1: \documentclass[12pt]{article}
2: \newcommand{\copyleft}{
3: GNU FDL\thanks{
4: Copyright (C) 1988 Peter G. Doyle.
5: Permission is granted to copy, distribute and/or modify this document
6: under the terms of the GNU Free Documentation License,
7: as published by the Free Software Foundation;
8: with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
9: }}
10: \title{Electric currents in infinite networks}
11: \author{Peter G. Doyle}
12: \date{Version dated 25 October 1988
13: \\ \copyleft
14: }
15:
16: \usepackage{epsfig}
17: %
18: % \fig
19: % followed by {tag}{caption}
20: %
21: %
22: %
23: \newcommand{\fig}[2]{
24: \begin{figure}
25: \psfig{figure=figures/#1.ps,width=370pt}
26: \caption{#2}
27: \label{#1}
28: \end{figure}
29: }
30: %
31: %
32: \newcommand{\energy}{{\rm energy}}
33: \newcommand{\resistance}{{\rm resistance}}
34: \newcommand{\oddresistance}{{\rm odd\_resistance}}
35: \newcommand{\evenresistance}{{\rm even\_resistance}}
36: \newcommand{\shortnetwork}{{\rm short\_network}}
37: \newcommand{\cutnetwork}{{\rm cut\_network}}
38: \newcommand{\shortgraph}{{\rm short\_graph}}
39: \newcommand{\cutgraph}{{\rm cut\_graph}}
40: \newcommand{\averageresistance}{{\rm average\_resistance}}
41: \newcommand{\averageevenresistance}{{\rm average\_even\_resistance}}
42: \newcommand{\averageoddresistance}{{\rm average\_odd\_resistance}}
43: \newcommand{\averagevalence}{{\rm average\_valence}}
44: \newcommand{\boundary}{{\rm boundary}}
45: \newcommand{\edgeboundary}{{\rm edge\_boundary}}
46: \newcommand{\valence}{{\rm valence}}
47: \newcommand{\valenceone}{{\rm valence1}}
48: \newcommand{\valencetwo}{{\rm valence2}}
49:
50: \newcommand{\og}{{O_{G}}}
51: \newcommand{\ohp}{{O_{HP}}}
52: \newcommand{\ohb}{{O_{HB}}}
53: \newcommand{\ohd}{{O_{HD}}}
54:
55: \newcommand{\qed}{\heartsuit}
56: \newcommand{\dispqed}{\;\;\qed}
57:
58:
59: \begin{document}
60:
61: \maketitle
62:
63: \section{Introduction.}
64:
65: In this survey, we will present the basic facts about conduction
66: in infinite networks.
67: This survey is based on the work of
68: Flanders
69: \cite{flanders:infinite,flanders:infiniteGrid},
70: Zemanian
71: \cite{zemanian:infinite},
72: and Thomassen
73: \cite{thomassen:infinite},
74: who developed the theory of infinite networks from scratch.
75: Here we will show how to get a more complete theory
76: by paralleling the well-developed theory of conduction
77: on open Riemann surfaces.
78: Like Flanders and Thomassen,
79: we will take as a test case for the theory
80: the problem of determining the resistance
81: across an edge of a $d$-dimensional grid of $1$ ohm resistors.
82: (See Figure 1.)
83: We will use our borrowed network theory
84: to unify, clarify and extend their work.
85:
86: \fig{grids}{Grids.}
87:
88: \section{The engineers and the grid.}
89:
90: Engineers have long known how to compute the resistance across
91: an edge of a $d$-dimensional grid of $1$ ohm resistors
92: using only the principles of symmetry and superposition:
93: Given two adjacent vertices $p$ and $q$,
94: the resistance across the edge from $p$ to $q$
95: is the voltage drop along the edge
96: when a $1$ amp current is injected at $p$ and withdrawn at $q$.
97: Whether or not $p$ and $q$ are adjacent,
98: the unit current flow from $p$ to $q$
99: can be written as the superposition
100: of the unit current flow from $p$ out to infinity
101: and the unit current flow from infinity into $q$.
102: By symmetry,
103: in the unit current flow from $p$ to infinity,
104: the flow out of $p$ is distributed equally
105: among the $2d$ edges going out of $p$,
106: so the flow along any one of them is $1 \over 2d$ amps.
107: Similarly, in the unit current flow into $q$,
108: the flow along each edge coming into $q$ is $1 \over 2d$.
109: If $p$ and $q$ are adjacent,
110: when the two flows are superimposed the flow along the edge from
111: $p$ to $q$ is $2 {{1}\over{2d}} = {{1}\over{d}}$
112: But since this edge has resistance $1$ ohm,
113: the voltage drop along it is also $1/d$,
114: so the effective resistance between $p$ and $q$ is $1/d$.
115:
116: A second justification for the answer $1/d$
117: was offered by Foster
118: \cite{foster:average},
119: based on his theorem that
120: the average of the resistances across all the edges of
121: a finite graph is
122: \[
123: {{n-1} \over {e}}
124: ,
125: \]
126: where $n$ is the number of vertices and $e$ the number of edges
127: of the graph.
128: (See Foster
129: \cite{foster:average,foster:extension},
130: Weinberg
131: \cite{weinberg:network} pp. 170--176.)
132: This theorem is an immediate consequence of Kirchhoff's Rule,
133: also known as Maxwell's Rule,
134: according to which
135: the resistance across an edge of a finite connected
136: network of $1$ ohm resistors
137: is the probability that a randomly selected spanning tree of the
138: network contains that edge.
139: (See Kirchhoff
140: \cite{kirchhoff:gleichungen},
141: Maxwell
142: \cite{maxwell:treatise} pp. 409--410,
143: Bollob\'{a}s
144: \cite{bollobas:graph},
145: Thomassen
146: \cite{thomassen:infinite}.)
147: If we confidently apply Foster's theorem to the $d$-grid,
148: we get the expected answer $1/d$ for the resistance across an edge.
149:
150: These two arguments could hardly be more elegant.
151: Mathematically speaking, though, they leave something to be desired.
152: As Flanders
153: \cite{flanders:infinite}
154: points out,
155: for an infinite network it is not even clear
156: what it means to talk about the effective resistance
157: between two nodes.
158: For a finite network,
159: Kirchhoff's laws determine a unique unit current
160: flow from $p$ to $q$,
161: but this clearly isn't true for an infinite network.
162: The engineers believe that in an infinite grid
163: there will still be a uniquely determined unit current flow
164: from $p$ to $q$,
165: and that this flow can be gotten by superimposing uniquely determined
166: symmetrical flows from $p$ to infinity and from infinity to $q$.
167: They offer various arguments to justify their beliefs,
168: based on approximating the infinite network by finite networks,
169: and the like.
170: The arguments I have heard are not rigorous,
171: and I am convinced that the true reason for the engineers' beliefs
172: is the analogy between conduction in the $d$-grid and classical
173: potential theory in $d$-dimensional Euclidean space.
174: Any property of conduction in Euclidean space should continue to hold
175: for the grid,
176: unless there is some obvious reason why it should not.
177:
178: The engineers are right, of course.
179: There is only one sensible definition
180: of the resistance across an edge of the $d$-grid,
181: or rather,
182: there are two sensible definitions,
183: which turn out to agree.
184: The answer $1/d$ is correct,
185: and the arguments they give to justify it are basically sound.
186: This correctness of the answer $1/d$ was shown
187: by Flanders
188: \cite{flanders:infiniteGrid}
189: in the case $d=2$,
190: using the symmetry-and-superposition method;
191: Thomassen
192: \cite{thomassen:infinite}
193: proved the result in general,
194: using Foster's method.
195: Once we have developed the theory of infinite networks,
196: we will go over the work of Flanders and Thomassen,
197: and show just how right the engineers are.
198:
199: \section{Networks as Riemann surfaces.}
200:
201: We can think of an infinite network, such as the $d$-grid,
202: as a degenerate Riemann surface,
203: where all of the tubes are very long and skinny.
204: One indication of this is that when
205: asked to consider conduction in the
206: three-dimensional grid, many people think immediately of
207: a jungle gym, even though it is made of hollow tubes.
208: Thus, if we need to develop the theory of infinite networks,
209: we may hope to get the theory as a limiting case of
210: the theory of conduction on Riemann surfaces,
211: or at least to be able to develop the network theory
212: along the same lines as the Riemann surface theory.
213: It is the latter approach that we will follow here.
214: The network theory we will develop follows the Riemann surface
215: theory as described in Ahlfors and Sario's standard text
216: \cite{ahlforsSario:riemannSurfaces};
217: see also Rodin and Sario
218: \cite{rodinSario:principal}.
219: We will not duplicate any of the proofs.
220: For network terminology and the basic theory of finite networks
221: see Doyle and Snell
222: \cite{doyleSnell:carus},
223: or the beautiful and concise treatment by Thomassen
224: \cite{thomassen:infinite}.
225:
226: \section{Infinite networks.}
227:
228: Consider a connected infinite electric network,
229: that is,
230: a connected infinite graph where every edge is assigned
231: a resistance.
232: We will assume that the network is {\em locally finite},
233: that is,
234: that each vertex has finite valence.
235: Multiple edges connecting two distinct vertices are allowed.
236: Any graph can be treated as
237: a network by assigning each of its edges a resistance of $1$ ohm.
238:
239: We will often want to approximate an infinite network
240: by a sequence of larger and larger finite networks.
241: Given a finite subset $S$ of the vertices of a network,
242: define $\cutnetwork (S)$ to be the network gotten by throwing
243: away all vertices that lie outside of $S$,
244: along with all edges that are incident with the discarded vertices.
245: Define $\shortnetwork (S)$ to be the network gotten by
246: identifying all the vertices outside of $S$ to yield a single
247: new vertex $\infty$.
248: Call a sequence
249: each member of which is a finite subset $T$ of the vertices
250: a {\em swelling sequence}
251: if any finite subset $S$ of the vertices is eventually
252: contained in $T$.
253: We will be particularly interested in properties of the infinite network
254: that can be expressed as limits for swelling $T$ of properties of
255: $\cutnetwork (T)$ or $\shortnetwork (T)$.
256:
257: \section{The even and odd flows.}
258:
259: Given an infinite network,
260: and any flow through the edges of the network,
261: we can compute the source strength
262: at each of the vertices of the network,
263: just as we would for a finite network.
264: Conversely, given a source distribution,
265: we would like to associate a flow to it.
266: We will assume that there are only a finite number of
267: vertices where the source strength is non-zero.
268: For the moment, we will also assume that
269: the source distribution is {\em balanced},
270: that is,
271: that the total source strength is zero.
272:
273: Given a balanced finite source distribution,
274: there are two canonical associated flows,
275: which we will call the {\em even flow} and the {\em odd flow}.
276: The even flow corresponds intuitively to a network
277: whose boundary at infinity is insulated.
278: It is the limit of the flows you get by cutting the network
279: down to a finite subset of the vertices,
280: and then letting the finite subset swell to fill up the
281: whole infinite network.
282: The odd flow corresponds to a network that is shorted together
283: at infinity.
284: It is the limit of the flows you get by shorting together
285: all the nodes outside of a finite subset.
286:
287: The odd flow can be characterized as the unique flow of minimum energy
288: among all flows having the specified source distribution.
289: The dual characterization of the even flow
290: describes it in terms of functions having prescribed values,
291: which is not quite what we want.
292: Instead,
293: as noted by Flanders
294: \cite{flanders:infinite},
295: the even flow can be characterized
296: as the unique flow of minimum energy
297: among all flows that are limits
298: (in the energy dissipation norm)
299: of flows having the specified source distribution.
