Skip to Content
You are currently on the new version of our website. Access the old version .
AxiomsAxioms
  • Article
  • Open Access

9 April 2026

15 Pages

On the Bounding Function of Cogirth and Supereulerian Regular Matroids

,
and
1
Department of Mathematics and Statistics, Qinghai Normal University, Xining 810008, China
2
School of Mathematics and Statistics, Qinghai Minzu University, Xining 810007, China
3
The State Key Laboratory of Tibetan Intelligent Information Processing and Application, Xining 810008, China
*
Author to whom correspondence should be addressed.

Abstract

Following Bauer’s 1985 question on how a graph’s minimum degree relates to it being supereulerian, Catlin solved the problem for graphs with a minimum degree of at least n 5 . Subsequently, Lai solved it for the case where the minimum degree is at least n 10 . Projecting this problem into the setting of matroids leads to the question of the relationship between a matroid being supereulerian and its cogirth. In 2025, Qin et al. showed that a connected simple regular matroid M with cogirth at max { r ( M ) + 1 15 , 9 } is supereulerian. In this paper, we prove that for any real number c with 15 < c < 20 , there exists a smallest integer f ( c ) such that every connected simple regular matroid M with cogirth at least max { r ( M ) + 1 c , f ( c ) } has a spanning Eulerian subset.
MSC:
05C07; 05C83

1. Introduction

The concept of Eulerian subgraphs and their generalizations has been a central theme in graph theory and combinatorial optimization, originating from the classical Chinese Postman Problem and Eulerian circuit problems. A graph is called supereulerian if it contains a spanning Eulerian subgraph, a notion extensively studied due to its connections to network design, cycle covers, and matroid theory. Early foundational work by Catlin [1] introduced reduction methods to characterize supereulerian graphs, while later developments by Lai [2] extended the study to Eulerian subgraphs containing prescribed edges. Recent advances have further expanded the theory to directed graphs [3] and reconfiguration settings [4]. These developments continue to drive both theoretical inquiries and practical applications in network optimization and combinatorial structures.
Beyond graph theory, the Euler characteristic has emerged as a powerful topological invariant with significant applications in data science, medical imaging, and physics. In topological data analysis, the Euler Characteristic Transform (ECT) provides a concise topological fingerprint for shape classification [5]. Recent applications include medical image segmentation, where Li et al. [6] introduced a fast Euler characteristic-based optimization method to enforce topological correctness in deep learning models. These interdisciplinary advances highlight the growing importance of Eulerian structures and their invariants in both theoretical and applied settings.
Recent research has increasingly explored connections between combinatorial structures and applications in computational, statistical, and network sciences. While this paper focuses on structural properties of regular matroids and supereulerian conditions, we briefly discuss several related developments. Gan [7] demonstrates how topology-aware construction improves large-scale graph applications, echoing the importance of structural information in algorithm design. Yang et al. [8] and Chen et al. [9] leverage matrix operations in peridynamics and circuit systems, respectively, a relevant point given that regular matroids admit totally unimodular matrix representations. Li et al. [10] and Ren et al. [11] address dependency structure identification in statistical models, conceptually linking it to independence systems and connectivity in matroids. These interdisciplinary perspectives suggest that the structural insights from regular matroids may find broader relevance beyond pure combinatorial theory.
We consider finite undirected graphs and matroids. Graph-theoretic terminology follows Bondy and Murty [12], with the exception of the terms “cycle” and “circuit,” while matroid terminology is taken from Oxley [13]. For a graph G, we write V ( G ) and E ( G ) for its vertex set and edge set, respectively, and denote the number of vertices by | V ( G ) | and the number of edges by | E ( G ) | . The degree of a vertex v ∈ V ( G ) is written as d G ( v ) (or simply d ( v ) when no confusion arises). The minimum and maximum degrees of G are denoted by δ ( G ) and Δ ( G ) , respectively. The edge connectivity of G is denoted by κ ′ ( G ) . To maintain consistent terminology across graph theory and matroid theory, we adopt the following conventions: a connected graph that is two-regular and nontrivial is called a circuit, and a disjoint union of circuits (as edge sets) is called a cycle. A connected graph G is Eulerian precisely when it forms a cycle. A connected graph G is said to be supereulerian if it admits a spanning Eulerian subgraph, i.e., a spanning cycle.
Let M = ( E , I ) be a matroid with ground set E, where I denotes its collection of independent sets. A circuit of M is defined as a minimal dependent set. The set of all circuits of M is written as C ( M ) , and M ∗ denotes the dual matroid. The rank, girth, and cogirth of M are denoted respectively by r ( M ) , g ( M ) , and g ∗ ( M ) . We define g ( M ) = min { | C | : C ∈ C ( M ) } when C ( M ) ≠ ∅ ; otherwise, g ( M ) = ∞ . A cycle in M is a set that can be expressed as a disjoint union of circuits of M. A cycle C is called spanning if r M ( C ) = r M ( E ( M ) ) , and M is said to be supereulerian whenever such a cycle exists.
The study of supereulerian graphs, which seeks to characterize graphs containing a spanning cycle, was initiated by Boesch, Suffel, and Tindell [14]. Pulleyblank [15] indicated that the problem of determining whether a graph had a spanning Eulerian subset remained NP-complete, even when restricted to planar graphs. For further literature on the supereulerian problem, Catlin’s survey [16] and its updated versions [17,18] are resourceful references.
The investigation of supereulerian matroids originated in [19]. In [20], Veblen established that a connected graph G was Eulerian precisely when its edge set E ( G ) could be expressed as a union of edge-disjoint circuits. Extending this idea to matroids, a subset X ⊆ E ( M ) is called an Eulerian subset (or simply Eulerian) if it forms an edge-disjoint union of circuits in M. Let E u ( M ) denote the collection of all Eulerian subsets of M. Then, C ( M ) consists of all nonempty minimal members in E u ( M ) . As C ( M ) uniquely determines the structure of M, the collection E u ( M ) also uniquely determines M. A matroid M is supereulerian if M has an Eulerian subset X with r M ( X ) = r ( M ) ; such an Eulerian subset is called a spanning Eulerian subset. There have been a number of studies on supereulerian matroids, as seen in [19,21,22,23], among others.
Due to Dirac’s theorem, which proves that every graph of order n with minimum degree greater than n / 2 necessarily contains a spanning Eulerian subgraph, many subsequent researchers have studied the condition where the minimum degree is greater than n / k , with a weaker k yielding stronger results. For the cycle matroid of a graph, a minimal cut corresponds to a cocircuit in the dual matroid; thus, a minimal cut with the smallest number of edges corresponds to a minimum cocircuit, i.e., the cogirth. Therefore,
δ ( G ) ≥ κ ′ ( G ) = g ∗ ( M ( G ) ) .
In a graph of order n, the maximum number of edges in a subgraph containing no cycle is the number of edges in a spanning tree, which is n − 1 . The corresponding cycle matroid has rank r ( M ( G ) ) , satisfying n − 1 = r , i.e., n = r + 1 . Since the minimum degree of a graph is at least its edge connectivity, and the set of edges incident to each vertex forms a minimal edge cut, we can replace the minimum degree by the size of a minimum edge cut (i.e., the edge connectivity) and replace n by r + 1 . Following Welsh’s consideration (see Conjecture 14.4.1 in [24]), we use cogirth g ∗ ( M ) in a matroid M to partially play the role of the minimum degree δ ( G ) in a graph G. Former studies on supereulerian matroids suggested the following problem.
Problem 1.
Let M denote a family of matroids. A pair of positive real numbers ( c , f ( c ) ) is an Eulerian pair in M if for any matroid M ∈ M , M has spanning Eulerian subset whenever g ∗ ( M ) ≥ max { ( r ( M ) + 1 ) c , f ( c ) } . Thus, for a given matroid family M , the following are of interests:
(i) 
Determine the existence of the Eulerian pairs in M .
(ii) 
If Eulerian pair ( c , f ( c ) ) exists in M , then we are to determine, for a given value c, the smallest value of f ( c ) .
Theorem 1
([25]). Let M = R be the family of all connected simple regular matroids. For any real number c with c ≥ 20 , there exists an integer f ( c ) such that ( c , f ( c ) ) has a spanning Eulerian subset.
Theorem 2.
Let M = R be the family of all connected simple regular matroids. For any real number c,
(1) 
If 15 < c ≤ 16 , ( c , 9 ) is Eulerian.
(2) 
If 16 < c < 20 , ( c , 10 ) is Eulerian.
Following Theorem 2, we have thus generalized Theorem 1.
Theorem 3.
Let M be a connected simple regular matroid. For any real number c with c ≥ 15 , there exists an integer f ( c ) such that if cogirth g ∗ ( M ) of M satisfies g ∗ ( M ) ≥ max { ( r ( M ) + 1 ) c , f ( c ) } , then M has a spanning Eulerian subset.
Proof. 
According to Theorem 2, when c is greater than 15 and less than or equal to 16, f ( c ) = 9 guarantees the existence of an Eulerian subset in M; when c is greater than 16 and less than 20, f ( c ) = 10 guarantees the existence of a spanning Eulerian subset in M. According to Theorem 1, when c is greater than 20, there exists f ( c ) such that M contains a spanning Eulerian subset. Hence, the conclusion holds. □

2. Preliminaries

For a matroid M and a subset X ⊆ E ( M ) , we write M ∖ X for the restriction M | ( E ( M ) − X ) . When X consists of a single element e, we denote M ∖ { e } simply by M ∖ e . The contraction of X in M, denoted M / X , is defined as ( M ∗ ∖ X ) ∗ . The circuits of M / X are precisely the minimal nonempty sets among { C − X : C ∈ C ( M ) } . A matroid N is said to be a minor of M if it can be expressed as M ∖ X / Y for some disjoint subsets X and Y of E ( M ) , where either set may be empty.
Jaeger [26] and Catlin [27] independently prove that
e v e r y   f o u r - e d g e - c o n n e c t e d   g r a p h   h a s   a   s p a n n i n g   E u l e r i a n   s u b s e t .
Let M be a class of matroids possessing a property P, and let O P denote an operation such as a contraction or deletion. We write O P ( M ) to indicate the result of applying O P to M. The class M is said to be closed under the operation O P if for every M ∈ M , the matroid O P ( M ) also satisfies property P. For instance, the class of regular matroids is closed under both deletion and contraction; consequently, every minor of a regular matroid remains regular. In what follows, we introduce several properties of matroids.

2.1. Seymour’s Decomposition of Regular Matroids

For two sets X and Y, their symmetric difference is given by X △ Y = ( X ∪ Y ) − ( X ∩ Y ) . We now define the i-sum and describe the cycle set arising from the i-sum of two matroids.
Definition 1
([13]). Let M 1 and M 2 be binary matroids, and set E = E ( M 1 ) △ E ( M 2 ) . Then, there exists a matroid M 1 △ M 2 with ground set E satisfying
C ( M 1 △ M 2 ) = { X 1 △ X 2 : X 1 ∈ C ( M 1 ) and X 2 ∈ C ( M 2 ) } ,
where C ( M ) denotes the set of cycles of M.
In Definition 2 below, to avoid triviality, we require that in the two-sum and three-sum cases, each summand M i has a sufficiently large number of elements.
Definition 2
([28,29]). For i ∈ { 1 , 2 } , let M i be a binary matroid with ground set E i = E ( M i ) .
(i) 
If E 1 ∩ E 2 = ∅ , and | E 1 | , | E 2 | < | E 1 △ E 2 | , then M 1 △ M 2 is called the one-sum of M 1 and M 2 , denoted M 1 ⊕ 1 M 2 .
(ii) 
If | E 1 ∩ E 2 | = 1 , write E 1 ∩ E 2 = { p } , where p is neither a loop nor a coloop in M 1 or M 2 , and | E 1 | , | E 2 | < | E 1 △ E 2 | , then M 1 △ M 2 is called the two-sum of M 1 and M 2 , denoted M 1 ⊕ 2 M 2 .
(iii) 
If | E 1 ∩ E 2 | = 3 , let E 1 ∩ E 2 = Z , where Z forms a circuit in both M 1 and M 2 , and Z contains no cocircuit of either M 1 or M 2 , and | E 1 | , | E 2 | < | E 1 △ E 2 | , then M 1 △ M 2 is called the three-sum of M 1 and M 2 , denoted M 1 ⊕ 3 M 2 .
Lemma 1
([13]). Let M = M 1 ⊕ i M 2 , i ∈ { 1 , 2 , 3 } . Then,
(1). 
If E ( M 1 ) ∩ E ( M 2 ) = ∅ , then C ( M 1 ⊕ 1 M 2 ) = C ( M 1 ) ∪ C ( M 2 ) .
(2). 
If E ( M 1 ) ∩ E ( M 2 ) = { p } , then C ( M 1 ⊕ 2 M 2 ) =   C ( M ∖ p ) ∪ C ( N ∖ p ) ∪ { ( C ∪ D ) − p : p ∈ C ∈ C ( M ) and p ∈ D ∈ C ( N ) } .
(3). 
If E ( M 1 ) ∩ E ( M 2 ) = T and T is a circuit of M 1 and M 2 , then C ( M 1 △ M 2 ) is the union of C ( M 1 ∖ T ) , C ( M 2 ∖ T ) , and the collection of minimal sets of the form C 1 △ C 2 − T where C i ∈ C ( M i ) such that C 1 ∩ T = C 2 ∩ T , and the last set has exactly one element.
Therefore, a connected matroid cannot be represented as a one-sum of its minors.
Lemma 2.
Let M, M 1 , M 2 be matroid. Each of the following holds.
(a) 
[13] Suppose that M = M 1 ⊕ 2 M 2 . Then, M is connected if and only if both M 1 and M 2 are connected.
(b) 
[30] Suppose that M be a simple three-connected binary matroid and M = M 1 ⊕ 3 M 2 is a three-sum of two matroids. Then, both M 1 and M 2 are connected. In particular, if either matroid is simple, then it is three-connected.
Lemma 3
([13] (Seymour’s Decomposition Theorem)). Every regular matroid M can be constructed by using direct sums, two-sums, and three-sums starting with matroids, each of which is either graphic, cographic, or isomorphic to R 10 , and each of which is isomorphic to a minor of M.
Lemma 4
([13]). Let M = M 1 ⊕ i M 2 , i ∈ { 1 , 2 , 3 } . Then, M ∗ = M 1 ∗ △ M 2 ∗ . In particular, M ∗ = M 1 ∗ ⊕ 2 M 2 ∗ for i = 1 , 2 .

