Hidden Clues in Ordinary Ions: PCA Turns Ubiquitous Hydrocarbon Fragments Into a Polymer Fingerprint

Every time a time-of-flight secondary ion mass spectrometer fires its bismuth ion beam at a polymer surface, the instrument records a cascade of negative hydrocarbon fragments labeled CnH−, where n runs from one to ten. For decades, analysts treated these ions as background noise, the chemical equivalent of static, because none of them is unique to any particular polymer. A new study published in Frontiers in Analytical Science by Heng-Yong Nie of Surface Science Western and the Department of Physics and Astronomy at the University of Western Ontario argues that this apparent weakness is actually the technique’s greatest untapped asset. By applying principal component analysis, or PCA, to the variability of these ubiquitous ions, Nie shows that the data can classify polymers, reveal their carbon density, and even track the degree of cross-linking in poly(methyl methacrylate), one of the most widely used acrylic materials in coatings and composites.

The core insight of the study is deceptively simple: when a material produces no diagnostic ions of its own, the relationships among its non-unique ions still encode structural information. Nie focused on four common polymers, polyethylene, polypropylene, polyisoprene, and polystyrene, whose negative ion mass spectra look remarkably similar. None of them generates a signature ion that would let an analyst say, at a glance, which polymer is on the surface. Instead of treating each CnH− intensity as a stand-alone number, the study examines how those intensities vary together across dozens of measurements. That covariance, rather than the identity of any single peak, turns out to be the fingerprint that separates the materials.

The key structural property driving this variability is carbon density, defined as the number of carbon atoms bonded to a given carbon atom in the polymer backbone. The concept is close to the carbon-to-hydrogen ratio but not identical. Polyethylene and polypropylene share the same C/H ratio at the monomer level, yet polypropylene has a higher carbon density because a methyl group replaces one hydrogen atom, giving some backbone carbons an extra carbon neighbor. Earlier work by the same author demonstrated that the intensity ratio between the C6H− and C4H− ions, denoted ρ, rises with carbon density, providing a quantitative handle on this structural parameter. The new study asks whether PCA, applied to the full set of ten CnH− intensities, can independently recover the same hidden structure.

PCA is a dimensionality reduction technique that transforms a matrix of measurements and variables into new orthogonal axes called principal components. The first principal component, PC1, captures the largest possible variance in the data; the second, PC2, captures the next largest, and so on. If the first two or three components explain most of the variance, the original high-dimensional dataset can be projected onto a two-dimensional plot without losing much information. In this work, a 48-by-10 data matrix, containing twelve measurements for each of the four polymers, was standardized and decomposed using the prcomp() function in the R programming language, which relies on singular value decomposition. The analysis was repeated for datasets normalized by the total ion intensity, by the sum of the ten CnH− intensities, and by each individual CnH− ion in turn.

The results were striking. Across every normalization scheme, the first two principal components explained between 87.3 percent and more than 98 percent of the total variance, meaning that a two-dimensional view was sufficient to represent the data faithfully. In the resulting biplots, where measurement scores appear as points and ion loadings as arrowed vectors, the four polymers clustered into distinct groups arranged along PC1 in order of increasing carbon density, from polyethylene through polypropylene and polyisoprene to polystyrene. Crucially, the PC1 scores correlated with ρ, the C6H− to C4H− ratio, confirming that both quantities reflect the same underlying structural property. PCA, in other words, independently rediscovered carbon density from raw ion intensities alone, without any prior chemical knowledge being fed into the model.

The biplot approach proved especially powerful for interpreting which ions matter. When the angle between two loading vectors is small, the corresponding ions are positively correlated; angles near 180 degrees indicate negative correlation, and angles near 90 degrees indicate none. The study found that C6H− through C10H− are tightly correlated with one another and point toward the scores of polystyrene, the polymer with the highest carbon density, while C2H− and C3H− point toward polyethylene and polypropylene. C4H− sits in between, contributing most strongly to polyisoprene, the medium-density polymer. This behavior explains why C4H− works so well as a reference ion: its intensity varies least across polymers of different carbon densities, so normalizing all other ions by C4H− removes variability unrelated to chemical structure while preserving the correlations that matter. The systematic comparison of all twelve normalization schemes showed that some ions are genuinely better than others at revealing structure, a conclusion that would have been invisible without the PCA framework.

