Legal textual entailment is a critical yet challenging task due to the complex structure and nuanced semantics of legal articles. This task requires not only comprehensive legal text understanding but also advanced reasoning capabilities for effective document structure analysis. In this paper, we leverage Large Language Models (LLMs) to analyze the causal relation structures within legal documents, thereby strengthening the reasoning ability for legal textual entailment through the generation of high-quality synthetic data. Our experimental results demonstrate that this approach outperforms previous methods, confirming the feasibility and effectiveness of using LLM-generated synthetic datasets. Moreover, variations in performance across test sets highlight the trade-offs between zero-shot and fine-tuning strategies, as well as the significant impact of prompt engineering on accuracy. These findings open up promising research directions for further improving reasoning and generalization in legal text processing tasks.