Particulate matter (PM) pollution in China is an emerging environmental issue which policy makers and the public have increasingly paid attention to. In order to investigate the characteristics, sources, and chemical processes of PM pollution in Guangzhou, field measurements were conducted from 20 November 2017 to 5 January 2018, with a time-of-flight aerosol chemical speciation monitor (ToF-ACSM) and other collocated instruments. Mass concentrations of non-refractory submicron particulate matter (NR-PM₁) measured by the ToF-ACSM correlated well with those of PM_2.5 or PM_1.1 measured by filter-based methods. The organic mass fraction increased from 45 % to 53 % when the air switched from non-pollution periods to pollution episodes (EPs), indicating significant roles of organic aerosols (OAs) during the whole study. Based on the mass spectra measured by the ToF-ACSM, positive matrix factorization (PMF) with the multilinear engine (ME-2) algorithm was performed to deconvolve OA into four factors, including hydrocarbon-like OA (HOA, 12 %), cooking OA (COA, 18 %), semi-volatile oxygenated OA (SVOOA, 30 %), and low-volatility oxygenated OA (LVOOA, 40 %). Furthermore, we found that SVOOA and nitrate were significantly contributed from local traffic emissions while sulfate and LVOOA were mostly attributed to regional pollutants. Comparisons between this work and other previous studies in China show that secondary organic aerosol (SOA) fraction in total OA increases spatially across China from the north to the south. Two distinctly opposite trends for NR-PM₁ formation were observed during non-pollution periods and pollution EPs. The ratio of secondary PM (SPM = SVOOA + LVOOA + sulfate + nitrate + ammonium) to primary PM (PPM = HOA + COA + chloride), together with peroxy radicals ${\mathrm{RO}}_{\mathrm{2}}^{\ast }$ and ozone, increased with increasing NR-PM₁ concentration during non-pollution periods, while an opposite trend of these three quantities was observed during pollution EPs. Furthermore, oxidation degrees of both OA and SOA were investigated using the f₄₄∕f₄₃ space and the results show that at least two OOA factors are needed to cover a large range of f₄₄ and f₄₃ in Guangzhou. Comparisons between our results and other laboratory studies imply that volatile organic compounds (VOCs) from traffic emissions, in particular from diesel combustion and aromatic compounds, are the most likely SOA precursors in Guangzhou. Peroxy radical ${\mathrm{RO}}_{\mathrm{2}}^{\ast }$ was used as a tracer for SOA formed through gas-phase oxidation. For non-pollution periods, SOA concentration was reasonably correlated with ${\mathrm{RO}}_{\mathrm{2}}^{\ast }$ concentration during both daytime and nighttime, suggesting that gas-phase oxidation was primarily responsible for SOA formation. However, there was no correlation between SOA and ${\mathrm{RO}}_{\mathrm{2}}^{\ast }$ in pollution EPs, suggesting a dramatically changed mechanism for SOA formation. This conclusion can also be supported by different features of SOA in a van Krevelen diagram between non-pollution periods and pollution EPs. Furthermore, for pollution EPs, when NR-PM₁ mass concentration was divided into six segments, in each segment except for the lowest one SOA concentration was correlated moderately with ${\mathrm{RO}}_{\mathrm{2}}^{\ast }$ concentration, suggesting that gas-phase oxidation still plays important roles in SOA formation. The intercepts of the above linear regressions, which likely correspond to the extent of other mechanisms (i.e., heterogeneous and multiphase reactions), increase with increasing NR-PM₁ mass concentration. Our results suggest that while gas-phase oxidation contributes predominantly to SOA formation during non-pollution periods, other mechanisms such as heterogeneous and multiphase reactions play more important roles in SOA formation during pollution EPs than gas-phase oxidation.