XPS

X-ray photoelectron spectroscopy is a surface analysis technique that identifies the elemental composition and chemical bonding states of the outermost few nanometers of a solid surface by measuring the kinetic energy of photoelectrons emitted under X-ray irradiation. When a monochromatic X-ray photon of known energy $h\nu$ strikes a surface atom, it can eject a core-level electron whose binding energy $E_B$ is characteristic of both the element and its chemical environment. The instrument measures the kinetic energy $E_K$ of the emitted photoelectron, and the binding energy is obtained from the photoelectric relation, $$ E_B = h\nu - E_K - \phi_{\text{spec}}, $$ where $\phi_{\text{spec}}$ is the work function of the spectrometer. Because each element has a unique set of core-level binding energies and because chemical bonding shifts those energies by measurable amounts, XPS provides both elemental identification and chemical-state information from the same measurement. The technique is surface-sensitive because photoelectrons generated deeper than a few inelastic mean free paths lose energy through scattering before reaching the detector, so the signal is dominated by atoms within roughly 1 to 10 nm of the surface depending on the material, the photoelectron kinetic energy, and the emission angle. In semiconductor manufacturing, XPS is used to characterize native oxides, interfacial layers, thin-film composition, plasma-etch residues, cleaning effectiveness, and the chemical state of gate dielectrics, barrier metals, and interconnect surfaces — measurements that inform process development, troubleshooting, and integration qualification. **The survey spectrum is commonly the first XPS acquisition and serves as an inventory of detectable elements, while chemical-state analysis generally requires higher-resolution scans of selected core levels.** The survey range and pass energy depend on the source, analyzer, and elements of interest; a broad, fast scan trades energy resolution for coverage and signal. Peak areas corrected with suitable relative sensitivity factors can estimate atomic fractions in a homogeneous sampling volume, but overlapping peaks, matrix effects, background choice, instrument transmission, and contamination limit accuracy. Conventional XPS also does not provide routine elemental detection of hydrogen or helium. **High-resolution core-level spectra are where XPS provides its most distinctive information — the ability to distinguish different chemical environments of the same element through binding-energy shifts caused by differences in oxidation state, bonding coordination, and nearest-neighbor electronegativity.** When a silicon atom is bonded to four oxygen atoms in stoichiometric SiO₂, its Si 2p binding energy shifts approximately 3.5 to 4 eV higher than elemental silicon because the oxygen neighbors withdraw electron density from the silicon core, increasing the energy required to remove a core electron. Intermediate oxidation states — Si₂O, SiO, Si₂O₃ — produce shifts between elemental and fully oxidized silicon, and the spectrum of a real SiO₂/Si interface contains a superposition of these contributions. Decomposing a measured envelope into its constituent components (peak fitting or curve fitting) requires assumptions about the line shape, width, and number of components, and different fitting choices can produce materially different chemical-state assignments from the same raw data. Peak fitting is therefore a modeling step, not a measurement, and results should be reported with the fitting parameters and constraints used. **Quantification in XPS converts peak areas into elemental or chemical-state concentrations, but the accuracy of the result depends on the sensitivity factors used, the background-subtraction method, the homogeneity of the sampling volume, and whether the sample geometry matches the assumptions of the quantification model.** The measured intensity $I_i$ from element $i$ in a homogeneous flat sample can be written as $$ I_i \propto n_i \sigma_i L_i T(E_K) \cos\theta, $$ where $n_i$ is atomic density, $\sigma_i$ is the photoionization cross-section, $L_i$ is the effective attenuation length appropriate to the material and geometry, $T(E_K)$ is analyzer transmission, and $\theta$ is measured from the surface normal. The omitted proportionality factors include photon flux, analyzed area, solid angle, and acquisition conditions. Relative sensitivity factors combine several terms so that a common homogeneous-sample estimate is $x_i = (I_i / S_i) / \sum_j (I_j / S_j)$. This model can misrepresent layered, rough, or laterally inhomogeneous structures, and quantitative work must keep acquisition and data-processing conventions consistent with those used to establish the sensitivity factors. **Angle-resolved XPS varies emission angle to change effective sampling depth without sputtering, providing a non-destructive probe of near-surface depth distributions.** In a simple straight-line attenuation model, increasing $\theta$ from the surface normal reduces the characteristic normal sampling depth approximately with $L\cos\theta$. A thin overlayer changes the angle-dependent overlayer-to-substrate intensity ratio, allowing thickness estimation when the layer model and attenuation data are appropriate. Elastic scattering, roughness, intermixing, shadowing, and lateral inhomogeneity complicate this approximation, so fitted thickness is model-dependent rather than a direct geometric measurement. **Sputter depth profiling combines XPS with ion-beam erosion to extend composition and chemical-state measurements beyond the native sampling depth, but the sputtering process itself can alter the chemistry and structure of the surface being analyzed.