X-Ray Photoelectron Spectroscopy (XPS) and Auger Electron Spectroscopy (AES)

Abstract

X-ray photoelectron spectroscopy (XPS), also known as electron spectroscopy for chemical analysis (ESCA), and Auger electron spectroscopy (AES) are widely used materials characterization techniques belonging to the general class of methods referred to as surface analysis. These non-destructive techniques provide, to varying degrees, semi-quantitative elemental, chemical-state and electronic-structure information from the top 10 nm of a material and are sensitive to elements Li and above. XPS and Auger have both found applications over a vast range of material classes; such as metallic, ceramic, polymeric, and composite; and technologies such as microelectronics, solar energy, and nanotechnology. Modern spectrometers are now not only capable of achieving high-energy resolution spectroscopy, but are also capable of 2-dimensional imaging.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

X-ray photoelectron spectroscopy (XPS), also known as electron spectroscopy for chemical analysis (ESCA), and Auger electron spectroscopy (AES) are widely used materials characterization techniques belonging to the general class of methods referred to as surface analysis. Each of these techniques provides, to varying degrees, semi-quantitative elemental, chemical-state and electronic-structure information from the top 10 nm of a material. Another widely used surface analytical method covered in this book is secondary ion mass spectrometry, Chap. 4.

Advances in instrumentation for XPS and AES during the time following their introduction, have led to the application of these methods to a vast range of material classes, such as metallic, ceramic, polymeric, and composite, and technologies such as microelectronics, solar energy, and nanotechnology. Modern photoelectron spectrometers are now capable of achieving energy resolutions <0.3 eV due to improved X-ray monochromator and energy analyzer design. The addition of field emission electron sources to Auger spectrometers has made spatial resolutions <10 nm possible. Developments in charge neutralization in both XPS and AES systems have made it possible to reliably collect data from most materials regardless of conductivity. Today, nearly any material that can be introduced into an ultra-high vacuum environment can be analyzed by XPS and AES.

This chapter begins by introducing XPS and AES with a brief discussion of the physical basis of each of the methods and then systematically develops the discussion with a description of how to extract and interpret the information contained within the data. As it is useful to combine multiple techniques in materials characterization, it is also useful to carry out combinations of different methods of the same technique in order to extract maximum information. Several examples of this approach will be presented in this chapter. In many cases, the same material will be presented using different characterization methods, demonstrating the range of possible information that is accessible by XPS and AES.

3.1 X-Ray Photoelectron Spectroscopy (XPS)

X-ray [1] photoelectron spectroscopy, is based on the photoelectric effect discovered by Heinrich Hertz in the late 1800s [2]. Later in 1905, Albert Einstein described the process in quantum mechanical terms [3]. It was not until the mid-1960s that Kai Siegbahn and his research group at the University of Uppsala, Sweden developed the technique into a practical analytical method [4].

3.2 Photoelectron and Auger Electron Emission

In the photoelectron emission process , an incident photon of energy hυ is absorbed by an atom. With that energy, a photoelectron is emitted with a kinetic energy equal to:

$$ KE= hv- BE-{\varPhi}_{spectrometer} $$

(3.1)

Where BE is equal to the binding energy of the electron and Φ_spectrometer is the work function of the spectrometer (see Fig. 3.1). In order to produce photoelectrons having discreet binding energies, a monoenergetic source of X-rays is necessary. As the kinetic energies of the photoelectrons are dependent on the X-ray source energy, photoelectron spectra are typically presented on a binding energy scale, Eq. 3.2, to make comparison of spectra collected with different sources more straightforward.

Fig. 3.1

Photoelectron and Auger electron emission process

$$ BE= hv- KE-{\varPhi}_{spectrometer} $$

(3.2)

Following the process of photoelectron emission, an excited state ion is created. The excess energy of this ion can be released through a relaxation process in which an electron from an upper shell fills the hole and then either an X-ray photon is released, X-ray fluorescence , or a second electron is emitted, Auger electron emission . In each case, energy is conserved in the relaxation and the X-ray photon or Auger electron is emitted with energy equal to the differences in energies of the orbitals, Eq. 3.3. When plotted on a binding energy scale, the positions of Auger lines will depend on the X-ray source energy. For this reason, X-ray induced Auger spectra are typically presented on a kinetic energy scale. This allows for straightforward comparison with electron induced Auger spectra.

$$ K{E}_{Auger}=B{E}_1-B{E}_2-B{E}_3 $$

(3.3)

3.3 The Fermi Level

In order to consistently report the binding energy of an electron, a universal energy reference must be defined. If a system of one orbital in the valence band of a metallic system having an infinite supply of electrons is considered, the energy of the orbital can be defined based on its occupancy [5]. According to the Pauli exclusion principle, the orbital can either be occupied by 0 or 1 electron. If the orbital is unoccupied, the energy of the orbital is 0. If the orbital is occupied, the energy is ε. The Gibbs energy, ζ, of the two states, N = 0 and N = 1, can be described then by ζ = 1 + λexp(-ε/T), where λ = exp(μ/T) and μ is the chemical potential of the electron and T is temperature. In condensed matter physics, the chemical potential of the electron is often referred to as the Fermi Level , E_f. The thermal average of the occupancy of the orbital can be described by the ratio of the Gibbs energy of the N = 1 state to the Gibbs energies of both the N = 0 and N = 1 states. In units of electron volts, eV, the ratio becomes:

$$ f(E)=\frac{1}{ \exp \left[\frac{E- Ef}{ kT}\right]+1} $$

(3.4)

Equation 3.4 is known as the Fermi-Dirac distribution function .

The Fermi-Dirac distribution function describes the average occupancy of an orbital with energy of E-E_f at temperature T. At T = 0 K, f(E) = 1 for E < E_f, and f(E) = 0 for E > E_f. For all values of T when E = E_f, f(E) = 0.5. The Fermi-Dirac distribution function plotted as a function of E-E_f for different temperatures is shown in Fig. 3.2. It is the convention in XPS that the Fermi Level is set to 0 eV on the binding energy scale, thereby defining the binding energy of an electron as the difference between the orbital energy of an electron and the Fermi Level.

Fig. 3.2

The Fermi-Dirac distribution function plotted as a function of E-E_f at 0, 300, and 2,000 K

At room temperature, kT = 0.025 eV which results in a nearly vertical Fermi level. This makes the Fermi level a useful means to measure the energy resolution of the spectrometer. As the energy resolution of most modern laboratory-based spectrometers, even those with X-ray monochromators, is limited by the X-ray source line width, the Fermi Level will be broadened to represent roughly a minimum electron temperature of 2,000 K (for a resolution of 0.4 eV).

3.4 The Finite Potential and Work Function

The most basic way to represent the electronic structure of a condensed phase is the jellium model in which the ion cores in an infinite metallic crystal are replaced by a uniform sea of positive charge. For the case where a surface is present, the surface represents the boundary of the positive charge. The simplest quantum mechanical method to model the surface is to assume an infinitely high potential barrier at the surface, known as the infinite potential, where ε = 0 inside of the metal and ε = ∞ outside of the metal (vacuum). A limitation of this model is that it ignores the periodic nature of the crystal and does not take into account variations in the properties of different crystal faces.

