The logistics industry is a major source of global carbon emissions and energy consumption. Cooperative logistics and split delivery are effective methods for reducing carbon emissions and improving distribution efficiency, respectively. However, prior research on pollution routing problems (PRP) has primarily focused on centralized cooperation and has not fully explored the potential of split delivery as a strategy for addressing PRPs. This study fills this gap by integrating decentralised cooperation and split delivery to the PRP domain, offering an innovative approach to the problem. A bi-objective mixed integer linear PRP model with split delivery and request selection is put forward for the problem. In addition, an $\varepsilon$ -constrained hybrid evolutionary algorithm is proposed herein, which combines a Greedy Randomized Adaptive Search Procedure-Evolutionary Local Search (GRASP-ELS) hybrid approach with the $\varepsilon$ -constrained method. The results of experiments demonstrate that the proposed algorithm outperforms the well-known multi-objective optimization algorithms: NSGA-II, MOPSO, and SPEA-II. The study also provides several managerial insights through sensitivity analysis. The proposed model and algorithm can provide a basis for decision making for logistics companies to improve logistics efficiency and reduce carbon emissions.