Theta Probe and Theta 300 instruments from Thermo Scientific have been utilized to define very thin layers of silicon dioxide (SiO2) and silicon oxynitride on silicon by means of parallel angle resolved XPS (PARXPS) method.
Comparison between PARXPS and ellipsometry demonstrated that the former method gives excellent accuracy, and a high level of precision was also shown through a dynamic stability test.
In addition, the PARXPS method gives non-destructive, reproducible depth profiles of the chemical conductions inside the layer. This article shows how PARXPS measurements of silicon oxynitride samples provide thickness and uniformity of films, chemical state data, nitrogen dose and distribution of nitrogen in each of its chemical states. It also describes the accuracy and precision of the information.
In this analysis, a set of SiO2 on silicon samples was investigated. The oxide's thickness was already determined through ellipsometry. The results demonstrate superior linearity with a unity gradient to within ~ 1% (Figure 1).
This offset is induced by surface impurity, which would not be possible to differentiate from the layer of oxide through ellipsometry. SIMS analysis reveals that there is a layer of impurity of about 0.8nm thick that correlates well with the PARXPS data.
Figure 1. Comparison of PARXPS and ellipsometry
To perform a dynamic repeatability test of Theta 300 instrument, two silicon oxynitride wafers were utilized. Measurements of both thickness and nitrogen dose were carried out at three sites on individual wafer (Figures 2 and 3).
The measurements were again performed 10 times and the wafers were taken out from the Theta 300 system between measurements. The mean (mn), standard deviation (SD) and relative standard deviation (RSD) are illustrated for nitrogen dose and thickness in Tables 1 and 2.
Figure 2. Dynamic repeatability tests showing the thickness of the oxynitride layer at three points on each of two wafers.
Figure 3. Dynamic repeatability tests showing the nitrogen dose in the oxynitride layer at three points on each of two wafers.
Table 1. Reproducibility of thickness measurements from dynamic repeatability test
Table 2. Reproducibility of nitrogen dose measurements from dynamic repeatability test
||(1014 atoms cm-2)
|SD (1014 atoms cm-2)
||Mean (1014 atoms cm-2)
|SD (1014 atoms cm-2)
A wafer map was obtained from half a 200mm wafer, and at each of the 522 mapping points the thickness of the oxide was determined from the PARXPS information and a thickness map was subsequently built (Figure 4). Table 3 shows the summary of the data. Ellipsometry was used to map the same wafer. Following this, line scans were obtained from both the ellipsometry and PARXPS maps. These are then compared which show close agreement (Figure 5).
Table 3. Summary of the data in the wafer map
|Number of mapping points
|Acquisition time per point
|Mean oxide thickness
|Relative standard deviation
Figure 4. 3D representation of oxide thickness on 200mm wafer from ARXPS data
Figure 5. Comparison of the thickness measured using PARXPS with that measured using ellipsometry
Figure 6 illustrates the N 1s spectra from a silicon oxynitride layer obtained at a bulk sensitive and a surface sensitive angle. The spectra reveal the presence of a minimum of two chemical environments for nitrogen (Na and Nb) (Figure 7).
PARXPS measurements taken from a silicon oxynitride layer on silicon reveal significant alterations in the nitrogen chemistry with angle (Figure 6). Na is a primary peak across the angular range, whilst Nb is reduced at the more grazing emission angles. This indicates that Nb is situated under the surface close to the interface of oxynitride and silicon.
Figure 6. N 1s spectra from an oxynitride layer from bulk and surface sensitive angles
Figure 7. N 1s binding energies for the chemical environments present in oxynitride layers
Relative Depth Plot
A relative depth plot displays the arrangement of the chemical species in the layer. This plot is acquired by determining the peak regions from a small array of surface sensitive angles, separating this by the intensity in a small range of massive sensitive angles and acquiring the natural logarithm. This process is again performed for individual species existing in the spectrum and the outcomes are plotted on a chart. An example of this is illustrated in Figure 8, which demonstrates the predicted layer of carbon present at the surface, the elemental silicon deepest in the layer, and the nitrogen species at varied depths inside the layer.
Figure 8. Relative depth plot from a silicon oxynitride layer
While the relative depth plot fails to give quantitative depth data, it gives a precise indication of the arrangement of layers. Independent of any model, this technique provides a useful check upon measuring the quantitative depth data.