A more realistic representation of the surface is to assume that the potential barrier is finite. Such a representation is known as the finite (or semi-infinite) potential . Here we define the potential as Φ, which is referred to as the work function , where Φ = (−ε−μ) and ε is the energy of the electron just outside of the surface and μ is the chemical potential of the electron at the surface. At semi-infinite distances from the surface, the energy of the electron is considered to be constant and is defined as the vacuum level , see Fig. 3.3. The work function represents the energy barrier that an electron must overcome in order to become a free electron. Values for Φ will vary as related to the chemical/electronic structure of the surface and can be influenced by the exposed face of the crystal surface and adsorbates on the crystal surface.

Fig. 3.3

The finite surface potential

3.5 Electron Attenuation and Surface Sensitivity

Electrons in the typical kinetic energy ranges studied by XPS (0–1,400 eV) and Auger (20–3,000 eV) are only able to travel finite distances through a condensed phase before they scatter either elastically (electron momentum may change, but energy is conserved) or inelastically (electron momentum and energy changes) resulting in the attenuation of a measure electron signal as a function of depth. The distance an electron travels through a condensed phase between scattering events is defined as either the elastic mean-free path λ _e , (EMFP) or the inelastic mean-free path (IMFP) , λ. When or inelastic scattering occurs, the electron loses kinetic energy and no longer contributes characteristic information regarding the binding energy of an emitted electron.

The attenuation of a characteristic electron signal can be described by the Beer-Lambert relationship .

$$ I={I}_0 \exp \left(\raisebox{1ex}{$-d$}\!\left/ \!\raisebox{-1ex}{$\lambda \cos \theta $}\right.\right) $$

(3.5)

Where I is the attenuated electron signal, I ₀ is the non-attenuated surface electron signal, d is the electron depth, λ is a value in length describing the attenuation of the electron, and θ is the electron emission angle. The Beer-Lambert relationship shows that the electron signal decays exponentially as a function and depth and emission angle.

Values for electron attenuation length , or escape depth , Λ, may be determined empirically by measuring the attenuation of a characteristic electron signal by overlayers of known thicknesses. This approached was used by Bain and Whitesides [6] where self-assembled monolayers of alkanethiols, HSC_nH_2n+1 on Au were studied. By plotting the natural log of the intensity of the Au 4p_3/2 (KE = 940 eV), the Au 4d_5/2 (KE = 1,151 eV), and the Au 4f_7/2 (KE = 1,402 eV) as a function of the carbon chain length n, the slope becomes nd/Λcosθ. Using 0.127 nm as the incremental length of a –CH₂- group and a 30° tilt angle vs. surface normal of the alkanethiol, d = 0.127cosθ. Values of Λ were thereby calculated to be 0.28, 0.34, and 0.42 nm for the Au 4p_3/2, the Au 4d_5/2, and the Au 4f_7/2 lines, respectively. An approach similar to this will be used later to determine overlayer thicknesses. The electron attenuation lengths Λ, determined in this study include contributions from both elastic and inelastic scattering.

Calculations for random scattering in crystalline materials suggest that λ _e  < λ [7]. Also, there is some degree of uncertainty in the mechanism of elastic, or quasieleastic (phonon) scattering for electrons in the range of energies studied by XPS and Auger, 50–1,500 eV which makes these processes difficult to model. For these reasons, the elastic scattering contributions to electron attenuation are often neglected in models describing the distance an electron travels through a condensed phase.

Values for inelastic mean-free paths in solids have been calculated by Tanuma, Powell, and Penn. By fitting calculated IMFPs to a modified form of the Bethe equation, Tanuma, Powell and Penn found the four parameters in the equation could be related to several material properties; atomic weight, density, number of valence electron per atom, and band gap energy [8]. The resulting equation was designated as TPP-2. TPP-2 was developed and tested with mainly high-density elements. Systematic differences of ~40 % were found for the IMFPs for lower-density organic compounds with IMFPs determined by TPP-2, resulting in the derivation of a modified expression known as TPP-2M [9, 10]. Values for IMFPs calculated by TPP-2M are shown in Fig. 3.4. The electron IMFPs calculated by TPP-2M, shown in Fig. 3.4, demonstrate that values of IMFPs increase systematically with kinetic energy, but do not necessarily show a relationship with atomic number. As can be observed in Fig. 3.4, values for IMFPs can vary a great deal from material to material.

Fig. 3.4

Values for inelastic mean-free paths in solid materials as a function of kinetic energy calculated using TPP-2M [10]

As the measured signal described by the Beer-Lambert relationship, I, will never reach 0 according to this relationship, an approach to consider the information (or sampling depth) of XPS and Auger may be defined as the point from which a specified fraction of detected electrons originate. A commonly used depth is one where I/I ₀ = 0.05. This occurs when λ = 3 and is generally described as the “95 % depth.”

3.6 Instrumentation

Instrumentation for X-ray photoelectron spectrometers can vary greatly in terms of sophistication. Figure 3.5 shows a schematic for a typical modern high resolution X-ray photoelectron spectrometer. The spectrometer consists of a monoenergetic X-ray source, electron energy analyzer, and detector. If light elements such as Mg or Al are used as X-ray source anode materials, an X-ray monochromator is not necessary as the characteristic X-ray emission from these elements are nearly monoenergetic. High intensity high resolution X-ray monchromators are now readily available and are generally included with modern spectrometers.

Fig. 3.5

Schematic of a typical modern high resolution X-ray photoelectron spectrometer

Once electrons are excited by the X-rays, these electrons are collected by an electron energy analyzer. The electron collection example shown consists of a set of extraction lenses, a hemispherical energy filter, and a detector. In this configuration, the sample, usually grounded, is placed next to the analyzer in a field-free region and electron focusing is done electrostatically. Here, only the electrons with trajectories allowed by the lens geometry are collected. An image of the electrons is first projected onto a plane where it passes through an aperture. This plane is often referred to as the selected area aperture plane as it usually consists of a variety of apertures of different size allowing one to change the area of the sample from which electrons are collected. The aperture diameter divided by the extraction lens magnification defines the analysis area. In order to achieve a high resolution which is constant over a wide binding energy range, the analyzer is usually operated in a constant pass energy fixed analyzer transmission (FAT) mode. In this mode, electron energies are retarded by the electric fields of the extraction optics to an energy known as the pass energy. Scanning of a range of electron energies is accomplished by variation of the retardation fields applied by the extraction lenses.

Next, the image of the electrons is projected to a second plane known as the analyzer’s entrance aperture plane. For high resolution spectroscopy, this aperture consists of a narrow slit in order to minimize the angular dispersion of the electrons as they enter the hemispherical energy filter. Energy filters can consist of either hemispherical sectors or full hemispheres as the electrostatic element. The former offers the advantage of a smaller size while the latter offers the advantage of enhanced electron transmission by avoiding distortions in the electric fields which result from the termination of the electric fields at the edges of a hemispherical sector. Energy filtering of the electrons is done by the application of potentials V ₁ to the inner hemisphere, and V ₂ to the outer hemisphere, the difference ΔV is determined by the pass energy. The potentials V₁ and V₂ float on the potential applied to the lens at the analyzer’s entrance aperture.

Finally, an image of the electrons is projected through the energy filter to a third plane referred to as the detection plane. Often, photoelectron spectrometers are equipped with either multiple discreet detectors or, in the example shown, a single position-sensitive detector in order to increase detection efficiency by counting the electron signal multiple times as it passes along the detection plane.

