View source for Line graph

{{Short description|Graph representing edges of another graph}}
{{about|the mathematical concept|the statistical presentations method|Line chart}}
{{distinguish|Path graph}}

In the [[mathematics|mathematical]] discipline of [[graph theory]], the '''line graph''' of an [[undirected graph]] {{mvar|G}} is another graph  {{math|L(''G'')}} that represents the adjacencies between [[edge (graph theory)|edges]] of {{mvar|G}}.  {{math|L(''G'')}} is constructed in the following way: for each edge in {{mvar|G}}, make a vertex in  {{math|L(''G'')}}; for every two edges in {{mvar|G}} that have a vertex in common, make an edge between their corresponding vertices in  {{math|L(''G'')}}.

The name ''line graph'' comes from a paper by {{harvtxt|Harary|Norman|1960}} although both {{harvtxt|Whitney|1932}} and {{harvtxt|Krausz|1943}} used the construction before this.{{sfnp|Hemminger|Beineke|1978|p=273}} Other terms used for the line graph include the '''covering graph''', the '''derivative''', the '''edge-to-vertex dual''', the '''conjugate''', the '''representative graph''', and the '''θ-obrazom''',{{sfnp|Hemminger|Beineke|1978|p=273}} as well as the '''edge graph''', the '''interchange graph''', the '''adjoint graph''', and the '''derived graph'''.<ref name="h72-71">{{harvtxt|Harary|1972}}, p. 71.</ref>

{{harvs|authorlink=Hassler Whitney|first=Hassler|last=Whitney|year=1932|txt}} proved that with one exceptional case the structure of a [[connected graph]] {{mvar|G}} can be recovered completely from its line graph.<ref name="whitney"/> Many other properties of line graphs follow by translating the properties of the underlying graph from vertices into edges, and by Whitney's theorem the same translation can also be done in the other direction. Line graphs are [[claw-free graph|claw-free]], and the line graphs of [[bipartite graph]]s are [[perfect graph|perfect]]. Line graphs are characterized by nine [[forbidden graph characterization|forbidden subgraphs]] and can be recognized in [[linear time]].

Various extensions of the concept of a line graph have been studied, including line graphs of line graphs, line graphs of multigraphs, [[Line graph of a hypergraph|line graphs of hypergraphs]], and line graphs of weighted graphs.

==Formal definition==
Given a graph {{mvar|G}}, its line graph {{math|''L''(''G'')}} is a graph such that
* each [[vertex (graph theory)|vertex]] of {{math|''L''(''G'')}} represents an edge of {{mvar|G}}; and
* two vertices of {{math|''L''(''G'')}} are [[Adjacent (graph theory)|adjacent]] if and only if their corresponding edges share a common endpoint ("are incident") in {{mvar|G}}.
That is, it is the [[intersection graph]] of the edges of {{mvar|G}}, representing each edge by the set of its two endpoints.<ref name="h72-71"/>

== Example ==

The following figures show a graph (left, with blue vertices) and its line graph (right, with green vertices). Each vertex of the line graph is shown labeled with the pair of endpoints of the corresponding edge in the original graph. For instance, the green vertex on the right labeled 1,3 corresponds to the edge on the left between the blue vertices 1 and 3. Green vertex 1,3 is adjacent to three other green vertices: 1,4 and 1,2 (corresponding to edges sharing the endpoint 1 in the blue graph) and 4,3 (corresponding to an edge sharing the endpoint 3 in the blue graph).

<gallery widths="140px" heights="140px" class="skin-invert-image">
File:Line graph construction 1.svg|Graph ''G''
File:Line graph construction 2.svg|Vertices in L(''G'') constructed from edges in ''G''
File:Line graph construction 3.svg|Added edges in L(''G'')
File:Line graph construction 4.svg|The line graph L(''G'')
</gallery>

== Properties ==

===Translated properties of the underlying graph===
Properties of a graph {{mvar|G}} that depend only on adjacency between edges may be translated into equivalent properties in {{math|''L''(''G'')}} that depend on adjacency between vertices. For instance, a [[Matching (graph theory)|matching]] in {{mvar|G}} is a set of edges no two of which are adjacent, and corresponds to a set of vertices in {{math|''L''(''G'')}} no two of which are adjacent, that is, an [[Independent set (graph theory)|independent set]].<ref name="paschos">{{citation|title=Combinatorial Optimization and Theoretical Computer Science: Interfaces and Perspectives|first=Vangelis Th.|last=Paschos|publisher=John Wiley & Sons|year=2010|isbn=978-0-470-39367-3|page=394|url=https://books.google.com/books?id=bOMH9fPOicgC&pg=PA394|quotation=Clearly, there is a one-to-one correspondence between the matchings of a graph and the independent sets of its line graph.}}</ref>

Thus,
* The line graph of a [[connected graph]] is connected. If {{mvar|G}} is connected, it contains a [[path (graph theory)|path]] connecting any two of its edges, which translates into a path in {{math|''L''(''G'')}} containing any two of the vertices of {{math|''L''(''G'')}}. However, a graph {{mvar|G}} that has some isolated vertices, and is therefore disconnected, may nevertheless have a connected line graph.<ref>The need to consider isolated vertices when considering the connectivity of line graphs is pointed out by {{harvtxt|Cvetković|Rowlinson|Simić|2004}}, [https://books.google.com/books?id=FA8SObZcbs4C&pg=PA32 p.&nbsp;32].</ref>
* A line graph has an [[articulation point]] if and only if the underlying graph has a [[bridge (graph theory)|bridge]] for which neither endpoint has degree one.<ref name="h72-71"/>
* For a graph {{mvar|G}} with {{mvar|n}} vertices and {{mvar|m}} edges, the number of vertices of the line graph {{math|''L''(''G'')}} is {{mvar|m}}, and the number of edges of {{math|''L''(''G'')}} is half the sum of the squares of the [[degree (graph theory)|degrees]] of the vertices in {{mvar|G}}, minus {{mvar|m}}.<ref>{{harvtxt|Harary|1972}}, Theorem 8.1, p.&nbsp;72.</ref>
* An [[Independent set (graph theory)|independent set]]  in {{math|''L''(''G'')}} corresponds to a [[Matching (graph theory)|matching]] in {{mvar|G}}. In particular, a [[maximum independent set problem|maximum independent set]] in {{math|''L''(''G'')}} corresponds to [[maximum matching]] in {{mvar|G}}. Since maximum matchings may be found in polynomial time, so may the maximum independent sets of line graphs, despite the hardness of the maximum independent set problem for more general families of graphs.<ref name="paschos"/> Similarly, a [[rainbow-independent set]] in {{math|''L''(''G'')}} corresponds to a [[rainbow matching]] in {{mvar|G}}.
* The [[edge chromatic number]] of a graph {{mvar|G}} is equal to the [[vertex chromatic number]] of its line graph {{math|''L''(''G'')}}.<ref>{{citation|title=Graph Theory|volume=173|series=Graduate Texts in Mathematics|first=Reinhard|last=Diestel|publisher=Springer|year=2006|isbn=978-3-540-26183-4|page=112|url=https://books.google.com/books?id=aR2TMYQr2CMC&pg=PA112}}. Also in [http://diestel-graph-theory.com/basic.html free online edition], Chapter 5 ("Colouring"), p.&nbsp;118.</ref>
* The line graph of an [[edge-transitive graph]] is [[vertex-transitive graph|vertex-transitive]]. This property can be used to generate families of graphs that (like the [[Petersen graph]]) are vertex-transitive but are not [[Cayley graph]]s: if {{mvar|G}} is an edge-transitive graph that has at least five vertices, is not bipartite, and has odd vertex degrees, then {{math|''L''(''G'')}} is a vertex-transitive non-Cayley graph.<ref>{{citation
 | last1 = Lauri | first1 = Josef
 | last2 = Scapellato | first2 = Raffaele
 | isbn = 0-521-82151-7
 | location = Cambridge
 | mr = 1971819
 | page = 44
 | publisher = Cambridge University Press
 | series = London Mathematical Society Student Texts
 | title = Topics in graph automorphisms and reconstruction
 | url = https://books.google.com/books?id=hsymFm0E0uIC&pg=PA44
 | volume = 54
 | year = 2003}}. Lauri and Scapellato credit this result to Mark Watkins.</ref>
* If a graph {{mvar|G}} has an [[Euler cycle]], that is, if {{mvar|G}} is connected and has an even number of edges at each vertex, then the line graph of {{mvar|G}} is [[Hamiltonian graph|Hamiltonian]]. However, not all Hamiltonian cycles in line graphs come from Euler cycles in this way; for instance, the line graph of a Hamiltonian graph {{mvar|G}} is itself Hamiltonian, regardless of whether {{mvar|G}} is also Eulerian.<ref>{{harvtxt|Harary|1972}}, Theorem 8.8, p.&nbsp;80.</ref>
* If two simple graphs are [[graph isomorphism|isomorphic]] then their line graphs are also isomorphic. The [[Line graph#Whitney isomorphism theorem|Whitney graph isomorphism theorem]] provides a converse to this for all but one pair of connected graphs.
* In the context of complex network theory, the line graph of a random network preserves many of the properties of the network such as the [[Small-world network|small-world property]] (the existence of short paths between all pairs of vertices) and the shape of its [[degree distribution]].{{sfnp|Ramezanpour|Karimipour|Mashaghi|2003}} {{harvtxt|Evans|Lambiotte|2009}} observe that any method for finding vertex clusters in a complex network can be applied to the line graph and used to cluster its edges instead.

===Whitney isomorphism theorem===
[[File:Diamond line graph.svg|thumb|The [[diamond graph]] (left) and its more-symmetric line graph (right), an exception to the strong Whitney theorem]]
If the line graphs of two [[connected graph]]s are isomorphic, then the underlying graphs are isomorphic, except in the case of the triangle graph {{math|''K''{{sub|3}}}} and the [[Claw (graph theory)|claw]] {{math|''K''{{sub|1,3}}}}, which have isomorphic line graphs but are not themselves isomorphic.<ref name="whitney">{{harvtxt|Whitney|1932}}; {{harvtxt|Krausz|1943}}; {{harvtxt|Harary|1972}}, Theorem 8.3, p.&nbsp;72. Harary gives a simplified proof of this theorem by {{harvtxt|Jung|1966}}.</ref>

As well as {{math|''K''{{sub|3}}}} and {{math|''K''{{sub|1,3}}}}, there are some other exceptional small graphs with the property that their line graph has a higher degree of symmetry than the graph itself. For instance, the [[diamond graph]] {{math|''K''{{sub|1,1,2}}}} (two triangles sharing an edge) has four [[graph automorphism]]s but its line graph {{math|''K''{{sub|1,2,2}}}} has eight. In the illustration of the diamond graph shown, rotating the graph by 90 degrees is not a symmetry of the graph, but is a symmetry of its line graph. However, all such exceptional cases have at most four vertices. A strengthened version of the Whitney isomorphism theorem states that, for connected graphs with more than four vertices, there is a one-to-one correspondence between isomorphisms of the graphs and isomorphisms of their line graphs.<ref>{{harvtxt|Jung|1966}}; {{harvtxt|Degiorgi|Simon|1995}}.</ref>

Analogues of the Whitney isomorphism theorem have been proven for the line graphs of [[multigraph]]s, but are more complicated in this case.<ref name="z97"/>

