Data Structures & AlgorithmsArticulation Points & Bridges

Articulation Points & Bridges

LevelAdvanced

Duration90 mins

TopicArticulation Points & Bridges

1 / 4

Understanding Articulation Points (Cut Vertices)

The Critical Nodes That Hold Networks Together

In the study of graph algorithms, we often focus on traversal, pathfinding, and optimization. But there's another class of problems that asks a fundamentally different question: What happens when part of the network fails?

Imagine a communication network connecting hundreds of cities. Data packets flow seamlessly between any two points through a web of routers and links. Now consider: Are there certain routers whose failure would completely disconnect parts of the network? Are there single points of failure that, if removed, would partition what was once a unified system into isolated fragments?

This question isn't academic—it's existential for network design, infrastructure planning, and system reliability. The vertices whose removal disconnects a graph are called articulation points (or cut vertices), and understanding them is essential for building robust systems.

What You Will Master

By the end of this page, you will deeply understand articulation points conceptually: what they are, why they exist, how to identify them through intuition, and why they represent critical vulnerabilities in any connected system. This foundation prepares you for the algorithmic techniques covered in subsequent pages.

Formal Definition of Articulation Points

Before we explore articulation points intuitively, let's establish the precise mathematical definition. Precision here prevents confusion later when we implement detection algorithms.

Definition:

Let G = (V, E) be an undirected connected graph where V is the set of vertices and E is the set of edges. A vertex v ∈ V is an articulation point (or cut vertex) if and only if removing v (and all edges incident to v) from G produces a graph that is disconnected.

Formally:

A vertex v is an articulation point if G' = (V - {v}, E') has more connected components than G, where E' contains all edges of E that do not involve v.

Connected Components Reminder

A connected component is a maximal subgraph in which any two vertices are connected by a path. In a connected graph, there is exactly one connected component (the entire graph). Removing an articulation point increases this count to two or more.

Key implications of this definition:

Only meaningful in connected graphs: If a graph is already disconnected, the concept of articulation points becomes more nuanced—we would analyze each connected component separately.
Removal includes incident edges: When we 'remove' a vertex, we also remove all edges connected to it. This is implicit in the definition but crucial for understanding.
Binary classification: Each vertex either is or isn't an articulation point—there's no gradient. However, some articulation points might be 'more critical' in practice (disconnecting larger portions).
At least two components result: By definition, removing an articulation point creates at least two disconnected components from what was previously one.

Original Connected Graph:
    
    A ─── B ─── C
          │
          D
         / \
        E   F
 
Vertex B is an articulation point because:
- Removing B disconnects A from the rest of the graph
- Result: {A} and {C, D, E, F} become separate components
 
After removing B:
    
    A       C
            │
            D
           / \
          E   F
 
Two disconnected components: {A} and {C, D, E, F}

Why Articulation Points Exist

Understanding why articulation points exist in graphs helps build intuition for detecting them. Articulation points emerge from a fundamental structural property: they serve as the only pathway between certain parts of the graph.

Consider it this way: If there's only one route to reach some portion of the graph, the vertices on that route become critical. If any of them fails, that portion becomes unreachable.

Structural Reasons for Articulation Points

•Tree-like structures: In a tree, every non-leaf vertex is an articulation point. Trees have no cycles, so removing any internal vertex disconnects its children from its parent.
•Linear chains: In a path graph A—B—C—D, vertices B and C are articulation points. The chain topology creates single points of failure.
•Hub nodes in sparse graphs: Vertices with high degree in sparse graphs often become articulation points because they're the sole connectors between regions.
•Graph 'necks': Narrow passages between densely connected regions create articulation points at the junction.

Contrast with robust graphs:

Graphs without articulation points are called 2-vertex-connected (or biconnected). In such graphs, there exist at least two vertex-disjoint paths between any pair of vertices. This redundancy means no single vertex failure can disconnect the graph.

The more 'cycle-rich' a graph is, the fewer articulation points it tends to have. Cycles create alternative paths, providing redundancy that eliminates single points of failure.

Graph Topology vs Articulation Point Prevalence
Graph Type	Articulation Points	Why
Complete graph (Kₙ)	None	Every pair of vertices is directly connected—maximum redundancy
Cycle graph (Cₙ)	None	Two paths exist between any pair—removing one vertex still leaves a path
Tree (n vertices)	n - (leaf count)	No cycles means no alternative paths
Path graph (n vertices)	n - 2	Linear structure; only endpoints aren't articulation points
Star graph (1 center + k leaves)	1 (the center)	Center connects all leaves; remove center and leaves become isolated
Grid graph	Varies by position	Interior vertices often aren't; edges of grid may have some

Visualizing Articulation Points Through Examples

Let's examine several graph examples to develop strong visual intuition for identifying articulation points. For each example, ask yourself: 'If I remove this vertex, does the graph split into disconnected pieces?'

1
2
3
4
5
6
7
8
9
10
11
12
    1 ─── 2 ─── 3
          │
          4 ─── 5
         
Analysis:
- Remove 1: Graph stays connected (2-3-4-5 still linked). NOT an articulation point.
- Remove 2: Components become {1} and {3, 4, 5}. ARTICULATION POINT.
- Remove 3: Graph stays connected (1-2-4-5 still linked). NOT an articulation point.
- Remove 4: Components become {1, 2, 3} and {5}. ARTICULATION POINT.
- Remove 5: Graph stays connected. NOT an articulation point.
 
Articulation points: {2, 4}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
        1
       / \
      2───3
      │
      4───5
      
Analysis:
- The cycle 1-2-3 has NO articulation points among 1,2,3 considered in isolation
  (removing any one still leaves a path between the other two).
- However, vertex 2 connects the cycle to the chain 4-5
- Remove 2: {1,3} and {4,5} become disconnected. ARTICULATION POINT.
- Remove 4: {1,2,3} and {5} become disconnected. ARTICULATION POINT.
 
Articulation points: {2, 4}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
    1───2       5───6
     \ /         \ /
      3───────────7
      
Analysis:
This graph has two cycles (1-2-3 and 5-6-7) connected through vertices 3 and 7.
 
- Within cycle 1-2-3: No articulation points (each vertex has a bypass).
- Within cycle 5-6-7: No articulation points.  
- But vertex 3 is the ONLY connector from {1,2} to {5,6,7}.
- And vertex 7 is the ONLY connector from {5,6} to {1,2,3}.
 
Remove 3: {1,2} disconnected from {5,6,7}. ARTICULATION POINT.
Remove 7: {5,6} disconnected from {1,2,3}. ARTICULATION POINT.
 
Articulation points: {3, 7}

The 'Pinch Point' Mental Model

Think of articulation points as 'pinch points' in a network—places where the graph narrows to a single vertex before expanding again. Like an hourglass shape, all traffic between the two bulbous ends must pass through the narrow middle.

Special Cases: Leaves, Roots, and Single-Vertex Graphs

Some vertex positions in graphs have special properties regarding articulation points. Understanding these edge cases clarifies the concept and prevents errors in algorithm implementation.

Special Cases Analysis

•Leaf vertices (degree 1): A leaf has only one neighbor. Removing a leaf never disconnects the graph—its single neighbor remains connected to everything it was before. Leaves are NEVER articulation points.
•Vertices of degree 0 (isolated): These don't exist in connected graphs by definition. In disconnected graphs, removing them doesn't change connectivity.
•The single vertex in a one-vertex graph: By convention, this is NOT considered an articulation point (removing it leaves an empty graph, not a disconnected one).
•Vertices in complete graphs (Kₙ, n ≥ 3): Since every vertex connects to every other, there's always an alternative path. Complete graphs have NO articulation points.

The degree observation:

While leaf vertices (degree 1) are never articulation points, high-degree vertices are not necessarily articulation points. Consider a complete graph: every vertex has maximum degree, yet none are articulation points because of the overwhelming redundancy.

Conversely, a low-degree vertex can be an articulation point. In a path graph 1—2—3—4—5, vertex 3 has degree 2 but is an articulation point.

The key insight: Articulation point status depends on the structure of connections, not simply the count of connections.

Common Misconception

Don't assume high-degree vertices are articulation points or that low-degree vertices aren't. A hub in a star graph (high degree) is an articulation point, but a hub in a well-connected mesh might not be. Always consider the global structure, not just local degree.

Articulation Points and Graph Connectivity

Articulation points are intimately connected to the broader concept of graph connectivity, a measure of how resilient a graph is to vertex or edge failures.

Levels of Vertex Connectivity

•1-vertex-connected (connected): The graph is connected—there exists a path between any two vertices. May or may not have articulation points.
•2-vertex-connected (biconnected): The graph remains connected after removing ANY single vertex. Has NO articulation points. Requires at least 3 vertices.
•k-vertex-connected: The graph remains connected after removing any (k-1) or fewer vertices. Higher k means more robust.

The vertex connectivity κ(G):

The vertex connectivity of a graph G, denoted κ(G), is the minimum number of vertices whose removal disconnects the graph (or reduces it to a single vertex for complete graphs).

If κ(G) = 1, the graph has at least one articulation point.
If κ(G) ≥ 2, the graph is biconnected (no articulation points).
If κ(G) = k, you need to remove at least k vertices to disconnect the graph.

Menger's Theorem connects this to path diversity: κ(G) equals the maximum number of internally vertex-disjoint paths between any pair of vertices. Higher connectivity means more redundant paths.

Connectivity Classes and Their Properties
Connectivity κ(G)	Graph Property	Articulation Points	Example
κ(G) = 0	Disconnected	N/A (already disconnected)	Two separate triangles
κ(G) = 1	Connected but vulnerable	At least one exists	Tree, path graph
κ(G) = 2	Biconnected	None	Cycle, complete graph K₃
κ(G) = k	k-connected	None (highly redundant)	Complete graph Kₙ has κ = n-1

Biconnected Components

Every graph can be decomposed into its biconnected components—maximal subgraphs without articulation points. Articulation points are the 'glue' that holds these biconnected components together. This decomposition is fundamental to understanding graph structure and is computed as a byproduct of articulation point detection.

Biconnected Components: The Building Blocks

Every undirected graph can be partitioned into biconnected components—maximal subgraphs that are themselves biconnected (containing no articulation points within them). Articulation points act as the interfaces between these components.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Original Graph:
    
    1───2      5───6
     \ /        \ /
      3──────────4
      │
      7───8
       \ /
        9
 
Biconnected Components:
  Component 1: {1, 2, 3}  (triangle - no internal articulation points)
  Component 2: {5, 6, 4}  (triangle - no internal articulation points)  
  Component 3: {3, 4}     (edge connecting two components)
  Component 4: {3, 7}     (edge connecting to lower structure)
  Component 5: {7, 8, 9}  (triangle - no internal articulation points)
 
Articulation Points: {3, 4, 7}
These vertices appear in multiple biconnected components!

Critical insight:

Articulation points are exactly the vertices that appear in more than one biconnected component. This makes sense: they're the connection points between otherwise independent structures.

The block-cut tree:

We can represent this decomposition as a tree (called the block-cut tree or BC-tree):

One node for each biconnected component (called a block)
One node for each articulation point (called a cut vertex)
An edge connects a block to a cut vertex if that cut vertex belongs to that block

This tree representation is invaluable for answering connectivity queries efficiently.

Why This Matters

Understanding biconnected components helps in designing redundant networks. Each biconnected component is internally robust—no single vertex failure within it causes disconnection. The vulnerabilities lie at the articulation points connecting these robust blocks.

Properties and Bounds on Articulation Points

How many articulation points can a graph have? Understanding the theoretical bounds helps verify algorithm correctness and provides context for graph structure.

Bounds on Articulation Point Count

•Minimum: 0 articulation points (achieved by biconnected graphs like cycles, complete graphs)
•Maximum for n vertices: n - 2 articulation points (achieved by path graphs: 1—2—3—...—n where only endpoints aren't articulation points)
•Trees with n vertices: Exactly n - (number of leaves) articulation points. Every non-leaf in a tree is an articulation point.
•Adding edges reduces count: Adding edges to a graph can only maintain or reduce the number of articulation points (never increase). Edges add redundancy.

Edge-articulation point relationship:

For a connected graph with n vertices and m edges:

If m = n - 1 (tree): Maximum potential articulation points
If m = n (one cycle added): At least one articulation point removed
If m = n(n-1)/2 (complete graph): Zero articulation points

General principle: The ratio m/n indicates redundancy. Higher edge density means more alternative paths and fewer articulation points. This is why network engineers add redundant links to eliminate single points of failure.

Articulation Point Counts for Common Graph Families
Graph Family	Vertices (n)	Edges (m)	Articulation Points
Path Pₙ	n	n-1	n-2
Cycle Cₙ	n	n	0
Complete Kₙ	n	n(n-1)/2	0
Star S₍ₙ₎	n+1	n	1 (center)
Binary tree (balanced)	2ˢ-1	2ˢ-2	2ˢ⁻¹-1 (all non-leaves)
Wheel Wₙ	n+1	2n	0 (if n≥3)

The Naive Detection Approach

Before exploring efficient algorithms, let's consider the straightforward approach to finding articulation points. This naive method, while inefficient, perfectly captures what we're trying to detect and provides a baseline for understanding the problem.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
function findArticulationPointsNaive(graph: Map<number, number[]>): number[] {
    const articulationPoints: number[] = [];
    const vertices = Array.from(graph.keys());
    
    for (const vertex of vertices) {
        // Remove this vertex and its edges
        const reducedGraph = removeVertex(graph, vertex);
        
        // Count connected components in reduced graph
        const originalComponents = countComponents(graph);
        const reducedComponents = countComponents(reducedGraph);
        
        // If removal increases component count, it's an articulation point
        // Note: originalComponents is 1 for connected graph
        // reducedComponents ≥ 2 means vertex was critical
        if (reducedComponents > 1) {
            articulationPoints.push(vertex);
        }
    }
    
    return articulationPoints;
}
 
function removeVertex(graph: Map<number, number[]>, vertex: number): Map<number, number[]> {
    const reduced = new Map<number, number[]>();
    for (const [v, neighbors] of graph) {
        if (v !== vertex) {
            // Keep only neighbors that aren't the removed vertex
            reduced.set(v, neighbors.filter(n => n !== vertex));
        }
    }
    return reduced;
}
 
function countComponents(graph: Map<number, number[]>): number {
    const visited = new Set<number>();
    let components = 0;
    
    for (const vertex of graph.keys()) {
        if (!visited.has(vertex)) {
            dfs(graph, vertex, visited);
            components++;
        }
    }
    
    return components;
}
 
function dfs(graph: Map<number, number[]>, start: number, visited: Set<number>): void {
    visited.add(start);
    for (const neighbor of (graph.get(start) || [])) {
        if (!visited.has(neighbor)) {
            dfs(graph, neighbor, visited);
        }
    }
}

Complexity analysis of the naive approach:

For each of the n vertices, we:
- Remove the vertex: O(n + m) to create new graph
- Count components via DFS: O(n + m)
Total time complexity: O(n × (n + m))

For dense graphs where m ≈ n², this becomes O(n³)—prohibitively slow for large graphs.

For sparse graphs where m ≈ n, this is O(n²)—still too slow for graphs with millions of vertices.

We need a better approach.

Why Naive Isn't Practical

A social network graph with 1 million users and 100 million connections would require approximately 10¹⁴ operations with the naive approach—years of computation. The efficient DFS-based algorithm (covered in Page 3) solves the same problem in O(V + E), making it practical for real-world networks.

Intuition for Efficient Detection

The naive approach's inefficiency comes from repeatedly traversing the entire graph. But what if we could determine articulation points during a single graph traversal?

The key insight comes from analyzing the DFS tree structure. When we perform depth-first search, we construct a spanning tree (the DFS tree) plus additional edges (back edges) that connect descendants to ancestors.

The critical observation:

Consider a vertex v in the DFS tree. Vertex v is an articulation point if and only if:

v is the root of the DFS tree AND has two or more children in the DFS tree.
- If the root has multiple children, removing it disconnects those subtrees.
v is NOT the root AND has a child u such that no vertex in the subtree rooted at u has a back edge to an ancestor of v.
- This means u's subtree's only connection to the rest of the graph is through v.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
DFS from vertex 1:
 
Original Graph:           DFS Tree:
    1───2                     1 (root)
    |\  |                     |
    | \ |                     2 (tree edge)
    |  \|                     |
    3───4                     4 (tree edge)
                              |
                              3 (tree edge)
                              
Back edges (shown as dashed): 
  1---3 (ancestor-descendant)
  2---4 (ancestor-descendant)
 
Analysis:
- Vertex 1 (root): Has only 1 DFS child (2). NOT an articulation point.
- Vertex 2: Child 4's subtree has back edge 1---3 reaching above 2. NOT an articulation point.
- Vertex 4: Child 3's subtree has back edge reaching to 1 (above 4). NOT an articulation point.
- Vertex 3: Leaf, NOT an articulation point.
 
