Resolvability and convexity properties in the Sierpi´nski product of graphs

Let G and H be graphs and let f : V ( G ) → V ( H ) be a function. The Sierpi´nski product of G and H with respect to f , denoted by G ⊗ f H , is deﬁned as the graph on the vertex set V ( G ) × V ( H ), consisting of | V ( G ) | copies of H ; for every edge gg ′ of G there is an edge between copies gH and g ′ H of H associated with the vertices g and g ′ of G , respectively, of the form ( g, f ( g ′ ))( g ′ , f ( g )). The Sierpi´nski metric dimension and the upper Sierpi´nski metric dimension of two graphs are determined. Closed formulas are determined for Sierpi´nski products of trees, and for Sierpi´nski products of two cycles where the second factor is a triangle. We also prove that the layers with respect to the second factor in a Sierpi´nski product graph are convex.


Introduction
Sierpiński graphs represent a very interesting and widely studied family of graphs.They were introduced in 1997 in the paper [21], where the primary motivation for their introduction was the intrinsic link to the Tower of Hanoi problem, for the latter problem see the book [16].Intensive research of Sierpiński graphs led to a review article [15] in which state of the art up to 2017 is summarized and unified approach to Sierpiński-type graph families is also proposed.Later research on Sierpiński graphs includes [3,7,8,9,19,25,32].
In this paper we study a recent generalization of Sierpiński graphs proposed by Kovič, Pisanski, Zemljič, and Žitnik in [23].Let G and H be graphs and let f : V (G) → V (H) be an arbitrary function.The Sierpiński product of graphs G and H with respect to f , denoted by G ⊗ f H, is defined as the graph on the vertex set V (G) × V (H) with edges of two types: • Type-1 edge: (g, h)(g, h ′ ) is an edge of G ⊗ f H for every vertex g ∈ V (G) and every edge hh ′ ∈ E(H), • Type-2 edge: (g, f (g ′ ))(g ′ , f (g)) is an edge of G ⊗ f H for every edge gg ′ ∈ E(G).
We observe that the edges of Type-1 induce n(G) = |V (G)| copies of the graph H in the Sierpiński product G ⊗ f H.For each vertex g ∈ V (G), we let gH be the copy of H corresponding to the vertex g.A Type-2 edge joins vertices from different copies of H in G ⊗ f H, and is called a connecting edge of G ⊗ f H.A vertex incident with a connecting edge is called a connecting vertex.We observe that two different copies of H in G ⊗ f H are joined by at most one edge.We denote by H G be the family of functions from V (G) to V (H).
It might be readily observed that the Sierpiński product is closely related to other product graphs.For instance, by considering a constant function f in the product, we obtain graphs which are indeed the same as the so-called rooted product graphs (see [12] for its definition).Also, selecting the identity function id ∈ G G , the Sierpiński product G ⊗ id G is the (first iteration of the) generalized Sierpiński graph in the sense of [13].Moreover, a Sierpiński product can also be considered as a subgraph of the (Cartesian, strong or lexicographic) product.Consequently, any contribution to the study of the Sierpiński product could give some more knowledge on these related products.
In the next two subsections we give motivation, basic terminology, and notation concerning the classical metric dimension of graphs, and introduce the study of the Sierpiński metric dimension and the upper Sierpiński metric dimension.Thereafter in Section 2 we determine the upper Sierpiński metric dimension for Sierpiński products of arbitrary trees.A general lower bound is established for the Sierpiński metric dimension for products of two trees, and an exact formula when the first factor is a path.In Section 3 a closed formula is determined for both dimensions when the first factor in the product is an arbitrary cycle and the second factor a triangle.In Section 4 we prove that the layers with respect to the second factor in a Sierpiński product graph are convex.In Section 5 we pose several open problems.