===Strongly regular and perfect line graphs===
[[File:Line perfect graph.svg|thumb|240px|A line perfect graph. The edges in each biconnected component are colored black if the component is bipartite, blue if the component is a tetrahedron, and red if the component is a book of triangles.]]
The line graph of the complete graph {{mvar|K{{sub|n}}}} is also known as the '''triangular graph''', the [[Johnson graph]] {{math|''J''(''n'', 2)}}, or the [[complement (graph theory)|complement]] of the [[Kneser graph]] {{math|''KG''{{sub|''n'',2}}}}.  Triangular graphs are characterized by their [[Spectral graph theory|spectra]], except for {{math|1=''n'' = 8}}.<ref>{{citation
 | last1 = van Dam | first1 = Edwin R.
 | last2 = Haemers | first2 = Willem H.
 | doi = 10.1016/S0024-3795(03)00483-X
 | journal = Linear Algebra and Its Applications
 | mr = 2022290
 | pages = 241–272
 | title = Which graphs are determined by their spectrum?
 | volume = 373
 | year = 2003
 | s2cid = 32070167
 | url = https://research.tilburguniversity.edu/en/publications/a9a1857e-13ce-4da7-bcca-b879f9f3215f
 | doi-access = free
 | url-access = subscription
 }}. See in particular Proposition 8, p.&nbsp;262.</ref> They may also be characterized (again with the exception of {{math|''K''{{sub|8}}}}) as the [[strongly regular graph]]s with parameters {{math|srg(''n''(''n'' − 1)/2, 2(''n'' − 2), ''n'' − 2, 4)}}.<ref>{{harvtxt|Harary|1972}}, Theorem 8.6, p.&nbsp;79. Harary credits this result to independent papers by L. C. Chang (1959) and [[Alan Hoffman (mathematician)|A. J. Hoffman]] (1960).</ref> The three strongly regular graphs with the same parameters and spectrum as {{math|''L''(''K''{{sub|8}})}} are the [[Chang graphs]], which may be obtained by [[Two-graph|graph switching]] from {{math|''L''(''K''{{sub|8}})}}.

The line graph of a [[bipartite graph]] is [[perfect graph|perfect]] (see [[Kőnig's theorem (graph theory)|Kőnig's theorem]]), but need not be bipartite as the example of the claw graph shows. The line graphs of bipartite graphs form one of the key building blocks of perfect graphs, used in the proof of the [[strong perfect graph theorem]].<ref>{{citation
 | last1 = Chudnovsky | first1 = Maria | author1-link = Maria Chudnovsky
 | last2 = Robertson | first2 = Neil | author2-link = Neil Robertson (mathematician)
 | last3 = Seymour | first3 = Paul | author3-link = Paul Seymour (mathematician)
 | last4 = Thomas | first4 = Robin | author4-link = Robin Thomas (mathematician)
 | doi = 10.4007/annals.2006.164.51
 | issue = 1
 | journal = [[Annals of Mathematics]]
 | pages = 51–229
 | title = The strong perfect graph theorem
 | url = http://annals.princeton.edu/annals/2006/164-1/p02.xhtml
 | volume = 164
 | year = 2006| arxiv = math/0212070| s2cid = 119151552
}}. See also {{citation
 | last1 = Roussel | first1 = F.
 | last2 = Rusu | first2 = I.
 | last3 = Thuillier | first3 = H.
 | doi = 10.1016/j.disc.2009.05.024
 | issue = 20
 | journal = [[Discrete Mathematics (journal)|Discrete Mathematics]]
 | mr = 2552645
 | pages = 6092–6113
 | title = The strong perfect graph conjecture: 40 years of attempts, and its resolution
 | volume = 309
 | year = 2009| s2cid = 16049392
 | doi-access = free
 }}.</ref> A special case of these graphs are the [[rook's graph]]s, line graphs of [[complete bipartite graph]]s. Like the line graphs of complete graphs, they can be characterized with one exception by their numbers of vertices, numbers of edges, and number of shared neighbors for adjacent and non-adjacent points. The one exceptional case is {{math|''L''(''K''{{sub|4,4}})}}, which shares its parameters with the [[Shrikhande graph]]. When both sides of the bipartition have the same number of vertices, these graphs are again strongly regular.<ref>{{harvtxt|Harary|1972}}, Theorem 8.7, p.&nbsp;79. Harary credits this characterization of line graphs of complete bipartite graphs to Moon and Hoffman. The case of equal numbers of vertices on both sides had previously been proven by Shrikhande.</ref> It has been shown that, except for   {{math|''C''{{sub|3}}}} ,  {{math|''C''{{sub|4}}}} , and  {{math|''C''{{sub|5}}}} , all connected strongly regular graphs can be made non-strongly regular within two line graph transformations.<ref>
{{cite arXiv
 | eprint       = 2410.16138
 | title        = Theoretical Insights into Line Graph Transformation on Graph Learning
 | last1        = Yang
 | first1       = Fan
 | last2        = Huang
 | first2       = Xingyue
 | year         = 2024
| class = cs.LG
 }}
</ref> The extension to disconnected graphs would require that the graph is not a disjoint union of  {{math|''C''{{sub|3}}}}.

More generally, a graph {{mvar|G}} is said to be a [[line perfect graph]] if {{math|''L''(''G'')}} is a [[perfect graph]]. The line perfect graphs are exactly the graphs that do not contain a [[cycle (graph theory)|simple cycle]] of odd length greater than three.<ref>{{harvtxt|Trotter|1977}}; {{harvtxt|de Werra|1978}}.</ref> Equivalently, a graph is line perfect if and only if each of its [[biconnected component]]s is either bipartite or of the form {{math|''K''{{sub|4}}}} (the tetrahedron) or {{math|''K''{{sub|1,1,''n''}}}} (a book of one or more triangles all sharing a common edge).{{sfnp|Maffray|1992}} Every line perfect graph is itself perfect.{{sfnp|Trotter|1977}}

===Other related graph families===
All line graphs are [[claw-free graph]]s, graphs without an [[induced subgraph]] in the form of a three-leaf tree.<ref name="h72-8.4"/> As with claw-free graphs more generally, every connected line graph {{math|''L''(''G'')}} with an even number of edges has a [[perfect matching]];<ref>{{Citation
 | last = Sumner | first = David P.
 | doi = 10.2307/2039666
 | mr = 0323648
 | journal = Proceedings of the American Mathematical Society
 | pages = 8–12
 | title = Graphs with 1-factors
 | volume = 42
 | year = 1974
 | issue = 1
 | publisher = American Mathematical Society
 | jstor = 2039666
}}. {{Citation
 | last = Las Vergnas | first = M. | author-link = Michel Las Vergnas
 | mr = 0412042
 | issue = 2–3–4
 | journal = Cahiers du Centre d'Études de Recherche Opérationnelle
 | pages = 257–260
 | title = A note on matchings in graphs
 | volume = 17
 | year = 1975
}}.</ref> equivalently, this means that if the underlying graph {{mvar|G}} has an even number of edges, its edges can be partitioned into two-edge paths.

The line graphs of [[tree (graph theory)|tree]]s are exactly the claw-free [[block graph]]s.<ref>{{harvtxt|Harary|1972}}, Theorem 8.5, p.&nbsp;78. Harary credits the result to [[Gary Chartrand]].</ref> These graphs have been used to solve a problem in [[extremal graph theory]], of constructing a graph with a given number of edges and vertices whose largest tree [[induced subgraph|induced as a subgraph]] is as small as possible.<ref>{{citation
 | last1 = Erdős | first1 = Paul | author1-link = Paul Erdős
 | last2 = Saks | first2 = Michael | author2-link = Michael Saks (mathematician)
 | last3 = Sós | first3 = Vera T. | author3-link = Vera T. Sós
 | doi = 10.1016/0095-8956(86)90028-6
 | issue = 1
 | journal = Journal of Combinatorial Theory, Series B
 | pages = 61–79
 | title = Maximum induced trees in graphs
 | volume = 41
 | year = 1986| doi-access = free
}}.</ref>

All [[eigenvalue]]s of the [[adjacency matrix]] {{mvar|A}} of a line graph are at least −2. The reason for this is that {{mvar|A}} can be written as <math>A = J^\mathsf{T}J - 2I</math>, where {{mvar|J}} is the signless incidence matrix of the pre-line graph and {{mvar|I}} is the identity. In particular, {{math|''A'' + 2''I''}} is the [[Gramian matrix]] of a system of vectors: all graphs with this property have been called generalized line graphs.{{sfnp|Cvetković|Rowlinson|Simić|2004}}

== Characterization and recognition ==

===Clique partition===
[[File:Line graph clique partition.svg|thumb|280px|Partition of a line graph into cliques]]
For an arbitrary graph {{mvar|G}}, and an arbitrary vertex {{mvar|v}} in {{mvar|G}}, the set of edges incident to {{mvar|v}} corresponds to a [[Clique (graph theory)|clique]] in the line graph {{math|''L''(''G'')}}. The cliques formed in this way partition the edges of {{math|''L''(''G'')}}. Each vertex of {{math|''L''(''G'')}} belongs to exactly two of them (the two cliques corresponding to the two endpoints of the corresponding edge in {{mvar|G}}).

The existence of such a partition into cliques can be used to characterize the line graphs: A graph {{mvar|L}} is the line graph of some other graph or multigraph if and only if it is possible to find a collection of cliques in {{mvar|L}} (allowing some of the cliques to be single vertices) that partition the edges of {{mvar|L}}, such that each vertex of {{mvar|L}} belongs to exactly two of the cliques.<ref name="h72-8.4">{{harvtxt|Harary|1972}}, Theorem 8.4, p.&nbsp;74, gives three equivalent characterizations of line graphs: the partition of the edges into cliques, the property of being [[Claw-free graph|claw-free]] and odd [[Diamond graph|diamond]]-free, and the nine forbidden graphs of Beineke.</ref> It is the line graph of a graph (rather than a multigraph) if this set of cliques satisfies the additional condition that no two vertices of {{mvar|L}} are both in the same two cliques. Given such a family of cliques, the underlying graph {{mvar|G}} for which {{mvar|L}} is the line graph can be recovered by making one vertex in {{mvar|G}} for each clique, and an edge in {{mvar|G}} for each vertex in {{mvar|L}} with its endpoints being the two cliques containing the vertex in {{mvar|L}}. By the strong version of Whitney's isomorphism theorem, if the underlying graph {{mvar|G}} has more than four vertices, there can be only one partition of this type.

For example, this characterization can be used to show that the following graph is not a line graph:
:[[File:LineGraphExampleA.svg|100px|class=skin-invert]]
In this example, the edges going upward, to the left, and to the right from the central degree-four vertex do not have any cliques in common. Therefore, any partition of the graph's edges into cliques would have to have at least one clique for each of these three edges, and these three cliques would all intersect in that central vertex, violating the requirement that each vertex appear in exactly two cliques. Thus, the graph shown is not a line graph.

===Forbidden subgraphs===
[[File:Forbidden line subgraphs.svg|thumb|300px|The nine minimal non-line graphs, from Beineke's forbidden-subgraph characterization of line graphs. A graph is a line graph if and only if it does not contain one of these nine graphs as an induced subgraph.]]
Another characterization of line graphs was proven in {{harvtxt|Beineke|1970}} (and reported earlier without proof by {{harvtxt|Beineke|1968}}). He showed that there are nine minimal graphs that are not line graphs, such that any graph that is not a line graph has one of these nine graphs as an [[induced subgraph]]. That is, a graph is a line graph if and only if no subset of its vertices induces one of these nine graphs. In the example above, the four topmost vertices induce a claw (that is, a [[complete bipartite graph]] {{math|''K''{{sub|1,3}}}}), shown on the top left of the illustration of forbidden subgraphs. Therefore, by Beineke's characterization, this example cannot be a line graph. For graphs with minimum degree at least 5, only the six subgraphs in the left and right columns of the figure are needed in the characterization.<ref name="mt97">{{harvtxt|Metelsky|Tyshkevich|1997}}</ref>

===Algorithms===
{{harvtxt|Roussopoulos|1973}} and {{harvtxt|Lehot|1974}} described linear time algorithms for recognizing line graphs and reconstructing their original graphs. {{harvtxt|Sysło|1982}} generalized these methods to [[directed graph]]s. {{harvtxt|Degiorgi|Simon|1995}} described an efficient data structure for maintaining a dynamic graph, subject to vertex insertions and deletions, and maintaining a representation of the input as a line graph (when it exists) in time proportional to the number of changed edges at each step.

