1: \begin{abstract}

2: As is well-known the Complex 

3: Langevin (CL) method sometimes fails to converge or converges to the wrong 

4: limit. We identified one reason for this long ago: insufficient decay of 

5: the probability density either near infinity or near poles of the drift, 

6: leading to boundary terms that spoil the formal argument for correctness. 

7: To gain a deeper understanding of this phenomenon, we analyze the 

8: emergence of such boundary terms thoroughly in a simple model, where 

9: analytic results can be compared with numerics.  We also show how some 

10: simple modification stabilizes the CL process in such a way that it can

11: produce results agreeing with direct integration.  Besides 

12: explicitly demonstrating the 

13: connection between boundary terms and correct convergence our

14: analysis also suggests a correctness criterion which could be 

15: applied in realistic lattice simulations.

16: 

17: \end{abstract}

18: