Fine particulate matter (particulate matter with a diameter of 2.5 μm or less; PM2.5) causes millions of premature deaths globally1, but not all particles are equally harmful2–4. Current air-pollution control strategies, prioritizing PM2.5 mass reduction, have provided considerable health benefits but further refinements based on differences in the toxicity of various emission sources may provide greater benefits5–7. Here we integrated field measurements with air-quality modelling to assess the unequal toxicities of PM2.5 from various anthropogenic sources. Our findings revealed that the toxicity per unit of PM2.5 mass differed substantially between major sources, differing by up to two orders of magnitude. PM2.5 from solid fuel combustion in residential stoves had the highest toxicity, followed by those from the metallurgy industry, brake wear, diesel vehicles, petrol vehicles, the cement industry and power plants. We further analysed the source contributions of toxicity-adjusted PM2.5 emissions and population exposures in China. From 2005 to 2021, both the PM2.5 mass and relative-potency-adjusted emissions substantially decreased. Although industrial sources contributed 57.5% to the reduction in PM2.5 mass emissions, the reduction in relative potency-adjusted emissions was driven by residential combustion (approximately 80%). Clean-air policies should consider the differing toxicities of PM2.5 when formulating source-specific emission control regulations. This study proposes a cellular toxicity-based framework for PM2.5 reduction that could address the specific health risks in diverse regions, but further epidemiological studies will be required to confirm their relevance to human health outcomes and their application to public policy.
