A Simple Test on 2-Vertex- and

2-Edge-Connectivity
Jens M. Schmidt
MPI fr Informatik, Saarbrcken
(jens.schmidt@mpi-inf.mpg.de)
Abstract
Testing a graph on 2-vertex- and 2-edge-connectivity are two funda-
mental algorithmic graph problems. For both problems, dierent linear-
time algorithms with simple implementations are known. Here, an even
simpler linear-time algorithm is presented that computes a structure from
which both the 2-vertex- and 2-edge-connectivity of a graph can be easily
read o. The algorithm computes all bridges and cut vertices of the
input graph in the same time.
1 Introduction
Testing a graph on 2-connectivity (i. e., 2-vertex-connectivity) and on 2-edge-
connectivity are fundamental algorithmic graph problems. Tarjan presented the
rst linear-time algorithms for these problems, respectively [11, 12]. Since then,
many linear-time algorithms have been given (e. g., [2, 3, 4, 5, 6, 13, 14, 15])
that compute structures which inherently characterize either the 2- or 2-edge-
connectivity of a graph. Examples include open ear decompositions [8, 16], block-
cut trees [7], bipolar orientations [2] and s-t-numberings [2] (all of which can be
used to determine 2-connectivity) and ear decompositions [8] (the existence of
which determines 2-edge-connectivity).
Most of the mentioned algorithms use a depth-rst search-tree (DFS-tree)
and compute so-called low-point values, which are dened in terms of a DFS-tree
(see [11] for a denition of low-points). This is a concept Tarjan introduced in his
rst algorithms and that has been applied successfully to many graph problems
later on. However, low-points do not always provide the most natural solution:
Brandes [2] and Gabow [6] gave considerably simpler algorithms for computing
most of the above-mentioned structures (and testing 2-connectivity) by using
simple path-generating rules instead of low-points; they call these algorithms
path-based.
The aim of this paper is a self-contained exposition of an even simpler linear-
time algorithm that tests both the 2- and 2-edge-connectivity of a graph. It is
suitable for teaching in introductory courses on algorithms. While Tarjans
two algorithms are currently the most popular ones used for teaching (see [6]
for a list of 11 text books in which they appear), in my teaching experience,
undergraduate students have diculties with the details of using low-points.
1
The algorithm presented here uses a very natural path-based approach in-
stead of low-points; similar approaches have been presented by Ramachan-
dran [10] and Tsin [14] in the context of parallel and distributed algorithms,
respectively. The approach is related to ear decompositions; in fact, it computes
an (open) ear decomposition if the input graph has appropriate connectivity.
Notation. We use standard graph-theoretic terminology from [1]. Let (G)
be the minimum degree of a graph G. A cut vertex is a vertex in a connected
graph that disconnects the graph upon deletion. Similarly, a bridge is an edge
in a connected graph that disconnects the graph upon deletion. A graph is
2-connected if it is connected and contains at least 3 vertices, but no cut ver-
tex. A graph is 2-edge-connected if it is connected and contains at least 2
vertices, but no bridge. Note that for very small graphs, dierent denitions of
(edge)connectivity are used in literature; here, we chose the common denition
that ensures consistency with Mengers Theorem [9]. It is easy to see that every
2-connected graph is 2-edge-connected, as otherwise any bridge in this graph on
at least 3 vertices would have an end point that is a cut vertex.
2 Decomposition into Chains
We will decompose the input graph into a set of paths and cycles, each of which
will be called a chain. Some easy-to-check properties on these chains will then
characterize both the 2- and 2-edge-connectivity of the graph. Let G = (V, E) be
the input graph and assume for convenience that G is simple and that |V | 3.
We rst perform a depth-rst search on G. This implicitly checks G on being
connected. If G is connected, we get a DFS-tree T that is rooted on a vertex r;
otherwise, we stop, as G is neither 2- nor 2-edge-connected. The DFS assigns
a depth-rst index (DFI) to every vertex. We assume that all tree edges (i. e.,
edges in T) are oriented towards r and all backedges (i. e., edges that are in G
but not in T) are oriented away from r. Thus, every backedge lies in exactly
one directed cycle C(e). Let every vertex be marked as unvisited.
We now decompose G into chains by applying the following procedure for
each vertex v in ascending DFI-order: For every backedge e that starts at v,
we traverse C(e), beginning with v, and stop at the rst vertex that is marked
as visited. During such a traversal, every traversed vertex is marked as visited.
Thus, a traversal stops at the latest at v and forms either a directed path or
cycle, beginning with v; we call this path or cycle a chain and identify it with
the list of vertices and edges in the order in which they were visited. The ith
chain found by this procedure is referred to as C
i
.
The chain C
1
, if exists, is a cycle, as every vertex is unvisited at the beginning
(note C
1
does not have to contain r). There are |E| |V | + 1 chains, as every
of the |E| |V | + 1 backedges creates exactly one chain. We call the set C =
{C
1
, . . . , C
|E||V |+1
} a chain decomposition; see Figure 1 for an example.
Clearly, a chain decomposition can be computed in linear time. This almost
concludes the algorithmic part; we now state easy-to-check conditions on C
that characterize 2- and 2-edge-connectivity. All proofs will be given in the
next section.
Theorem 1. Let C be a chain decomposition of a simple connected graph G.
Then G is 2-edge-connected if and only if the chains in C partition E.
2
(a) An input graph G.
v
10
v
4
v
1
v
6
v
5
v
9
v
3
v
2
v
7
v
8
(b) A DFS-tree of G (de-
picted with straight-lines)
and the edge-orientation it
imposes. There are |E|
|V | + 1 = 5 backedges.
&
&
&
&
&
v
10
v
4
v
1
v
6
v
5
v
9
v
3
v
2
v
7
v
8
(c) A chain decomposition
C = {C
1
, . . . , C
5
} of G.
The chains C
2
and C
3
are
paths; all other chains are
cycles. The edge v
6
v
5
is
not contained in any chain
and therefore a bridge.
Since (G) 2 and C \ C
1
contains a cycle, G con-
tains a cut vertex (in fact,
v
5
and v
6
are cut vertices).
Figure 1: A graph G, its DFS-tree and a chain decomposition of G.
Theorem 2. Let C be a chain decomposition of a simple 2-edge-connected graph
G. Then G is 2-connected if and only if C
1
is the only cycle in C.
The properties in Theorems 1 and 2 can be eciently tested: In order to
check whether C partitions E, we mark every edge that is traversed by the chain
decomposition. If G is 2-edge-connected, every C
i
can be checked on forming
a cycle by comparing its rst and last vertex on identity. For pseudo-code, see
Algorithm 1.
Algorithm 1 Check(graph G) G is simple and connected with |V | 3
1: Compute a DFS-tree T of G
2: Compute a chain decomposition C; mark every visited edge
3: if G contains an unvisited edge then
4: output not 2-edge-connected
5: else if there is a cycle in C dierent from C
1
then
6: output 2-edge-connected but not 2-connected
7: else
8: output 2-connected
We state a variant of Theorem 2, which does not rely on edge-connectivity.
Its proof is very similar to the one of Theorem 2.
Theorem 3. Let C be a chain decomposition of a simple connected graph G.
Then G is 2-connected if and only if (G) 2 and C
1
is the only cycle in C.
3
3 Proofs
It remains to give the proofs of Theorems 1 and 2. For a tree T rooted at r
and a vertex x in T, let T(x) be the subtree of T that consists of x and all
descendants of x (independent of the edge orientations of T). Theorem 1 is
immediately implied by the following lemma.
Lemma 4. Let C be a chain decomposition of a simple connected graph G. An
edge e in G is a bridge if and only if e is not contained in any chain in C.
Proof. Let e be a bridge and assume to the contrary that e is contained in a
chain whose rst edge (i. e., whose backedge) is b. The bridge e is not contained
in any cycle of G, as otherwise the end points of e would still be connected when
deleting e, contradicting that e is a bridge. This contradicts the fact that e is
contained in the cycle C(b).
Let e be an edge that is not contained in any chain in C. Let T be the
DFS-tree that was used for computing C and let x be the end point of e that is
farthest away from the root r of T, in particular x = r. Then e is a tree-edge,
as otherwise e would be contained in a chain. For the same reason, there is no
backedge with exactly one end point in T(x). Deleting e therefore disconnects
all vertices in T(x) from r. Hence, e is a bridge.
The following lemma implies Theorem 2, as every 2-edge-connected graph
has minimum degree 2.
Lemma 5. Let C be a chain decomposition of a simple connected graph G with
(G) 2. A vertex v in G is a cut vertex if and only if v is incident to a bridge
or v is the rst vertex of a cycle in C \ C
1
.
Proof. Let v be a cut vertex in G; we may assume that v is not incident to a
bridge. Let X and Y be connected components of G\v. Then X and Y have to
contain at least two neighbors of v in G, respectively. Let X
+v
and Y
+v
denote
the subgraphs of G that are induced by X v and Y v, respectively. Both
X
+v
and Y
+v
contain a cycle through v, as both X and Y are connected. It
follows that C
1
exists; assume w. l. o. g. that C
1
/ X
+v
. Then there is at least
one backedge in X
+v
that starts at v. When the rst such backedge is traversed
in the chain decomposition, every vertex in X is still unvisited. The traversal
therefore closes a cycle that starts at v and is dierent from C
1
, as C
1
/ X
+v
.
If v is incident to a bridge, (G) 2 implies that v is a cut vertex. Let v
be the rst vertex of a cycle C
i
= C
1
in C. If v is the root r of the DFS-tree T
that was used for computing C, both cycles C
1
and C
i
end at v. Thus, v has at
least two children in T and v must be a cut vertex. Otherwise v = r; let wv be
the last edge in C
i
. Then no backedge starts at a vertex with smaller DFI than
v and ends at a vertex in T(w), as otherwise vw would not be contained in C
i
.
Thus, deleting v separates r from all vertices in T(w) and v is a cut vertex.