The algorithms of {{harvtxt|Roussopoulos|1973}} and {{harvtxt|Lehot|1974}} are based on characterizations of line graphs involving odd triangles (triangles in the line graph with the property that there exists another vertex adjacent to an odd number of triangle vertices). However, the algorithm of {{harvtxt|Degiorgi|Simon|1995}} uses only Whitney's isomorphism theorem.  It is complicated by the need to recognize deletions that cause the remaining graph to become a line graph, but when specialized to the static recognition problem only insertions need to be performed, and the algorithm performs the following steps:
*Construct the input graph {{mvar|L}} by adding vertices one at a time, at each step choosing a vertex to add that is adjacent to at least one previously-added vertex. While adding vertices to {{mvar|L}}, maintain a graph {{mvar|G}} for which {{math|1=''L'' = ''L''(''G'')}}; if the algorithm ever fails to find an appropriate graph {{mvar|G}}, then the input is not a line graph and the algorithm terminates.
*When adding a vertex {{mvar|v}} to a graph {{math|''L''(''G'')}} for which {{mvar|G}} has four or fewer vertices, it might be the case that  the line graph representation is not unique. But in this case, the augmented graph is small enough that a representation of it as a line graph can be found by a [[brute force search]] in constant time.
*When adding a vertex {{mvar|v}} to a larger graph {{mvar|L}} that equals the line graph of another graph {{mvar|G}}, let {{mvar|S}} be the subgraph of {{mvar|G}} formed by the edges that correspond to the neighbors of {{mvar|v}} in {{mvar|L}}. Check that {{mvar|S}} has a [[vertex cover]] consisting of one vertex or two non-adjacent vertices. If there are two vertices in the cover, augment {{mvar|G}} by adding an edge (corresponding to {{mvar|v}}) that connects these two vertices. If there is only one vertex in the cover, then add a new vertex to {{mvar|G}}, adjacent to this vertex.
Each step either takes constant time, or involves finding a vertex cover of constant size within a graph {{mvar|S}} whose size is proportional to the number of neighbors of {{mvar|v}}. Thus, the total time for the whole algorithm is proportional to the sum of the numbers of neighbors of all vertices, which (by the [[handshaking lemma]]) is proportional to the number of input edges.

==Iterating the line graph operator==
{{harvtxt|van Rooij|Wilf|1965}} consider the sequence of graphs
:<math>G, L(G), L(L(G)), L(L(L(G))), \dots.\ </math>
They show that, when {{mvar|G}} is a finite [[connected graph]], only four behaviors are possible for this sequence:
*If {{mvar|G}} is a [[cycle graph]] then {{math|''L''(''G'')}} and each subsequent graph in this sequence are [[graph isomorphism|isomorphic]] to {{mvar|G}} itself.  These are the only connected graphs for which {{math|''L''(''G'')}} is isomorphic to {{mvar|G}}.<ref>This result is also Theorem 8.2 of {{harvtxt|Harary|1972}}.</ref>
*If {{mvar|G}} is a claw {{math|''K''{{sub|1,3}}}}, then {{math|''L''(''G'')}} and all subsequent graphs in the sequence are triangles.
*If {{mvar|G}} is a [[path graph]] then each subsequent graph in the sequence is a shorter path until eventually the sequence terminates with an [[empty graph]].
*In all remaining cases, the sizes of the graphs in this sequence eventually increase without bound.
If {{mvar|G}} is not connected, this classification applies separately to each component of {{mvar|G}}.

For connected graphs that are not paths, all sufficiently high numbers of iteration of the line graph operation produce graphs that are Hamiltonian.<ref>{{harvtxt|Harary|1972}}, Theorem 8.11, p.&nbsp;81. Harary credits this result to [[Gary Chartrand]].</ref>

== Generalizations ==

=== Medial graphs and convex polyhedra ===
{{main article|Medial graph}}
When a [[planar graph]] {{mvar|G}} has maximum [[degree (graph theory)|vertex degree]] three, its line graph is planar, and every planar embedding of {{mvar|G}} can be extended to an embedding of {{math|''L''(''G'')}}. However, there exist planar graphs with higher degree whose line graphs are nonplanar. These include, for example, the 5-star {{math|''K''{{sub|1,5}}}}, the [[gem graph]] formed by adding two non-crossing diagonals within a regular pentagon, and all [[convex polyhedron|convex polyhedra]] with a vertex of degree four or more.<ref>{{harvtxt|Sedláček|1964}}; {{harvtxt|Greenwell|Hemminger|1972}}.</ref>

An alternative construction, the [[medial graph]], coincides with the line graph for planar graphs with maximum degree three, but is always planar. It has the same vertices as the line graph, but potentially fewer edges: two vertices of the medial graph are adjacent if and only if the corresponding two edges are consecutive on some face of the planar embedding. The medial graph of the [[dual graph]] of a plane graph is the same as the medial graph of the original plane graph.<ref>{{citation
 | last = Archdeacon | first = Dan | author-link = Dan Archdeacon
 | doi = 10.1016/0012-365X(92)90328-D
 | issue = 2
 | journal = [[Discrete Mathematics (journal)|Discrete Mathematics]]
 | mr = 1172842
 | pages = 111–141
 | title = The medial graph and voltage-current duality
 | volume = 104
 | year = 1992| doi-access = free
 }}.</ref>

For [[regular polyhedra]] or simple polyhedra, the medial graph operation can be represented geometrically by the operation of cutting off each vertex of the polyhedron by a plane through the midpoints of all its incident edges.<ref>{{citation
 | last = McKee | first = T. A.
 | contribution = Graph-theoretic model of geographic duality
 | doi = 10.1111/j.1749-6632.1989.tb22465.x
 | location = New York
 | mr = 1018637
 | pages = 310–315
 | publisher = New York Acad. Sci.
 | series = Ann. New York Acad. Sci.
 | title = Combinatorial Mathematics: Proceedings of the Third International Conference (New York, 1985)
 | volume = 555
 | year = 1989| issue = 1
 | bibcode = 1989NYASA.555..310M
 | s2cid = 86300941
 }}.</ref> This operation is known variously as the second truncation,<ref>{{citation|title=Polyhedra: A Visual Approach|first=Anthony|last=Pugh|publisher=University of California Press|year=1976|isbn=978-0-520-03056-5}}.</ref> degenerate truncation,<ref>{{citation|title=Space Structures—their Harmony and Counterpoint|first=Arthur Lee|last=Loeb|edition=5th|publisher=Birkhäuser|year=1991|isbn=978-3-7643-3588-5}}.</ref> or [[rectification (geometry)|rectification]].<ref>{{mathworld|title=Rectification|id=Rectification}}</ref>

