Atmospheric aerosols influence climate and human health, yet the mechanisms governing new particle formation (NPF) and early growth remain incompletely understood, in part because ultrafine particle composition and volatility are difficult to measure and interpret. Here we combine size- and temperature-resolved Thermal Desorption Chemical Ionization Mass Spectrometer measurements with positive matrix factorization (PMF) to reduce chemical complexity in particles formed during α-pinene ozonolysis in a continuous-flow chamber. PMF resolves seven factors: six parent ion-dominated factors spanning semi-volatile to low-volatility behavior with factor log C* ranging from 2.48 to -3.67, and one decomposition factor dominated by thermal and ionization fragments. Representative ions indicate a systematic volatility progression from semi-volatile carboxylic acids to low-volatility multifunctional acids and diacids. An absorptive partitioning model using size distribution-derived total particulate mass shows that low-volatility factors remain effectively in the particle phase across the experiment, while semi-volatile factors increase their particle-phase fraction as condensed mass increases. Size-resolved factor contributions across sampled volume mean diameters 30-130 nm show that particles below 75 nm are enriched in low-volatility factors, whereas semi-volatile compounds become important only after substantial growth to ~75-100 nm, consistent with reduced Kelvin limitations at larger sizes. This framework provides volatility- and size-resolved constraints on nanoparticle chemical evolution, advancing NPF understanding by identifying which volatility regimes control growth as particles transition from nucleation to accumulation of condensable mass.

The chemical composition of particles generated from α-pinene ozonolysis was characterized by Thermal Desorption Chemical Ionization Mass Spectrometry (TDCIMS). For these experiments, the TDCIMS was operated in negative ion mode, and assigned ions were assumed to have the form [M−H]-. Particles were collected on a filament and then thermally desorbed during a linear temperature ramp to approximately 800 °C. After each particle collection, a background analysis was performed. PMF analysis was performed using EPA PMF 5.0 on the resulting background-subtracted thermograms, defined as collection intensity minus background intensity.

Each desorption thermogram consists of 350 points spanning 0–70 s. Because the PMF input matrix required a single monotonically increasing independent variable, an arbitrary sequential index was assigned to the rows of the concentration and uncertainty matrices and stored in the first column labeled time. This index was used only to preserve the ordering of points across multiple collection-background thermogram pairs and does not have physical meaning as time or temperature.

After PMF analysis, each sequential 350-point block was mapped back to the original 0–70 s desorption-time axis for interpretation. Because the desorption ramp was linear, this mapping also corresponds to the relative temperature position within each thermogram. Thus, although the PMF input files use an arbitrary sequential index, the resulting factor profiles and contributions were interpreted in desorption-time and temperature space.

PMF does not require prior knowledge of source profiles and can be applied to additive, non-negative datasets using any ordered independent variable. In this study, PMF was used to deconvolute overlapping thermogram signals and group ions with similar desorption behavior, enabling interpretation in terms of apparent volatility.