A Method for Mapping Graphs to Real Metric

Space
Philipp Harland
June 2025

Abstract
In this paper, we will be detailing a method to embed graphs into the
2-dimensional real plane, R. We will also extend the result to hypergraphs.
We will also be examining some of the immediate theory that comes from
this.

Email: [email protected]

Contents
1 Introduction 1

2 Generalization – Central Results 2

2.1 Ordinary Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Hypergraphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1 Introduction
Def. 1.1. A graph, G = (G, I) is a set of n points, which we individually call
”nodes”, and notate G; and, an ”adjacency matrix”, I, which is a diagonally
symmetric element of Mn (Z2 ) from which we generate an ”adjacency”. That
is, nodes ai and aj are adjacent if Iij = 1.
Def. 1.2. Given 2 graphs, G and H, G ∼ = H, or G is isomorphic to H, if there
exists a bijection, f, s.t. u adj v ⇐⇒ f (u) adj f (v).
In the context of classical graph theory, one defines graphs as purely topologi-
cal objects that exist in abstract spaces with no notion of metric. We seek to
generalize this definition in this paper.

1
2 Generalization – Central Results
2.1 Ordinary Graphs
We consider the category Graphn / ∼, the category of graphs of a fixed size n,
where G ∼ H =⇒ G ∼ = H. For a y element, F, in this category, we consider a
set of n disjoint points in R2 , which we will denote MF .
We let J(F ) be a bijection between the nodes of F and {1...n}, and let J(MF )
be the same. This forms the ”pullback”:

MF F

Z+
≤n

Now, we can define the ”concretization” of F, or, the embedding of F into R2

as follows:
Def. 2.1.1. The concretization of F is the tuple (MF , K), where K is a set of
functions in the function space Z≤n × Z≤n → Z2 . These functions are defined
such that when a set of n nodes, S, with no edges is given any bijective labeling,
and, these functions are assigned as adjacencies in S whenever they are equal
to 1 w.r.t. the labels on S, S ′ ∼
= F.
This completes our triangular ”pullback” below:

MF F

Z+
≤n

We may relax the set’s conditions, such that it is possible for it to be a multiset. This
allows for certain graph operations such as contracting nodes, in a way.
Now, we can begin to build up some of the immediate implications of our theory, first
by generalizing graph isomorphism to our ”concretized” graphs:
Def. 2.1.2. 2 concretized graphs, S and T, are isomorphic if their reconstructed
graphs (graphs constructed from the process defined in Def. 2.1.) are isomorphic.
Def. 2.1.3. 2 concretized graphs, S and T, have the property of Rec(S) ⊂Graph
Rec(T ), where Rec(G) is the reconstruction of a graph, G.
For that matter, any relation in Graph can be generalized to ConcGraph. Specifi-
cally, if ∼∈ Rel(Graph), ∼Conc is defined as S ∼Conc T =⇒ Rec(S) ∼ Rec(T ).
We also know because of the categorical properties that functors to and from Graph and
ConcGraph, that equivalence relations in Graph translate to ones in ConcGraph.
In fact, these two categories are isomorphic given the functors that map A → Rec(A)

2
and B → Conc(B). These become functors when applied to all elements, and when
the functors are composed in either way, we gain 1Graph and 1ConcGraph .
Thm. 2.1.1. All graphs have a concretized graph.
Proof. To first prove that every graph has a concretized graph, we first acknowledge
that ∀G ∈ Graphn , G ⊂ Kn , where Kn is the complete graph on n vertices.
We also know that the complete graph Kn has a concretized graph, which is
simply a set of n points with all such functions equal to 1 in its K-set for
all pairs of node labels. From then on, the proof is trivial.

2.2 Hypergraphs
This case is slightly less convenient to define, as we cannot easily define concretizations
for hypergraphs with adjacency tables without our constructions becoming convoluted.
Instead, we rely on the classical notion of a set of ordered pairs with their projections
representing nodes.
We provide the following construction which is analogous to that defined in Section
2.1:
We have a hypergraph, C, and a set of points, MC , which we have an assigned set of
functions, L, which map any number of labels on C, albeit they are fixed per function,
to a unique number that is positive if they are connected in C, and 0 if they are not
connected. We let J(C) and J(MC ) be any bijective labelings.
Now, similarly to Def. 2.1.1, we define the concretization of a hypergraph:
Def. 2.2.1. Given a hypergraph, M of k nodes – the concretization of M, Conc(M ),
is the set of points in R2 with the unique assigned K-set such that when the inverses of
positive integers w.r.t. K are taken to be hyperedges on a set of k nodes with no edges
of any labeling, the resulting hypergraph is isomorphic to M.
Thm. 2.2.1. All hypergraphs have a concretization.
Proof. This is trivial to prove, as it is a direct consequence of Thm. 2.1.1..
We also immediately obtain the definitions of isomorphism in ConcHypGraph, and
subgraphs in the context of concrete hypergraphs as immediate consequences of the
results obtained in 2.1. That is, since there are the same isomorphisms between
ConcHypGraph and HypGraph as there are between ConcGraph and Graph. This
is trivial to verify.
In fact, we have the relations Graph ⊂ HypGraph and ConcGraph ⊂ ConcHypGraph,
which immediately spring from the fact that hypergraphs are an enlargement of the idea
of a graph as described in classical theory. When we couple these relations with the
isomorphism relations we have obtained, we get the following diagram:

Graph ⇐⇒ ConcGraph

↓ ↓

HypGraph ⇐⇒ ConcHypGraph