 and 11.5, these layers dissolve within a few hours (Figure 2 a). Furthermore, the SiO₂ layer “SiO₂_160Ar” shows a removal rate in the aqueous solution with a pH-value between 11.0 and 11.5 (Figure 2 b), even though this removal rate is significantly lower than the one of the Al₂O₃ layers. The SiO₂ layers sputtered with a lower supply of Ar and the Si₃N₄ do not show a significant decrease in layer thicknesses due to exposure to the aqueous solution with a pH-value between 11.0 and 11.5.

Figure 3 shows the decrease of layer thickness for an aqueous solution (pH-value between 12.5 and 13). The SiO₂-layer sputtered with an Ar supply of 35 sccm in the transition mode is less stable than the SiO₂ sputtered with an Ar supply of 20 sccm in the compound mode. However, the least stable layer of SiO₂ is still more stable than the best Al₂O₃-layer. Furthermore, the least stable of the Si₃N₄ layers (Si₃N₄_3) is still more stable than the best SiO₂-layer (SiO₂_20Ar).

3.3

Mechanical stress in reactive sputtered layers

The layer stress (compressive stress) of the three investigated Al₂O₃ layers is between -145 MPa and -210 MPa (Figure 3 a). The layers did not show a significant change of layer stress (above 10 MPa) in a time period of more than 3500 h.

The SiO₂ layers show a significant change of layer stress (figure 3 b). The layer “SiO₂_35Ar” shows a crucial increase of compressive stress of more than 120 MPa. The sample “SiO₂_160Ar” shows a variation of the compressive stress of ~ 40 MPa and thus also a significant change of layer stress. The sample “SiO₂_20Ar” has the most stable SiO₂-layer. However, a small increase in the compressive stress of ~ 10 MPa over a period of > 3500 h should be considered.

The Si₃N₄ layers show huge differences with regard to the mechanical stress. The Si₃N₄ layers sputtered with an Ar supply of 20 sccm have a high compressive stress of approximately -1490 MPa (Si₃N₄_1) and approximately -670 MPa (Si₃N₄_2). In contrast, the Si₃N₄ layers sputtered with an Ar supply of 80 sccm (Si₃N₄_3 and Si₃N₄_4) have a low compressive stress between -20 and -90 MPa. In case of the Si₃N₄, the layer Si₃N₄_4 shows a significant change of layer stress (figure 4 c and d).

Figure 4:

Progress of the compressive stresses of sputtered Al₂O₃ (a), SiO₂ (b), and Si₃N₄ (c and d).

4.1

Process curves and sample deposition

S-shaped process curves for reactive sputtering can be generated as described in the literature [21, 25, 26], whereby an interpretation of the influence of the sputter parameter is possible. The lower deposition rate for oxides (SiO₂), sputtered with an operating point in the compound mode instead of in the transition mode stands in accordance with the literature [29]. The reason for this behavior is that the sputter yield of the compound material (e.g. SiO₂) is often lower than the sputter yield of the pure (target-) material (e.g. Si) [26]. A lower sputter yield for the compound material as for the pure (target-) material could also explain why Si₃N₄ sputtered with a higher supply of N₂ has a lower deposition rate than Si₃N₄ sputtered with a low supply of N₂.

4.2

Solubility behavior of reactive sputtered layers

With regard to the solubility behavior of the sputtered layers, an influence of the intrinsic material properties and of the deposition parameters are visible.

Since the most stable Al₂O₃ layer (lowest removal rate) is less stable than the most unstable SiO₂ layer, and the most stable SiO₂ layer is less stable than the most unstable Si₃N₄ layer, the influence of the intrinsic material properties is evident. These significant differences can be explained by the different bonds of the materials. Due to the application of “Fanjans rules” , it can be estimated that the fraction of covalent bonds increases from Al₂O₃ to SiO₂ to Si₃N₄. These rules consider how far the cation polarizes the anion and thereby increases the fraction of covalent bonds which leads to a higher resistance in polar solutions [30]. A small radius (r_cation) and a high charge (z_cation) of cations, as well as a large radius (r_anion) and high charges (z_anion) of anions lead to an increase of the covalent fraction. To estimate and compare the potential of the building of covalent bonds, calculations are possible [30, 31]. The potential of the building of covalent bonds has been calculated for the three tested materials (Al₂O₃, SiO₂ and Si₃N₄) with the following equation:

Assuming the charges and ionic radii given in table 2, the potential for the building of covalent bonds is as follows: Al₂O₃ = 13, SiO₂ = 21 and Si₃N₄ = 34. Thus, the influence of the intrinsic material properties with regards to solubility behavior can be explained.

Table 2:

Binding partner and corresponding ionic radius [30].

In addition to the intrinsic material properties the deposition parameters have a significant influence on the solubility behaviour of the investigated layers. For Al₂O₃ and for SiO₂, the most unstable layer is the one sputtered with the highest supply of Ar (160 sccm). Since the pumping of the coating-chambers was done with a constant rate in all deposition runs, a higher supply of Ar leads to a higher pressure in the chamber. The higher pressure leads to a higher probability of inelastic scattering between Ar and sputtered adatoms, whereby the kinetic energy of the adatoms is reduced. A reduced kinetic energy of the adatoms can lead to an porose layer structure with an higher specific surface [32-34], whereby a higher removal rate of layer-thickness in an aqueous solution can be explained.

4.3

Mechanical stress in reactive sputtered layers

In contrast to these Al₂O₃ layers, the SiO₂ layers show a significant change of mechanical stress due to storage at ambient conditions. The change of stress can result from an interaction between water from the environment and the SiO₂-layers. The penetration of water into a layer can lead to water-induced stresses [19] which can be tensile stresses [35] or compressive stresses [19, 36]. For (vapor-deposited) SiO₂ layers in particular, the occurrence of water-induced stresses has been published [19, 36–39]. Water-induced stress could also be the reason for the unstable stress of the layer Si3N4_4.

The high compressive stress for sputtered Si₃N₄ layers (circa -1,5 GPa) is in accordance with results found in the literature [40]. In both cases, our findings and the findings descripted in the literature, the high compressive stress results from sputtering with a low working pressure or with a low Ar-supply, respectively [40]. Like discussed above, a higher working pressure leads to a higher probability of inelastic scattering between Ar and sputtered adatoms, whereby the kinetic energy of the adatoms is reduced and the microstructure of the layer becomes more porous [32-34]. A more porous microstructure can lead to lower compressive stress or higher tensile stress. The correlation between a low working pressure during sputtering and a compact and dense microstructure, which often goes together with compressive stress, is already described for Si₃N₄ [41]. Thus, the low stress for the Si₃N₄ layers sputtered with an Ar-supply of 80 sccm (Si₃N₄_3 and Si₃N₄_4) can be explained.