3.7 Elemental and Chemical Shifts

The most fundamental capability of XPS is the ability to identify the elements (Li and above). The relationship between the measured binding energies of the orbitals and atomic number is shown in Fig. 3.6 [11]. This figure shows a clear trend that for a given orbital there is an increase in the electron’s binding energy. This is expected as the addition of electrons increases the separation between the core level binding energy and the Fermi Level. Initially, the change in binding energy with atomic number is quite large, as is shown with the 1 s level by Fig. 3.6. With Al K_α (hυ = 1,486.6 eV) the limit of the source’s ability to excite 1 s electrons is reached at Mg (binding energy = 1,303 eV). As we go to the upper shells, the rate of increase of the binding energies slows. With typical XPS systems having either Mg K_α (hυ = 1,253.6 eV) or Al K_α sources, there will be a limit in terms of the orbitals the which can be excited. As can be seen in the figure, when the limit to excite a given orbital is reached, there is a higher level orbital available to excite within the range of available excitation energies.

Fig. 3.6

The relationship between orbital binding energies and atomic number [11]

Focusing now upon the binding energies of the first row transition metals, as shown in Table 3.1 [11], the magnitude of the increase in binding energy with atomic number is much larger for the 2p_3/2 orbital as compared to the 3p. Due the different rates at which the binding energies change with atomic number, the differences in the energy of the core level orbitals increase as well. It is because of this that we are able to extract elementally characteristic information from Auger electron spectroscopy.

Table 3.1 First-row transition metal binding energies

An example of elemental analysis by XPS of first-row transition metal nitrides is shown by the survey spectra in Fig. 3.7 [12–15]. We can clearly see the distinct shifts in the binding energies of the metal 2p and 2s lines from ScN to CrN. Less apparent on this energy scale are the shifts of the metal 3p and 3s line. The survey spectra also show a shift of the binding energies of the metal Auger lines as would be expected according to Table 3.1. Generally, survey spectra are collected over a broad range of binding energies in order increase the likelihood of collecting a characteristic signal from all of the elements present in a material and at a low energy resolution in order to maximize the electron signal, thus minimizing data collection time. If one chooses an energy range which is wide enough, it should be possible to detect all elements (except hydrogen and helium). As is the case with most elements, the first-row transition metal nitride survey spectra have multiple lines present for each element.

Fig. 3.7

XPS survey spectra of first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ [12–15]

While it is clear that the metal lines shift between the various nitrides, the N 1s line appears to be stationary in the survey spectra. If spectra are collected at both higher energy resolution and higher data density, one can look for small shifts in energy indicating changes in the chemical/electronic environment of the material. High resolution N 1s spectra of the first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ are shown in Fig. 3.8 and the positions of the N 1s lines are summarized in Table 3.2 [12–15]. When data are collected at high resolution, differences in the position of the N 1s line can be observed. The N 1s line for TiN(001) appears at the highest binding energy, 397.3 eV, while the N 1s line for ScN(001) appears at the lowest, 396.1 eV, with VN(001), and CrN(001) appearing in between. The position of the N 1s is influenced by the degree of electron sharing by the nitrogen and the metal. Core level orbitals shift to higher values during photoelectron emission to compensate for a loss in electron screening caused by the chemical bond while they shift to lower values to compensate for an excess in electron screening. The higher binding energy of the N 1s in TiN, as compared to the other first-row transition metal nitrides, suggests electron density of the chemical bond is not localized around the N atom, but rather is shared between the Ti and N (covalent bond). The lower binding energy of the N 1s of ScN suggests that in this case, the electron density of the chemical bond is more localized around the N atom indicating a larger degree of charge separation (ionic bond).

Fig. 3.8

XPS N 1s spectra of first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ [12–15]

Table 3.2 N1s binding energy of first-row transition metal nitrides ScN, TiN, VN, and CrN

Binding energy shifts due to chemical effects can also be explained by differences in the electronegativities of the elements participating in the chemical bond. If an element with a higher electronegativity is bonded to an element with a lower electronegativity, there will be a shift in electron density at the valence level towards the more electronegative element. This results in a shift in of the core level orbitals to higher binding energies to account for the loss of screening provided by the valence electrons. Table 3.3 shows the position of the C 1s line for various functional groups [11]. As carbon is bonded to nitrogen, being slightly more electronegative that carbon, the C 1s line is shifted 1 eV. Oxygen and chlorine both shift the C 1s line 1.5 eV and fluorine shifts the line 2.8 eV. A carbon-oxygen double bond shifts the C 1s line equal to that of two oxygen bonds.

Table 3.3 C 1s binding energies for various functional groups

An example of how chemical shifts influence photoelectron spectra is shown in Fig. 3.9 with XPS survey, and high resolution O 1s and C 1s spectra of poly-methylmethacrylate (PMMA) (The binding energy scales of the high resolution spectra were referenced by setting the position of the type 1 carbon (sp³ backbone and pendant C–CH₃ groups) to 285.0 eV). The survey spectrum indicates the presence of oxygen and carbon. With the high-resolution spectra, the different chemical environments for the oxygen and carbon of the function groups present in PMMA can be observed. Through the use of a modeling technique known as curve, or peak fitting, the binding energies and relative amounts of the various functional groups present may be ascertained. With curve fitting, a set of synthetic curves representing a specific chemical/electronic environment for that element are assembled so that the sum of the intensities of the curves matches that of the experimentally collected spectrum. Commercially available curve-fitting software packages provide the ability to create synthetic peaks and backgrounds having a variety of theoretical shapes.

Fig. 3.9

Chemical structure and X-ray photoelectron spectra of poly-methylmethacrylate (PMMA)

The ester group of poly-methylmethacrylate exists as a dipolar resonance structure as shown in Fig. 3.9, where on average, type 1 oxygen exists with a partial negative charge and type 2 oxygen exists with a partial positive charge, resulting in two distinct chemical environments. The two charge states of oxygen are reflected in the high resolution XPS spectrum by the two lines at 532.1 and 533.6 eV. Type 1 oxygen, having more surrounding electron density than type 2 oxygen, shows a lower binding energy indicative of an excess in electronic screening as compared to the type 2 oxygen.

Carbon exists in four different chemical environments, which is indicated by the four C 1s lines at 285, 285.7, 286.8, and 289.1 eV. The position of the type 2 carbon (sp³ backbone C–CH₃ group) being slightly higher in binding energy than the type 1 carbon, reflects the electron withdrawing nature of the type 4 carbon (sp² –C = O(O)- ester group). The type 3 carbon (sp³ O-CH3 group) shows a binding energy slightly higher than a typical alcohol groups (as shown in Table 3.3), a result of the partial positive charge on the type 2 oxygen having electron withdrawing character.

3.8 Spin-Orbit Splitting

Another factor which influences the structure of photoelectron lines results from the coupling of the spin of the electron (s = ± ½) with an orbital’s angular momentum (l = 0, 1, 2, 3… for s, p, d, f orbitals). This coupling is known as spin-orbit or l-s splitting . For all orbitals having a non-zero orbital angular momentum, the orbital binding energies have a different value for each of the two spin-split states. Photoelectron spectra for s, p, d, and f orbitals are shown in Fig. 3.10. These two states are identified by the total angular momentum quantum number j, where j = |l ± s|. The population of each line is defined by the secondary total angular momentum quantum number, m _j , which varies by integer steps from –j to j generating 2j + 1 different values of m _j . Using the p orbital as an example, j has values of 3/2 and 1/2 which results in 4 different states for j = 3/2 and two different states for j = 1/2 giving a 3/2-to-1/2 area ratio of 2-to-1. The d and f orbitals will therefore have area ratios for the two spin-split states of 3-to-2 and 4-to-3, respectively.

