Background: Many low- and middle-income countries are experiencing high and increasing exposure to ambient fine particulate air pollution (PM _2.5). The effect of PM _2.5 on infant and child mortality is usually modelled using concentration response curves extrapolated from studies conducted in settings with low ambient air pollution, which may not capture its full effect. Methods: We pool data on more than half a million births from 69 nationally representative Demographic and Health Surveys that were conducted in 43 low- and middle-income countries between 1998 and 2014, and we calculate early-life exposure (exposure in utero and post partum) to ambient PM _2.5 using high-resolution calibrated satellite data matched to the child's place of residence. We estimate the association between the log of early-life PM _2.5 exposure, both overall and separated by type, and the odds of neonatal and infant mortality, adjusting for child-level, parent-level and household-level characteristics. Results: We find little evidence that early-life exposure to overall PM _2.5 is associated with higher odds of mortality relative to low exposure to PM _2.5. However, about half of PM _2.5 is naturally occurring dust and sea-salt whereas half is from other sources, comprising mainly carbon-based compounds, which are mostly due to human activity. We find a very strong association between exposure to carbonaceous PM _2.5 and infant mortality, particularly neonatal mortality, i.e. mortality in the first 28 days after birth. We estimate that, at the mean level of exposure in the sample to carbonaceous PM _2.5 - 10.9 μg/m ³ - the odds of neonatal mortality are over 50% higher than in the absence of pollution. Conclusion: Our results suggest that the current World Health Organization guideline of limiting the overall ambient PM _2.5 level to less than 10 μg/m should be augmented with a lower limit for harmful carbonaceous PM _2.5.