7A Prange in Spectrochim Acta Vol 44B 1989 p 437

8. A.W. Denier van der Gon, R.M. Tromp, and M.C. Reuter, in Thin Solid Films, No. 236, 1993, p 140 Electron Spectroscopies (AES and XPS)

AES and XPS determine the energies of electrons emitted from a surface. Those with specific energies usually are photoelectrons or Auger electrons. For detailed information, the reader is referred to Ref 3, 5, and 6. Principles for basic understanding are briefly characterized in the following and illustrated by some examples.

X-ray Photoelectron Spectroscopy. Irradiation of the sample with x-rays of energy hu (e.g., characteristic radiation of AlKa: hv = 1486.6 eV; or MgKa: hv = 1253.6 eV, often combined with a monochromator for narrower line width) causes emission of photoelectrons with kinetic energy Ekin according to:

where EB is the binding energy of the respective electron level and j A is the work function of the electron energy analyzer (Fig. 3). Because h u is the energy of the x-ray source used, the binding energy can be determined directly in most usual instruments if j A and the analyzed energy Ekin are empirically calibrated with standard samples. Tables of binding energies are available (Ref 9).

The terminology of XPS follows that of atomic physics. Each electronic level is characterized by its orbital number n (= 1, 2, 3, 4 . . .), the orbital momentum m (- 5,p, d, f), and the total spin quantum number I (- 1/2, 3/2, 5/2 . . .).

Chemical Effects and Compound Analysis. In general, chemical bonding changes the electron binding energy of valence band and core levels, which for core levels is recognized in XPS by a distinct "chemical shift" of the elemental peak with respect to the pure element (Ref 3). An example is shown in Fig. 4 for the 2p 3/2 and 2p 1/2 doublet of the Ti-2p XPS spectrum of a native oxide layer on a titanium nitride (TiN) coating (Ref 10). If the characteristic binding energies are known from measurements of standard samples or, often less precisely, from data banks (e.g., the National Institute of Standards and Technology the XPS database in Ref 9), the relative amount of a compound can be directly determined by peak fitting of peaks at the respective energies and subsequent determination of the Ti-2p 3/2, 1/2 respective peak areas. This was done for the Ti-2p 3/2, 1/2 doublet in Fig. 4, which is decomposed into two doublets, one for TiO2 and the other for TiN. Comparison of the peak areas in this case gives X TiO2/XTiN - 0.49. With the emission angle 9- 45° and an electron attenuation length of 10 - 1.2 nm, this value corresponds to a 1.8 nm thick TiO2 layer on TiN.


Fig. 4 X-ray photoelectron spectroscopy of the Ti-2p 1/2, 3/2 doublet in TiN and TiO2 obtained with a thin oxide layer on TiN. Source: Ref 10


Fig. 4 X-ray photoelectron spectroscopy of the Ti-2p 1/2, 3/2 doublet in TiN and TiO2 obtained with a thin oxide layer on TiN. Source: Ref 10

Usually, the chemical shift of compounds such as nitrides and oxides is of the order of a few electron volts (see Fig. 4) (Ref 3, 9). In many compounds, different valency states have a characteristic shift of the core level. For example, the Ta-4f doublet in Ta2O5 shows a shift of the binding energy of 5 eV from Ta0 to Ta5+, whereas for TaO a shift of 2 eV (corresponding to Ta2+) is obtained (Ref 11).

Auger Electron Spectroscopy. The most important method for chemical microanalyses of a surface is electron-beam-excited AES, which is based on the Auger effect (Ref 3, 5, 6). It is a radiationless transition, competing with x-ray fluorescence when an atom is ionized (e.g., by electron impact) in an inner shell (e.g., ^-shell). The generated electron vacancy can be filled by an electron from an outer shell (e.g., Z-shell), and the energy gain of this transition causes the emission of another electron from the outer shell. According to the net energy of the described Auger transition involving K, Li, L2 levels, the kinetic energy of the emitted Auger electron is given by:

where EK, ELi, and EL2 are the respective electron binding energies, AE is a small correction term for the relaxation of the twofold ionized state, and j A is the work function of the analyzer.

Equation 2 gives a characteristic kinetic energy—independent of the excitation energy--for any specific Auger transition of an element and therefore is the basis of qualitative analysis. Because three electron levels are involved, hydrogen and helium cannot be detected. For elements with higher atomic number, transitions such as LMM-, MNN-, and so on are used for Auger analysis. In general, the derivative spectrum d[N(E) • E]/dE is recorded as shown in Fig. 5 for TiN (Ref 10).

Fig. 5 Auger electron spectroscopy of a TiN coating in the derivative mode. At 383 eV, one of the major titanium peaks overlaps with the nitrogen (381 eV) peak. Source: Ref 10

Chemical Effects and Compound Analysis. Because of the three electron levels involved in AES, the chemical bonding effect is much more complex than in XPS and is most often recognized as a peak shape change that impedes elemental quantification (Ref 3). However, the knowledge of standard spectra of the relevant factors (or principal components) allows the decomposition of a measured depth profile into quantitative contributions, as shown in Fig. 6 for Cr2O3 and CrN in an oxidized (Cr,Pd)N coating (Ref 12).

