This study employs panel data from 1884 listed manufacturing companies in China (2009–2021) to investigate the environmental effects of path dependence on atmospheric pollution emissions. Using dictionary-based textual analysis of annual reports, we measure three dimensions of path dependence—technological, institutional, and managerial—and examine their non-linear relationships with sulfur dioxide emissions through fixed-effects models. Our findings reveal consistent U-shaped patterns across all dependence types: moderate levels initially reduce emissions (the “honey phase”) while excessive reliance leads to increased pollution (the “arsenic phase”). The analysis demonstrates that technological path dependence operates through sunk costs and learning effects, institutional dependence reflects regulatory inertia, and managerial dependence stems from organizational routines. Robustness tests using alternative pollution measures and instrumental variable approaches confirm these relationships. The study identifies significant heterogeneity in these effects. Non-state-owned enterprises exhibit stronger path dependence impacts due to greater flexibility, while high-maturity firms show amplified U-curves reflecting their accumulated experience. Conversely, capital-intensive enterprises display attenuated effects, suggesting diminishing returns to scale in pollution control. These findings highlight the dual nature of path dependence as both a stability mechanism and potential barrier to innovation. The policy implication is that manufacturing pollution control strategies should account for both dependence levels and firm-specific characteristics, maintaining path dependence within optimal ranges to harness stabilization benefits without impeding technological transitions. This research contributes to environmental governance literature by extending path dependence theory to pollution control and offering a multidimensional analytical framework for sustainable manufacturing transformation.