1: \begin{abstract}%

2: Physics-informed neural networks (PINNs) have recently emerged as effective methods for solving partial differential equations (PDEs) in various problems.

3: Substantial research focuses on the failure modes of PINNs due to their frequent inaccuracies in predictions.

4: However, most are based on the premise that minimizing the loss function to zero causes the network to converge to a solution of the governing PDE. 

5: In this study, we prove that PINNs encounter a fundamental issue that the premise is invalid. 

6: We also reveal that this issue stems from the inability to regulate the behavior of the derivatives of the predicted solution.

7: Inspired by the \textit{derivative pathology} of PINNs, we propose a \textit{variable splitting} strategy that addresses this issue by parameterizing the gradient of the solution as an auxiliary variable. 

8: We demonstrate that using the auxiliary variable eludes derivative pathology by enabling direct monitoring and regulation of the gradient of the predicted solution.

9: Moreover, we prove that the proposed method guarantees convergence to a generalized solution for second-order linear PDEs, indicating its applicability to various problems.

10: \end{abstract}

11: