2.5.1 Energy Dispersive Spectroscopy (EDS)
Thin film atomic concentrations were characterized using two main techniques: (1) Energy Dispersive Spectroscopy (EDS) and X-ray Photoelectron Spectroscopy (XPS). The former analyzes the bulk concentration down to at least 100 nm while the latter measures the elemental concentration down to ~2 nm, or the first few surface monolayers. A diagram of the principles involved can be seen in Figure 21.
EDS measures atomic concentration by subjecting the sample to high-energy electrons in an SEM (~5 – 30 KeV). These excite X-ray lines characteristic of the elements in the material. The initial high-energy electrons generated in the SEM are accelerated towards the sample, impacting with high enough velocities to transfer sufficient energy for inner shell electrons up to a few microns deep to leave their
47
respective atoms. Once these inner-shell electrons are ejected, electrons from the outer shells drop down to fill up the now-empty states. The difference in energy between the former outer electrons and the now-inner electrons is released at wavelengths in the X- ray regime, which are characteristic of the atoms. A solid-state detector can measure the intensity of these X-rays over a range of wavelengths, translating the data into an estimate of the bulk atomic concentration in a thin film.
This particular study used a JEOL 8900 Scanning Probe SEM with a Beryllium window in front of a SiL detector. The scanning area was typically a square, ~ 1mm on a side, and at the 100x magnification level. Due the low-energy X-ray absorbance properties of the detector and window, elements with atomic numbers below that of Mg had trouble being measured. In addition, carbon contamination on the window reduced resolution. Other problems impacting accuracy include secondary electrons generated from X-rays coming from deep within the sample (although initial impacting electron energies were always kept above 2x the expected energy needed to knock out inner electrons to reduce this problem), charging on the surface (although most were metallic), overlap between peaks within 100 meV of each other (wavelength dispersive spectroscopy, with a crystal such as LiF diffracting wavelengths close to each other for easier analysis, could be used), and a non-standard calibration applied to convert the characteristic X-ray lines to composition. In all scans, the author collected X-ray lines for at least 60 seconds using initial electron energies from 7 to 20 KeV, resolving atomic composition data within 1-2% precision. It was estimated that the accuracy of this technique to determine bulk thin film composition was +/- 5%.
48
Figure 21. Principles of a) EDS and b) XPS for a P atom. For EDS, 1) a high-energy electron 2) knocks an inner shell electron out, resulting in 3) a more energetic outer shell electron falling into its spot and 4) releasing X-ray radiation characteristic of the atom. For XPS, 1) X-rays from an Al source 2) knock (usually valence) electrons out, with a kinetic energy (KE) calculated from the energy of the initial X-ray (ħν) – the binding energy of the electron (BE) – the work function of the detector (Φ) [76].
49
2.5.2 X-Ray Photoelectron Spectroscopy (XPS)
XPS was used to analyze compositions typically 1 – 4 nm from the surface. As such, it could only exploit a technique where characteristic properties were detected from atoms near the surface. In order to do this, it probes the samples with X-rays. The X-rays (typically around 1 KeV) penetrate through much of the bulk, knocking off electrons with binding energies less than this. The kinetic energy of such ‘photoelectrons’, equal to the incident energy of the X-ray minus the binding energy, is read by a detector that applies a magnetic field to separate the negatively-charged electrons around a turn; more kinetically energetic photoelectrons travel faster and have less time to bend in a field before striking a solid state detector. By backing out the initial X-ray energy and known work function of the detector, the intensities of photoelectrons coming off at different binding energies could be tabulated. Using a few ‘fudge factors’ (e.g. the relative sensitivity factor), this can be converted into atomic percentages based on binding energies characteristic to the constituents. Since the kinetic energies of these photoelectrons can reach up to approximately 1 KeV, photoelectrons produced down in the bulk do not have sufficiently kinetic energy to escape to the surface. Thus, even though the initial X-rays travel through much of the bulk, only signal within the first few nanometers is actually picked up by the detector. In addition to determining the composition, XPS can also be used to estimate the oxidation states and neighboring environment of the constituent atoms since the binding energies of photoelectrons are somewhat dependent on them. For instance, a photoelectron coming from a Ti4+ ion will have had a higher binding energy than that from Ti3+ ion due to less electron shielding from the attraction of the atomic nucleus. Some ions, such as Pb+4 and Pb+2, do not follow this trend due to relativistic effects.
50
In this study, a Surface Science Instruments (SSI) model SSX-100 was used, with an Al Kαsource generating X-rays having an energy of 1.4866 KeV over a spot size with a diameter ~ 800 μm. The hemispherical analyzer which ‘bends’ the photoelectrons to separate them into bins can operate over ~50 to 1100 eV. Operating pressures are at UHV pressures ~2 x 10-9 torr, about a few orders of magnitude greater than the EDS microprobe so the photoelectrons generated have a large mean free path to the detector. The listed detection limits were from .1 to 1% with accuracies within a few percent. Possible sources of error include sample and photoelectron charging leading to intensity peak broadening, plasmons and auger electrons introducing unwanted peaks, and contamination having a much greater effect on the low-intensity photoelectron signal compared (usually) to the much higher count ratio seen in EDS. Efforts at ‘cleaning’ the surface by sputtering off unwanted atoms (e.g. carbon-base molecules) with Ar+ ions beforehand were limited when oxidation states were desired. Theses Ar+ had a tendency to change the oxidation states of some atoms (e.g. Ru, Pb) noticeably after impact.