This graph is biconnected - no articulation points!

The Discovery/Low Time Framework

Tarjan's algorithm formalizes this intuition using two values for each vertex: discovery time (when DFS first visits it) and low time (the minimum discovery time reachable through the subtree and back edges). We'll cover this elegant algorithm in detail in Page 3.

Preview of the efficient approach:

Perform a single DFS traversal, recording:
- discovery time: When each vertex is first visited
- low value: The earliest discovered vertex reachable from the subtree
For non-root vertex v with child u:
- If low[u] ≥ discovery[v], vertex v is an articulation point
- This means u's subtree can't 'escape' above v via back edges
For the root:
- It's an articulation point if and only if it has ≥ 2 DFS children

This single-pass algorithm achieves O(V + E) time complexity—optimal for this problem.

Summary: Articulation Points Defined

We've established a comprehensive conceptual foundation for articulation points. Let's consolidate the key insights before moving to bridges and algorithmic implementation.

Key Takeaways

•Articulation points are vertices whose removal disconnects the graph — They're single points of failure in network connectivity.
•They exist because of structural bottlenecks — Places where all paths between certain regions must pass through one vertex.
•Leaf vertices are never articulation points — They don't lie on any path between other vertices.
•Trees maximize articulation points; complete graphs have none — Edge density creates redundancy that eliminates cut vertices.
•Biconnected components are the robust building blocks — Articulation points connect these internally-redundant structures.
•Naive detection is O(n(n+m)); efficient DFS-based detection is O(n+m) — Understanding the DFS tree structure enables linear-time algorithms.

What's next:

The next page explores bridges (cut edges)—the edge analog of articulation points. Bridges are edges whose removal disconnects the graph. The concepts are closely related, and the detection algorithms share the same elegant DFS-based framework.

Page Complete

You now have a deep conceptual understanding of articulation points: what they are, why they matter, when they exist, and intuition for efficient detection. Next, we'll explore bridges—edges that, like articulation points for vertices, represent critical links in graph connectivity.

1 / 4

Loading learning content...

Data Structures & AlgorithmsArticulation Points & Bridges

Articulation Points & Bridges

LevelAdvanced

Duration90 mins

TopicArticulation Points & Bridges

1 / 4

Understanding Articulation Points (Cut Vertices)

The Critical Nodes That Hold Networks Together

What You Will Master

Formal Definition of Articulation Points

Before we explore articulation points intuitively, let's establish the precise mathematical definition. Precision here prevents confusion later when we implement detection algorithms.

Definition:

Formally:

A vertex v is an articulation point if G' = (V - {v}, E') has more connected components than G, where E' contains all edges of E that do not involve v.

Connected Components Reminder

Key implications of this definition:

Only meaningful in connected graphs: If a graph is already disconnected, the concept of articulation points becomes more nuanced—we would analyze each connected component separately.
Removal includes incident edges: When we 'remove' a vertex, we also remove all edges connected to it. This is implicit in the definition but crucial for understanding.
Binary classification: Each vertex either is or isn't an articulation point—there's no gradient. However, some articulation points might be 'more critical' in practice (disconnecting larger portions).
At least two components result: By definition, removing an articulation point creates at least two disconnected components from what was previously one.

Original Connected Graph:
    
    A ─── B ─── C
          │
          D
         / \
        E   F
 
Vertex B is an articulation point because:
- Removing B disconnects A from the rest of the graph
- Result: {A} and {C, D, E, F} become separate components
 
After removing B:
    
    A       C
            │
            D
           / \
          E   F
 
Two disconnected components: {A} and {C, D, E, F}

Why Articulation Points Exist

Consider it this way: If there's only one route to reach some portion of the graph, the vertices on that route become critical. If any of them fails, that portion becomes unreachable.

Structural Reasons for Articulation Points

•Tree-like structures: In a tree, every non-leaf vertex is an articulation point. Trees have no cycles, so removing any internal vertex disconnects its children from its parent.
•Linear chains: In a path graph A—B—C—D, vertices B and C are articulation points. The chain topology creates single points of failure.
•Hub nodes in sparse graphs: Vertices with high degree in sparse graphs often become articulation points because they're the sole connectors between regions.
•Graph 'necks': Narrow passages between densely connected regions create articulation points at the junction.