A non-destructive concentration depth profile can be constructed from PARXPS information using the method of maximum entropy; such a profile is shown in Figure 9. In this profile, the primary nitrogen species persists across the oxynitride layer, while the minor species (Nb) is present close to the silicon interface. The depth profile in Figure 9 correlates with the relative depth plot (Figure 8), but gives more critical data. A summary of the thickness and dose data derived from Figure 9 is shown in Table 4.
Figure 9. Concentration depth profile through a silicon oxynitride layer constructed using the Maximum Entropy method. Note that concentration axes do not apply to Si and C profiles.
Table 4. Quantitative information extracted from the depth profile in Figure 9
||1.8 x 1015 atoms cm-2
||1.6 x 1014 atoms cm-2
Comparison with Sputtering
Using low energy ion beams, it is possible to analyze oxynitride layers by sputter profiling. Figure 10 illustrates a case wherein argon ions with an energy of 500eV were utilized.
Figure 10. Sputter profile through the same oxynitride layer as that shown in Figure 9. The ion energy used for this profile is 500 eV. (a) is the full profile and (b) is an enlargement of the early part of the profile.
The results of sputter profile revile that Nb vanishes during the early stages of the profile, suggesting that Nb is situated near the surface. Based on the PARXPS information, this interpretation is not accurate. The right explanation of the sputter profile is that under the effect of the ion beam, Nb is being changed into Na. This change does not occur during the course of PARXPS analysis. Analogous results were acquired using an ion energy of 250eV.
Reproducibility of Profiles
To analyze the reproducibility of the data as well as the Maximum Entropy approach for profile creation, data from an oxynitride layer were acquired thrice from the same point on a wafer. Figure 11 shows these results. In this sample, nitrogen existed only in a single chemical state.
Figure 11. Comparison of three depth profiles generated from the same point on a wafer using the Maximum Entropy method. The data show excellent reproducibility.
Dose and Thickness Measurement
Dose as well as thickness can be quantified at different points on a wafer to create a map or profile. These can be created from complete wafers by means of the Theta 300 instrument. A 49-point oxynitride thickness map obtained from a 300mm wafer is illustrated in Figure 12. 1.733nm is the mean thickness of this layer, with a standard deviation of 0.79% and an overall thickness range of 0.054nm.
Figure 12. A 49-point oxynitride thickness map from the whole of a 300mm wafer.
It is possible to create a nitrogen dose map from the same data set (Figure 13). Nitrogen dose is measured once its depth distribution is corrected. In this example, the mean dose is 2.43 x 1014 atoms/cm2, with a standard deviation of 2.15% and an overall variation of dose across the entire wafer of 0.32 x 1014 atoms/cm2.
Figure 13. A 49-point nitrogen dose map from a 300mm wafer
For measuring film thickness, PARXPS is found to be excellent in terms of accuracy and precision when compared to ellipsometry. Unlike ellipsometry, PARXPS is capable of differentiating the chemical states of nitrogen in the layer and gives their depth and distribution. A thin layer of contamination will exist at the wafer's surface, but ellipsometry will not take account of this contamination.
There is a risk that the quantified layer thickness will comprise the layer of contamination. In PARXPS, the layer of contamination is not included in the determination of oxide or oxynitride thickness. As dielectric layers become thinner and thinner, the layer of contamination becomes a major part of the layer thickness. It becomes critical that the dielectric layer is differentiated from the contaminant layer.
It is possible to construct thickness maps and chemical state maps of a wafer from the same data set. With the help of PARXPS, an oxidized form of nitrogen was observed to be present at the silicon interface. Upon analyzing this layer through using sputter depth profiling, the oxidized form of nitrogen is removed during the preliminary stages of profile acquisition.
This would lead to an inaccurate conclusion that the oxidized nitrogen is not at the interface between the silicon and the oxynitride, but at the layer's surface. This is because the range of the argon ions utilized for sputtering is analogous to the thickness of the layer.
Hence, the ion beam would interact with the layer's interface region from the beginning of the profiling. In this example, the outcome of the interaction is to bring down the oxidized nitrogen to the lower binding energy.
This information has been sourced, reviewed and adapted from materials provided by Thermo Scientific – X-Ray Photoelectron Spectroscopy.
For more information on this source, please visit Thermo Scientific – X-Ray Photoelectron Spectroscopy.