** An ion beam — typically argon, but cluster ions such as argon clusters or C₆₀ are used for organic and polymeric materials — removes material layer by layer, and after each sputter interval the XPS spectrum is acquired from the freshly exposed surface. The resulting depth profile shows composition versus sputter time (or estimated depth if the sputter rate is calibrated). Artifacts inherent to sputter profiling include preferential sputtering (one element is removed faster than another, distorting the measured composition), ion-beam-induced mixing (knock-on of surface atoms into deeper layers, broadening interfaces), chemical changes (reduction of oxides, implantation of the sputter species), and roughening of the crater floor. These artifacts mean that a sputter depth profile is not a simple cross-section of the original composition — it is a convolution of the original structure with the damage function of the ion beam. Cluster-ion sources reduce some of these artifacts for certain materials but do not eliminate them universally. **In semiconductor process control, XPS measurements are used to verify interfacial chemical states, thin-film composition, contamination, and the effectiveness of cleaning or surface preparation steps, but translating an XPS result into a process decision requires understanding what the measurement does and does not represent.** A native oxide thickness measured by XPS at a single emission angle is a model-dependent number that assumes a sharp interface and uniform density; the physical film may be graded or rough. A carbon concentration at the surface may include adventitious contamination from air exposure between process and measurement, not only process-generated carbon. The distinction between surface-adventitious and process-relevant signals requires control experiments or in-situ measurement. A copper surface analyzed after a cleaning step may show Cu 2p peaks with satellite features indicating Cu(OH)₂ or CuO, but the chemical state measured in the spectrometer may differ from the state that existed at the process step if the sample was exposed to air, moisture, or different temperature between the process chamber and the XPS instrument. Transfer conditions and their potential effect on the measured chemistry should be reported alongside the spectral results. | Measurement mode | Information provided | Depth range | Destructive | Key limitation | |---|---|---|---|---| | Survey spectrum | Elemental inventory, rough composition | Top 1-10 nm | No | Cannot resolve chemical states | | High-resolution scan | Chemical-state assignment, relative composition | Top 1-10 nm | No | Peak fitting is model-dependent | | Angle-resolved XPS | Non-destructive depth distribution of states | Within escape depth | No | Assumes flat, uniform layers | | Sputter depth profile | Composition vs. depth beyond escape depth | Tens to hundreds of nm | Yes | Sputtering artifacts alter chemistry | | Small-spot / imaging XPS | Lateral composition mapping | Top 1-10 nm per analysis point | No | Spatial resolution and sensitivity are instrument-dependent | ```flowchart Define the surface or interface to be characterized and the chemical-state question the measurement must answer → Select the appropriate X-ray source and verify that the photon energy provides adequate separation of the core levels of interest → Acquire a survey spectrum to identify all elements present and check for unexpected contamination → Select core levels for high-resolution scanning based on the survey results and the process question → Acquire high-resolution spectra with pass energy and step size sufficient to resolve the expected chemical shifts → Perform peak fitting with physically justified constraints, reporting the line shape, width, and number of components used → If depth information is needed, choose angle-resolved XPS for non-destructive near-surface profiling or sputter depth profiling for deeper structures, and report the artifacts and assumptions of each → Quantify using sensitivity factors matched to the instrument and source, and state the quantification model and its assumptions → Compare results against process specifications, accounting for adventitious contamination and any transfer-induced chemistry changes → Archive the raw spectra alongside the processed results so that future reanalysis with updated fitting or sensitivity factors is possible ``` Read XPS through a spectroscopic-fingerprint lens: the binding energy of a core-level photoelectron encodes both the identity of the emitting atom and the electronic influence of its chemical neighbors, so every peak position, shape, and shift is a fingerprint of the local bonding environment at the surface — but extracting a reliable process answer from that fingerprint requires controlling the measurement conditions, choosing physically grounded fitting models, and distinguishing the chemistry of the process from the chemistry introduced by sample handling and analysis.