Contrast with robust graphs:

The more 'cycle-rich' a graph is, the fewer articulation points it tends to have. Cycles create alternative paths, providing redundancy that eliminates single points of failure.

Graph Topology vs Articulation Point Prevalence
Graph Type	Articulation Points	Why
Complete graph (Kₙ)	None	Every pair of vertices is directly connected—maximum redundancy
Cycle graph (Cₙ)	None	Two paths exist between any pair—removing one vertex still leaves a path
Tree (n vertices)	n - (leaf count)	No cycles means no alternative paths
Path graph (n vertices)	n - 2	Linear structure; only endpoints aren't articulation points
Star graph (1 center + k leaves)	1 (the center)	Center connects all leaves; remove center and leaves become isolated
Grid graph	Varies by position	Interior vertices often aren't; edges of grid may have some

Visualizing Articulation Points Through Examples

1
2
3
4
5
6
7
8
9
10
11
12
    1 ─── 2 ─── 3
          │
          4 ─── 5
         
Analysis:
- Remove 1: Graph stays connected (2-3-4-5 still linked). NOT an articulation point.
- Remove 2: Components become {1} and {3, 4, 5}. ARTICULATION POINT.
- Remove 3: Graph stays connected (1-2-4-5 still linked). NOT an articulation point.
- Remove 4: Components become {1, 2, 3} and {5}. ARTICULATION POINT.
- Remove 5: Graph stays connected. NOT an articulation point.
 
Articulation points: {2, 4}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
        1
       / \
      2───3
      │
      4───5
      
Analysis:
- The cycle 1-2-3 has NO articulation points among 1,2,3 considered in isolation
  (removing any one still leaves a path between the other two).
- However, vertex 2 connects the cycle to the chain 4-5
- Remove 2: {1,3} and {4,5} become disconnected. ARTICULATION POINT.
- Remove 4: {1,2,3} and {5} become disconnected. ARTICULATION POINT.
 
Articulation points: {2, 4}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
    1───2       5───6
     \ /         \ /
      3───────────7
      
Analysis:
This graph has two cycles (1-2-3 and 5-6-7) connected through vertices 3 and 7.
 
- Within cycle 1-2-3: No articulation points (each vertex has a bypass).
- Within cycle 5-6-7: No articulation points.  
- But vertex 3 is the ONLY connector from {1,2} to {5,6,7}.
- And vertex 7 is the ONLY connector from {5,6} to {1,2,3}.
 
Remove 3: {1,2} disconnected from {5,6,7}. ARTICULATION POINT.
Remove 7: {5,6} disconnected from {1,2,3}. ARTICULATION POINT.
 
Articulation points: {3, 7}

The 'Pinch Point' Mental Model

Special Cases: Leaves, Roots, and Single-Vertex Graphs

Some vertex positions in graphs have special properties regarding articulation points. Understanding these edge cases clarifies the concept and prevents errors in algorithm implementation.

Special Cases Analysis

•Leaf vertices (degree 1): A leaf has only one neighbor. Removing a leaf never disconnects the graph—its single neighbor remains connected to everything it was before. Leaves are NEVER articulation points.
•Vertices of degree 0 (isolated): These don't exist in connected graphs by definition. In disconnected graphs, removing them doesn't change connectivity.
•The single vertex in a one-vertex graph: By convention, this is NOT considered an articulation point (removing it leaves an empty graph, not a disconnected one).
•Vertices in complete graphs (Kₙ, n ≥ 3): Since every vertex connects to every other, there's always an alternative path. Complete graphs have NO articulation points.

The degree observation:

Conversely, a low-degree vertex can be an articulation point. In a path graph 1—2—3—4—5, vertex 3 has degree 2 but is an articulation point.

The key insight: Articulation point status depends on the structure of connections, not simply the count of connections.

Common Misconception

Articulation Points and Graph Connectivity

Articulation points are intimately connected to the broader concept of graph connectivity, a measure of how resilient a graph is to vertex or edge failures.

Levels of Vertex Connectivity

•1-vertex-connected (connected): The graph is connected—there exists a path between any two vertices. May or may not have articulation points.
•2-vertex-connected (biconnected): The graph remains connected after removing ANY single vertex. Has NO articulation points. Requires at least 3 vertices.
•k-vertex-connected: The graph remains connected after removing any (k-1) or fewer vertices. Higher k means more robust.