Fig. 3.10

Examples of core level photoelectron spectra for the s, p, d, and f orbitals

3.9 Energy Loss and Final-State Structures

In many cases, chemical information may be extracted from changes in spectral line shapes. The changes in spectral line shape can be related to electron interactions within the solid and to the resulting electronic configuration of an ion following photoelectron emission.

When electrons scatter inelastically within a solid, energy is transferred from the exiting electron to another electron(s). In one case, transfer of energy can result in individual electrons being excited into discrete or continuum states above the Fermi Level. An example of excitation into continuum states is given by the peak asymmetry often observed in the photoelectron spectra from metallic solids, see Fig. 3.10 where the asymmetry is most apparent in the C 1s spectrum of graphite. The degree of peak asymmetry is proportional to the density-of-states (DOS) near the Fermi Level.

In another example of energy transfer, an individual electron can transfer energy to several electrons. Such an example is the excitation of a plasmon , a collective oscillation of electrons in the conduction band, known as a plasmon-loss . The shape of the plasmon-loss feature is directly related to the electron structure of the solid. Examples of plasmon-loss structures for silicon with a native oxide and SiO₂ are shown in Fig. 3.11. Silicon, being semi-metallic exhibits very sharp plasmon-loss features with a smaller separation (hω₁) as compared to the broader features of SiO₂, which is a wide bandgap insulator.

Fig. 3.11

A comparison of plasmon-loss structures for silicon with a native oxide and SiO₂

The following example describes peak structures which occur as a result of the electronic configuration of the atom following photoionization. The photoelectron’s binding energy is based on the final-state configuration of the atom undergoing photoionization. The most basic way to describe photoionization is given by the Sudden Approximation which assumes the orbitals not involved in the photoionization process remain at the same energies in the final-state as they were in the initial-state (frozen orbital approximation). Under this approximation, the XPS energies are the negative Hartree-Fock orbital energies (Koopman’s binding energy ), E_B,K ≈ -e_B,K. Actual binding energies will represent the readjustment of the N-1 charges to minimize energy (relaxation), E _B  = E _f ^N-1 – E _i ^N.

With certain inorganic compounds and metal-adsorbate systems, ionization of the core subshell results in a strong coulombic perturbation responsible for producing localized states. Final-state screening occurs when electron density is transferred to localized states that screen the core hole. Such a case produces a screened final-state where polarization of electron density is towards the core-ionized atom. In certain cases an unscreened final-state is produced when polarization of electron density is away from the core-ionized atom. An example of final-state structures is shown in Fig. 3.12 [16–21].

Fig. 3.12

Ti 2p photoelectron spectra of Si₃N₄ deposited on TiN(001) at substrate biases of -7 V (floating), -150, and 250 V [16–21]. Surface science spectra by American Vacuum Society; American Institute of Physics. Reproduced with permission of American Institute of Physics (A P I) in the format reprint in a book/ebook via Copyright Clearance Center

Figure 3.12 shows that when Si₃N₄ is deposited on TiN(001) there is an enhancement of the satellites peaks at ~457 and ~464 eV which increases with substrate bias during growth. For Si₃N₄ overlayers grown with V _b  = -250 V, the Ti satellite peak is nearly as intense as the main core peak. On the nonpolar TiN(001) 1 × 1 surface, each Ti is surrounded by five N atoms, and vice versa. The deposition of Si₃N₄ results in an increased nitrogen concentration around surface and near surface Ti atoms, resulting in negative polarization due to the higher electronegativity of N than Ti. The structures at ~457 and ~464 represent unscreened final-states as they occur at higher binding energies than the screened final-states at ~455 and ~462 eV. The increase in the intensity of unscreened final-states indicates electron density is being directed away from TiN towards the Si₃N₄ interface which increases with substrate bias voltage.

3.10 Ultraviolet Photoelectron Spectroscopy (UPS)

The probability to excite a bound electron is related to the degree of overlap between the initial state wave function of the bound electron to the final state wave function of the free electron (transition matrix element). For valence orbitals with low angular momentum, the excitation probability is very low when an X-ray photon is used as the excitation source. That probability can be increased if a photon source with lower energy is used instead. One such method known as ultraviolet photoelectron spectroscopy (UPS) makes use of ultraviolet light as the excitation source. The process of photoelectron emission is identical to that of XPS with the exception that the technique generally involves only valence orbitals (binding energy < 10 eV) and low binding energy core levels.

A very useful aspect of being able to observe valence level electron is that it offers insight into the electronic structure of the material. Such an example is shown in Fig. 3.13 [12–15]. Here are shown UP spectra of the same series of first-row transition metal nitrides as presented in earlier examples collected with He I (21.2 eV) radiation. As the number of electrons per unit cell increases, from ScN to CrN, the metal 3d bands begin to fill pushing the Fermi level above the minimum in the density-of-states (DOS). The Fermi edge then becomes more dominant and the N 2p bands move to higher binding energy. UP spectra often appear to be rather complex as the hybridized valence states produce spectra containing multiple peaks spanning a range of energies. For this reason, valence band studies are combined with other types of spectroscopies (XPS, optical, etc.…) and band structure calculations [22, 23].

Fig. 3.13

He I ultraviolet photoelectron spectra of first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ [12–15]

3.11 Work Function Measurement

With UPS, it is possible to measure the work function of a material by observing the onset of photoelectron emission from the material, as shown by Fig. 3.14. As discussed earlier, the work function represents the energy barrier that an electron must overcome in order to become a free electron. Electrons excited with energies less than the work function cannot overcome this barrier and will not escape as free electrons. Therefore, by observing the onset of photoelectron emission, one is able to determine the minimum energy of this barrier. The intense feature in Fig. 3.14 at ~16 eV, dominated of secondary electrons represents the first electrons with sufficient energies to overcome the barrier. Placing the Fermi Level at 0 eV, the work function is determined by subtracting energy of the onset of photoelectron emission from the photon energy. The small feature at ~25 eV represents secondary electron emission from the detector. As this emission is independent of the energies of the electrons passing through the analyzer, it can be separated from the rest of the spectrum by applying a negative bias to the sample. In this example, a 9 V battery was used.

Fig. 3.14

Work function measure of polycrystalline Ag using ultraviolet photoelectron spectroscopy

3.12 Quantitative Analysis

As is the case with other materials characterization methods, it is desirable to have the capability to relate the measured signal to the quantity of various species present in the material being studied. With XPS, the collected signal is related to a variety of parameters, one of which is the number of atoms emitting photoelectrons per unit volume.