Fig. 6 Factor analysis in Auger electron spectroscopy depth profiling of an oxidized (Cr,Pd)N coating. (a)

Standard spectra of the components chromium nitride and chromium oxide. (b) Comparison of measured and synthesized spectra. (c) Depth profile of the principal components. Source: Ref 12

Applying the mathematical method of factor analysis (Ref 13) to a series of spectra, for example in a depth profile (Ref 14, 15) enables the assessment of the number of principal components and a test of the validity of the otherwise obtained standard spectra, as well as the determination of one unknown component n if n-1 components are known. For more details of this powerful method, the reader is referred to Ref 13, 14, and 15 (see also the section "Thin Film and Interface Analysis" in this article).

Information Depth. Whenever an Auger electron or photoelectron is generated, its characteristic energy can be analyzed outside the sample only if it leaves the surface without any inelastic scattering event by which it loses some amount of energy and contributes then only to the lower-energy background of the spectrum. The probability of inelastic scattering increases exponentially with the traveled distance and is described by the inelastic mean free path 10 or, more precisely, attenuation length, which additionally includes elastic scattering (Ref 16). Depending on electron energy and material, the attenuation length is typically between 0.4 and 3 nm and increases with the kinetic energy. Although physical theories for predictions of 10 have recently been developed (Ref 17), the semiempirical relation of Seah and Dench (Ref 18) is still useful and gives the right order of magnitude for elements as well as for many inorganic compounds for the energy range of 30 to 3000 eV:

where 10 is in nanometers, E is the kinetic electron energy in electron volts, and a is the mean atomic distance in nanometers. The atomic mass number, M, the density, p, and Avogadro's number, N0, determine a = [M/(p • N0)]1/3. Values of 10 are typically between 0.4 nm (E »100 eV) and 2 nm (E »1500 eV), meaning that 63% of the measured intensity of Auger electrons or photoelectrons stems from a surface layer with the thickness 10 for emission perpendicular to the surface (9 = 0). This fact is the reason for the surface-specific information of the electron spectroscopies. If the emission angle 9 increases, the electron escape depth 1 is smaller than 10 and given by 1= 10 • cos 9(Ref 18).

Quantification Principles of Electron Spectroscopies. The signal intensity, given by the number of Auger electrons or photoelectrons in an elemental peak, is a measure of the number of atoms of this element in the analyzed volume and therefore of its concentration. Usually, the signal intensity is given by the area under the measured peak [N(E)] after background subtraction (Ref 3). Whereas background subtraction is generally used in XPS, the much larger electron background in AES causes relatively large errors. Therefore it is still customary to use the Auger peak-to-peak height in the differentiated [d[N(E) • E]/dE] spectrum in AES to characterize the elemental signal intensity Ii, as shown in Fig. 5 for the Auger spectrum of TiN (Ref 10). For both AES and XPS we may write:

Ii = k0 • (1 + rB) • Si • Xi • l02 • cos 9

where k0 is an instrumental constant given by the analyzed area, excitation intensity (primary current or x-ray intensity), and the total analyzer transmission; rB is the backscattering factor in AES, mainly dependent on the atomic number, and is typically between 0.1 and 1.5 (however, rB = 0 for XPS); Sj is the relative elemental sensitivity factor (generally defined with respect to pure silver in AES and with respect to pure carbon in XPS); Xi is the mole fraction in the analyzed volume (analyzed area times l2); l2 is the inelastic mean free path for the peak energy of element i or, more precisely, the attenuation length (Ref 16, 17), and 9 is the angle of emission of the detected electrons with the normal to the sample surface.

For a complete spectrum with Ii the intensity of the most intense peak of every detected element i, k0, and 9 are constant. Assuming that rB (in AES) and 1 are approximately constant, it follows from Eq 4 that:

which gives the mole fraction X, for the chosen element i of a total of n elements. Note that Eq 5 is only a first-order approximation, because the matrix dependence of 102, and of rB in AES, are neglected. This expected deviation can be taken into account by a correction factor FAB of element A in matrix B, by which every IA/SA has to be multiplied in Eq 3 (Ref 3, 10, 19):

where 10A and 10B are the electron attenuation lengths and r£ and rB are the AES electron backscattering coefficients of the pure element A and of element A in matrix B, respectively. The FAB correction factor depends on the difference, mainly with respect to density and mean atomic number, of matrix and pure element. In favorable cases, FAB is of the order of a few percent; for strong differences, FAB can be between 0.5 and 2 (Ref 3). According to Eq 5a, these values are a measure of the error margin contained in quantification of homogeneous samples using Eq 5.