The vertex connectivity κ(G):

The vertex connectivity of a graph G, denoted κ(G), is the minimum number of vertices whose removal disconnects the graph (or reduces it to a single vertex for complete graphs).

If κ(G) = 1, the graph has at least one articulation point.
If κ(G) ≥ 2, the graph is biconnected (no articulation points).
If κ(G) = k, you need to remove at least k vertices to disconnect the graph.

Menger's Theorem connects this to path diversity: κ(G) equals the maximum number of internally vertex-disjoint paths between any pair of vertices. Higher connectivity means more redundant paths.

Connectivity Classes and Their Properties
Connectivity κ(G)	Graph Property	Articulation Points	Example
κ(G) = 0	Disconnected	N/A (already disconnected)	Two separate triangles
κ(G) = 1	Connected but vulnerable	At least one exists	Tree, path graph
κ(G) = 2	Biconnected	None	Cycle, complete graph K₃
κ(G) = k	k-connected	None (highly redundant)	Complete graph Kₙ has κ = n-1

Biconnected Components

Biconnected Components: The Building Blocks

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Original Graph:
    
    1───2      5───6
     \ /        \ /
      3──────────4
      │
      7───8
       \ /
        9
 
Biconnected Components:
  Component 1: {1, 2, 3}  (triangle - no internal articulation points)
  Component 2: {5, 6, 4}  (triangle - no internal articulation points)  
  Component 3: {3, 4}     (edge connecting two components)
  Component 4: {3, 7}     (edge connecting to lower structure)
  Component 5: {7, 8, 9}  (triangle - no internal articulation points)
 
Articulation Points: {3, 4, 7}
These vertices appear in multiple biconnected components!

Critical insight:

Articulation points are exactly the vertices that appear in more than one biconnected component. This makes sense: they're the connection points between otherwise independent structures.

The block-cut tree:

We can represent this decomposition as a tree (called the block-cut tree or BC-tree):

One node for each biconnected component (called a block)
One node for each articulation point (called a cut vertex)
An edge connects a block to a cut vertex if that cut vertex belongs to that block

This tree representation is invaluable for answering connectivity queries efficiently.

Why This Matters

Properties and Bounds on Articulation Points

How many articulation points can a graph have? Understanding the theoretical bounds helps verify algorithm correctness and provides context for graph structure.

Bounds on Articulation Point Count

•Minimum: 0 articulation points (achieved by biconnected graphs like cycles, complete graphs)
•Maximum for n vertices: n - 2 articulation points (achieved by path graphs: 1—2—3—...—n where only endpoints aren't articulation points)
•Trees with n vertices: Exactly n - (number of leaves) articulation points. Every non-leaf in a tree is an articulation point.
•Adding edges reduces count: Adding edges to a graph can only maintain or reduce the number of articulation points (never increase). Edges add redundancy.

Edge-articulation point relationship:

For a connected graph with n vertices and m edges:

If m = n - 1 (tree): Maximum potential articulation points
If m = n (one cycle added): At least one articulation point removed
If m = n(n-1)/2 (complete graph): Zero articulation points

Articulation Point Counts for Common Graph Families
Graph Family	Vertices (n)	Edges (m)	Articulation Points
Path Pₙ	n	n-1	n-2
Cycle Cₙ	n	n	0
Complete Kₙ	n	n(n-1)/2	0
Star S₍ₙ₎	n+1	n	1 (center)
Binary tree (balanced)	2ˢ-1	2ˢ-2	2ˢ⁻¹-1 (all non-leaves)
Wheel Wₙ	n+1	2n	0 (if n≥3)

The Naive Detection Approach

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
function findArticulationPointsNaive(graph: Map<number, number[]>): number[] {
    const articulationPoints: number[] = [];
    const vertices = Array.from(graph.keys());
    
    for (const vertex of vertices) {
        // Remove this vertex and its edges
        const reducedGraph = removeVertex(graph, vertex);
        
        // Count connected components in reduced graph
        const originalComponents = countComponents(graph);
        const reducedComponents = countComponents(reducedGraph);
        