===Total graphs===
The total graph {{math|''T''(''G'')}} of a graph {{mvar|G}} has as its vertices the elements (vertices or edges) of {{mvar|G}}, and has an edge between two elements whenever they are either incident or adjacent. The total graph may also be obtained by subdividing each edge of {{mvar|G}} and then taking the [[Graph power|square]] of the subdivided graph.<ref>{{harvtxt|Harary|1972}}, p.&nbsp;82.</ref>

===Multigraphs===
The concept of the line graph of {{mvar|G}} may naturally be extended to the case where {{mvar|G}} is a multigraph. In this case, the characterizations of these graphs can be simplified: the characterization in terms of clique partitions no longer needs to prevent two vertices from belonging to the same to cliques, and the characterization by forbidden graphs has seven forbidden graphs instead of nine.{{sfnp|Ryjáček|Vrána|2011}}

However, for multigraphs, there are larger numbers of pairs of non-isomorphic graphs that have the same line graphs. For instance a complete bipartite graph {{math|''K''{{sub|1,''n''}}}} has the same line graph as the [[dipole graph]] and [[Shannon multigraph]] with the same number of edges. Nevertheless, analogues to Whitney's isomorphism theorem can still be derived in this case.<ref name="z97">{{harvtxt|Zverovich|1997}}</ref>

===Line digraphs{{anchor|Line digraph}}===
[[File:DeBruijn-as-line-digraph.svg|thumb|360px|Construction of the [[de Bruijn graph]]s as iterated line digraphs]]
It is also possible to generalize line graphs to directed graphs.{{sfnp|Harary|Norman|1960}} If {{mvar|G}} is a directed graph, its '''directed line graph''' or '''line digraph''' has one vertex for each edge of {{mvar|G}}. Two vertices representing directed edges from {{mvar|u}} to {{mvar|v}} and from {{mvar|w}} to {{mvar|x}} in {{mvar|G}} are connected by an edge from {{mvar|uv}} to {{mvar|wx}} in the line digraph when {{math|1=''v'' = ''w''}}. That is, each edge in the line digraph of {{mvar|G}} represents a length-two directed path in {{mvar|G}}.  The [[de Bruijn graph]]s may be formed by repeating this process of forming directed line graphs, starting from a [[complete graph|complete directed graph]].{{sfnp|Zhang|Lin|1987}}

===Weighted line graphs===
In a line graph {{math|''L''(''G'')}}, each vertex of degree {{mvar|k}} in the original graph {{mvar|G}} creates {{math|''k''(''k'' − 1)/2}} edges in the line graph.  For many types of analysis this means high-degree nodes in {{mvar|G}} are over-represented in the line graph {{math|''L''(''G'')}}.  For instance, consider a [[random walk]] on the vertices of the original graph {{mvar|G}}.  This will pass along some edge {{mvar|e}} with some frequency {{mvar|f}}. On the other hand, this edge {{mvar|e}} is mapped to a unique vertex, say {{mvar|v}}, in the line graph {{math|''L''(''G'')}}.  If we now perform the same type of random walk on the vertices of the line graph, the frequency with which {{mvar|v}} is visited can be completely different from ''f''.  If our edge {{mvar|e}} in {{mvar|G}} was connected to nodes of degree {{math|''O''(''k'')}}, it will be traversed {{math|''O''(''k''{{sup|2}})}} more frequently in the line graph {{math|''L''(''G'')}}.  Put another way, the Whitney graph isomorphism theorem guarantees that the line graph almost always encodes the topology of the original graph {{mvar|G}} faithfully but it does not guarantee that dynamics on these two graphs have a simple relationship.  One solution is to construct a weighted line graph, that is, a line graph with [[Weighted graph|weighted edges]].  There are several natural ways to do this.{{sfnp|Evans|Lambiotte|2009}} For instance if edges {{mvar|d}} and {{mvar|e}} in the graph {{mvar|G}} are incident at a vertex {{mvar|v}} with degree {{mvar|k}}, then in the line graph {{math|''L''(''G'')}} the edge connecting the two vertices {{mvar|d}} and {{mvar|e}} can be given weight {{math|1/(''k'' − 1)}}.  In this way every edge in {{mvar|G}} (provided neither end is connected to a vertex of degree 1) will have strength 2 in the line graph {{math|''L''(''G'')}} corresponding to the two ends that the edge has in {{mvar|G}}. It is straightforward to extend this definition of a weighted line graph to cases where the original graph {{mvar|G}} was directed or even weighted.{{sfnp|Evans|Lambiotte|2010}}  The principle in all cases is to ensure the line graph {{math|''L''(''G'')}} reflects the dynamics as well as the topology of the original graph {{mvar|G}}.

===Line graphs of hypergraphs===
{{main article|Line graph of a hypergraph}}
The edges of a [[hypergraph]] may form an arbitrary [[family of sets]], so the [[line graph of a hypergraph]] is the same as the [[intersection graph]] of the sets from the family.

=== Disjointness graph ===
The '''disjointness graph''' of {{mvar|G}}, denoted {{math|''D''(''G'')}}, is constructed in the following way: for each edge in {{mvar|G}}, make a vertex in {{math|''D''(''G'')}}; for every two edges in {{mvar|G}} that do ''not'' have a vertex in common, make an edge between their corresponding vertices in {{math|''D''(''G'')}}.<ref>{{Cite journal|last=Meshulam|first=Roy|date=2001-01-01|title=The Clique Complex and Hypergraph Matching|journal=Combinatorica|language=en|volume=21|issue=1|pages=89–94|doi=10.1007/s004930170006|s2cid=207006642|issn=1439-6912}}</ref> In other words, {{math|''D''(''G'')}} is the [[complement graph]] of {{math|''L''(''G'')}}. A [[Clique (graph theory)|clique]] in {{math|''D''(''G'')}} corresponds to an independent set in {{math|''L''(''G'')}}, and vice versa.