The metric dimension of graphs
The distance between two vertices u and v in a connected graph G, denoted d G (u, v), is the number of edges in a shortest path from u to v, that is, If every two distinct vertices of G have different metric S-representations, then the set S is called a resolving set of G (also called a metric generator ).The metric dimension of G, denoted by dim(G), is the cardinality of a smallest possible resolving set in G.A metric basis of G is a resolving set of cardinality dim(G).A vertex v in a graph G is said to distinguish (or resolve) two vertices x and The concept of the metric dimension of a graph was birthed independently by Harary and Melter [14] in 1976 and by Slater [29] in 1975, and is now well studied in graph theory.To date MathSciNet lists over 380 papers on metric dimension in graphs, covering a large number of different investigations dealing with theoretical and applied results on such parameter.
According to the structural properties of resolving sets in graphs, they can easily be used to model several practical situations in which uniquely recognizing points or locations is required.That was precisely one of the motivations of the seminal works [14] and [29], where resolving sets appeared to be used for the location of intruders in networks.Further on, some other related models and applications have appeared here and there.Among them, we remark the recent work [31], where the authors presented a connection between some metric dimension parameter and the representation of genomic sequences.Among the theoretical studies on this topic, the literature contains a wide range of different contributions, some recent and remarkable articles are for instance [6,11,28].For more information on investigations on the classical version we suggest the fairly complete survey [30].
With respect to the theoretical studies, the metric dimension of graph products and graph operations has attracted the attention of several investigations.In this sense, we mention a few interesting contributions related with this exposition due to the relationship between the Sierpiński product and some other products previously mentioned.The metric dimension of Cartesian product graphs has been considered in several works like [4], for the general case, and among other ones, in [5,18,20] for some particular examples of Cartesian products.The lexicographic product of graphs has been studied with respect to its metric dimension in [17,27], while the strong product has been considered in [1,26].On the other hand, the metric dimension of the rooted product has been dealt with in [10,24].

Sierpiński metric dimension
Let G and H be graphs and H G be the family of functions from V (G) to V (H).We introduce new types of metric dimension, the Sierpiński metric dimension, denoted by dim S (G, H), as the minimum over all functions f from H G of the metric dimension of the Sierpiński product with respect to f , and upper Sierpiński metric dimension, denoted by Dim S (G, H), as the maximum over all functions f ∈ H G of the metric dimension of the Sierpiński product with respect to f .That is, dim S (G, H) := min We might remark that the classical metric dimension of Sierpiński graphs was already studied in [22], as well as, that of the generalized Sierpiński graphs over stars was considered in [2].

Sierpiński products of trees
A vertex of degree at least 3 in a tree T is called a branch vertex (also called a major vertex in the literature).A leaf u of T is called a terminal leaf of a branch vertex v of T if d T (u, v) < d T (u, w) for every other branch vertex w of T .The terminal degree of a branch vertex v is the number of terminal leaves associated with v.A branch vertex v of T is an exterior branch vertex of T if it has positive terminal degree.The path from a terminal leaf to the vertex immediately preceding the branch vertex that it is closest to is called a terminal path.Thus, every vertex on a terminal path in T is either a leaf of T or has degree 2 in T .A vertex on a terminal path that has degree 2 in T is called an internal terminal vertex.Equivalently, every vertex on a terminal path that is not a terminal leaf, is an internal terminal vertex.Thus if u is an internal terminal vertex in T , then the vertex u is an internal vertex of a path P that joins a leaf and a branch vertex closest to that leaf in T where every internal vertex of P has degree 2 in T .
Let n 1 (T ) denote the number of leaves of T , and let ex(T ) denote the number of exterior branch vertices of T .The formula for the metric dimension of a tree reads as follows.
Theorem 2.1.( [14,29]) If T is a tree that is not a path, then It is clear that dim(P n ) = 1.Combining this fact with Theorem 2.1 yields the following consequence.
Let T be a tree that is not a path, and let v 1 , . . ., v k be the exterior branch vertices in T that have terminal degree at least 2. If the exterior branch vertex v i has terminal degree ℓ i ≥ 2 and if L i is a set consisting of all terminal leaves but one associated with v i for all i ∈ [k], then (1) can be equivalently stated as: and the set is a metric basis of T (of cardinality dim(T )).We call the basis B(T ) a standard metric basis of T .Thus, every vertex in a standard metric basis of a tree T is a leaf, and such a basis contains all but one selected fixed leaf associated with the exterior branch vertex of terminal degree at least 2 in T .

Upper Sierpiński metric dimension in trees
In this section we determine the upper Sierpiński metric dimension of the Sierpiński product of trees.Notice that for any trees T 1 and T 2 and any function f ∈ H G , the Sierpiński product Proof.Let w be a branch vertex of T 2 , and let f w : V (T 1 ) → V (T 2 ) be the constant function defined by f w (v) = w for every vertex v ∈ V (T 1 ).The exterior branch vertices in T 1 ⊗ fw T 2 are precisely the exterior branch vertices in each of the copies of T 2 , and so ex(T 1 ⊗ fw T 2 ) = n(T 1 )ex(T 2 ).Moreover, the leaves in T 1 ⊗ fw T 2 are precisely the leaves in each of the copies of T 2 , and so n 1 (T 1 ⊗ fw T 2 ) = n(T 1 )n 1 (T 2 ).Therefore, as T 1 ⊗ fw T 2 is a tree, by (1) we have We next show that Dim Let us first consider that T 2 is a path P n with n ≥ 3. Hence, we here indeed need to prove that dim(T 1 ⊗ f P n ) ≤ n(T 1 ) since dim(P n ) = 1.Suppose to the contrary that dim(T 1 ⊗ f P n ) > n(T 1 ).Thus, from (1) we have that n which means there is a positive integer k such that ex(T − k exterior branch vertices, then there must be at least k copies of P n in T 1 ⊗ f P n not containing any exterior branch vertex.This situation can only happen when the connecting edges of T 1 ⊗ f P n in such copies of P n are incident with at least one leaf of each of these copies.Consequently, we deduce that n which is a contradiction with our assumption, and so dim(T 1 ⊗ f P n ) ≤ n(T 1 ) as required.
We next consider the case when T 2 is not a path.Let E c = {e 1 , e 2 , . . ., e m(T 1 ) } be the set of connecting edges in T 1 ⊗ f T 2 .We order these connecting edges and define e i as the ith connecting edge of T 1 ⊗ f T 2 for i ∈ [m(T 1 )].We next define forests X 0 , X 1 , . . ., X m(T 1 ) as follows.Let X 0 be obtained from the tree T 1 ⊗ f T 2 by removing the connecting edges in E c .We note that X 0 is the disjoint union of n(T 1 ) copies of the tree T 2 .Applying (1) to each component of the forest We now define the forests X 1 , . . ., X m(T 1 ) as follows.For i ∈ [m(T 1 )], let X i be the forest obtained from X i−1 by adding the ith connecting edge, that is, .
Applying (1) to each component of the forest which represents the number of vertices of degree 1 in X i−1 which are of degree at least 2 in X i .Roughly speaking, Φ 1 (X i ) is the number of degree 1 vertices in X i−1 "destroyed" by adding the ith connecting edges e i to X i−1 when constructing X i .Set further where ex(X i−1 ) − ex(X i ) is the difference between the number of exterior branch vertices in X i−1 and the number of exterior branch vertices in X i .We note that dim for all i ∈ [m(T 1 )].We would like to show that for all i ∈ [m(T 1 )], which would imply that dim Hence to prove the theorem, it remains to show that (3) holds.For this purpose, let the ith connecting edge e i join vertices x i and y i in X i−1 when constructing X i .
Suppose that x i is neither an internal terminal vertex nor a leaf in X i−1 .In this case, the vertex x i contributes 0 to both terms Φ 1 (X i ) and Φ 2 (X i ).
Suppose that x i is an internal terminal vertex in X i−1 .Let w i be the exterior branch vertex associated with the vertex x i in X i−1 .In this case, when e i is added to X i−1 , the vertex x i becomes an exterior branch vertex in X i , while the vertex w i may no longer be an exterior branch vertex, implying that the vertex x i contributes 0 to the term Φ 1 (X i ), and contributes at least 1 − 1 = 0 to the term Φ 2 (X i ).
Suppose that x i is a leaf in X i−1 .As before, let w i be the exterior branch vertex associated with the vertex x i in X i−1 .In this case, when e i is added to X i−1 , the leaf x i in X i−1 is not a leaf in X i , and therefore the vertex x i contributes 1 to the term Φ 1 (X i ).Moreover, the effect of adding e i is that the vertex w i may no longer be an exterior branch vertex, implying that the vertex x i contributes at least −1 to the term Φ 2 (X i ).In all of the above three cases, the contribution of x i to Φ 1 (X i ) + Φ 2 (X i ) is at least 0. Analogous arguments hold for the vertex y i , showing that the contribution of y i to Φ 1 (X i ) + Φ 2 (X i ) is at least 0. Therefore, (3) holds.This completes the proof of Theorem 2.3.✷

Sierpiński metric dimension in trees
In this section we study the Sierpiński metric dimension of the Sierpiński product of trees.The Sierpiński metric dimension of two paths is given by the following result.Proposition 2.4.If T 1 and T 2 are both paths, then dim S (T 1 , T 2 ) = 1.
Proof.Let T 1 = P n and let T 2 = P m .If n = 1 or m = 1, then T 1 ⊗ f T 2 is a path, and so by Corollary 2.2, dim S (T 1 , T 2 ) = 1.Hence we may assume that n ≥ 2 and m ≥ 2, for otherwise the result is immediate.Let the path T 2 be an x, y-path that starts at vertex x and ends at vertex y, and let T 1 be the path v 1 v 2 . . .v n .Let f : V (T 1 ) → V (T 2 ) be the function defined by x; i (mod 4) ∈ {1, 2}, y; otherwise.
for all i ∈ [n].In this case the Sierpiński product T 1 ⊗ f T 2 is a path P nm , and so by Corollary 2.2, dim In view of Proposition 2.4 it is only of interest to study the Sierpiński product of two trees with at least one of the trees not a path.In this case, we shall establish the following lower bound on the Sierpiński metric dimension, where we use the notation d T (v) to represent the degree of a vertex v in T .Lemma 2.5.If T 1 and T 2 are trees, where T 2 is not a path, then Proof.Let f : V (T 1 ) → V (T 2 ) be an arbitrary function.Recall that for each vertex v ∈ V (T 1 ), vT 2 denotes the copy of the tree T 2 in T 1 ⊗ f T 2 corresponding to the vertex v.We let C v be the set of vertices in vT 2 that are connecting vertices in T 1 ⊗ f T 2 .Thus, each vertex in C v is incident with a connecting edge in T 1 ⊗ f T 2 that joins that vertex to a vertex in a copy of T 2 different from vT 2 .We note that since every edge incident with v in the tree T 1 is associated with a connecting edge in T 1 ⊗ f T 2 that is incident with a vertex in vT 2 .Let B be a standard metric basis of T 1 ⊗ f T 2 , and so dim(T 1 ⊗ f T 2 ) = |B| and the basis B contains all but one leaf associated with the terminal vertices of degree at least 2 in T 1 ⊗ f T 2 .Let B v be the restriction of B to vT 2 , that is, for every vertex v ∈ V (T 1 ).The set B v ∪ C v is a resolving set in the tree vT 2 , and so and so This establishes the desired lower bound in the statement of the theorem.✷ Using Lemma 2.5, we have the following result.
Theorem 2.6.For n ≥ 2, if T 1 = P n and T 2 is a tree that is not a path, then Proof.Let T 1 = P n and let T 2 be a tree that is not a path.By Corollary 2.2, dim(T 2 ) ≥ 2. By Lemma 2.5, dim noting that T 1 contains two vertices of degree 1 and n − 2 vertices of degree 2. Hence, it suffices for us to show that By assumption, T 2 is not a path.Hence, T 2 contains at least one exterior branch vertex with terminal degree at least 2. If T 2 contains two distinct exterior branch vertices both with terminal degree at least 2, then let u 1 and u 2 be two selected (terminal) leaves associated with these two exterior branch vertices.If T 2 contains only one exterior branch vertex, then T 2 is a star or a subdivided star with terminal degree at least 3.In this case, let u 1 and u 2 be two arbitrary leaves in T 2 .Let T 1 be the path v 1 v 2 . . .v n , and define the function for all i ∈ [n].For notational simplicity, instead of v i T i 2 we simply write iT 2 for the copy of T 2 in the Sierpiński product T 1 ⊗ f T 2 that corresponds to the vertex v i for all i ∈ [n].We note that the copy 1T 2 and nT 2 both contain one less leaf in the product T 1 ⊗ f T 2 , while every copy iT 2 where i ∈ [n − 1] \ {1} contains two fewer leaves in the product On the other hand, by our choice of the vertices u 1 and u 2 , the number of exterior branch vertices in each copy iT 2 where i ∈ [n] remains unchanged in the product completing the proof of Theorem 2.6.✷

Sierpiński products of cycles
In this section we study the Sierpiński metric dimension of the Sierpiński product of cycles, and proved closed formulas for the cases in which at least one of the factors is a triangle.
, and let the vertices of the cycle G be g 1 , g 2 , . . ., g n in the natural order of adjacencies.Let f : V (G) → V (H) be an arbitrary function and consider G ⊗ f H.
For notational simplicity, let iH denote g i H for all i ∈ [n], that is, iH is the ith copy of H corresponding to the vertex g i of G. Let x i y i+1 be the connecting edge between iH and (i + 1)H, i ∈ [n], where the sum is made mod n.Thus, x i = (g i , f (g i+1 )) and y i+1 = (g i+1 , f (g i )).Set further w i be a vertex in iH different from x i and y i .Note that if x i = y i , then V (iH) = {x i , y i , w i }.In case x i = y i , then let w ′ i be the third vertex of V (iH), that is, in this case we have Consequently, we now obtain a contradiction from each situation in which we would suppose that d(x, u 2(⌈k/2⌉−1) ) = d(y, u 2(⌈k/2⌉−1) ), for any possible assumption taken for x, y as considered before.Therefore, x, y are identified by u 2(⌈k/2⌉−1) , and so, S is a resolving set.Since F k is not a path, then S it is indeed a metric basis, as claimed.✷ Proof.Clearly, dim S (C n , C 3 ) ≥ 2, hence we only need to prove that dim S (C n , C 3 ) ≤ 2. We use the notation from the first two paragraphs of the proof of Theorem 3.1.In particular, G = C n and H = C 3 .
Case 1: n is odd.Let B be the sequence 3, 1, 2, 3. Let f n be defined as follows.
Case 2: n is even.Let C be the sequence 2, 2, 3, 3. Let f n be defined as follows.
It is straightforward to verify that for every n ≥ 3, the Sierpiński product G ⊗ fn H has the following structure.For each i we have x i = y i , and hence w i is the third vertex from V (iH), see Fig. 1 where G ⊗ f 5 H is drawn in two different ways.It is now clear that G ⊗ fn H ∼ = F n and hence Lemma 3.2 completes the argument.✷

Convexity property of Sierpiński products
In this section, we establish a distance convex property of the Sierpiński product of two graphs.Recall that a subgraph H of a graph G is convex if whenever u, v ∈ V (H) and P is a shortest u, v-path in G, then P lies completely in H.  Proof.Throughout the proof, let X = G ⊗ f H. Suppose on the contrary that there exists vertices u, v ∈ V (gH) such that d X (u, v) < d gH (u, v).Note that this does not happen in trees, hence in the rest we may assume that G contains cycles.Suppose now that u, v, and gH are selected such that d X (u, v) is as small as possible among all such counterexamples.Let u = (g, h), v = (g, h ′ ), and let P be a shortest u, v-path in X. Set further g 1 = g.
Claim.The shape of the path P is as follows.Let g 1 , . . ., g k , k ≥ 2, be the vertices of G ordered such that P passes through g 1 H, . . ., g k H in that order.Then P starts with the connecting edge (g 1 , f (g 2 ))(g 2 , f (g 1 )), proceeds with a geodesic P 2 in g 2 H between (g 2 , f (g 1 )) and (g 2 , f (g 3 )), then continuing with the connecting edge (g 2 , f (g 3 ))(g 3 , f (g 2 )), and so on.Finally P arrives at g k H, proceeds along a geodesic in G k between (g k , f (g k−1 )) and (g k , f (g 1 )), and ends with the connecting edge (g k , f (g 1 ))(g 1 , f (g k )), where f (g k ) = h ′ .See Fig. 2.
Note first that k ≥ 3 because k = 2 would imply that there are two connecting edges between g 1 H and g 2 H.
We next show that the vertices g 1 , . . ., g k are pairwise different.Suppose on the contrary that there exist i and j, such that 2 ≤ i < j < k and g j+1 = g i .Let P ′ be the subpath of P between the vertices u ′ = (g i , f (g i+1 )) and v ′ = (g j+1 , f (g j )) = (g i , f (g j )).As P is a geodesic in X, Bellman's principle of optimality implies that P ′ is also a geodesic.In addition to its connecting edges, P ′ contains geodesics P i , P i+1 , . . ., P j , which are respectively, projected onto H, geodesics • between f (g i−1 ) and f (g i+1 ), • between f (g i ) and f (g i+2 ), . . .
• between f (g j−2 ) and f (g j ), and • between f (g j−1 ) and f (g i ). ( The shape of P Suppose first that j − i is odd.Then we have This is a contradiction with the selection of u and v as a minimal counterexample.Suppose second that j − i is even.Then we have ≥ d H (f (g i−1 ), f (g i )) + d H (f (g i ), f (g j )) ≥ d H (f (g i−1 ), f (g j )) = d g i H (u ′ , v ′ ) .
Hence we get the same contradiction as in the previous case.
We have thus proved that the vertices g 1 , . . ., g k are pairwise different.To complete the proof of the claim, we need to verify that P starts and ends with a connecting edge.Suppose on the contrary that P starts with a subpath in g 1 H from u to w and then proceed along the connecting edge between g 1 H and g 2 H.By the minimality assumption on u and v we have d X (u, w) = d g 1 H (u, w).We now have: which yields d X (w, v) < d g 1 H (w, v).This contradiction proves that P indeed starts with a connecting edge.A parallel argument yields that P also ends with a connecting edge.This proves the claim.
To conclude the proof, let P 2 , P 3 , . . ., P k be the sections of P restricted to g 2 H, g 3 H, . . ., g k H, respectively.Then we proceed similarly as we did for the subpath P ′ above.More precisely, if k − 1 is odd, then

Theorem 4 . 1 .
If G and H be connected graphs, f : V (G) → V (H), and g ∈ V (G), then gH is a convex subgraph of G ⊗ f H.