Assuming a homogeneous sample, our detector count rate can be separated into terms describing the number of electrons emitted per unit volume and the volume of the material analyzed. Some of the terms are sample dependent, and some are instrument dependent.

$$ {A}_i=\left({N}_i{\sigma}_i\left(\gamma \right) JT\left({E}_i\right)\right)\left(a{\lambda}_i\left({E}_i\right) \cos \theta \right) $$

(3.6)

• A _i  = detector count rate

• A _i  = (electrons/volume)(volume)

Sample dependent terms

• where: N = atoms/cm³

• σ(γ) = photoelectric (scattering) cross-section, cm²

• λ(E _i ) = inelastic electron mean-free path, cm

Instrument dependent terms

• J = X-ray flux, photon/cm²-s

• T(E _i ) = analyzer transmission function

• a = analysis area, cm²

• θ = photoelectron emission angle

By assuming the concentration to be a relative ratio of atoms, we can neglect the terms that depend only on the instrument:

$$ {N}_i={A}_i/{\sigma}_iT\left({E}_i\right){\lambda}_i\left({E}_i\right) $$

In this case, differences in photoelectric (scattering) cross-section, σ(γ), can be considered on a relative scale by a term knows as the relative sensitivity factor (RSF) , S_i. These factors are tabulated by the instrument manufacturer and are specific to the particular configuration of the spectrometer, and its mode of operation. As in most cases, the exact chemical and physical nature is unknown prior to the experiment thus it is difficult to accurately determine the inelastic mean-free paths (IMFPs), λ_i, of the electrons being analyzed, so they are usually assumed to be identical for all of the electrons collected. Through calibration, modern acquisition and analysis software can account for the transmission function of the spectrometer. The values of S are determined theoretically or empirically with standards. Here, the number of emitting atoms per unit volume is determined by dividing our signal, the number of electrons emitted (usually by means of the area of the measured photoelectron line) by the relative sensitivity factor. Now the corrected signal can be compared to the total of the corrected signals measured, and a composition is determined. Values determined by this method are referred to as atomic concentrations.

$$ {N}_i={A}_i/{S}_i $$

$$ {C}_i=\frac{A_i/{S}_i}{{\displaystyle {\sum}_i^j}{A}_{i,j}/{S}_{i,j}} $$

(3.7)

Quantitative analysis by XPS of the first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ is shown in Table 3.4 [12–15, 24–27]. The data show compositions of these films to be quite similar to bulk compositions measured by Rutherford backscattering spectroscopy (RBS). Because XPS is an extremely surface sensitive technique, analyzing as little as the top 10 nm of a sample, the composition of the surface is in most cases altered by exposure to air demonstrating an advantage of being able to do in situ growth/analysis. This will be discussed in more detail in Sect. 3.14.

Table 3.4 Quantitative XPS analysis of first-row transition metal nitrides ScN, TiN, VN, and CrN

3.13 Angle-Resolved XPS and Thickness Measurement

If one assumes that the electrons travel only in a straight line and do not change directions prior to exiting the solid, then one can vary the amount of material probed simply by changing its angle relative to the spectrometer. An experiment where XPS data are collected at different emission angles is known as angle-resolved XPS (ARXPS) . If the compound being studied has a uniform composition that does not vary with depth, one would not expect to see differences in the ratios of the characteristic photoelectron signals at different emission angles. However, if there are variations in composition with depth that are on the order of the sampling depth of XPS, one should be able to observe this with an ARXPS study. Si 2p X-ray photoelectron spectra from silicon with a native oxide collected at emission angles of 0 and 75° are shown in Fig. 3.15. At an emission angle of 75°, there is a distinct increase of the signal representing the oxide layer at 103.3 eV as compared to the data collected at an emission angle of 0°. Although it is obvious to expect that the native oxide, which forms due to air exposure, lies on top of the Si, this example demonstrates the ability of an ARXPS to be able to differentiate species that reside at the surface from those in the near-surface region.

Fig. 3.15

Si 2p X-ray photoelectron spectra from silicon with a native oxide collected at emission angles of 0 and 75°

By making use of the Beer-Lambert relationship Eq. 3.5, which describes the photoelectron signal intensity as a function of depth and emission angle, a mathematical model can be developed to estimate the thickness of a thin overlayer. Using a simple system consisting of an overlayer a with thickness d on top of a substrate b, referred to as the two layer model , one can define a ratio I _a /I _b where the signal from layer a comes from depths of 0 to d, and the signal from layer b comes from depths of d to ∞.

$$ \frac{I_a}{I_b}=\frac{I_a^0{\displaystyle {\int}_0^d} \exp \left(-t/\lambda cos\theta \right) dt}{I_b^0{\displaystyle {\int}_d^{\infty }} \exp \left(-t/\lambda cos\theta \right) dt} $$

(3.8)

Let:

$$ {I}_a^0={n}_a{\rho}_a{S}_a/M{W}_a,{I}_b^0={n}_b{\rho}_b{S}_b/M{W}_b $$

$$ {N}_a={I}_a/{S}_a,{N}_b={I}_b/{S}_b $$

Where $\eta$ is the number of atoms per molecule unit, $\rho$ is the molecular density, S is the relative sensitivity factor, MW is the molecular weight, and N is the atomic concentration. Following integration and rearrangement, we arrive at the expression given by Eq. 3.9:

$$ ln\left(\frac{N_a{n}_b{\rho}_b{\lambda}_bM{W}_a}{N_b{n}_a{\rho}_a{\lambda}_aM{W}_b}+1\right)=\frac{d}{\lambda_a \cos \theta } $$

(3.9)

With this equation, the thickness d can be calculated at one emission angle, or if the left-hand side of Eq. 3.9 is plotted as a function of 1/cosθ, the slope is d/λ _a .

The two-layer model was used to determine the thickness of graphene sheets grown epitaxially in multilayer stacks on SiC(0001) that were transferred layer-by-layer to SiO₂/Si substrates [28]. Figure 3.16 shows the high resolution C 1s spectra collected at one emission angle from the as-grown surface and the surface following the first and second transfer. The spectra shows a peak at 284.5 eV indicating the epitaxial graphene layers and a peak 283.7 eV which represents the SiC(0001) substrate. As graphene sheets are removed, the area of the graphene peak decreases while the area for SiC increases. Using the two-layer model, the thickness of the as-grown surface was determined to decrease from 2.1 to 1.53 nm following the first transfer, and 0.68 nm following the second transfer. The consistent change in thickness demonstrated that graphene transfers occurred in a layer-by-layer mode.

Fig. 3.16

C 1s X-ray photoelectron spectra from graphene sheets grown epitaxially in multilayer stacks on SiC(0001). Spectra are shown for the as-grown surface and the surface following two transfers. Reprinted from Unarunotai, S., Koepke, J. C., Tsai, C. –L., Du, F., Chialvo, C. E., Murata, Y., Haasch, R., Petrov, I., Mason, N., Shim, M., Lyding, J., Rogers, J. A.: Layer-by-Layer Transfer of Multiple, Large Area Sheets of Graphene Grown in Multilayer Stacks on a Single SiC Wafer. ACS Nano, 4, 5591 (2010). Copyright 2010 American Chemical Society

An example of the use of angle-resolved XPS to determine overlayer thickness is shown in Fig. 3.17. Here data from the same native oxide on silicon as shown in Fig. 3.15, which also includes data collected at a 45° emission angle. Atomic concentrations for SiO₂ and Si were determined by the areas of each respective peak. In this example, the left-hand side of Eq. 3.9 is plotted as a function of 1/cosθ. Using the slope, d/λ _a , the thickness was calculated to be 1.7 nm.