==Notes==
{{reflist|30em}}

== References ==
{{refbegin|30em}}
*{{citation
 | last = Beineke | first = L. W.
 | contribution = Derived graphs of digraphs
 | editor1-last = Sachs | editor1-first = H.
 | editor2-last = Voss | editor2-first = H.-J.
 | editor3-last = Walter | editor3-first = H.-J.
 | location = Leipzig
 | pages = 17–33
 | publisher = Teubner
 | title = Beiträge zur Graphentheorie
 | year = 1968}}.
*{{citation
 | last = Beineke | first = L. W.
 | doi = 10.1016/S0021-9800(70)80019-9
 | journal = [[Journal of Combinatorial Theory]]
 | pages = 129–135
 | title = Characterizations of derived graphs
 | volume = 9
 | year = 1970
 | mr = 0262097
 | issue = 2| doi-access = free
 }}.
*{{citation
 | last1 = Cvetković | first1 = Dragoš
 | last2 = Rowlinson | first2 = Peter
 | last3 = Simić | first3 = Slobodan
 | doi = 10.1017/CBO9780511751752
 | isbn = 0-521-83663-8
 | location = Cambridge
 | mr = 2120511
 | publisher = Cambridge University Press
 | series = London Mathematical Society Lecture Note Series
 | title = Spectral generalizations of line graphs
 | volume = 314
 | year = 2004}}.
*{{citation
 | last1 = Degiorgi | first1 = Daniele Giorgio
 | last2 = Simon | first2 = Klaus
 | contribution = A dynamic algorithm for line graph recognition
 | doi = 10.1007/3-540-60618-1_64
 | location = Berlin
 | mr = 1400011
 | pages = 37–48
 | publisher = Springer
 | series = Lecture Notes in Computer Science
 | title = Graph-theoretic concepts in computer science (Aachen, 1995)
 | volume = 1017
 | year = 1995| isbn = 978-3-540-60618-5
 }}.
*{{citation
 | last1 = Evans | first1 = T.S.
 | last2 = Lambiotte | first2 = R. 
 | arxiv = 0903.2181
 |  title = Line graphs, link partitions and overlapping communities
 | journal = [[Physical Review E]]
 | year = 2009
 | volume = 80
 | issue = 1
 | article-number = 016105
 | doi = 10.1103/PhysRevE.80.016105| pmid = 19658772
 | bibcode = 2009PhRvE..80a6105E}}.
*{{citation
 | last1 = Evans | first1 = T.S.
 | last2 = Lambiotte | first2 = R. 
 | arxiv = 0912.4389
 |  title = Line Graphs of Weighted Networks for Overlapping Communities
 | journal = [[European Physical Journal B]]
 | year = 2010
 | volume = 77
 | issue = 2
 | pages = 265–272
 | doi = 10.1140/epjb/e2010-00261-8| bibcode = 2010EPJB...77..265E| s2cid = 119504507
 }}.
*{{citation
 | last1 = Greenwell | first1 = D. L.
 | last2 = Hemminger | first2 = Robert L.
 | doi = 10.1016/0012-365X(72)90058-1
 | journal = Discrete Mathematics
 | mr = 0297604
 | pages = 31–34
 | title = Forbidden subgraphs for graphs with planar line graphs
 | volume = 2
 | year = 1972| doi-access = free
 }}.
*{{citation
 | last1 = Harary | first1 = F. | author1-link = Frank Harary
 | last2 = Norman | first2 = R. Z.
 | journal = [[Rendiconti del Circolo Matematico di Palermo]]
 | pages = 161–169
 | title = Some properties of line digraphs
 | volume = 9
 | year = 1960
 | doi = 10.1007/BF02854581
 | issue = 2| hdl = 10338.dmlcz/128114
 | s2cid = 122473974 | hdl-access = free
 }}.
*{{citation|first=F.|last=Harary|author-link=Frank Harary|title=Graph Theory|publisher=Addison-Wesley|location=Massachusetts|year=1972|url=http://www.dtic.mil/dtic/tr/fulltext/u2/705364.pdf|contribution=8. Line Graphs|pages=71–83|access-date=2013-11-08|archive-date=2017-02-07|archive-url=https://web.archive.org/web/20170207091815/http://www.dtic.mil/dtic/tr/fulltext/u2/705364.pdf}}.
*{{citation
 | last1 = Hemminger | first1 = R. L.
 | last2 = Beineke | first2 = L. W.
 | contribution = Line graphs and line digraphs
 | editor1-last = Beineke | editor1-first = L. W.
 | editor2-last = Wilson | editor2-first = R. J.
 | pages = 271–305
 | publisher = Academic Press Inc.
 | title = Selected Topics in Graph Theory
 | year = 1978}}.
*{{citation
 | last = Jung | first = H. A.
 | doi = 10.1007/BF01360250
 | journal = [[Mathematische Annalen]]
 | language = German
 | mr = 0197353
 | pages = 270–271
 | title = Zu einem Isomorphiesatz von H. Whitney für Graphen
 | volume = 164
 | year = 1966| issue = 3
 | s2cid = 119898359
 }}.
*{{citation
 | last = Krausz | first = J.
 | journal = Mat. Fiz. Lapok
 | pages = 75–85
 | title = Démonstration nouvelle d'un théorème de Whitney sur les réseaux
 | volume = 50
 | year = 1943
 | mr = 0018403}}.
*{{citation
 | last = Lehot | first = Philippe G. H.
 | doi = 10.1145/321850.321853
 | journal = [[Journal of the ACM]]
 | mr = 0347690
 | pages = 569–575
 | title = An optimal algorithm to detect a line graph and output its root graph
 | volume = 21
 | year = 1974| issue = 4
 | s2cid = 15036484
 | doi-access = free
 }}.
*{{citation
 | last = Maffray | first = Frédéric
 | doi = 10.1016/0095-8956(92)90028-V
 | issue = 1
 | journal = [[Journal of Combinatorial Theory]]
 | mr = 1159851
 | pages = 1–8
 | series = Series B
 | title = Kernels in perfect line-graphs
 | volume = 55
 | year = 1992| doi-access = free
 }}.
*{{citation
 | last1 = Metelsky | first1 = Yury
 | last2 = Tyshkevich | first2 = Regina | author2-link = Regina Tyshkevich
 | doi = 10.1002/(SICI)1097-0118(199708)25:4<243::AID-JGT1>3.0.CO;2-K
 | journal = Journal of Graph Theory
 | pages = 243–251
 | title = On line graphs of linear 3-uniform hypergraphs
 | volume = 25
 | year = 1997
 | issue = 4}}.
*{{citation
 | last1 = Ramezanpour
 | first1 = A.
 | last2 = Karimipour
 | first2 = V.
 | last3 = Mashaghi
 | first3 = A.
 | journal = Phys. Rev. E
 | article-number = 046107
 | title = Generating correlated networks from uncorrelated ones
 | url = http://pre.aps.org/abstract/PRE/v67/i4/e046107
 | volume = 67
 | year = 2003
 | issue = 4
 | doi = 10.1103/physreve.67.046107
 | arxiv = cond-mat/0212469
 | bibcode = 2003PhRvE..67d6107R
 | pmid = 12786436
 | s2cid = 33054818
 }}.
*{{citation
 | last1 = van Rooij | first1 = A. C. M.
 | last2 = Wilf | first2 = H. S. | author2-link = Herbert Wilf
 | doi = 10.1007/BF01904834 | doi-access = free
 | issue = 3–4
 | journal = [[Acta Mathematica Hungarica]]
 | pages = 263–269
 | title = The interchange graph of a finite graph
 | volume = 16
 | year = 1965| hdl = 10338.dmlcz/140421
 | s2cid = 122866512
 | hdl-access = free
 }}.
*{{citation
 | last = Roussopoulos | first = N. D.
 | doi = 10.1016/0020-0190(73)90029-X
 | mr = 0424435
 | issue = 4
 | journal = [[Information Processing Letters]]
 | pages = 108–112
 | title = A max {''m'',''n''} algorithm for determining the graph ''H'' from its line graph ''G''
 | volume = 2
 | year = 1973}}.
*{{citation
 | last1 = Ryjáček | first1 = Zdeněk
 | last2 = Vrána | first2 = Petr
 | doi = 10.1002/jgt.20498
 | issue = 2
 | journal = Journal of Graph Theory
 | mr = 2778727
 | pages = 152–173
 | title = Line graphs of multigraphs and Hamilton-connectedness of claw-free graphs
 | volume = 66
 | year = 2011| s2cid = 8880045
 }}.
*{{citation
 | last = Sedláček | first = J.
 | contribution = Some properties of interchange graphs
 | mr = 0173255
 | pages = 145–150
 | publisher = Publ. House Czechoslovak Acad. Sci., Prague
 | title = Theory of Graphs and its Applications (Proc. Sympos. Smolenice, 1963)
 | year = 1964}}.
*{{citation
 | last = Sysło | first = Maciej M.
 | doi = 10.1016/0020-0190(82)90080-1
 | issue = 1
 | journal = [[Information Processing Letters]]
 | mr = 678028
 | pages = 28–30
 | title = A labeling algorithm to recognize a line digraph and output its root graph
 | volume = 15
 | year = 1982}}.
*{{citation
 | last = Trotter | first = L. E. Jr.
 | doi = 10.1007/BF01593791
 | issue = 2
 | journal = Mathematical Programming
 | mr = 0457293
 | pages = 255–259
 | title = Line perfect graphs
 | volume = 12
 | year = 1977| s2cid = 38906333
 }}.
*{{citation
 | last = de Werra | first = D.
 | doi = 10.1007/BF01609025
 | issue = 2
 | journal = Mathematical Programming
 | mr = 509968
 | pages = 236–238
 | title = On line perfect graphs
 | volume = 15
 | year = 1978| s2cid = 37062237
 | url = http://infoscience.epfl.ch/record/88504
 }}.
*{{citation
 | last = Whitney | first = H. | author-link = Hassler Whitney
 | doi = 10.2307/2371086
 | journal = American Journal of Mathematics
 | pages = 150–168
 | title = Congruent graphs and the connectivity of graphs
 | volume = 54
 | year = 1932
 | issue = 1
 | jstor = 2371086| hdl = 10338.dmlcz/101067
 | hdl-access = free
 }}.
*{{citation
 | last1 = Zhang | first1 = Fu Ji
 | last2 = Lin | first2 = Guo Ning
 | issue = 2
 | journal = Acta Math. Sinica
 | pages = 195–205
 | title = On the de Bruijn–Good graphs
 | volume = 30
 | year = 1987
 | mr = 0891925}}.
*{{citation
 | last = Зверович | first = И. Э.
 | doi = 10.4213/dm478
 | issue = 2
 | journal = Diskretnaya Matematika
 | mr = 1468075
 | pages = 98–105
 | script-title=ru:Аналог теоремы Уитни для реберных графов мультиграфов и реберные мультиграфы
 | volume = 9
 | year = 1997| language = Russian
 | doi-access = free
 }}. Translated into English as {{citation
 | last = Zverovich | first = I. È.
 | doi = 10.1515/dma.1997.7.3.287
 | issue = 3
 | journal = Discrete Mathematics and Applications
 | pages = 287–294
 | title = An analogue of the Whitney theorem for edge graphs of multigraphs, and edge multigraphs
 | volume = 7
 | year = 1997| s2cid = 120525090
 }}.
{{refend}}

== External links ==
*[http://graphclasses.org/classes/gc_249.html Line graphs], [http://graphclasses.org/index.html Information System on Graph Class Inclusions]
*{{mathworld | urlname = LineGraph | title = Line Graph}}


{{DEFAULTSORT:Line Graph}}
[[Category:Graph families]]
[[Category:Intersection classes of graphs]]
[[Category:Graph operations]]