4 Extensions
We state how some additional structures can be computed from a chain decom-
position. Note that Lemmas 4 and 5 can be used to compute all bridges and cut
vertices of G in linear time. Having these, the 2-connected components (i. e., the
4
maximal 2-connected subgraphs) of G and the 2-edge-connected components
(i. e., the maximal 2-edge-connected subgraphs) of G can be easily obtained.
This gives the so-called block-cut tree [7] of G, which represents the dependence
of the 2-connected components and cut vertices in G in a tree (it gives also the
corresponding tree representing the 2-edge-connected components and bridges
of G).
Additionally, the set of chains C computed by our algorithm is an ear de-
composition if G is 2-edge-connected and an open ear decomposition if G is
2-connected. Note that C is not an arbitrary (open) ear decomposition, as it
depends on the DFS-tree. The existence of these ear decompositions character-
ize the 2-(edge-)connectivity of a graph [8, 16]; Brandes [2] gives a simple linear-
time transformation that computes a bipolar orientation and a s-t-numbering
from such an open ear decomposition.
References
[1] J. A. Bondy and U. S. R. Murty. Graph Theory. Springer, 2008.
[2] U. Brandes. Eager st-Ordering. In Proceedings of the 10th European Symposium
of Algorithms (ESA02), pages 247256, 2002.
[3] J. Ebert. st-Ordering the vertices of biconnected graphs. Computing, 30:1933,
1983.
[4] S. Even and R. E. Tarjan. Computing an st-Numbering. Theor. Comput. Sci.,
2(3):339344, 1976.
[5] S. Even and R. E. Tarjan. Corrigendum: Computing an st-Numbering (TCS
2(1976):339-344). Theor. Comput. Sci., 4(1):123, 1977.
[6] H. N. Gabow. Path-based depth-rst search for strong and biconnected compo-
nents. Inf. Process. Lett., 74(3-4):107114, 2000.
[7] F. Harary and G. Prins. The block-cutpoint-tree of a graph. Publ. Math. Debrecen,
13:103107, 1966.
[8] L. Lovsz. Computing ears and branchings in parallel. In Proceedings of the
26th Annual Symposium on Foundations of Computer Science (FOCS85), pages
464467, 1985.
[9] K. Menger. Zur allgemeinen Kurventheorie. Fund. Math., 10:96115, 1927.
[10] V. Ramachandran. Parallel open ear decomposition with applications to graph
biconnectivity and triconnectivity. In Synthesis of Paral lel Algorithms, pages
275340, 1993.
[11] R. E. Tarjan. Depth-rst search and linear graph algorithms. SIAM Journal on
Computing, 1(2):146160, 1972.
[12] R. E. Tarjan. A note on nding the bridges of a graph. Information Processing
Letters, 2(6):160161, 1974.
[13] R. E. Tarjan. Two streamlined depth-rst search algorithms. Fund. Inf., 9:8594,
1986.
[14] Y. H. Tsin. On nding an ear decomposition of an undirected graph distributively.
Inf. Process. Lett., 91:147153, 2004.
[15] Y. H. Tsin and F. Y. Chin. A general program scheme for nding bridges. Infor-
mation Processing Letters, 17(5):269272, 1983.
[16] H. Whitney. Non-separable and planar graphs. Trans. Amer. Math. Soc.,
34(1):339362, 1932.
5
A Appendix
We omitted the proof of Theorem 3, as it is very similar to the one of Theorem 2. For
completeness, we give the proof here.
Proof of Theorem 3: Let T be the DFS-tree that was used for computing C and let
r be its root. First, let G be 2-connected. Clearly, this implies (G) 2. Moreover,
r has exactly one child, as otherwise r would be a cut vertex. Thus, r is incident to
a backedge, which implies that C
1
exists and is a cycle that starts at r. Assume to
the contrary that v is the rst vertex of a cycle C
i
= C
1
. If v = r, both cycles C
1
and C
i
end at v. Thus, v has at least two children in T. This implies that v is a cut
vertex, which contradicts the 2-connectivity of G. If v = r, let wv be the last edge in
C
i
. There is no backedge that starts at a vertex with smaller DFI than v and ends at
a vertex in T(w), as otherwise wv would be contained in a chain C
j
with j < i. Thus,
deleting v disconnects r from all vertices in T(w), which contradicts the 2-connectivity
of G.
Let (G) 2 and C
1
be the only cycle in C and assume to the contrary that G
is not 2-connected. Then G contains a cut vertex v, as (G) 2 implies |V | 3.
Clearly, C
1
can intersect with at most one connected component of G \ v. Let X be
a connected component of G \ v that does not contain any vertex of C
1
. Let X
+v
be the subgraph of G that is induced by X v. There must be a cycle in X
+v
, as
otherwise X
+v
would be a tree, whose leafs would contradict (G) 2. Hence, X
+v
contains at least one backedge; let b be the rst backedge in X
+v
that is traversed by
the chain decomposition. As r / X, all vertices in D(b) except the start point w of b
have greater DFIs than w. Thus, the traversal on b computes a chain C
i
X
+v
that
is a cycle and that is distinct from C
1
, as X does not contain any vertex of C
1
.
6