To demonstrate practical utility, Nie turned to cross-linked poly(methyl methacrylate). Cross-linking, which locks polymer chains together to improve mechanical strength and thermal stability, is essentially a process of hydrogen removal, and removing hydrogen raises carbon density. Thin PMMA films were treated with hyperthermal hydrogen induced cross-linking, a technique in which neutral H2 projectiles with kinetic energies of 5 to 20 electronvolts cleave C–H bonds on the surface while preserving chemical functionalities. Treatment times of 10, 100, and 500 seconds drove the ρ value from 32.2 percent for the pristine control up to 44.7, 55.1, and 64.7 percent respectively. When the PMMA data were combined with the four-polymer dataset and analyzed by PCA, the cross-linked samples fell along the same carbon-density axis defined by the reference polymers, with the heavily cross-linked films aligning with the C6H− to C10H− loading vectors. This offers a route to assessing cross-linking in extremely thin films, where conventional characterization techniques struggle or fail entirely.

Reproducibility was tested with a second experiment conducted roughly ten years after the first, using different sample forms: polyethylene particles, polypropylene and polystyrene pellets, liquid polyisoprene on aluminum foil, and a polypropylene-polystyrene mixture film. The ρ values from the two experiments agreed strikingly well, at 19.4 versus 19.8 percent for polyethylene, 23.9 versus 23.4 percent for polypropylene, 27.3 versus 26.4 percent for polyisoprene, and 54.1 versus 54.2 percent for polystyrene. Even more telling, the 2:3 polypropylene-polystyrene mixture yielded a measured ρ of 42.4 percent, closely matching the calculated value of 41.9 percent derived from the component ratios and their weight fractions. In the PCA biplot, the mixture’s scores landed neatly between those of its two constituents, demonstrating that the approach handles polymer blends as naturally as pure materials.

The study is candid about limitations. The C6H− intensity used in ρ can be inflated by interfering species with overlapping mass-to-charge ratios, such as C3H2OF−, SiC2H5O−, C2H3NO2−, or C3H5O2−, and any such interference propagates into both the ρ values and the PCA results. The method also depends on careful instrumentation: spectra were collected with a pulsed 25 kiloelectronvolt Bi3+ primary ion beam at a dosage of 2 × 10^11 ions per square centimeter, well below the static limit, with Poisson correction applied for detector dead time and mass resolution reaching 4800 for C6H−. And while PCA reveals trends, it does not by itself quantify properties; ρ remains the quantitative measure, with PCA providing classification and structural insight.

The broader implication reaches beyond polymers. For any family of chemically similar materials that lack diagnostic ions, the information needed to tell them apart may be hiding in the variability of their spectra rather than in the identity of their peaks. Pairing a surface-sensitive technique like ToF-SIMS with data-driven tools such as PCA, and visualizing the results through covariance biplots, offers a general strategy for extracting structural clues that no single peak can provide. As analytical instruments continue to generate ever richer spectra faster than humans can interpret them, this study makes a compelling case that the future of chemical identification may lie not in finding the perfect ion, but in learning to read the patterns written across all of them.

Subject of Research: Principal component analysis of hydrocarbon secondary ions from polymers measured by ToF-SIMS to determine carbon density and cross-linking degree

Article Title: The variability in hydrocarbon ions (CnH−) of polymers detected by ToF-SIMS: principal component analysis on carbon density and cross-linking degree

Article References: Nie, H.-Y. (2025). The variability in hydrocarbon ions (CnH−) of polymers detected by ToF-SIMS: principal component analysis on carbon density and cross-linking degree. Frontiers in Analytical Science, 5, Article 1512520. https://doi.org/10.3389/frans.2025.1512520

Image Credits: AI Generated

DOI: 10.3389/frans.2025.1512520

Keywords: ToF-SIMS, principal component analysis, polymers, hydrocarbon ions, carbon density, cross-linking, PMMA, biplot, surface analysis, mass spectrometry, polyethylene, polystyrene