Place recognition is a fundamental task in robotics and autonomous navigation, playing a key role in intelligent driving and environmental perception. However, existing approaches often rely on complex architectures, leading to high computational demands and energy consumption, which hinder efficient hardware deployment. Spiking neural networks (SNNs), due to their asynchronous and event-driven computation, hold great promise for low-power and high-speed inference. Motivated by this, we propose an end-to-end SNN model for LiDAR-based place recognition (LPR). Unlike conventional SNN methods that rely on repeated static inputs to simulate temporal dynamics, our model directly processes LiDAR scan sequences by assigning individual frames to separate time steps. This strategy maximizes the utilization of SNN’s temporal dimension, avoids redundancy from repeated inputs, and captures richer dynamic information. Additionally, we design a spatiotemporal fusion module to overcome the limitations of previous SNNs, where pooling operations failed to effectively capture temporal feature dynamics. Experiments on real-world benchmark datasets demonstrate that our method improves recognition accuracy while substantially reducing energy consumption. Overall, this work presents an energy-efficient solution for LiDAR-based place recognition using spiking neural networks.