300:
301: The names even and odd come from the method of images
302: in classical electrical theory.
303: Imagine doubling the network by taking two copies of it and glueing
304: them together along their ideal boundaries at infinity---whatever
305: that might mean---just as you
306: would double the unit disk to get a sphere.
307: We will call the two sides of the double
308: the {\em bright side} and the {\em dark side}.
309: The even flow corresponds to extending the source distribution
310: on the bright side symmetrically to the dark side,
311: the odd flow to extending it antisymmetrically.
312:
313: \section{The even and odd resistances.}
314:
315: Given any two distinct points $p$ and $q$,
316: place a source of strength $+1$ at $p$
317: and a source of strength $-1$ at $q$.
318: Call the resulting flows the even and odd flows from $p$ to $q$.
319: Call the dissipation of the even flow the even resistance
320: between $p$ and $q$,
321: and similarly for the odd resistance.
322:
323: The odd flow from $p$ to $q$ is the unique unit flow from $p$ to $q$
324: of minimum energy,
325: so the odd resistance is always less than or equal to the even resistance,
326: and the two resistances are equal if and only if
327: the even and odd flows agree.
328: As before, the even flow is the limit of flows of compact support,
329: and minimizes energy among unit flows from $p$ to $q$ with this
330: property.
331: It is also characterized as the unit flow that is proportional
332: to the flow of the unique harmonic function that has minimum
333: energy among all functions taking values $1$ at p and $0$ at $q$.
334:
335: \section{Shorting and cutting.}
336:
337: For finite networks,
338: Rayleigh's cutting law states that for any two vertices $p$ and $q$,
339: cutting away part of the network can only make the resistance between
340: $p$ and $q$ bigger.
341: By the same token, shorting parts of the network together can only
342: make the resistance smaller.
343:
344: These laws continue to hold for infinite networks
345: to the fullest extent possible.
346: In particular, for any finite subset $S$ containing $p$ and $q$
347: we have
348: {\samepage
349: \begin{eqnarray*}
350: \resistance (p,q,\shortnetwork (S))
351: \leq
352: \oddresistance (p,q)
353: \\
354: \leq
355: \evenresistance (p.q)
356: \leq
357: \resistance (p,q,\cutnetwork (S))
358: .
359: \end{eqnarray*}
360: }
361:
362: \section{The ghost flow.}
363:
364: For any balanced finite source distribution,
365: the difference between the even flow and the odd flow
366: is the flow of a harmonic function,
367: and in particular is sourceless.
368: It is the flow on the bright side of the double that results from
369: placing sources of twice the specified strength
370: at the corresponding places on the dark side.
371: Let's call this flow the {\em ghost flow}.
372:
373: The ghost flow ``from $p$ to $q$'' has the following remarkable properties:
374: Its energy is the difference of the even and odd resistances,
375: so it vanishes if and only if the even and odd resistances agree.
376: Among all flows of harmonic functions $u$ of finite energy,
377: the ghost flow is the unique flow that minimizes the quantity
378: \[
379: \energy (u) - (u(p)-u(q))
380: .
381: \]
382: Thus if all finite energy harmonic functions $u$ have
383: $u(p)=u(q)$,
384: and in particular if the network has no non-constant
385: finite energy harmonic functions,
386: then the ghost flow vanishes.
387: Otherwise,
388: there is an essentially unique harmonic function of unit energy
389: that maximizes the difference
390: $u(p)-u(q)$,
391: and the ghost flow is proportional to the flow of this function.
392:
393: From all of this,
394: we conclude that the even and odd resistances between $p$ and $q$ agree
395: if and only if
396: all finite energy harmonic functions $u$ have
397: $u(p)=u(q)$,
398: and that the even and odd resistances agree
399: for all pairs of points $p$ and $q$
400: if and only if
401: there are no non-constant finite energy harmonic functions,
402: Note that to check that all of the even and odd resistances agree,
403: we need only check the cases where $p$ and $q$ are adjacent.