Depth-Dependent Composition. Homogeneous composition in electron spectroscopies means constant composition within about 51 from the surface, because the intensities contributing to the signal decay exponentially with depth z, that is, Ii = I0 exp (-z/1). For z = 51, Ii is about 0.7% of I° of a pure element at the surface and is therefore approximately at the limit of detection for most elements. Because 10 is between 0.4 and 2 nm, the above condition means a required homogeneity in a range between 2 and 10 nm. This is the maximum information depth or the intrinsic sampling depth in AES and XPS.

For example, if 10 is 2 nm (10 = 1 for emission perpendicular to the surface, 9= 0) in an AB composition and the first monolayer d = 0.25 nm composed of pure A (XA = 1) with the rest being pure B, then IA/ fA = 1 - exp (-0.25/2) = 0.12 and IA/I°B = exp (-0.25/2) = 0.88, which corresponds to XA = 0.12 homogeneously distributed in a region of > 10 nm. From the measured intensity alone, both possibilities or intermediate cases are indistinguishable. If 10 is 0.4 nm, XA = 0.46 is obtained. Figure 7 shows the relation between the ratio X A (5 • 1)/XA(d), of the measured X A (5 • 1) and the true XA(d) homogeneously distributed within a layer of thickness d. It is easily recognized from Fig. 7 that because of matrix effects with respect to the elemental standard, changing 1 changes the sampling depth and therefore the sensitivity factor, as outlined above (Ref 19).

Fig. 7 Ratio of the average concentration X A, assuming a constant X A with depth of an A-B alloy, to the concentration XA(d) of a thin layer of thickness d of A on a substrate B

Because the relative change in 1 with composition should be the same for any elemental peak, the sensitivity factor, «SA/SB, which is decisive for correct quantification of homogeneous samples, should not change with changing composition (except for rB in AES, Eq 5a). Therefore, any matrix has a characteristic set of elemental sensitivity factors that can be best evaluated by the use of standards (Ref 3, 4, 19, 20).

Microanalysis and Lateral Resolution. The technique of scanning a focused electron beam and simultaneous plotting the peak-to-background intensity of an elemental signal using scanning Auger microscopy (SAM) allows elemental mapping (Ref 21). A secondary electron image of an electronic thin-film device is shown in Fig. 8(a), whereas Fig. 8(b, c, d) show the silicon, aluminum, and fluorine maps. In this case, the device was etched by a fluorine-containing reactive gas. It is clearly recognized that fluorine remained only at the aluminum contact layers and not on the silicon structures (Ref 22).

Fig. 8 Integrated circuit after cleaning treatment with a fluorine-containing compound analyzed with Auger electron spectroscopy/scanning Auger microscopy. (a) Secondary electron image. (b) Chemical map of aluminum. (c) Chemical map of silicon. (d) Chemical map of fluorine. Note that fluorine is removed by electron-stimulated desorption from focusing spots of the electron beam. Source: Ref 22

Fig. 8 Integrated circuit after cleaning treatment with a fluorine-containing compound analyzed with Auger electron spectroscopy/scanning Auger microscopy. (a) Secondary electron image. (b) Chemical map of aluminum. (c) Chemical map of silicon. (d) Chemical map of fluorine. Note that fluorine is removed by electron-stimulated desorption from focusing spots of the electron beam. Source: Ref 22

A lateral resolution in SAM of about 0.1 pm can be routinely obtained. Modern instruments with field emission cathodes provide primary beam diameters of 15 nm at 1 nA beam current, suitable for the analysis of microelectronics devices. Limitations in quantification and detrimental effects such as beam heating are discussed in Ref 22.

XPS offers somewhat less spatial resolution than SAM, about 70 to 5 pm. Small-spot XPS can be performed either by excitation of a restricted sample area (e.g., a focused x-ray beam by a bent monochromator crystal) or by an electron optical lens in front of the analyzer, which selects a limited area of the sample for analysis (Ref 23). Imaging XPS uses monochromatic electrons and a multichannel parallel detection device to generate an elemental image (Ref 23).

Limitations and Special Problems. High-spatial-resolution AES of small particles, protrusions, or precipitates in micrometer dimensions is difficult to analyze quantitatively because of distortions by backscattered primary electrons that excite Auger electrons from the surrounding material. The achievable resolution is limited by the amount of current in a small-diameter beam that is necessary for a sufficient signal-to-noise ratio and by the temperature increase that is due to the high current density (up to 103 A/cm2), which may cause diffusion and even evaporation processes in thin films on substrates with low thermal conductivity (Ref 21, 22). Furthermore, electron-stimulated desorption may lead to decomposition of compounds at the surface. For both cases of sample damage, lowering the current density (e.g., by enlarging the beam diameter or scanning the electron beam over a certain area) is a remedy, but at the cost of decreasing spatial resolution. The usually much lower power density greatly reduces this problem in XPS.

Charging of insulating materials is a severe problem in both AES and XPS. However, whereas negative charging of the sample in AES is difficult to overcome in practical analysis, even though a number of special techniques can help (Ref 24, 25), XPS causes positive surface charging, which can often be successfully compensated by low-energy electron flooding with an auxiliary electron source (Ref 26).

0 0

Post a comment