Fig. 3.17

Two layer model determination of the thickness of silicon with a native oxide

3.14 Ion Sputtering and Depth Profiling

As XPS is an extremely surface sensitive technique, information is only obtained from the near-surface region of the sample. In order to probe deeper, XPS data collection may be combined with bombardment by an energetic ion beam through a process known as ion sputtering . In certain cases, ion sputtering is used to only remove thin contamination or oxide layers. In other cases, a sequence of XPS data are collected following periods of ion sputtering, either continuous or alternating, allowing the investigation of the sample at different depths. Such an experiment is known as a depth profile .

An example of a depth profile of 100 nm SiO₂/Si sample is shown in Fig. 3.18. The top panel shows a cascade plot of the Si 2p region as a function of depth (surface is front). The difference between the oxide signal at 103.3 eV and the substrate Si⁰ signal at 99.3 eV is large enough that the areas each of two chemical states of Si are able to be determined throughout the profile. The atomic concentrations of the Si⁰, SiO₂ and O are shown in the bottom panel.

Fig. 3.18

Si 2p XPS depth and atomic concentration profiles from 100 nm SiO₂/Si

The atomic concentration profile shows an O/Si ratio of about 1.5, which is lower than the expected O/Si ratio of 2. Standard relative sensitivity factors (RSFs) were used in the calculation in order to demonstrate a drawback of ion sputtering, where one species is removed at a faster rate as compared to another. This effect referred to as preferential sputtering usually results in the lighter atoms being removed from the material by the ion beam as compared to the heavier ones. When this occurs, the composition of the surface exposed by the ion beam is altered until a composition is reached where all of the atoms present in the material are removed at an equal rate. Such a layer is referred to as an altered layer . Provided that the relative composition of the original film does not vary as a function of depth, once the altered layer is reached, the relative ratios of the elements should then remain constant throughout the profile. When this is the case and it is possible to determine the film composition by another technique, then the RSFs may be adjusted to give the correct ratio of elements in the atomic concentration profile.

A second example of preferential sputtering is shown by the XPS quantitative analysis data of the first-row transition metal grown and analyzed in situ (Table 3.4). In this case, samples were sputtered under conditions and times typical for sample cleaning in order to demonstrate the extent of preferential sputter with these materials (5 min., 3 KeV, 0.0043 mA/cm²). For all four of the nitrides examined, the N/metal ratio decreased following ion bombardment and in particular, the N/V ratio for VN decreased from 1.02 to 0.46.

In addition to preferential sputtering, ion bombardment can also alter the electronic and physical structure of the material being sputtered, an example of this is shown in Fig. 3.19 [13, 25]. The spectra show the intensity at ~463 and ~457 eV related to an unscreened final-state diminishes following ion bombardment. It was shown earlier that the N/Ti ratio decreased from 1.00 to 0.73 following ion bombardment. When the N/Ti ratio drops below 1, Ti is reduced in order to maintain charge balance by the addition of 3d electrons. Porte [29] studied a series of TiN films of varying stoichiometry and found that for TiN_x with x < 0.8, the increased concentration of electrons near the Fermi Level enhances the screening process of the Ti 2p core hole, hence diminishing the relative intensity of the intensity of the unscreened final-state peaks.

Fig. 3.19

The Ti 2p X-ray photoelectron spectra of as deposited and Ar⁺ sputtered TiN(001) [13, 25]

A similar effect is shown in Fig. 3.20 with the UPS valence band spectra for ScN(001) from as-deposited and Ar⁺ ion bombarded surfaces [12, 24] The surface spectrum from as-deposited ScN(001), shows a diminishing density-of-states (DOS) at the Fermi Level. ScN was shown by Gall, et al., to be a semiconductor with an indirect Γ-X bandgap of 1.3 ± 0.3 eV and a direct X-point gap of 2.4 ± 0.3 eV [22]. Following ion bombardment, there is a clear increase in Sc 3d electron density near the Fermi Level of the ScN with the spectrum even showing the development of a visible Fermi Edge suggesting the material is becoming more metallic. Reduction of Sc occurs to maintain charge balance in the material following the preferential removal of N caused by the sputtering process.

Fig. 3.20

The ultraviolet photoelectron spectra of as deposited and Ar⁺ sputtered ScN(001) [12, 24]

3.15 XPS Imaging and Spectroscopy

If so equipped, modern X-ray photoelectron spectrometers are capable of collecting 2-dimensional photoelectron images. This can either be done by collecting a scanned image using a rastered X-ray beam, microprobe imaging , or by collecting a 2-dimensional image using a stationary X-ray beam, parallel imaging . Photoelectron images are collected by fixing the energy of the analyzer to a core level peak and counting only those electrons. In both cases, microprobe and parallel imaging mode, photoelectron spectra may be collected from selected areas within the image for site-specific characterization of the surface.

An example of the combination of XPS imaging and spectroscopy is shown in Fig. 3.21. This example shows an intermediate step in the fabrication process of photoluminescent patterned porous Si pixel arrays [30]. A pattern of 150 μm pixels was first transferred to p-type Si(100) by micro-contact printing monolayers of octadecyltrichlorosilane (OTS), the pixels consisting of areas of the Si not covered by the OTS monolayer. Next, a Pt⁰-divinyltetramethyldisiloxane Si etching catalyst was selectively deposited inside the 150 μm circles. XPS imaging combined with spectroscopy was used to confirm the selected deposition of the Pt catalyst. A C 1s photoelectron image representing the OTS covered area and a Pt 4f photoelectron image representing the catalyst complex were collected. The image clearly shows enhanced Pt 4f intensity within the pixels. Once the location of the pixels has been confirmed by the images, photoelectron spectra from within a pixel and the OTS coated area can be collected. The spectrum from within the pixel shows a nearly tenfold greater Pt content as compared to the OTS coated area to further confirm the selective deposition process.

Fig. 3.21

Superimposed C 1s and Pt 4f X-ray photoelectron images (top) and the Pt 4f high resolution X-ray photoelectron spectra corresponding to spots a and b (bottom). Top image reprinted from Harada, Y., Li, X., Bohn, P. W., Nuzzo, R. G.: Catalytic Amplification of the Soft Lithographic Patterning of Si. Nonelectrochemical Orthogonal Fabrication of Photoluminescent Porous Si Pixel Arrays. J. Am. Chem., Soc., 123, 8709 (2001). Copyright 2001 American Chemical Society

3.16 Auger Electron Spectroscopy (AES)

The electron emission process from which Auger electron spectroscopy (AES) is based, was first discovered in 1922 by Lise Meitner [31] and later independently discovered in 1925 by Pierre Auger [32]. In 1953, J. J. Lander reported characteristic Auger energies from a variety of materials excited by low energy electrons [33]. Lander was first to point out that “excitation of Auger peaks by a beam of low velocity electrons provides an interesting technique for surface analysis.” It was not until the work of Larry Harris at General Electric in the mid 1960s that Auger electron spectroscopy became a practical analytical technique [34, 35].

3.17 Instrumentation

As is the case with photoelectron spectrometers, Auger systems are available in varying configurations. A schematic of a scanning Auger microprobe is shown in Fig. 3.22. This system is based on a concentric single-pass cylindrical mirror analyzer (CMA) [36] and scanning electron gun design. This configuration offers the advantage of both the analyzer and electron gun having the same direct line-of-sight to the specimen, avoiding the difficulties arising from shadowing. Because of space limitations, this advantage comes at the expense of the limited energy resolution offered by the single-pass CMA as compared to higher resolution designs such as hemispherical analyzers used in X-ray photoelectron spectrometers.

