1: \begin{abstract}

2: We establish the existence of wave-like solutions to

3: spatially coupled graphical models which, in the large size limit,

4: can be characterized by a one-dimensional real-valued state.

5: This is extended to a proof of the threshold saturation phenomenon for all

6: such models, which

7: includes spatially coupled irregular LDPC codes over the BEC, but

8: also addresses hard-decision decoding for transmission over general

9: channels, the CDMA multiple-access problem, comressed sensing, and some statistical

10: physics models.

11: 

12: For iterative coding systems with two components and

13: transmission over the BEC, the asymptotic convergence behavior is

14: completely characterized by the EXIT curves of the components. More

15: precisely, the system converges to the desired fixed point corresponding

16: to perfect decoding if and only if the two EXIT functions describing

17: the components do not cross.

18: For spatially coupled systems whose state is one-dimensional a

19: closely related graphical criterion applies. Now the curves are

20: allowed to cross, but not by too much.  More precisely, we show

21: that the threshold saturation phenomenon is related to the positivity

22: of the (signed) area enclosed by two EXIT-like functions associated to the

23: component systems, a very intuitive and easy-to-use graphical

24: characterization.

25: 

26: In the spirit of EXIT functions and Gaussian approximations, we

27: also show how to apply the technique to higher dimensional and even

28: infinite-dimensional cases. In these scenarios the method is no

29: longer rigorous, but it typically gives accurate predictions.  To

30: demonstrate this application, we discuss transmission over general

31: channels using both the belief-propagation as well as the min-sum

32: decoder.  \end{abstract}