2.2. Supereulerian Binary Matroids

Let G be a graph and denote by M ( G ) its cycle matroid. According to Proposition 2.3.1 in [13], the dual matroid M ∗ ( G ) has as its circuits precisely the minimal edge cuts of G. The edge arboricity a ( G ) of G is defined as the smallest number of edge-disjoint forests needed to cover E ( G ) . Supereulerian cographic matroids admit a characterization in terms of edge arboricity.
Lemma 5
([22]). Let G be a graph and M = M ( G ) . The following are equivalent:
(1)
M ∗ has a spanning Eulerian subset.
(2)
a ( G ) ≤ 2 .
Kronk and Mitchem [31] proved a relationship between a ( G ) and Δ ( G ) , the maximum degree of a graph G.
Lemma 6
([31]). If G is connected, non-complete and a ( G ) = k ≥ 3 , then Δ ( G ) ≥ 2 k − 1 .
Let M denote a collection of matroids. A matroid N is an M -supereulerian contractible configuration, or simply contractible in M , if N ∈ M and for any M ∈ M containing N as a restriction,
M   h a s   a   s p a n n i n g   E u l e r i a n   s u b s e t   i f   a n d   o n l y   i f   M / N   h a s   a   s p a n n i n g   E u l e r i a n   s u b s e t .
For a matroid M, define τ ( M ) to be the maximum number of mutually disjoint bases in M. Let M 2 denote the family of all binary matroids. In [19], it is shown that M 2 -supereulerian contractible configurations exist. In particular, any binary matroid N with τ ( N ) ≥ 2 or satisfying E ( N ) ∈ C ( N ) and | N | ≤ 3 is contractible in M 2 . These suggest the following binary reductions.
Definition 3
([21]). Let M be a binary matroid. A subset N is called eligible in M if either τ ( M | N ) ≥ 2 , or N ∈ C ( M ) with | N | ≤ 3 . A matroid M is said to be eligible if it can be spanned by some eligible subset. We apply the following procedure to a binary matroid M with r ( M ) > 0 .
(i) 
Initialize M 0 : = M .
(ii) 
For each i = 0 , 1 , … , perform the following steps.
(ii-1) 
If M i contains a subset N that is eligible in M i and satisfies r ( N ) < r ( M i ) , then set M i + 1 : = M i / N and proceed to the next iteration.
(ii-2) 
If M i has no eligible subsets, or every eligible subset N in M i satisfies r ( M i ) = r ( N ) , then terminate.
When this procedure halts, the resulting matroid is called a binary reduction of M.
By definition, the following was observed in [21]:
E v e r y   e l i g i b l e   b i n a r y   m a t r o i d   h a s   a   s p a n n i n g   E u l e r i a n   s u b s e t .
In [21], binary reductions and contractions are shown to have interesting properties.
Lemma 7
([21]). Let M be a binary matroid and let M ′ be a binary reduction of M. Then, the following statements hold:
( a )
M has a Eulerian subset if and only if M ′ has a Eulerian subset.
( b )
Either M ′ is spanned by some subset N that is eligible in M ′ , or g ( M ′ ) ≥ 4 .
( c )
g ∗ ( M ) ≤ g ∗ ( M ′ ) .
Lemma 8
([23]). Let M be a regular matroid such that M = M 1 ⊕ i M 2 with g ( M ) ≥ 4 , where i ∈ { 2 , 3 } and Z = E ( M 1 ) ∩ E ( M 2 ) . If N is an eligible set in M 2 ∖ Z , then N is also eligible in M and satisfies r ( N ) < r ( M ) .

3. Main Results

Lemma 9
([23]). Let M be a simple regular matroid satisfying g ∗ ( M ) ≥ r ( M ) + 1 c , f ( c ) , where c > 1 . The following statements are equivalent:
(i) 
M has a spanning Eulerian subset.
(ii) 
Under the additional condition g ( M ) ≥ 4 , M has a spanning Eulerian subset.
Let R 10 be the vector matroid represented over G F ( 2 ) by the matrix below:
R 10 = 1 0 0 0 0 1 1 0 0 1 0 1 0 0 0 1 1 1 0 0 0 0 1 0 0 0 1 1 1 0 0 0 0 1 0 0 0 1 1 1 0 0 0 0 1 1 0 0 1 1
Lemma 10.
Each of the following holds:
(i) 
R 10 ∗ ≅ R 10 and as (see, for example, [22]) E ( R 10 ) is a disjoint union of a four-circuit and six-circuit, E ( R 10 ) is an Eulerian subset in both R 10 and R 10 ∗ .
(ii) 
[25] g ∗ ( R 10 ) = 4 , and so M cannot be expressed as a three-sum of a minor M 1 and an R 10 .
To prepare for the proof of Theorem 2, we first present some preliminary lemmas.

3.1. Cographic Matroids

Lemma 11.
Let G be a simple connected graph on n vertices and m edges, with minimum degree δ ( G ) ≥ 4 . If the girth satisfies g ( G ) ≥ max m − n + 2 c , 9 for some constant c with 15 < c < 20 , then G is four-regular.
Proof. 
Let maximum degree Δ ( G ) = t , D i ( G ) = { v : d ( v ) = i , v ∈ V ( G ) } , R = ⋃ i ≥ 5 D i ( G ) and | R | = r . Since δ ( G ) ≥ 4 , it follows that D i ( G ) = ∅ for 0 ≤ i ≤ 3 . Observe that
2 m = Σ v ∈ V ( G ) d G ( v ) = Σ i ≥ 4 i | D i ( G ) | = 4 n + Σ i ≥ 5 ( i − 4 ) | D i ( G ) | ≥ 4 n + r .
By g ( G ) ≥ m − n + 2 c > m − n + 2 20 , we derive
2 n < 40 g ( G ) − 4 − r .
Suppose G is not four-regular. Then, t ≥ 5 , r ≥ 1 . Let v ∈ V ( G ) satisfy d G ( v ) = t . In the following, we analyze the cases according to whether g ( G ) is even or odd.
  • Case 1. g ( G ) is odd.
Let g ( G ) = 2 s + 1 . Since g ( G ) ≥ 9 , it follows that s ≥ 4 . We employ Breadth-First Search (BFS) starting from vertex v to count the number of vertices in G. Let v represent the first layer. Then, vertices at a distance of k from v reside in the ( k + 1 ) th layer. Since the girth g = 2 s + 1 , it implies that for any two vertices at layer k, their neighboring vertices at layer k + 1 are distinct, where k ≤ s . Therefore, G contains at least 1 + t + 3 t + ⋯ + 3 s − 1 t = 1 + t 2 ( 3 s − 1 ) vertices within distance s of v. Thus, n ≥ 1 + t 2 ( 3 s − 1 ) . Substituting n ≥ 1 + t 2 ( 3 s − 1 ) and g ( G ) = 2 s + 1 into (4) yields 40 ( 2 s + 1 ) − 4 − r > 2 n ≥ 2 + t ( 3 s − 1 ) . Since t ≥ 5 and r ≥ 1 , it follows that
3 s < 16 s + 38 5 .
Let ϕ ( s ) = 3 s − 16 s − 38 5 . Then, ϕ ( s ) < 0 for s ∈ [ 4 , ∞ ) from (5). But the derivative is ϕ ′ ( s ) = 3 s ln 3 − 16 , which is non-negative for s ≥ 4 . Hence, ϕ ( s ) is increasing on [ 4 , + ∞ ) . Then, ϕ ( s ) ≥ ϕ ( 4 ) = 47 5 > 0 ; therefore, 3 s > 16 s + 38 5 , which contradicts (5).
  • Case 2. g ( G ) is even.
Let g ( G ) = 2 s and u v ∈ E ( G ) . Since g ( G ) ≥ 9 , it follows that s ≥ 5 . Similarly, use BFS starting from u and v to count the number of vertices in G. Let u , v represent the first layer. Then, G has at least 2 + ( t + 2 ) + 3 ( t + 2 ) + ⋯ + 3 s − 2 ( t + 2 ) = 2 + t + 2 2 ( 3 s − 1 − 1 ) vertices of distance at most s − 1 from v or u. Then, n ≥ 2 + t + 2 2 ( 3 s − 1 − 1 ) . From (4) and n ≥ 2 + t + 2 2 ( 3 s − 1 − 1 ) , we get 80 s − 4 − r > 2 n ≥ 4 + ( t + 2 ) ( 3 s − 1 − 1 ) . Since t ≥ 5 and r ≥ 1 , it follows that
3 s − 1 ≤ 80 s 7 − 2 7 .
Let ϕ ( s ) = 3 s − 1 − 80 s 7 + 2 7 . Then, ϕ ′ ( s ) = 3 s − 1 ln 3 − 80 7 , and ϕ ′ ( s ) ≥ 0 for s ≥ 5 . Hence, ϕ ( s ) is an increasing function in [ 5 , + ∞ ) . Then, ϕ ( s ) ≥ ϕ ( 5 ) = 169 7 , hence 3 s − 1 > 80 s 7 − 2 7 , which contradicts (6). □
Theorem 4.
For any simple connected cographic matroid M. If girth g ( M ) ≥ 4 and cogirth
g ∗ ( M ) ≥ max { r ( M ) + 1 c , 9 } ,
where 15 < c < 20 , then M has a spanning Eulerian subset.
Proof. 
Since M is cographic, there exists a connected graph G such that M ∗ = M ( G ) , and consequently, M = M ∗ ( G ) . Observe that g ( G ) = g ( M ( G ) ) = g ( M ∗ ) = g ∗ ( M ) ≥ max r ( M ) + 1 c , 9 , and δ ( G ) ≥ κ ′ ( G ) = g ∗ ( M ( G ) ) = g ( M ∗ ( G ) ) = g ( M ) ≥ 4 . From r ( M ) = r ( M ∗ ( G ) ) = m − n + 1 , we obtain g ( G ) ≥ max m − n + 2 c , 9 . By Lemma 11, G is four-regular. Because g ( G ) ≥ 9 , G cannot be complete; hence, Lemma 6 implies a ( G ) ≤ 2 . Thus, by Lemma 5, M has a spanning Eulerian subset. □

3.2. Fractional Arboricity and Strength of Matroids

The present work makes use of several established results from [21,22,32] as technical tools and also relies on fundamental concepts such as fractional arboricity and matroid strength. The notion of matroid strength was introduced by Cunningham in [33]. Further discussion regarding its properties can be found in [22,34]. Let M be a binary matroid. A maximal independent set in M is referred to as a base of M. For any nonempty subset X ⊆ E ( M ) with r ( X ) > 0 , we define the density function as d M ( X ) = | X | r ( X ) . When the matroid M is clear from context, we simply write d ( X ) . The quantities
η ( M ) = min X ⊆ E ( M ) r ( X ) < r ( M ) d ( M / X ) and γ ( M ) = max X ≠ ∅ X ⊆ E ( M ) d M ( X )
are known as the strength and the fractional arboricity of M, respectively.
Lemma 12.
Let p and q be integers satisfying p > q > 0 , and let M be a matroid with r ( M ) > 0 . Then, the following statements hold:
( i )
[32] γ ( M ) ≥ | E ( M ) | r ( M ) ≥ η ( M ) , and τ ( M ) = ⌊ η ( M ) ⌋ .
( i i )
[21] If there exists a subset X ⊆ E ( M ) (possibly empty) such that η ( M ∗ / X ) ≤ p q , then there exists a nonempty subset N ⊂ E ( M ) satisfying η ( M | N ) ≥ p p − q .
Lemma 13
([25]). Let M be a matroid with rank r ( M ) > 0 . If for some rational number q, | E ( M ) | r ( M ) = q , then there exists a subset X with 0 ≠ X ⊆ E ( M ) such that η ( M | X ) ≥ q .
Let X be a subset of edges in G. The graph obtained from G by removing all edges in X is written as G − X . When X consists of a single edge e, we denote G − { e } simply by G − e . For a vertex u, let E G ( u ) denote the set of edges incident to u in G. Define F ( g ) as the collection of all two-connected, non-planar simple graphs satisfying the following condition: G ∈ F ( g ) if and only if G contains a minimal three-edge-cut Z such that every edge cut X of G with | X | ≤ 3 intersects Z, and moreover, g ( G − Z ) ≥ g . Such a Z is called a special three-edge-cut of G. Clearly, for any graph G ∈ F ( g ) and any vertex v ∈ V ( G ) − V ( Z ) , we have d G ( v ) ≥ 4 .
Lemma 14
([25]). Let G ∈ F ( g ) where g ≥ 9 . Then,
| V ( G ) | ≥ 4 + 8 × 3 h − 2 , if g = 2 h + 1 ; 4 + 5 × 3 h − 2 , if g = 2 h .
Lemma 15.
Let G ∈ F ( g ) and Z be a special cut of G. If
g ≥ max { 2 ( | E ( G ) | − | V ( G ) | ) + 1 c , 9 } , 15 < c ≤ 16 ,
then
(1) 
η ( M ( G ) / Z ) ≤ 2 .
(2) 
M ∗ ( G ) ∖ Z contains a nonempty eligible proper subset.
Proof. 
Assume η ( M ( G ) / Z ) > 2 ; by Lemma 12, we have 2 < η ( M ( G ) / Z ) ≤ | E ( M ( G ) / Z ) | r ( M ( G ) / Z ) . As is well known, r ( M ( G ) / Z ) = r ( M ) − r ( M | Z ) , so 2 < | E ( M ( G ) / Z ) | r ( M ( G ) ) − r ( Z ) ≤ | E ( G ) | − 3 ( | V ( G ) | − 1 ) − 3 , which implies
| E ( G ) | − | V ( G ) | ≥ | V ( G ) | − 4 .
If g is odd, let g = 2 h + 1 . By g ≥ 9 , we have h ≥ 4 . From g ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 c ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 16 , we derive 2 h + 1 ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 16 , which implies | E ( G ) | − | V ( G ) | ≤ 16 ( 2 h + 1 ) 2 − 1 2 . By Lemma 14, we have | V ( G ) | − 4 ≥ 8 × 3 h − 2 . Substitute back into (8), we have
16 h + 15 2 ≥ 8 × 3 h − 2 .
Let φ ( h ) = 8 × 3 h − 2 − 16 h − 15 2 . Clearly, φ ′ ( h ) = 8 × 3 h − 2 ln 3 − 16 > 0 for h ≥ 4 . Then, φ ( h ) ≥ φ ( 4 ) = 1 2 > 0 , hence 16 h + 15 2 < 8 × 3 h − 2 for h ≥ 4 , which contradicts (9).
If g is even, let g = 2 h . By g ≥ 9 , we have h ≥ 5 . Form g ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 16 , we have 2 h ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 16 , which implies | V ( G ) | − 4 ≤ | E ( G ) | − | V ( G ) | ≤ 16 h − 1 2 . By Lemma 14, we have | V ( G ) | − 4 ≥ 5 × 3 h − 2 . Then,
16 h − 1 2 ≥ 5 × 3 h − 2 .
Let ϕ ( h ) = 5 × 3 h − 2 − 16 h + 1 2 . Clearly, ϕ ′ ( h ) = 5 × 3 h − 2 ln 3 − 16 > 0 for h ≥ 5 . Hence, ϕ ( h ) ≥ ϕ ( 5 ) = 111 2 > 0 , then 16 h − 1 2 < 5 × 3 h − 2 for h ≥ 5 , which contradicts (10). Then, η ( M ( G ) / Z ) ≤ 2 .
Take p = 2 and q = 1 . By applying Lemma 12(ii) with M ∗ = M ∗ ( G ) ∖ Z and X = ∅ , we obtain a nonempty subset N ⊂ E ( M ∗ ( G ) ∖ Z ) satisfying
η ( M ∗ ( G ) ∖ Z ) | N ≥ 2 and τ ( M ∗ ( G ) ∖ Z ) | N = η ( M ∗ ( G ) ∖ Z ) | N ≥ 2 .
Consequently, N forms a nonempty eligible proper subset of M ∗ ( G ) ∖ Z . □
Lemma 16.
Let G ∈ F ( g ) and Z be a special cut of G. If
g ≥ max { 2 ( | E ( G ) | − | V ( G ) | ) + 1 c , 10 } , 16 < c < 20 .
Then,
(1) 
η ( M ( G ) / Z ) ≤ 2 .
(2) 
M ∗ ( G ) ∖ Z contains a nonempty eligible proper subset.
Proof. 
Assume η ( M ( G ) / Z ) > 2 , by Lemma 12, we have 2 < η ( M ( G ) / Z ) ≤ | E ( M ( G ) / Z ) | r ( M ( G ) / Z ) . As is well known, r ( M ( G ) / Z ) = r ( M ) − r ( M | Z ) , so 2 < | E ( M ( G ) / Z ) | r ( M ( G ) ) − r ( Z ) ≤ | E ( G ) | − 3 ( | V ( G ) | − 1 ) − 3 , which implies