Fig. 3.22

Schematic of a scanning Auger microprobe

With the system shown by Fig. 3.22, an electron beam is generated by either a LaB₆ emission source in older systems, or a Schottky field emission source in modern systems. A condenser lens focuses the electron beam to an objective aperture, controlling both the spot size and current density of the electron beam. Most systems are equipped with a number of different sized apertures allowing the system to be operated through a range of spot sizes and current densities. Next, the electron beam passes through an objective lens where it is focused into a small probe. Scanning of the beam is accomplished by deflector plates at the exit of the electron gun.

In this design, the sample is positioned in a field-free region of the analyzer where the electrons to be analyzed are excited. Electrons are emitted from the sample in all directions and only those with trajectories allowed by the geometry of the system are collected. In most applications of this design, retarding fields are not employed, resulting in the electrons traveling through the analyzer at the energies which they were emitted. This results in the z-dependence of the sample position with the kinetic energies measured by the double-focusing CMA analyzer. Positioning of the sample is accomplished by setting the energy of the analyzer to scan a small range of energies around the primary beam energy, and then translating the sample along the z-axis of the analyzer until the energy of the elastically scattered primary electron’s energy matches that of the incident energy. This process is referred to as elastic peak or z-axis alignment .

The energies of the electrons being collected are filtered by applying different potentials to the inner and outer cylinders. The potentials of the cylinders focus the electrons to a plane where a slit aperture is located. The slit aperture controls the energy resolution of the analyzer and in the case where retardation fields are not employed, resolutions follow an E/ΔE relationship. Beyond the analyzer resolution adjustment aperture is the electron detector. Modern Auger systems make use of multiple detectors in order to increase the electron counting efficiency of the instrument in a similar fashion to those found in XPS systems.

In the case of scanning Auger spectrometers, a secondary electron detector is placed close to the sample in order to collect the electrons used to generate secondary electron images. These detectors are of identical design to those found in scanning electron microscopes (SEMs).

3.18 Elemental Shifts

As described earlier, the kinetic energies of Auger lines depend on the differences in the energies of the orbitals involved in the Auger transition process. Because the number of electrons involved in the Auger emission process is three, an element must have three or more electrons in order for there to be an Auger transition. Therefore, the lightest element that can be detected by Auger spectroscopy is Li. As the orbitals fill with increasing atomic number, binding energies of the core and valence level electrons increase together with the energy separation of these orbitals. Table 3.1 shows the magnitude of the increase in orbital separation for the first-row transition metals.

The relationship between the kinetic energy of Auger transitions and atomic number is shown in Fig. 3.23. The figure demonstrates the trend of increasing kinetic energy with atomic number. As was the case with XPS, accessible Auger transitions are limited by the energy of the primary electron beam. Although primary beam energies of up to 20 keV are available for scanning Auger systems, data from Auger transitions with kinetic energies <2,500 eV are collected for the most part.

Fig. 3.23

The relationship between the kinetic energy of Auger transitions and atomic number. (Reproduced by permission of Physical Electronics USA)

One aspect of Auger electron spectroscopy is immediately apparent in Fig. 3.23, as the atomic number increases, the number of possible Auger transitions increases (each dot in Fig. 3.23 represents an individual Auger transition). This complexity is related to the fact that an Auger transition is a process involving three electrons. As the atomic number increases, the number of orbitals satisfying the selection rules for a given transition increases. As would be expected, the complex nature of Auger transitions results in the possibility of overlapping lines in spectra collected from materials made up of several elements.

Examples of Auger survey spectra of first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ are shown in Fig. 3.24 [37–40]. Modern Auger spectrometers are able to collect Auger spectra as counts as a function of kinetic energy, Fig. 3.24a. This mode is referred to as N(E) mode . Auger N(E) data appear as small peaks superimposed on an intense and fairly linear background. The positions for the Auger lines in N(E) data are reported as the energy the peak maximum. Auger spectra displayed as N(E) data share the appearance of the X-ray induced Auger transitions observed in XP spectra, allowing for straight-forward comparison. To enhance the prominence of the Auger data, it is typical to present the spectra as a first derivative of counts as a function of kinetic energy, Fig. 3.24b. This mode is referred to as dN(E) mode . The positions for the Auger lines in dN(E) data are reported as the energy of the peak minimum.

Fig. 3.24

Auger survey spectra of first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ (a) N(E) and (b) dN(E) (′ = M_2,3M₄M₄, ″ = M₁M₄M₄, ′″ = L₃M_2,3M_2,3, ″″ = L₃M_2,3M_4,5 ′″″ = L₃M_4,5M_4,5) [37–40]

It is apparent in the Auger spectra presented in Fig. 3.24 that Auger transitions span a wide range of kinetic energies and are fairly complex in structure. Also apparent in the spectra is the issue of peak overlap. This overlap is especially significant in the spectra of ScN and TiN where the N KLL and directly overlap a major metal LMM line.

3.19 Chemical State/Matrix Effects

If one or both of the upper level orbitals that are involved in the Auger transition lie in the valence band, chemical/electronic information contained within the valence band will be reflected in the Auger spectrum [41]. Other chemical/electronic effects which influence the Auger line shape are closely lying energy loss peaks such as plasmons as described earlier. An example the effect of the chemical/electronic environment on Auger spectra is shown in Fig. 3.24. Here the C KLL Auger spectra of highly-ordered pyrolytic graphite (HOPG) and SiC imbedded in Cu are shown (Fig. 3.25). The most notable of the differences in the two spectra are reflected in the Auger pre-peak region making each spectrum distinct in appearance and suggest the possibility of being able to identify different chemical states from Auger spectra. However, the density-of-states (DOS) in the valence band can be quite complex, as was shown to be the case in Sect. 3.10, which means that extracting information regarding the chemical/electronic environment of a material will not be as straightforward as with XPS.

Fig. 3.25

C KLL Auger spectra of highly-ordered pyrolytic graphite (HOPG) and SiC imbedded in Cu

3.20 Quantitative Analysis

As was the case with X-ray photoelectron spectra, the collected signal is related to a variety of parameters, one of which is the number of atoms emitting Auger electrons per unit volume. Once again, assuming a homogeneous sample, our detector count rate can be separated into terms describing the number of electrons emitted per unit volume and the volume of the material analyzed.

$$ {A}_i=\left({N}_i{\sigma}_i\left(\gamma \right){\chi}_i\left(1+r\right) JT\left({E}_i\right)\right)\left(a{\lambda}_i\left({E}_i\right) \cos \theta \right) $$

(3.10)

• A_i = detector count rate

• A_i = (electrons/volume)(volume)

Sample dependent terms

• where: N = atoms/cm³

• σ(γ) = ionization (scattering) cross-section, cm²

• χ_i = Auger transition probability

• r = secondary ionization coefficient

• λ(E_i) = inelastic electron mean-free path, cm

Instrument dependent terms

• J = Electron flux, electron/cm²-s

• T(E_i) = analyzer transmission function

• a = analysis area, cm²

• θ = Auger electron emission angle

By assuming the concentration to be a relative ratio of atoms, we can neglect the terms that depend only on the instrument:

$$ {\mathrm{N}}_{\mathrm{i}}=\mathrm{Ai}/{\upsigma}_{\mathrm{i}}{\upchi}_{\mathrm{i}}\left(1+\mathrm{r}\right)\mathrm{T}\left({\mathrm{E}}_{\mathrm{i}}\right){\uplambda}_{\mathrm{i}}\left({\mathrm{E}}_{\mathrm{i}}\right) $$

