The XPS peak that probably isn't sp3 carbon

My research background is microwave spectroscopy, which means I have spent a lot of time on the least glamorous part of the job: deciding what a peak is. A spectrometer does not hand you chemistry. It hands you a position, a width and an area, and every statement past that point is an interpretation you made, or more often an interpretation you inherited. Assignments propagate. Someone publishes one, the next paper cites it as established, the software's default fitting template ships with the label in it, and twenty years later the label is doing work that nobody has re-derived.

That is the story in Yamada et al. 2026, published open access in the Journal of Materials Science on 29 June 2026. It is a long, unglamorous paper about carbon fiber. Inside it is a claim that touches a very large amount of published work.

The assignment in question

X-ray photoelectron spectroscopy measures the binding energy of core electrons. For carbon, the C1s spectrum of a graphitic material has a main peak from aromatic C=C, and then a shoulder at slightly higher binding energy that essentially every carbon paper has to account for. The standard move is to fit a component near 285 eV and label it sp3 carbon: diamond-like, tetrahedral, non-graphitic. It is intuitive, since sp3 carbon does sit higher than sp2, and it gives you a satisfying number to report as a "degree of disorder."

The paper's central objection is two sentences long and does most of the damage:

Crucially, while sp3C is widely assigned to ~ 285 eV, both computations and experimental C1s XPS spectra of uncharged sp3C peaks show much lower peak positions, near sp2C. Because sp3C is thermally unstable, assigning this peak at ca. 285 eV as sp3C in carbon materials prepared at high temperatures is highly controversial; consequently, many past sp3C assignments may be incorrect.

Take those in order, because they are separate problems.

First, the energetic argument. When you measure sp3 carbon that is not charging up, its C1s peak does not appear at 285 eV. It appears much closer to sp2. So the feature at 285 eV was never sitting where uncharged sp3 carbon actually sits. Charging matters here because carbon materials are often poorly conductive; as photoelectrons leave, the surface goes positive, and the whole spectrum, or parts of it, shifts to higher apparent binding energy. A shifted peak from an artifact can land exactly where the textbook says sp3 lives.

Second, the thermodynamic argument, which is the one I find hardest to argue with. sp3 carbon is thermally unstable. The materials in question were heated to 1473 K and above. If you claim a substantial sp3 population in something annealed at 1200 °C or hotter, you are claiming a phase survived conditions that should have converted it. That is not impossible, but it is a claim requiring support, and it is almost never offered as one. It arrives as a fitting label.

What else that peak could be

The useful thing about this paper is that it does not simply say the assignment is wrong. It enumerates the competing origins, and the list is uncomfortable because most of the entries are not chemistry at all.

By the authors' account, a peak at roughly 285 eV in a carbon C1s spectrum may be charged-up sp3 carbon (an artifact of sample conductivity), C–N bonding (if there is any nitrogen present, and there frequently is), adventitious carbon (surface contamination from air exposure, not part of your material), or the structural origin this paper argues for. Historically, the paper notes, peaks around 285 eV were attributed to adventitious carbon in the first place. The interpretation drifted toward sp3 over the following decades.

Three of those four candidates say nothing about the material you synthesized. Two are contamination or instrument effects. That is the situation before you get to the paper's own contribution.

The rule they found

Here is the part I think is genuinely elegant, and it reframes what the measurement is capable of telling you.

The team optimized 34 large graphene-based model structures containing edges, oxygen functional groups, sp3 carbon, vacancy defects and non-hexagonal rings, using density functional theory at the B3LYP/6-31G(d) level in Gaussian 16, then simulated the spectra. Plotting calculated C1s binding energy against structure produced a clean correlation, and it is not about hybridization. It is about the rings surrounding the atom you are measuring.

Every carbon atom in the interior of a graphene sheet sits at the junction of three rings. In pristine graphite those are three hexagons, so the summed ring number is 6 + 6 + 6 = 18. Introduce a defect and that sum changes. The paper finds that binding energy shifts systematically higher as the summed ring number increases. A carbon in a 6-6-8 environment, summing to 20, reaches 285.1 eV. Using their fitted relationship, a 5-5-6 carbon (summing to 16) is estimated at 283.0 eV, and a 6-7-8 carbon (summing to 21) at 285.9 eV. The same trend holds for N1s in tertiary nitrogen.

So the abstract's conclusion is that the ca. 285 eV peak originates from carbon atoms surrounded by three rings including at least one heptagon, one octagon, or an even larger vacancy defect, conditional on the peak not being charging, C–N, or adventitious carbon.

Sit with what that changes. The conventional label says the peak reports a bonding type: this carbon is tetrahedral rather than trigonal. The new assignment says it reports a neighborhood: this carbon is perfectly ordinary sp2 carbon that happens to be adjacent to a ring that is not a hexagon. Same peak, same energy, completely different physical picture, and a completely different set of conclusions about your material. One story says you have diamond-like domains. The other says you have topological defects in a graphitic sheet. Those imply different synthesis problems and different properties.

It also explains why the assignment was sticky. A peak at higher binding energy than C=C genuinely does mean "something is disturbed here." The error was in naming which disturbance, and the wrong name happened to be the one with an obvious chemical story attached.

What the paper actually did, and what it did not

Some scoping, because the headline is strong and the evidence has a specific shape.

This is a computational assignment anchored to one material system, not a direct measurement of ring topology. The experimental side is isotropic pitch-based carbon fiber, heat-treated across a range up to 3173 K, with HOPG as reference, characterized by Raman, IR and XPS plus XRD and elemental analysis. The computational side supplies the interpretation. The conclusion that the main defects in the fiber heated at 1873 K and above are edges, cyclic ethers and non-hexagonal rings follows from matching the two.

Second, and this matters for anyone who wants to apply the numbers: calculated absolute binding energies are not trustworthy on their own, and the authors say so. Peak positions in these calculations are heavily influenced by model size, edge type and calculation conditions, so it is necessary to adjust calculated positions to experimental positions using scaling factors. On the experimental side the C=C peak top was set to 284.3 eV to match. The differences between structures are the result; the absolute values inherit a calibration choice. The regression estimates for unmeasured defect types, like the 283.0 eV and 285.9 eV figures above, are extrapolations from a fitted line and are labelled as estimates in the paper.

Third, the claim is scoped to high-temperature materials. The thermal-instability argument does not apply to a polymer-derived carbon made at 400 °C, or to a genuinely diamond-like film. If your material could plausibly hold sp3 carbon, sp3 carbon remains on the table. The target here is the reflexive use of the label in graphitic materials processed hot enough that it should have annealed out.

Fourth, this is a hypothesis with mechanistic support, not a settled correction of the literature. The paper's own wording is that many past assignments "may be incorrect." B3LYP/6-31G(d) is a reasonable and well-tested choice for these systems, and the authors justify it, but it is one level of theory. The honest status is: the conventional assignment now has a specific, testable competitor, and the burden has shifted.

Why this pattern keeps recurring

I have written some version of this report before, about different instruments. In Report 013 the issue was Raman band ratios treated as a direct readout of disorder. In Report 086 it was the five very different things "identified by mass spectrometry" can mean. In Report 062 it was a non-detection reported as an absence.

The common structure: an instrument produces a robust, reproducible number, and a layer of interpretation sits on top that is far less robust and is not re-examined because the number underneath keeps behaving. Nobody in this story measured badly. The 285 eV feature is real and repeatable. Every deconvolution that fitted it got a defensible area. The failure is entirely in the sentence that converts a binding energy into a claim about bonding, and that sentence is usually the one that makes it into the abstract.

There is a reason it survives. An assignment that is wrong but consistent still produces internally consistent trends. If your "sp3 fraction" is really a non-hexagonal-ring fraction, it will still go up when you damage the material and down when you anneal it. The correlations hold, the paper passes review, and the mislabel is invisible precisely because the underlying quantity is tracking something real. It only breaks when you try to reason from the label to a mechanism, or when someone computes what the peak should actually be.

What to ask

If you make, buy, or read characterization of carbon materials, three questions.

First, ask what the sample's processing temperature was before accepting an sp3 assignment. This is a one-line check with real discriminating power. If the material saw 1473 K or more, an sp3 claim is asserting that a thermally unstable form survived annealing, and that deserves a supporting argument rather than a fitting label.

Second, ask how charging was handled and what the peak was referenced to. Charging shifts spectra of poorly conductive samples to higher apparent binding energy, which is the exact direction that manufactures this peak. What the C=C position was set to, and how, determines where every other component appears to sit. It is a methods-section detail that decides the result.

Third, ask whether nitrogen or air exposure could account for it, before reaching for structure. C–N overlaps this region, and adventitious carbon was the original assignment for peaks around 285 eV. If a sample has been in air and contains any nitrogen, two artifactual explanations are live before you reach any conclusion about the material itself.

The broader habit is the one worth keeping. A peak position is data. A peak label is a hypothesis someone published, and it can be inherited for fifty years without anyone re-testing whether the physics supports it. When a group finally does the calculation, sometimes the answer is that the feature was never reporting what its name says. It was reporting the shape of the rings next door.