| E ( G ) | − | V ( G ) | ≥ | V ( G ) | − 4 .
If g is odd, let g = 2 h + 1 . By g ≥ 10 , we have h ≥ 5 . From g ≥ 2 ( | E ( G ) | − | V ( G ) | ) + 1 c > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , we derive 2 h + 1 > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , which implies | E ( G ) | − | V ( G ) | < 20 ( 2 h + 1 ) − 1 2 . By Lemma 14, we have | V ( G ) | − 4 ≥ 8 × 3 h − 2 . Substituting back into (12), we have
20 h + 19 2 > 8 × 3 h − 2 .
Let φ ( h ) = 8 × 3 h − 2 − 20 h − 19 2 . Clearly, φ ′ ( h ) = 8 × 3 h − 2 ln 3 − 20 > 0 for h ≥ 5 . Then, φ ( h ) ≥ φ ( 5 ) = 106.5 > 0 , hence 20 h + 19 2 < 8 × 3 h − 2 for h ≥ 4 , which contradicts (13).
If g is even, let g = 2 h . By g ≥ 10 , we have h ≥ 5 . From g > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , we have 2 h > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , which implies | V ( G ) | − 4 ≤ | E ( G ) | − | V ( G ) | ≤ 20 h − 1 2 . By Lemma 14, we have | V ( G ) | − 4 ≥ 5 × 3 h − 2 . Then,
20 h − 1 2 ≥ 5 × 3 h − 2 .
Let ϕ ( h ) = 5 × 3 h − 2 − 20 h + 1 2 . Clearly, ϕ ′ ( h ) = 5 × 3 h − 2 ln 3 − 20 > 0 for h ≥ 5 . Hence, ϕ ( h ) ≥ ϕ ( 5 ) = 71 2 > 0 , then 20 h − 1 2 < 5 × 3 h − 2 for h ≥ 5 , which contradicts (14). Then, η ( M ( G ) / Z ) ≤ 2 .
Set p = 2 and q = 1 . Applying Lemma 12(ii) with M ∗ = M ∗ ( G ) ∖ Z and X = ∅ , we find a nonempty subset N ⊂ E ( M ∗ ( G ) ∖ Z ) such that
η ( M ∗ ( G ) ∖ Z ) | N ≥ 2 and τ ( M ∗ ( G ) ∖ Z ) | N = η ( M ∗ ( G ) ∖ Z ) | N ≥ 2 .
Thus, N is a nonempty eligible proper subset of M ∗ ( G ) ∖ Z . □
Let F ′ ( g ) be a family of graphs such that G ∈ F ′ ( g ) if and only if G is a two-connected graph and there is an edge e of G such that e ∈ X for any edge-cut X with | X | ≤ 3 and g ( G − e ) ≥ g . We call such e a special edge of G.
Lemma 17.
Let G ∈ F ′ ( g ) , and e be a special edge of G. If g ≥ max { 2 ( | E ( G ) | − | V ( G ) | ) + 1 c , 9 } , 15 < c < 20 , then
(1) 
η ( M ( G ) / e ) ≤ 2 .
(2) 
M ∗ ( G ) ∖ e contains a nonempty eligible proper subset.
Proof. 
Let e = u 1 u 2 . Since G is two-connected and e ∈ X for any edge-cut X with | X | ≤ 3 , it follows that for any u ∈ V ( G ) − { u 1 , u 2 } ,
d G ( u ) = d G − e ( u ) ≥ 4 and d G − e ( u 1 ) + d G − e ( u 2 ) ≥ 4 .
Hence, G − e contains cycles
Assume that η ( M ( G ) / e ) > 2 . By Lemma 12, we have
2 < η ( M ( G ) / e ) ≤ | E ( M ( G ) / e ) | r ( M ( G ) / e ) = | E ( G ) | − 1 ( | V ( G ) | − 1 ) − 1 ,
which implies
| E ( G ) | − | V ( G ) | ≥ | V ( G ) | − 2 .
Let the minimum degree in G excluding u 1 and u 2 be δ , which is evidently greater than or equal to four, and let d G ( v ) = Δ ( G − e ) = t ≥ 4 for some v ∈ V ( G ) .
If g is odd, let g = 2 h + 1 . By g ( G − e ) ≥ g ≥ 9 , we have h ≥ 4 . Let V i denote the set of all vertices at distance i − 1 from vertex v. By the uniqueness of distances, for any i ≠ j , V i and V j are disjoint sets. Since each vertex in a layer contributes one degree to the layer above, and every vertex in the layer below is a neighbor of some vertex in the layer above, we have V i + 1 = N u ∈ V i − V i . It follows that
| V i + 1 | = ∑ u ∈ V i ( d ( u ) − 1 ) .
Therefore, the closer u 1 and u 2 are to v, the fewer vertices there are at distance at most h + 1 from v. Hence, we assume that u 1 , u 2 ∈ V 2 . Then, | V 1 | = 1 , | V 2 | = t , | V 3 | ≥ ( t − 2 ) ( δ − 1 ) + d u 1 + d u 2 − 2 ≥ ( t + δ − 2 ) ( δ − 1 ) − ( 2 δ − 4 ) , By (16), | V k | = | V 3 | ( δ − 1 ) k − 3 , k ∈ { 4 , 5 , ⋯ , h + 1 } . Then,
| V ( G ) | ≥ | V 1 | + ⋯ + | V h + 1 | ≥ 1 + t + t ( δ − 1 ) − ( 2 δ − 4 ) + [ t ( δ − 1 ) − ( 2 δ − 4 ) ] ( δ − 1 ) + ⋯ + [ t ( δ − 1 ) − ( 2 δ − 4 ) ] ( δ − 1 ) h − 2 ≥ 3 + δ ( δ − 1 ) h − 1 δ − 2 − 2 ( δ − 1 ) h − 1
Let φ ( δ ) = 3 + δ ( δ − 1 ) h − 1 δ − 2 − 2 ( δ − 1 ) h − 1 . Since φ ′ ( δ ) > 0 for δ ≥ 4 , it follows that φ ( δ ) ≥ φ ( 4 ) = 1 + 4 × 3 h − 1 , hence | V ( G ) | − 2 ≥ φ ( δ ) − 2 ≥ 4 × 3 h − 1 − 1 . By 2 h + 1 > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , we have 20 h + 19 2 ≥ | E ( G ) | − | V ( G ) | . Then,
20 h + 19 2 ≥ 4 × 3 h − 1 − 1 .
Let ϕ ( h ) = 4 × 3 h − 1 − 1 − 20 h − 19 2 . Since ϕ ′ ( h ) = 4 × 3 h − 1 ln 3 − 20 > 0 for h ≥ 4 , it follows that ϕ ( h ) ≥ ϕ ( 4 ) = 17.5 > 0 , then 20 h + 19 2 < 4 × 3 h − 1 − 1 for h ≥ 4 , which contradicts (17).
If g is even, let g = 2 h . By g ( G − e ) ≥ g ≥ 9 , we have h ≥ 5 . Let u v ∈ E ( G ) and u ∉ { u 1 , u 2 } . Let V i ′ denote the set of all vertices whose minimum distance to either v or u is i − 1 . By the uniqueness of distances, for any i ≠ j , V i ′ and V j ′ are disjoint sets. Since each vertex in a layer contributes one degree to the layer above, and every vertex in the layer below is a neighbor of some vertex in the layer above, we have V i + 1 ′ = N u ∈ V i ′ − V i ′ . It follows that
| V i + 1 ′ | = ∑ u ′ ∈ V i ′ ( d ( u ′ ) − 1 ) .
Therefore, the closer u 1 and u 2 are to v, the fewer vertices there are at distance at most h from v or v. Hence, we assume that u 1 , u 2 ∈ V 2 . Then, | V 1 ′ | = 2 , | V 2 ′ | = t + δ − 2 , | V 3 | ≥ ( t + δ − 4 ) ( δ − 1 ) + d u 1 + d u 2 − 2 ≥ ( t + δ − 2 ) ( δ − 1 ) − ( 2 δ − 4 ) , By (18), | V k ′ | = | V 3 ′ | ( δ − 1 ) k − 3 , k ∈ { 4 , 5 , ⋯ , h } . Then,
| V ( G ) | ≥ | V 1 | + | V 2 | + ⋯ + | V h | ≥ 2 + ( t + δ − 2 ) + ( t + δ − 2 ) ( δ − 1 ) − ( 2 δ − 4 ) + [ ( t + δ − 2 ) ( δ − 1 ) − ( 2 δ − 4 ) ] ( δ − 1 ) + ⋯ + [ ( t + δ − 2 ) ( δ − 1 ) − ( 2 δ − 4 ) ] ( δ − 1 ) h − 3 ≥ 4 + ( 2 δ − 2 ) ( δ − 1 ) h − 1 − 1 δ − 2 − 2 ( δ − 1 ) h − 2
Let φ ( δ ) = 4 + ( 2 δ − 2 ) ( δ − 1 ) h − 1 − 1 δ − 2 − 2 ( δ − 1 ) h − 2 . Since φ ′ ( δ ) > 0 for δ ≥ 4 , it follows that φ ( δ ) ≥ φ ( 4 ) = 1 + 7 × 3 h − 2 , hence | V ( G ) | − 2 ≥ φ ( δ ) − 2 ≥ 7 × 3 h − 2 − 1 . By g > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , we have 2 h > 2 ( | E ( G ) | − | V ( G ) | ) + 1 20 , hence 20 h − 1 2 ≥ | E ( G ) | − | V ( G ) | . Then
20 h − 1 2 ≥ 7 × 3 h − 2 − 1 .
Let ϕ ( h ) = 7 × 3 h − 2 − 1 − 20 h + 1 2 . Since ϕ ′ ( h ) = 7 × 3 h − 2 ln 3 − 20 > 0 for h ≥ 5 , it follows that ϕ ( h ) ≥ ϕ ( 5 ) = 88.5 , hence 20 h − 1 2 < 7 × 3 h − 2 − 1 for h ≥ 5 , which contradicts (19). Then, η ( M ( G ) / e ) ≤ 2 .
Take p = 2 and q = 1 . By Lemma 12(ii), with M ∗ = M ∗ ( G ) ∖ e and X = ∅ , there exists a nonempty subset N ⊂ E ( M ∗ ( G ) ∖ e ) such that
η ( M ∗ ( G ) ∖ e ) | N ≥ 2 and τ ( M ∗ ( G ) ∖ e ) | N = η ( M ∗ ( G ) ∖ e ) | N ≥ 2 .
Consequently, N is a nonempty eligible proper subset of M ∗ ( G ) ∖ e . □
Theorem 5.
For any simple connected regular matroid M = M 1 ⊕ i M 2 , i ∈ { 2 , 3 } , r ( M 2 ) ≤ r ( M 1 ) , M 2 is cographic but not graphic. If girth g ( M ) ≥ 4 and cogirth
g ∗ ( M ) ≥ max { ( r ( M ) + 1 ) c , 9 } , i f 15 < c ≤ 16 ; max { ( r ( M ) + 1 ) c , 10 } , i f 16 < c < 20 ,
then M has a spanning Eulerian subset.
Proof. 
By Lemma 2, M 2 is connected and free of loops. It is well known that a matroid M is connected if and only if its dual M ∗ is connected, and every graphic matroid is isomorphic to the cycle matroid of some connected graph. Since M 2 is both cographic and connected, it follows that M 2 ∗ is connected and there exists a connected loopless graph G such that M 2 ∗ = M ( G ) . Because the cycle matroid of a planar graph has a graphic dual, and M 2 = M ∗ ( G ) is non-graphic, G cannot be planar; consequently n ≥ 5 , and G is two-connected. Thus, M 2 = M ∗ ( G ) and r ( M 2 ) = r ( M ∗ ( G ) ) = | E ( G ) | − | V ( G ) | + 1 .
Let g satisfy (7) and (11). Observe that g ( G − Z ) = g ( M ( G ) ∖ Z ) = g ( M 2 ∗ ∖ Z ) . From Definition 1 together with Lemmas 1 and 4, we obtain
C ( M 2 ∗ ∖ Z ) ⊆ C ( M ∗ ) .
Consequently, g ( M 2 ∗ ∖ Z ) ≥ g ( M ∗ ) = g ∗ ( M ) , and therefore g ( G − Z ) ≥ g ∗ ( M ) . Since r ( M ) ≥ 2 r ( M 2 ) − i + 1 = 2 | E ( G ) | − | V ( G ) | + 1 − i + 1 ≥ 2 | E ( G ) | − | V ( G ) | , it follows that g ( G − Z ) ≥ g .
Since G is two-connected, each set E G ( v ) forms a minimal edge cut and thus corresponds to a circuit in M ∗ ( G ) = M 2 . Given that g ( M ) ≥ 4 and M = M 1 ⊕ i M 2 with i ∈ { 2 , 3 } , every cycle of size at most three in M 2 must have nonempty intersection with Z. If | Z | = 1 , then G ∈ F ′ ( g ) . By Lemma 17, M 2 ∖ Z = M ∗ ( G ) ∖ Z contains a nonempty eligible proper subset. If | Z | = 3 , then Z ∈ C ( M 2 ) implies that Z is a minimal three-edge-cut of G; consequently, Z qualifies as a special three-edge-cut of G, and G ∈ F ( g ) . By Lemmas 15 and 16, it follows again that M 2 ∖ Z = M ∗ ( G ) ∖ Z contains a nonempty eligible proper subset.
Let N be a nonempty eligible subset of M 2 . By Lemma 8, N is also a nonempty eligible subset of M and satisfies r ( N ) < r ( M ) . Let M ′ denote a binary reduction of M. Because N is a nonempty eligible subset of M with r ( N ) < r ( M ) , we have | E ( M ′ ) | < | E ( M ) | . If M ′ is eligible, then, by (3), it has spanning Eulerian subset. If M ′ is not eligible, Lemma 7(b) implies g ( M ′ ) ≥ 4 . Observe that for any subset T of the ground set, g ( M ∗ ) ≤ g ( M ∗ ∖ T ) ; consequently, g ∗ ( M ) ≤ g ∗ ( M / T ) . But since M ′ is the result of applying a series of contraction operations to M, then g ∗ ( M ′ ) ≥ g ∗ ( M ) . It follows that M ′ without a coloop and M ′ can be decomposed into connected components H 1 , H 2 , ⋯, H c within M ′ (where M ′ itself might be a single connected component), then M ′ = H 1 ⊕ H 2 ⊕ ⋯ ⊕ H c . For any connected component H in M ′ , by Lemma 4, g ∗ ( H ) = g ( H ∗ ) ≥ g ( M ′ ∗ ) = g ∗ ( M ′ ) = g ∗ ( M ) . By g ( H ) ≥ g ( M ′ ) ≥ 4 , and since all minors of any regular matroid are regular, H is also regular. Since M is the minimal counterexample, it follows that H has a spanning Eulerian subset. Then, M ′ has a spanning Eulerian subset. By Lemma 7(a), M has a spanning Eulerian subset, which is a contradiction. □

4. Proof of Theorem 2

By Lemma 9, we may restrict our attention to matroids whose girth is greater than three. Let M be a counterexample of minimum ground set | E ( M ) | size, and suppose g ( M ) ≥ 4 . Seymour’s Decomposition Theorem (Lemma 3) yields a decomposition M = M 1 ⊕ i M 2 , where i ∈ { 1 , 2 , 3 } and each M 1 , M 2 is either graphic, cographic, or isomorphic to R 10 . Because M is connected, the case i = 1 forces M to be graphic, cographic, or M ≅ R 10 . When i = 2 or 3, we have | E ( M 1 ) | , | E ( M 2 ) | ≥ 1 , so M 1 and M 2 are proper minors of M. The condition g ∗ ( M ) ≥ 9 together with Lemma 10 rules out R 10 as a possibility for either M 1 or M 2 . Additionally, it is well known that graphic matroids are closed under i-sums. Thus, either M itself falls into the graphic or cographic case, or else M = M 1 ⊕ i M 2 with i ∈ { 2 , 3 } where at least one of M 1 , M 2 is cographic but not graphic.
Let M be a counterexample with the minimum cardinality of the ground set and r ( M 2 ) ≤ r ( M 1 ) . By Lemma 10(i), Theorems 4 and 5, M is graphic or M 2 is graphic. If M is graphic, there is a connected graph G such that M = M ( G ) . Since κ ′ ( G ) = g ∗ ( M ( G ) ) = g ∗ ( M ) ≥ 9 and (1), M has a spanning Eulerian subset, which is a contradiction. Hence, M 2 is graphic.
Since M 2 is graphic, there is a connected graph G 1 such that M 2 ≅ M ( G 1 ) . By Lemma 4, we have M ∗ = M 1 ∗ △ M ∗ ( G 1 ) . Then, C ( M ∗ ( G 1 ) ∖ Z ) ⊆ C ( M ∗ ) from Definition 1. Thus, every minimal edge cut of G 1 is a circuit of M ∗ . Note that C ( M ∗ ( G 1 ) ∖ Z ) = C ( M ∗ ( G 1 / Z ) ) , and for any v ∈ V ( G ) − V ( Z ) , E G 1 ( v ) is a bond of M ( G 1 / Z ) . Hence, | E G 1 ( v ) | ≥ g ( M ∗ ( G 1 / Z ) ) ≥ g ( M ∗ ) ≥ 9 . Therefore, d G 1 ( v ) ≥ 9 for any v ∈ V ( G 1 ) − V ( Z ) .
If | Z | = 1 , let Z = { e } . By Definition 2, e is neither a loop nor a coloop of M ( G ) and | E ( M 2 ) | ≥ 3 . Since g ( M ) ≥ 4 , C ( M ( G ) ∖ Z ) ⊆ C ( M ) and G is connected, it follows that V ( G 1 ) − V ( e ) ≠ ∅ . Since d G 1 ( v ) ≥ 9 for any v ∈ V ( G 1 ) − V ( Z ) , we have | V ( G 1 ) | ≥ 10 . Therefore, 2 | E ( G 1 − Z ) | ≥ ∑ v ∈ V ( G 1 − Z ) − V ( e ) d G 1 − Z ( v ) ≥ 9 ( | V ( G 1 − Z ) | − 2 ) , and so | E ( G 1 − Z ) | | V ( G 1 − Z ) | − 1 > 2 . By Lemma 13, M 2 ∖ Z = M ( G 1 ) ∖ Z contains a nonempty subset N 1 with η ( N 1 ) ≥ 2 . Thus, N 1 is a nonempty eligible subset of M. Let M ′ be a binary reduction of M. By Lemma 8, | E ( M ′ ) | < | E ( M ) | . Since M ′ is a restriction of M, it follows that g ∗ ( M ′ ) ≥ g ∗ ( M ) . By the minimality of M and Lemma 7, M ′ has a spanning Eulerian subset, and so M has a spanning Eulerian subset, a contradiction.
If | Z | = 3 , since g ( M ) ≥ 4 and C ( M ( G 1 ) ∖ Z ) ⊆ C ( M ) , we have g ( M ( G 1 ) ∖ Z ) ≥ 4 . By Definition 2 and g ( M ( G 1 ) ∖ Z ) ≥ 4 , we have V ( G 1 ) − V ( Z ) ≠ ∅ . Since d G 1 ( v ) ≥ 9 for any v ∈ V ( G 1 ) − V ( Z ) , then | V ( G 1 ) | ≥ 10 . Hence, 2 | E ( G 1 − Z ) | = ∑ v ∈ V ( G 1 − Z ) − V ( Z ) d G 1 − Z ( v ) ≥ 9 ( | V ( G 1 − Z ) | − 3 ) , and so | E ( G 1 − Z ) | | V ( G 1 − Z ) | − 1 > 2 . By Lemma 13, M 2 ∖ Z = M ( G 1 ) ∖ Z contains a nonempty subset N 2 with η ( N 2 ) ≥ 2 . Thus, N 2 is a nonempty eligible subset of M. Let M ′ be a binary reduction of M. By Lemma 8, | E ( M ′ ) | < | E ( M ) | . Since M ′ is a restriction of M, it follows that g ∗ ( M ′ ) ≥ g ∗ ( M ) . By the minimality of M and Lemma 7, M ′ has a spanning Eulerian subset, and so M has a spanning Eulerian subset, a contradiction.

5. Open Problems

In Theorem 2, we only provided an explicit f ( c ) , but we did not prove that this f ( c ) was the minimum value for which the result still holds. Therefore, we may consider the following problems:
  • Problem 1.
For a given integer c, what is the minimum possible value of f ( c ) such that the pair ( c , f ( c ) ) is an Eulerian pair for regular matroids?
  • Problem 2.
Can the result of Theorem 2, which holds for regular matroids, be extended to binary matroids, or more generally to arbitrary matroids?

Author Contributions

Conceptualization, X.Q. and B.H.; software, F.Y.; writing—original draft preparation, X.Q.; writing—review and editing, X.Q. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by National Natural Science Foundation of China (No. 11961055).

Data Availability Statement

The research reported in this article did not involve the use of any data.

Conflicts of Interest

The authors state that they have no financial or personal relationships that could have inappropriately affected or influenced the content of this article.

References

  1. Catlin, P.A. Spanning Eulerian subgraphs and matchings. Discret. Math. 1989, 76, 95–116. [Google Scholar] [CrossRef]
  2. Lai, H.-J. Eulerian subgraphs containing given edges. Discret. Math. 2001, 230, 63–69. [Google Scholar] [CrossRef][Green Version]
  3. Meng, J.X.; Dong, C.; Liu, J. A new condition on dominated pair degree sum for a digraph to be supereulerian. Discret. Appl. Math. 2025, 362, 389–396. [Google Scholar]
  4. Messinger, M.E.; Porter, A. Eulerian k-dominating reconfiguration graphs. Discret. Math. Theor. Comput. Sci. 2025, 27, 2. [Google Scholar] [CrossRef]
  5. Ji, M.; Meng, K.; Ding, K. Euler characteristics and homotopy types of definable sublevel sets, with applications to topological data analysis. arXiv 2023, arXiv:2309.03142. [Google Scholar] [CrossRef]
  6. Li, L.; Ma, Q.; Ouyang, C.; Paetzold, J.C.; Rueckert, D.; Kainz, B. Topology optimization in medical image segmentation with fast χ Euler characteristic. IEEE Trans. Med. Imaging 2025, 44, 5221–5232. [Google Scholar] [CrossRef] [PubMed]
  7. Gan, X. GraphService: Topology-aware constructor for large-scale graph applications. ACM Trans. Archit. Code Optim. 2025, 22, 2. [Google Scholar] [CrossRef]
  8. Yang, J.; Gao, C.; Zhou, Z.; Zhang, D.; Bu, F.; Chen, X.; Song, J. Three-dimensional peridynamics based on matrix operation and its application in rock mass compression failure simulation. Comput. Geotech. 2025, 185, 107354. [Google Scholar] [CrossRef]
  9. Chen, Y.; Li, H.; Song, Y.; Zhu, X. Recoding hybrid stochastic numbers for preventing bit width accumulation and fault tolerance. IEEE Trans. Circuits Syst. I 2025, 72, 1243–1255. [Google Scholar] [CrossRef]
  10. Li, L.; Xia, Y.; Ren, S.; Yang, X. Homogeneity pursuit in the functional-coefficient quantile regression model for panel data with censored data. Stud. Nonlinear Dyn. Econom. 2025, 29, 323–348. [Google Scholar] [CrossRef]
  11. Ren, Y.; Zhang, J.; Xia, Y.; Wang, R.; Xie, F.; Guan, J.; Zhang, H.; Zhou, S. Regression-based conditional independence test with adaptive kernels. Artif. Intell. 2025, 347, 104391. [Google Scholar] [CrossRef]
  12. Bondy, J.A.; Murty, U.S.R. Graph Theory; GTM 244; Springer: New York, NY, USA, 2008. [Google Scholar]
  13. Oxley, J.G. Matroid Theory, 2nd ed.; Oxford University Press: New York, NY, USA, 2011. [Google Scholar]
  14. Boesch, F.T.; Suffel, C.; Tindell, R. The spanning subgraphs of eulerian graphs. J. Graph Theory 1977, 1, 79–84. [Google Scholar] [CrossRef]
  15. Pulleyblank, W.R. A note on graphs spanned by eulerian graphs. J. Graph Theory 1979, 3, 309–310. [Google Scholar] [CrossRef]
  16. Catlin, P.A. Supereulerian graphs, a survey. J. Graph Theory 1992, 16, 177–196. [Google Scholar] [CrossRef]
  17. Chen, Z.H.; Lai, H.-J. Reduction techniques for super-Eulerian graphs and related topics: A survey. In Combinatorics and Graph Theory 95; World Scientific Publishing: River Edge, NJ, USA, 1995; Volume I, pp. 53–69. [Google Scholar]
  18. Lai, H.-J.; Shao, Y.; Yan, H. An update on supereulerian graphs. WSEAS Trans. Math. 2013, 12, 926–940. [Google Scholar]
  19. Lai, H.-J.; Liu, B.; Liu, Y.; Shao, Y. Spanning cycles in regular matroids without M*(K5) minors. Eur. J. Comb. 2008, 29, 298–310. [Google Scholar] [CrossRef]
  20. Veblen, O. An application of modular equations in analysis situs. Ann. Math. 1912, 14, 86–94. [Google Scholar] [CrossRef]
  21. Huo, B.F.; Du, Q.S.; Li, P.; Wu, Y.; Yin, J.; Lai, H.-J. Supereulerian regular matroids without small cocircuits. J. Graph Theory 2023, 102, 107–127. [Google Scholar] [CrossRef]
  22. Li, P.; Lai, H.-J.; Shao, Y.; Zhan, M. Spanning cycles in regular matroids without small cocircuits. Eur. J. Comb. 2012, 33, 1765–1776. [Google Scholar] [CrossRef]
  23. Qin, X.X.; Zhao, F.Y.; Lai, H.-J.; Huo, B.F. Supereulerian of regular matroids with cogirth conditions. Discret. Appl. Math. 2025, 377, 327–334. [Google Scholar] [CrossRef]
  24. Oxley, J.G. Matroid Theory; Oxford University Press: New York, NY, USA, 1992. [Google Scholar]
  25. Qin, X.X.; Huo, B.F.; Ye, F.L.; Lai, H.-J. Supereulerian of regular matroids with cogirth conditions II. under review.
  26. Jaeger, F. A note on subeulerian graphs. J. Graph Theory 1979, 3, 91–93. [Google Scholar] [CrossRef]
  27. Catlin, P.A. A reduction method to find spanning Eulerian subgraphs. J. Graph Theory 1988, 12, 29–44. [Google Scholar] [CrossRef]
  28. Seymour, P.D. Decomposition of regular matroids. J. Combin. Theory Ser. B 1980, 28, 305–359. [Google Scholar] [CrossRef]
  29. Seymour, P.D. Matroids and multicommodity flows. Eur. J. Comb. Theory Ser. B 1981, 2, 257–290. [Google Scholar] [CrossRef][Green Version]
  30. Yin, J.; Huo, B.; Lai, H.-J. Note on the Connectivity of Binary Matroids. Discontinuity Nonlinearity Complex. 2022, 11, 405–408. [Google Scholar] [CrossRef]
  31. Kronk, H.V.; Mitchem, J. Critical point arboritic graphs. J. Lond. Math. Soc. 1975, 3, 459–466. [Google Scholar] [CrossRef]
  32. Catlin, P.A.; Grossman, J.W.; Hobbs, A.M.; Lai, H.-J. Fractional arboricity, strength and principal partitions in graphs and matroids. Discret. Appl. Math. 1992, 40, 285–302. [Google Scholar] [CrossRef]
  33. Cunningham, W.H. Optimal attack and reinforcement of a network. J. Assoc. Comp. Mach. 1985, 32, 549–561. [Google Scholar] [CrossRef]
  34. Li, P. Bases and Cycles in Graphs and Matroids. Doctoral Dissertation, West Virginia University, Morgantown, WV, USA, 2012. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.