        // If removal increases component count, it's an articulation point
        // Note: originalComponents is 1 for connected graph
        // reducedComponents ≥ 2 means vertex was critical
        if (reducedComponents > 1) {
            articulationPoints.push(vertex);
        }
    }
    
    return articulationPoints;
}
 
function removeVertex(graph: Map<number, number[]>, vertex: number): Map<number, number[]> {
    const reduced = new Map<number, number[]>();
    for (const [v, neighbors] of graph) {
        if (v !== vertex) {
            // Keep only neighbors that aren't the removed vertex
            reduced.set(v, neighbors.filter(n => n !== vertex));
        }
    }
    return reduced;
}
 
function countComponents(graph: Map<number, number[]>): number {
    const visited = new Set<number>();
    let components = 0;
    
    for (const vertex of graph.keys()) {
        if (!visited.has(vertex)) {
            dfs(graph, vertex, visited);
            components++;
        }
    }
    
    return components;
}
 
function dfs(graph: Map<number, number[]>, start: number, visited: Set<number>): void {
    visited.add(start);
    for (const neighbor of (graph.get(start) || [])) {
        if (!visited.has(neighbor)) {
            dfs(graph, neighbor, visited);
        }
    }
}

Complexity analysis of the naive approach:

For each of the n vertices, we:
- Remove the vertex: O(n + m) to create new graph
- Count components via DFS: O(n + m)
Total time complexity: O(n × (n + m))

For dense graphs where m ≈ n², this becomes O(n³)—prohibitively slow for large graphs.

For sparse graphs where m ≈ n, this is O(n²)—still too slow for graphs with millions of vertices.

We need a better approach.

Why Naive Isn't Practical

Intuition for Efficient Detection

The naive approach's inefficiency comes from repeatedly traversing the entire graph. But what if we could determine articulation points during a single graph traversal?

The critical observation:

Consider a vertex v in the DFS tree. Vertex v is an articulation point if and only if:

v is the root of the DFS tree AND has two or more children in the DFS tree.
- If the root has multiple children, removing it disconnects those subtrees.
v is NOT the root AND has a child u such that no vertex in the subtree rooted at u has a back edge to an ancestor of v.
- This means u's subtree's only connection to the rest of the graph is through v.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
DFS from vertex 1:
 
Original Graph:           DFS Tree:
    1───2                     1 (root)
    |\  |                     |
    | \ |                     2 (tree edge)
    |  \|                     |
    3───4                     4 (tree edge)
                              |
                              3 (tree edge)
                              
Back edges (shown as dashed): 
  1---3 (ancestor-descendant)
  2---4 (ancestor-descendant)
 
Analysis:
- Vertex 1 (root): Has only 1 DFS child (2). NOT an articulation point.
- Vertex 2: Child 4's subtree has back edge 1---3 reaching above 2. NOT an articulation point.
- Vertex 4: Child 3's subtree has back edge reaching to 1 (above 4). NOT an articulation point.
- Vertex 3: Leaf, NOT an articulation point.
 
This graph is biconnected - no articulation points!

The Discovery/Low Time Framework

Preview of the efficient approach:

Perform a single DFS traversal, recording:
- discovery time: When each vertex is first visited
- low value: The earliest discovered vertex reachable from the subtree
For non-root vertex v with child u:
- If low[u] ≥ discovery[v], vertex v is an articulation point
- This means u's subtree can't 'escape' above v via back edges
For the root:
- It's an articulation point if and only if it has ≥ 2 DFS children

This single-pass algorithm achieves O(V + E) time complexity—optimal for this problem.

Summary: Articulation Points Defined

We've established a comprehensive conceptual foundation for articulation points. Let's consolidate the key insights before moving to bridges and algorithmic implementation.

Key Takeaways

•Articulation points are vertices whose removal disconnects the graph — They're single points of failure in network connectivity.
•They exist because of structural bottlenecks — Places where all paths between certain regions must pass through one vertex.
•Leaf vertices are never articulation points — They don't lie on any path between other vertices.
•Trees maximize articulation points; complete graphs have none — Edge density creates redundancy that eliminates cut vertices.
•Biconnected components are the robust building blocks — Articulation points connect these internally-redundant structures.
•Naive detection is O(n(n+m)); efficient DFS-based detection is O(n+m) — Understanding the DFS tree structure enables linear-time algorithms.

What's next:

Page Complete

1 / 4