The form of Eq. 3.10 is nearly identical to that of Eq. 3.5 with the exception of two additional terms, χ_i the Auger transition probability and r the secondary ionization, or backscatter coefficient. For simplicity, it is typical to combine σ and χ into a single term that can be considered on a relative scale by a term knows as the relative sensitivity factor (RSF) , S_i. As is the case with XPS, these factors are also tabulated by the instrument manufacturer and are specific to the particular configuration of the spectrometer, and its mode of operation. Likewise, in most cases the exact chemical and physical nature is unknown prior to the experiment, it is difficult to accurately determine the inelastic mean-free paths (IMFPs) λ_i, of the electrons being analyzed and secondary ionization coefficient r so they are usually assumed to be identical for all the electrons collected. Through calibration, modern acquisition and analysis software can account for the transmission function of the spectrometer. The values of S are determined theoretically or empirically with standards. Here, the number of emitting atoms per unit volume is determined by dividing our signal, the number of electrons emitted (usually by means of the peak-to-peak height of the measured Auger electron line) by the relative sensitivity factor. Now the corrected signal can be compared to the total of the corrected signals measured, and a composition is determined. Values determined by this method are referred to as atomic concentrations.

$$ {N}_i={A}_i/{S}_i $$

$$ {C}_i=\frac{A_i/{S}_i}{{\displaystyle {\sum}_i^j}{A}_{i,j}/{S}_i} $$

(3.11)

Quantitative analysis by Auger electron spectroscopy of the first-row transition metal nitrides ScN(001), TiN(001), VN(001), and CrN(001) grown and analyzed in situ is shown in Table 3.5 [37–40]. Significant overlap of the N KL_2,3L_2,3 with the Sc L₃M_4,5M_4,5 and the Ti L₃M_2,3M_2,3 make accurate quantitative analysis of ScN and TiN extremely challenging (see Fig. 3.24). The ratios of I_γ/I_β for ScN and TiN reflect this overlap as the intensity of the N KL_2,3L_2,3 is actually a combination of both N and metal intensity. Another difficulty found with Auger spectroscopy is the fact the peak shapes and intensities can vary from one chemical environment to another as discussed in the previous section which also complicates quantitative analysis. As can be seen in Table 3.5, the ratios of I_γ/I_α and I_γ/I_β for VN and CrN are significantly higher than the expected N/Me values of 1.00. This discrepancy is due to changes in the relative sensitivity factors which arise due to chemical state induced changes in peak shape (to demonstrate this point, standard manufacturer relative sensitivity factors were used). When it is possible to use a second analytical technique for quantitation, such as RBS, the Auger relative sensitivity factors may be adjusted. As was the case with XPS, the N/Me ratios decrease following ion bombardment, although in this case because of the peak overlap, it is not clear for ScN and TiN which species is preferentially removed. Despite these difficulties in quantitative analysis, Auger spectroscopy can still be used to make qualitative comparisons of atomic concentrations.

Table 3.5 Quantitative Auger analysis of first-row transition metal nitrides ScN, TiN, VN, and CrN

3.21 Depth Profiling

Like XPS, Auger electron spectroscopy is an extremely surface sensitive technique, information is only obtained from the near-surface region of the sample. In order to probe deeper, Auger data collection may be combined with bombardment by an energetic ion beam through a process known as ion sputtering . In certain cases, ion sputtering is used to only remove thin contamination or oxide layers. In other cases, a sequence of Auger data are collected following periods of ion sputtering, either continuous or alternating, allowing the investigation of the sample at different depths. Such an experiment is known as a depth profile .

An example of an Auger depth profile of a multilayer HfAlSiN thin film is shown in Fig. 3.26. In this example multiple layers of HfAlSiN are grown by reactive magnetron sputtering under conditions of varying ion flux to the sample in order to understand the growth mechanism of these films. By growing the film as a series of multiple layers of known thicknesses, comparison of the compositional analysis of the Auger profile can be made with microstructure studies by cross-section transmission electron microscopy (TEM).

Fig. 3.26

Auger depth profile of a multilayer HfAlSiN thin film

3.22 Imaging and Spectroscopy

A useful capability of scanning Auger systems is the ability to direct the primary electron beam to specific areas of a sample. Once an area of interest is identified in an SEM image, the electron beam can be directed to that area to collect spectra only from that area. An example of SEM imaging combined with an Auger spectroscopic study of a chromium carbon nickel alloy is shown in Fig. 3.27. The secondary electron image of the alloy shows image contrast suggesting the presence of multiple phases. Site specific Auger survey collected from a light region, Fig. 3.27a, and a dark region of this alloy, Fig. 3.27b, shows elemental segregation in this material. The light region consists of a nickel rich phase while the dark region consists of a chromium rich phase. This experimental approach is similar to the combination of SEM imaging with EDS (energy-dispersed spectroscopy). Auger however, offers the advantages of being more sensitive to light elements like carbon, having higher energy resolution which allows more accurate elemental identification and quantification, and being surface sensitive.

Fig. 3.27

Auger SEM image and survey spectra collected from two regions in chromium carbon nickel alloy

3.23 Imaging and Auger Mapping

Secondary electron imaging can also be combined with Auger elemental mapping in the characterization of a specimen. In this approach, areas of interest identified by SEM imaging and selected area spectroscopy may be further characterized using the Auger energy analyzer to collect 2-dimensional images (or maps) from specific Auger transitions. The combination of SEM imaging and Auger elemental mapping of a microelectronic device is shown in Fig. 3.28. Such a combination of data collection methods is helpful in device process development and characterization as well as defect review. Like the example of the combination of SEM imaging with Auger spectroscopy shown in the previous example, this experimental approach is similar to the combination of SEM imaging with EDS. Auger mapping does offer the advantage of being more sensitive to the light elements and being more sensitive to species that reside only at the surface.

Fig. 3.28

Auger SEM image and elemental maps of Al and Si

3.24 Summary

X-ray photoelectron spectroscopy is a non-destructive, semi-quantitative method capable of providing elemental and chemical-state information from elements Li and above. It is capable of detecting elements down to 0.1–1 atomic percent from depths of 0.5–10 nm. In addition to elemental and chemical-state information, XPS can also provide useful information about the electronic structure of a material through observation of energy loss and final-state structures. Certain instruments are capable of producing 2-dimensional photoelectron images.

Ultraviolet photoelectron spectroscopy is capable of providing chemical state and electronic structure information from materials. However, due to the complex nature of the density-of-states (DOS) in the valence band, it is more difficult to extract this information, as compared to XPS, usually requiring band-structure calculations and other spectroscopies. By observation of the onset of photoelectron emission, work function measurements may be made using UPS. Like XPS, UPS is non-destructive. However, UPS cannot typically provide quantitative information.

Auger electron spectroscopy is a non-destructive, semi-quantitative method capable of providing elemental and some chemical-state information from elements Li and above. It has detection limits and escape depths similar to those of XPS. With the use of a scanning electron beam as the probe, useful combinations of Auger spectroscopy, SEM imaging and Auger mapping can be made, extending the range of information which can be extracted from the materials analyzed.