404:
405: \section{Networks with no non-constant finite energy harmonic functions.}
406:
407: A network having no non-constant finite energy harmonic functions
408: is said to belong to the class $\ohd$,
409: or to be $\ohd$.
410: For an $\ohd$ network all even and odd resistances agree,
411: and by superposition,
412: the even and odd flows agree for any balanced finite source distribution.
413: This property characterized $\ohd$ networks,
414: so if you want to avoid getting into arguments about how to define
415: the flow corresponding to a specified source distribution,
416: you had best stick to $\ohd$ networks.
417:
418: The class $\ohd$ fits into the hierarchy
419: \[
420: \og \subset \ohp \subset \ohb \subset \ohd
421: ,
422: \]
423: Here $P$ and $B$ stand for `positive' and `bounded,'
424: and $\og$ refers to the class of surfaces
425: having no positive Green's function.
426: A network is $\og$ if and only if it has
427: no flow out to infinity of finite energy,
428: by which we mean a flow of finite energy whose source strength
429: is non-negative everywhere and positive somewhere.
430: If a network is $\og$ we call it {\em recurrent};
431: otherwise we call it {\em transient}.
432: The terminology comes from probability theory:
433: If you carry out a random walk on the vertices
434: of the network,
435: where at each step
436: you walk along one of the edges leaving the vertex you're at,
437: with probability proportional to the conductance of the edge,
438: then on a recurrent network you're certain to return eventually to your
439: starting point,
440: but on a transient network there is a positive probability that you
441: will wander off and never return.
442:
443: The inclusions in this hierarchy imply that if you know that a network is
444: recurrent, or at least that it has no non-constant positive harmonic
445: functions,
446: then you know that it has no non-constant finite energy harmonic functions,
447: and thus that even and odd flows and resistance always agree.
448:
449: \section{Flows to infinity.}
450:
451: So far,
452: we have been dealing only with source distributions that are balanced.
453: For transient networks
454: we can relax this condition,
455: though now we can only consider the odd flow.
456: As before,
457: this flow is the limit of flows in the networks obtained by shorting
458: together nodes outside of a finite set.
459: As before,
460: it is the unique energy minimizing flow having the specified
461: source distribution.
462:
463: \section{The $d$-grid.}
464:
465: From probability theory,
466: we have the following standard facts
467: (see Spitzer
468: \cite{spitzer:principles},
469: and also Avez
470: \cite{avez:ohp}):
471: For $d=1,2$, the $d$-grid is recurrent,
472: i.e. $\og$.
473: For $d \geq 3$ the $d$-grid is transient,
474: but lies in $\ohp - \og$.
475: It follows that any grid is $\ohd$.
476: (We will give an independent proof of this fact later on.)
477: Thus for the $d$-grid
478: even and odd resistances agree,
479: and there is no argument about what resistance means.
480:
481: \section{Symmetry and superposition.}
482:
483: The symmetry-and-superposition computation
484: of the effective resistance across
485: an edge of the $d$-grid depends on being able to represent the
486: unit flow from $p$ to $q$ as the superposition of symmetrical
487: flows out of $p$ and into $q$.
488: If $d \geq 3$,
489: so that the grid is transient,
490: we can write the flow from $p$ to $q$ as the superposition
491: of the odd unit flow out of $p$ and the odd unit flow into $q$,
492: and these flows must be symmetrical by uniqueness.
493: So far as the case $d \geq 3$ goes,
494: then,
495: the engineers are completely vindicated.
496:
497: \section{The 2-grid.}
498:
499: That leaves the $2$-grid.
500: Why can the flow from $p$ to $q$ be written as the superposition
501: of symmetrical flows out of $p$ and into $q$?
502: Again, the engineers' explanations vary,
503: but they clearly all believe that
504: even though the network is recurrent you can define a unique unit current
505: flow from any vertex $p$ out to infinity as a limit of the flows
506: you get as follows:
507: Choose a large finite subset of the vertices,
508: and define a source distribution
509: consisting of a source of strength $+1$ at $p$ and a finite
510: number of negative sources (i.e. sinks) at points outside of the
511: specified finite subset.
512: As the finite subset swells, the flow thus determined should
513: converge to a uniquely determined flow from $p$ to infinity.
514:
515: Let's say that a network having this property
516: has a {\em good enough Green's function}.
517: On such a network you can work with flows out to infinity
518: much as you would on a transient network,
519: where you have a bona fide Green's function.
520: In particular,
521: you can get the flow from $p$ to infinity as the limit for swelling $T$
522: of the flow from $p$ to $\infty$ in $\shortnetwork (T)$,
523: or as the limit as $q$ marches off to infinity
524: of the flow from $p$ to $q$ in the infinite graph.
525:
526: Note that the $1$-grid does not have a good enough Green's function:
527: If you short together the nodes outside a finite set,
528: and let the set go to infinity in a lopsided way,
529: you get a lopsided limit,
530: if indeed you get a limit at all.
531: Similarly, the flow from $p$ to $q$ changes abruptly when
532: $q$ jumps from one side of $p$ to the other.
533:
534: For the $2$-grid you can actually write down the Green's function,
535: and then make explicit estimates to show that it is good enough.
536: This was the method used by Flanders
537: \cite{flanders:infiniteGrid}
538: to demonstrate
539: that it is valid to apply the symmetry-and-superposition method
540: to the $2$-grid.
541: This is all right as far as it goes,
542: but it would be nice to have a more conceptual proof,
543: based on the analogy between the grid and the plane.
544: After all, the reason everyone believes that the $2$-grid has
545: a good enough Green's function is that the grid looks just like the plane,
546: and the plane has a good enough Green's function.
547: If we can somehow make sense of this argument,
548: then without further ado we ought to be able to carry over
549: our results for the $2$-grid
550: to the symmetrical graphs shown in Figure 2.
551: \ldots
552:
553: \fig{lookalikes}{Symmetrical graphs that look just like the plane.}
554:
555: \section{Foster's method.}
556:
557: We turn now to Foster's method of computing
558: the resistance across an edge of the $d$-grid.
559: Thomassen
560: \cite{thomassen:infinite}
561: shows how to make rigorous the application of
562: Foster's theorem for finite networks to smallish symmetrical infinite
563: networks like the $d$-grid.
564: We will review Thomassen's analysis,
565: and extend his results in the light of the network theory
566: we have developed.
567:
568: Given a finite subset $S$ of the vertices of a graph $G$,
569: define $\boundary (S)$ to consist of those vertices of $S$ that
570: are connected by an edge to a vertex outside $S$.
571: Define $\edgeboundary (S)$
572: to consist of all edges connecting a vertex in $S$ to a vertex
573: outside $S$.
574: Call a swelling sequence along which
575: \[
576: {{|\edgeboundary (T)|}\over{|T|}}
577: \]
578: goes to $0$ a {\em pinching sequence}.
579: Call a graph {\em smallish} if it has
580: a pinching sequence.
581: We will only be considering graphs of bounded valence;
582: such a graph is smallish
583: if and only if it has a swelling sequence
584: along which
585: \[
586: {{|\boundary (T)|}\over{|T|}}
587: \]
588: goes to $0$.
589:
590: Say that two vertices $v, v'$ are {\em of the same kind}
591: if $v$ can be mapped to $v'$ by a symmetry of the graph.
592: Similarly for edges.
593: Call a graph {\em symmetrical}
594: if it has only a finite number of different kinds of vertices
595: (or edges---it makes no difference).
596:
597: In a smallish symmetrical graph,
598: the vertices have well-defined relative frequencies,
599: which are positive rational numbers summing to $1$.
600: Similarly for edges.
601: These relative frequencies are the limits of the actual
602: relative frequencies
603: in $\cutgraph (T)$ or $\shortgraph (T)$,
604: where $T$ runs out along any pinching sequence.
605: Define the average of any function that depends only on the
606: kind of a vertex (or edge)
607: with respect to these relative frequencies.
608: This average will be the limit of the average in
609: $\cutgraph (T)$ or $\shortgraph (T)$ as $T$ runs out along
610: any pinching sequence.
611:
612: {\bf Theorem.}
613: In a smallish symmetrical graph,
614: \begin{eqnarray*}
615: \averageevenresistance (G)
616: =
617: \averageoddresistance (G)
618: \\
619: =
620: {2 \over {\averagevalence (G)}}
621: .
622: \end{eqnarray*}
623:
624: {\bf Proof.}
625: For any finite set of vertices $T$,
626: Rayleigh's cutting law implies that
627: the even resistance in the infinite graph
628: across any edge that remains in $\cutgraph (T)$ is
629: less than or equal to the (plain old) resistance
630: across that edge in $\cutgraph (T)$.
631: But by Foster's theorem,
632: when $|T|$ is large the average resistance
633: across the edges of $\cutgraph (T)$
634: is very nearly
635: \[
636: {2 \over {\averagevalence ( \cutgraph (T) )}}
637: .
638: \]
639: Taking limits along a pinching sequence,
640: we find that
641: \[
642: \averageevenresistance (G) \leq 2 / \averagevalence (G)
643: .
644: \]
645: An analogous shorting argument yields
646: \[
647: \averageoddresistance (G) \geq 2 / \averagevalence (G)
648: .
649: \]
650: But even resistances are always at least as big as odd resistances,
651: so
652: \[
653: \averageevenresistance (G) \geq \averageoddresistance (G)
654: .
655: \]
656: This closes up the circle of inequalities,
657: and the theorem follows.
658: $\qed$
659:
660: {\bf Corollary.}
661: In a smallish symmetrical graph $G$,
662: all even and odd resistances agree,
663: and $G$ is $\ohd$.
664: $\qed$
665:
666: {\bf Note.}
667: Actually,
668: we expect that a smallish symmetrical graph is $\ohp$.
669: But Geoff Mess tells me that this isn't true
670: \ldots
671:
672: {\bf Corollary.}
673: In a smallish vertex-transitive graph $G$,
674: the average of the resistances across the edges emanating from any given vertex
675: is
676: \[
677: 2 / \valence (G)
678: . \dispqed
679: \]
680:
681: In an edge-transitive graph, there can be either one type of vertex
682: or two types.
683: If there are two types,
684: their relative frequencies are inversely proportional to their valences,
685: which we denote by $\valenceone (G)$ and $\valencetwo (G)$.
686:
687: {\bf Corollary.}
688: Let $G$ be a smallish edge-transitive graph.
689: If $G$ has one type of vertex,
690: the resistance across any edge is
691: \[
692: {2 \over {\valence (G)}}
693: .
694: \]
695: If $G$ has two types of vertices,
696: the resistance across any edge is
697: \[
698: {{\valenceone (G) + \valencetwo (G)} \over {\valenceone (G) \valencetwo (G)}}
699: . \dispqed
700: \]
701:
702: {\bf Note.}
703: This result follows from the symmetry-and-superposition argument
704: as well,
705: though we still haven't justified the use of this method if the
706: graph is recurrent.
707: We expect that a smallish symmetrical graph can only be recurrent
708: if it looks like the $2$-grid.
709: Geoff Mess tells me that this follows from the work of
710: Gromov
711: \cite{gromov:polynomial}
712: and Varopoulos
713: \cite{varopoulos:polya}.
714: \ldots
715:
716: \newpage
717:
718: \bibliography{net}
719: \bibliographystyle{plain}
720:
721: \end{document}
722: