XPS for Nanomaterials: Surface Chemistry, Oxidation State and What 10 nm Means

The question XPS answers that nothing else does

Electron microscopy shows you shape. Diffraction shows you crystal structure. Light scattering shows you hydrodynamic size. None of them tells you what the outermost few nanometres are chemically made of — and for a functionalised nanoparticle, a coated pigment, a catalyst or a graphene derivative, that is the property being bought and sold.

X-ray photoelectron spectroscopy gives elemental composition, chemical state and oxidation state of the near-surface region, quantitatively, for every element except hydrogen and helium. It is the reference method for surface chemistry, and it is also the technique whose results are most frequently over-claimed on a certificate of analysis.

How it works

Photoemission in one equation

A monochromatic X-ray source illuminates the sample. A core-level electron absorbs a photon and is ejected. The instrument measures the kinetic energy of the ejected electron, and binding energy follows from the photon energy minus the kinetic energy minus the spectrometer work function.

Because core-level binding energies are characteristic of the element, the spectrum identifies what is present. Because they shift by a fraction of an electronvolt to a few electronvolts with chemical environment, the spectrum also identifies the state it is present in — the chemical shift, which is where the technique earns its keep.

The sources

Laboratory instruments almost always use aluminium K-alpha at 1486.6 eV, sometimes magnesium K-alpha at 1253.6 eV. Monochromated aluminium is the modern default because it removes satellite lines and narrows the intrinsic linewidth, which matters when you are trying to separate chemical states half an electronvolt apart.

What "surface sensitive" means in numbers

X-rays penetrate microns into a solid, so photoelectrons are generated throughout that depth. Almost none of them escape. The surface sensitivity comes entirely from the inelastic mean free path of the electron on its way out: an electron that loses energy in a collision no longer carries its element's binding-energy signature and contributes only to the background.

For the kinetic energies typical of laboratory XPS, the inelastic mean free path in a solid is on the order of one to three nanometres. The conventional sampling depth is about three times that — so, in round terms, the top five to ten nanometres, with the signal weighted exponentially towards the outermost layers. Roughly two-thirds of the signal comes from the first mean free path.

Why this matters on a nanoparticle

Here is the misreading the field lives with. On a bulk substrate, a sampling depth of, say, eight nanometres is unambiguously a surface measurement — it is a vanishing fraction of the sample. On a 20 nm particle the same eight nanometres reaches a substantial fraction of the way to the centre, and integrated over a sphere it constitutes a large proportion of the total volume. The measurement is closer to a whole-particle composition than to a surface composition.

Two practical consequences. First, a supplier reporting XPS on small particles as evidence of a monolayer coating has not demonstrated what they think: the same spectrum is consistent with the coating material being distributed through the particle. Second, the reverse error is also common — a thin shell on a large core is over-represented in XPS relative to its mass fraction, so a small XPS signal for a coating element does not mean a small amount of coating.

Where the geometry has to be quantified, the routes are angle-resolved XPS on flat samples, and modelling the sphere explicitly. Both are beyond a routine certificate, and that is the point: a routine certificate cannot settle a core-shell question with XPS alone, and should not be read as though it has. Cross-check against microscopy, which does resolve shells directly — see the TEM and SEM comparison.

Survey and high-resolution scans

A survey scan sweeps a wide binding-energy range at low resolution and answers "what elements are here". It is the first thing to look at, and it is where unexpected contamination shows up.

High-resolution scans over a narrow window around one core level, at higher energy resolution, are where chemical state comes from. A certificate quoting only atomic percentages from a survey has given you composition and not chemistry; the oxidation-state claim requires the high-resolution regions and the fits.

Quantification, and its real error bars

Peak areas are converted to atomic concentrations using relative sensitivity factors that account for photoionisation cross-section, analyser transmission and escape depth. The method is sound and it is routinely presented with more precision than it has.

• Sensitivity factors are instrument- and library-dependent. Two laboratories can report meaningfully different atomic percentages from the same sample.
• The standard treatment assumes a homogeneous, flat sample. A nanoparticle powder is neither, and the depth weighting interacts with the geometry.
• Hydrogen is invisible, so all percentages are on a hydrogen-free basis. For organic coatings this matters.
• Adventitious carbon — the thin hydrocarbon layer that forms on any surface exposed to air — contributes to every carbon signal and inflates the apparent carbon content.

Relative comparisons on the same instrument, same day, same preparation are far more reliable than absolute numbers. A supplier's carbon-to-oxygen ratio for a graphene oxide is useful for ranking their own batches, and should be treated cautiously when compared against a different laboratory's number.

Charge referencing, and the argument about it

Insulating samples charge under X-ray irradiation, shifting every peak. The energy scale must therefore be referenced to something. The near-universal convention is to set the adventitious carbon C 1s peak to 284.8 eV.

The convention is convenient and it is contested, for good reasons: the adventitious layer is not a defined chemical species, its position genuinely varies with substrate, and on a carbon-based material it is inseparable from the sample's own carbon. Alternatives include referencing to a sputter-deposited noble metal, to the Fermi edge on conductors, or to an internal standard of known state. Whichever is used, it must be reported — ISO 19318 addresses exactly this, the reporting of methods used for charge control and charge correction. A binding energy quoted without its referencing method is quoted to about an electronvolt at best, which is the same order as the chemical shifts being claimed.

Instrument energy-scale calibration is a separate matter, specified in ISO 15472 and performed against clean metal reference lines — gold, silver and copper — at values that are well established and should be checked periodically rather than assumed.

Peak fitting and how it is abused

A high-resolution envelope is deconvolved into components representing distinct chemical states. Done well, this is the most informative operation in the technique. Done badly, it is curve-drawing.

Signals of a good fit: a stated background model, physically justified constraints on component widths, correct spin-orbit doublet separations and area ratios where applicable, and the same fitting model applied consistently across a sample series. Signals of a bad one: many components with unconstrained widths and positions, no residual shown, and component assignments made only by proximity to a literature value. A useful discipline is to ask how the conclusion changes if one component is removed — if it does not, the component was decoration.

Two examples worth knowing. In graphene oxide, the C 1s envelope contains contributions from sp² carbon, carbon singly bonded to oxygen in hydroxyl and epoxide environments, carbonyl, and carboxyl, at progressively higher binding energy; the relative areas are the standard measure of oxidation, and the relative order is reliable even where absolute positions vary between laboratories. In titanium dioxide, the Ti 2p doublet position identifies the fully oxidised state, and reduced titanium appears as a low-binding-energy shoulder — a small, real, and easily over-fitted feature.

Oxidation state: the real payoff

Where XPS is genuinely irreplaceable is in distinguishing oxidation states that no other routine technique separates. Metallic silver versus silver oxide, metallic versus oxidised iron, the presence of reduced titanium in a photocatalyst, sulfur speciation in a functionalised particle, nitrogen environments in a doped carbon. For a silver nanomaterial sold on antimicrobial performance, the fraction of the surface that is metallic rather than oxidised is closer to the functional property than the total silver content is.

One subtlety that catches people out: chemical shifts do not always move in the intuitive direction. Silver is the standard counterexample — its core level shifts to lower binding energy on oxidation, opposite to the usual pattern for metals. Where a shift is small and the direction matters, the modified Auger parameter, which combines a photoelectron line with an Auger line, is more robust because it is insensitive to charge referencing.

Sample handling

• Ultra-high vacuum is required, so volatile and wet samples need drying, and anything that outgasses badly may be refused by the facility.
• Contamination is the dominant error source. Fingerprints, silicone from tubing and glove powder are all clearly visible and all avoidable.
• Powders are usually pressed into indium foil or onto conductive tape. Tape can contribute its own signal; ask whether it did.
• Beam damage is real for organics, some oxides and halides. Reduction of metal ions under prolonged X-ray exposure has been observed often enough that a long acquisition on a sensitive material should be checked against a short one.
• Air exposure changes the answer. A sample stored in air for a month is not the sample that left the reactor, and for oxidation-state work the storage history belongs in the report.

What to ask for on a certificate

• The survey and the high-resolution regions, not only atomic percentages.
• The charge referencing method and the value used.
• The fitting model — background type, components, constraints.
• Whether the sample was exposed to air, and for how long, before measurement.
• For any core-shell or coating claim, what independent measurement supports it, given the sampling-depth geometry above.

The general version of that discipline is how to read a certificate of analysis.

Standards and reference data

ISO 15472 specifies calibration of the binding-energy scale for X-ray photoelectron spectrometers. ISO 19318 covers the reporting of methods used for charge control and charge correction. The NIST X-ray Photoelectron Spectroscopy Database is the standard free reference for binding energies and is a better first stop than a search engine, because it carries the measurement conditions alongside the values.

Related guides

• Where surface analysis sits in the workflow: the characterisation workflow.
• Surface chemistry from the mass-loss side: TGA for nanomaterials.
• The material where XPS carbon-oxygen ratios do most of the work: graphene, GO or rGO.