1: \section{Introduction}
2: It seems a paradox - and is just a typical process in mathematical analysis
3: that a structure turns simple in a more general setting which at the same time
4: widens its range of application. Accordingly - ~not very surprising~ - some
5: ill-famed 'intricacies of the propagator approach to TLM' (sic. Rebel
6: \cite{diss}, p. 5) virtually vanish if some of its elements are taken as the
7: building blocks of a more general scheme.
8: In fact, constructing the latter on essentially these elements in a quasi
9: axiomatic manner will prove such intricacies to be mere artefacts of an
10: inadequate framework.\\
11: The choice of elements proposed in this paper 'generalizes' the Johns
12: algorithm in two directions.
13: In the first place, abstract scattering channels replace transmission lines,
14: which have some unpleasent properties (section \nolinebreak \ref{S:sec2}).
15: Secondly, non-trivial cell boundary (\emph{interface\/}) scattering
16: is permitted during the connection cycle.
17: The schemes thus obtained are characterized by a non trivial two-step
18: (\emph{connection-reflection\/}) cycle of iteration which exhibits
19: certain duality relations - ~whence their name.
20:
21: When P.B. Johns and co-workers introduced the transmission line matrix (TLM)
22: numerical method in the 1970s \cite{JoBe} it was almost instantaneously
23: assimilated by the microwave engineering community. In the same audience
24: the method remained until today subject of assiduous study and extensive
25: publication. Three conferences explicitely focussed on TLM \cite {A,B,C},
26: and the monographs of Christopoulos \cite{Ch1} and de Cogan \cite{Co1} deal
27: in detail with the original ideas as well as with classical applications.
28:
29: Familiarity with the transmission line picture, and the well known
30: scattering concept, certainly fostered the acceptance of the TLM method
31: among microwave engineers.
32: On the other hand, just so the primary interest turned of course on
33: applications in their own discipline, rather than onto the
34: %applications in their disciplines rather than onto the
35: inner algorithmic structure as an object of mathematical analysis.
36: Over the years still a node more was routinely invented,
37: with new dispersion characteristics and/or equipped with still another
38: ingenious stub, designed to model special propagation or transport
39: phenomena, in varied geometries or boundary conditions.
40: \cite{Trenkic1} stands somewhat exemplary for this line of research.
41:
42: Mathematical questions addressing the inner structure of the TLM
43: algorithm and its potential generalizations have thus apparently
44: been for a long time of secondary interest.
45: They have yet not been left completely out of view.
46: Chen, Ney and Hoefer \cite{Chen} proved equivalence of the original
47: (expanded) TLM node without stubs \cite{JoBe,Johns0} to the Yee
48: finite-difference grid \cite{Yee1,KuLue}.
49: Recently, the non-trivial question of consistence of Johns'
50: symmetrical condensed node (SCN), cf. Johns \cite{Johns2},
51: with Maxwell's equations and the intimately related problem
52: of convergence to a smooth solution for decreasing time step and grid
53: spacing have been tackled, and in parts solved, by Rebel \cite{diss}.
54: His thesis presents, by the way, a thorough survey over the ramifications
55: of TLM until that time (year 2000), without perhaps spending
56: sufficient attention to its non-orthogonal mesh extensions.
57:
58: From a quite general viewpoint, viz. widely independent of any
59: particular physical interpretation, the structure of the stub loaded
60: (deflected) non-orthogonal TLM algorithm has been analysed in \cite{He5}.
61: The present paper goes even further and challenges the transmission line
62: picture at all. The latter, in universally imposing free wave propagation
63: between cells (with great benefit, at times), induces modeling
64: limitations under circumstances that are outlined in section
65: \nolinebreak \ref{S:sec2}.
66: Many restrictions can be by-passed by replacing transmission lines with
67: abstract scattering channels in terms of 'paired' distributions.
68:
69: Dual scattering channel schemes are characterized by a two-step
70: updating cycle with certain duality relations between the two steps.
71: The TLM method with its familiar connection-reflection cycle is
72: trivial \emph{as a dual scheme} in that the connection map
73: reduces essentially to identity (viz. pure transmission or total
74: reflection) - \nolinebreak again with modeling limitations.
75: These can be raised, anew, in permitting non trivial cell interface
76: scattering during the connection step of iteration.
77:
78: One major merit of the transmission line method is unconditional
79: stability ~\cite{Johns3}.
80: Since this property is usually drawn upon the passivity of linear
81: transmission line networks, the question of stability needs a proper
82: investigation for DSC schemes that do not use lines.
83: That problem is studied in~\cite{He7}, where it is shown that a wide
84: class of DSC schemes are in fact unconditional stable.
85: Due to the convolution type updating scheme
86: (\emph{Johns' cycle}; cf. section ~\ref{S:sec3}) it is sufficient for
87: stability that the reflection and connection maps share simple
88: contraction properties (paraphrased as $\,\alpha$-\emph{passivity}
89: in ~\cite{He7}).
90: In summary, DSC schemes are unconditinally stable under quite general
91: circumstances, and they are conceptually simple, though a set of
92: technical definitions is of course ineluctable in a neat theory.
93: Last but not least, nothing obscure nor 'intricate' should be associated
94: anymore with the propagator approach.
95: