Reconstruction conjecture
Template:Short description Template:Unsolved Informally, the reconstruction conjecture in graph theory says that graphs are determined uniquely by their subgraphs. It is due to Kelly<ref name=Kelly57>Kelly, P. J., A congruence theorem for trees, Pacific J. Math. 7 (1957), 961–968.</ref> and Ulam.<ref name=Ulam60>Ulam, S. M., A collection of mathematical problems, Wiley, New York, 1960.</ref><ref>Template:Cite journal</ref>
Formal statementsEdit
Given a graph <math>G = (V,E)</math>, a vertex-deleted subgraph of <math>G</math> is a subgraph formed by deleting exactly one vertex from <math>G</math>. By definition, it is an induced subgraph of <math>G</math>.
For a graph <math>G</math>, the deck of G, denoted <math>D(G)</math>, is the multiset of isomorphism classes of all vertex-deleted subgraphs of <math>G</math>. Each graph in <math>D(G)</math> is called a card. Two graphs that have the same deck are said to be hypomorphic.
With these definitions, the conjecture can be stated as:
- Reconstruction Conjecture: Any two hypomorphic graphs on at least three vertices are isomorphic.
- (The requirement that the graphs have at least three vertices is necessary because both graphs on two vertices have the same decks.)
Harary<ref name="Harary64">Harary, F., On the reconstruction of a graph from a collection of subgraphs. In Theory of Graphs and its Applications (Proc. Sympos. Smolenice, 1963). Publ. House Czechoslovak Acad. Sci., Prague, 1964, pp. 47–52.</ref> suggested a stronger version of the conjecture:
- Set Reconstruction Conjecture: Any two graphs on at least four vertices with the same sets of vertex-deleted subgraphs are isomorphic.
Given a graph <math>G = (V,E)</math>, an edge-deleted subgraph of <math>G</math> is a subgraph formed by deleting exactly one edge from <math>G</math>.
For a graph <math>G</math>, the edge-deck of G, denoted <math>ED(G)</math>, is the multiset of all isomorphism classes of edge-deleted subgraphs of <math>G</math>. Each graph in <math>ED(G)</math> is called an edge-card.
- Edge Reconstruction Conjecture: (Harary, 1964)<ref name="Harary64"/> Any two graphs with at least four edges and having the same edge-decks are isomorphic.
Recognizable propertiesEdit
In context of the reconstruction conjecture, a graph property is called recognizable if one can determine the property from the deck of a graph. The following properties of graphs are recognizable:
- Order of the graph – The order of a graph <math>G</math>, <math>|V(G)|</math> is recognizable from <math>D(G)</math> as the multiset <math>D(G)</math> contains each subgraph of <math>G</math> created by deleting one vertex of <math>G</math>. Hence <math>|V(G)| = |D(G)|</math> <ref name="Wall"/>
- Number of edges of the graph – The number of edges in a graph <math>G</math> with <math>n</math> vertices, <math>|E(G)|</math> is recognizable. First note that each edge of <math>G</math> occurs in <math>n-2</math> members of <math>D(G)</math>. This is true by the definition of <math>D(G)</math> which ensures that each edge is included every time that each of the vertices it is incident with is included in a member of <math>D(G)</math>, so an edge will occur in every member of <math>D(G)</math> except for the two in which its endpoints are deleted. Hence, <math>|E(G)| = \sum \frac{q_i}{n-2}</math> where <math>q_i</math> is the number of edges in the ith member of <math>D(G)</math>.<ref name="Wall"/>
- Degree sequence – The degree sequence of a graph <math>G</math> is recognizable because the degree of every vertex is recognizable. To find the degree of a vertex <math>v_i</math>—the vertex absent from the ith member of <math>D(G)</math>—, we will examine the graph created by deleting it, <math>G_i</math>. This graph contains all of the edges not incident with <math>v_i</math>, so if <math>q_i</math> is the number of edges in <math>G_i</math>, then <math>|E(G)| - q_i = \deg(v_i)</math>. If we can tell the degree of every vertex in the graph, we can tell the degree sequence of the graph.<ref name="Wall"/>
- (Vertex-)Connectivity – By definition, a graph is <math>n</math>-vertex-connected when deleting any vertex creates a <math>n-1</math>-vertex-connected graph; thus, if every card is a <math>n-1</math>-vertex-connected graph, we know the original graph was <math>n</math>-vertex-connected. We can also determine if the original graph was connected, as this is equivalent to having any two of the <math>G_i</math> being connected.<ref name="Wall"/>
- Tutte polynomial
- Characteristic polynomial
- Planarity
- The number of spanning trees in a graph
- Chromatic polynomial
- Being a perfect graph or an interval graph, or certain other subclasses of perfect graphs<ref name=rim>von Rimscha, M.: Reconstructibility and perfect graphs. Discrete Mathematics 47, 283–291 (1983)</ref>
VerificationEdit
Both the reconstruction and set reconstruction conjectures have been verified for all graphs with at most 13 vertices by Brendan McKay.<ref name=McKay97>McKay, B. D., Small graphs are reconstructible, Australas. J. Combin. 15 (1997), 123–126.</ref><ref>Template:Cite journal</ref>
In a probabilistic sense, it has been shown by Béla Bollobás that almost all graphs are reconstructible.<ref name=Bollobas90>Bollobás, B., Almost every graph has reconstruction number three, J. Graph Theory 14 (1990), 1–4.</ref> This means that the probability that a randomly chosen graph on <math>n</math> vertices is not reconstructible goes to 0 as <math>n</math> goes to infinity. In fact, it was shown that not only are almost all graphs reconstructible, but in fact that the entire deck is not necessary to reconstruct them — almost all graphs have the property that there exist three cards in their deck that uniquely determine the graph.
Reconstructible graph familiesEdit
The conjecture has been verified for a number of infinite classes of graphs (and, trivially, their complements).
- Regular graphs<ref name=h74>Template:Citation</ref> - Regular Graphs are reconstructible by direct application of some of the facts that can be recognized from the deck of a graph. Given an <math>n</math>-regular graph <math>G</math> and its deck <math>D(G)</math>, we can recognize that the deck is of a regular graph by recognizing its degree sequence. Let us now examine one member of the deck <math>D(G)</math>, <math>G_i</math>. This graph contains some number of vertices with a degree of <math>n</math> and <math>n</math> vertices with a degree of <math>n-1</math>. We can add a vertex to this graph and then connect it to the <math>n</math> vertices of degree <math>n-1</math> to create an <math>n</math>-regular graph which is isomorphic to the graph which we started with. Therefore, all regular graphs are reconstructible from their decks. A particular type of regular graph which is interesting is the complete graph.<ref name="Wall">{{#invoke:citation/CS1|citation
|CitationClass=web }}</ref>
- Trees<ref name=h74/>
- Disconnected graphs<ref name=h74/>
- Unit interval graphs <ref name=rim/>
- Separable graphs without end vertices<ref name=Bondy>Template:Cite journal</ref>
- Maximal planar graphs
- Maximal outerplanar graphs
- Outerplanar graphs
- Critical blocks
ReductionEdit
The reconstruction conjecture is true if all 2-connected graphs are reconstructible.<ref name=yang>Yang Yongzhi:The reconstruction conjecture is true if all 2-connected graphs are reconstructible. Journal of graph theory 12, 237–243 (1988)</ref>
DualityEdit
The vertex reconstruction conjecture obeys the duality that if <math>G</math> can be reconstructed from its vertex deck <math>D(G)</math>, then its complement <math>G'</math> can be reconstructed from <math>D(G')</math> as follows: Start with <math>D(G')</math>, take the complement of every card in it to get <math>D(G)</math>, use this to reconstruct <math>G</math>, then take the complement again to get <math>G'</math>.
Edge reconstruction does not obey any such duality: Indeed, for some classes of edge-reconstructible graphs it is not known if their complements are edge reconstructible.
Other structuresEdit
It has been shown that the following are not in general reconstructible:
- Digraphs: Infinite families of non-reconstructible digraphs are known, including tournaments (Stockmeyer<ref name=Stockmeyer77>Stockmeyer, P. K., The falsity of the reconstruction conjecture for tournaments, J. Graph Theory 1 (1977), 19–25.</ref>) and non-tournaments (Stockmeyer<ref name=Stockmeyer81>Stockmeyer, P. K., A census of non-reconstructable digraphs, I: six related families, J. Combin. Theory Ser. B 31 (1981), 232–239.</ref>). A tournament is reconstructible if it is not strongly connected.<ref name=HararyPalmer>Harary, F. and Palmer, E., On the problem of reconstructing a tournament from sub-tournaments, Monatsh. Math. 71 (1967), 14–23.</ref> A weaker version of the reconstruction conjecture has been conjectured for digraphs, see new digraph reconstruction conjecture.
- Hypergraphs (Kocay<ref name=Kocay87>Kocay, W. L., A family of nonreconstructible hypergraphs, J. Combin. Theory Ser. B 42 (1987), 46–63.</ref>).
- Infinite graphs. If T is the tree where every vertex has countably infinite degree, then the union of two disjoint copies of T is hypomorphic, but not isomorphic, to T.(Fisher<ref name=Fisher69>Fisher, Joshua, A counterexample to the countable version of a conjecture of Ulam, J. Combin. Theory 7 (4) (1969), 364–365.</ref>) <ref>Template:Cite journal</ref><ref>Template:Cite journal</ref>
- Locally finite graphs, which are graphs where every vertex has finite degree. The question of reconstructibility for locally finite infinite trees (the Harary-Schwenk-Scott conjecture from 1972) was a longstanding open problem until 2017, when a non-reconstructible tree of maximum degree 3 was found by Bowler et al.<ref>Bowler, N., Erde, J., Heinig, P., Lehner, F. and Pitz, M. (2017), A counterexample to the reconstruction conjecture for locally finite trees. Bull. London Math. Soc.. {{#invoke:doi|main}}</ref>
See alsoEdit
Further readingEdit
For further information on this topic, see the survey by Nash-Williams.<ref name=NashWilliams78>Nash-Williams, C. St. J. A., The Reconstruction Problem, in Selected topics in graph theory, 205–236 (1978).</ref>