MPEMathematical Problems in Engineering1563-51471024-123XHindawi10.1155/2020/58987355898735Research ArticleNordhaus–Gaddum-Type Relations for Arithmetic-Geometric Spectral Radius and EnergyWangYajing1https://orcid.org/0000-0002-3333-4002GaoYubin2LoiseauJean Jacques1School of Data Science and TechnologyNorth University of ChinaTaiyuanShanxi 030051Chinanuc.edu.cn2Department of MathematicsNorth University of ChinaTaiyuanShanxi 030051Chinanuc.edu.cn202016720202020240220202306202016720202020Copyright © 2020 Yajing Wang and Yubin Gao.This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Spectral graph theory plays an important role in engineering. Let G be a simple graph of order n with vertex set V=v1,v2,,vn. For viV, the degree of the vertex vi, denoted by di, is the number of the vertices adjacent to vi. The arithmetic-geometric adjacency matrix AagG of G is defined as the n×n matrix whose i,j entry is equal to di+dj/2didj if the vertices vi and vj are adjacent and 0 otherwise. The arithmetic-geometric spectral radius and arithmetic-geometric energy of G are the spectral radius and energy of its arithmetic-geometric adjacency matrix, respectively. In this paper, some new upper bounds on arithmetic-geometric energy are obtained. In addition, we present the Nordhaus–Gaddum-type relations for arithmetic-geometric spectral radius and arithmetic-geometric energy and characterize corresponding extremal graphs.

Shanxi Scholarship Council of China201901D211227
1. Introduction

Spectral properties of graphs have been widely applied in the field of engineering technology. In systems engineering, complex networks are often used as abstract models to study and reveal the relationship among system elements. The mathematical study of complex network depends on graph theory. In network dynamics, the analysis of threshold in the epidemic model and synchronization condition of coupling oscillators is essentially the study of spectral of the corresponding graph. Naturally, the development of systems engineering depends on the results and conclusions of basic studies in graph spectral theory. Similarly, in the chemical engineering, a chemical molecule can also be abstracted as a graph. The combinatorial forms of spectral of weighted graphs, based on topological indices, are closely related to the molecular orbital energy levels of π-electrons in conjugated hydrocarbons, as well as the spectrum properties. The combinatorial form of the spectrum, significantly correlated with the stability of conjugated alternant molecules, is defined as the sum of the absolute values of the eigenvalues of graphs. In this paper, we further study the arithmetic-geometric spectral radius and arithmetic-geometric energy.

Let G=V,E be a simple graph with vertex set V=v1,v2,,vn and edge set E, and let E=m. An edge eE with end vertices vi and vj is denoted by vivj. The degree of the vertex viV of G for i=1,2,,n is denoted by di. Let Δ and δ represent the maximum and minimum degrees of G, respectively. The graph G¯ is called the complement to the graph G, if its vertex set is the same as V, and vi and vji,j=1,2,,n,ij are adjacent in G¯ if and only if they are not adjacent in G.

The adjacency matrix of a graph G is the matrix aijn×n, denoted by AG, where aij=1 if vi and vj are adjacent; if not, aij=0. All eigenvalues of AG are denoted by the nonincreasing sequence λ1,λ2,,λn. The spectral radius of G is the greatest eigenvalue λ1. The energy of G is defined as(1)G=i=1nλiG.

As a graph invariant, topological indices are used to understand physicochemical properties of chemical compounds since they capture some properties of molecules. In 2015, the paper  proposed the arithmetic-geometric index of a graph G and defined the arithmetic-geometric adjacency matrix (AG matrix) of G, denoted by AagG. An element of AG matrix is defined in the following manner:(2)gij=di+dj2didj,if vivjE,0,otherwise.

The AG eigenvalues of G are the eigenvalues of its corresponding AG matrix. Note that AG matrix is real and symmetric so that all eigenvalues of AagG are real, which can be recorded as η1η2ηn. Similarly, the greatest AG eigenvalue η1 is called the arithmetic-geometric spectral radius (AG spectral radius) of G. In addition, the arithmetic-geometric energy (AG energy) of G is defined in an analogue way as(3)agG=i=1nηi.

The first Zagreb index M1G is a kind of important topological indices, which is defined in  as(4)M1G=i=1ndi2=vivjEGdi+dj.

A graph G is said to be strongly regular with parameters n,r,μ,ν if it is rregular, every pair of adjacent vertices has μ0 common neighbors, and every pair of distinct nonadjacent vertices has ν0 common neighbors . If ν=0, then G is a disjoint union of complete graphs, whereas if ν1 and G is noncomplete, then the eigenvalues of G are r (the trivial eigenvalue) and the roots x1 and x2 (the nontrivial eigenvalues) of the quadratic equation:(5)x2+νμx+νr=0.

It is straightforward to show that the complement of a strong regular graph with parameters n,r,μ,ν is still a strong regular graph with parameters n,n1r,n22r+ν,n2r+μ. Recently, some bounds on AG spectral radius and AG energy were obtained in a couple of papers [4, 5]. Until now, hundreds of such “energy” have been introduced, such as inverse sum index energy , distance energy [7, 8], ABC energy [9, 10], matching energy [11, 12], and Randić energy [13, 14]. Recently, there are many studies, such as , on the energy of graphs with given parameters.

In the paper of Nordhaus and Gaddum , the lower and upper bounds on χG+χG¯ and χGχG¯ were given, where χG and χG¯ were the chromatic number of a graph G and its complement G¯, separately. Since then, in terms of an invariant ψG of graph G, any bound on ψG+ψG¯ or ψGψG¯ is referred to as a Nordhaus–Gaddum-type inequality or relation. And Nordhaus–Gaddum-type relations have received wide attention (e.g., ).

We write G1G2 if the graphs G1 and G2 are isomorphic. As usual, let Kn denote the complete graph of order n and Kp1,p2,,pn the complete multipartite graph, having pi vertices in the ith partite set for each i=1,2,,n. Thus, KnKp1,p2,,pn, where p1=p2==pn=1. For other undefined notions and terminologies from graph theory, the readers are referred to [20, 21]. The rest of the paper is structured as follows: in Section 2, we give some useful lemmas; in Section 3, we get some new bounds on the AG energy; and in Sections 4 and 5, we obtain the Nordhaus–Gaddum-type relations for the AG spectral radius and AG energy of graph G, respectively.

2. PreliminariesLemma 1.

(see ). If B is an n×n real symmetric matrix with eigenvalues λ1λ2λn, then for any 0xRn, xTBxλ1xTx. Equality holds if and only if x is an eigenvector of B corresponding to λ1.

Lemma 2.

(see ). Let G be a graph of order n and size m with maximum degree Δ and minimum degree δ1. Then,(6)λ12mδn1+δ1Δ,with equality holding if and only if G is regular, a star plus copies of K2, or a complete graph plus a regular graph with a smaller degree of vertices.

Lemma 3.

(see ). Let G be a graph of order n with the maximum degree Δ and minimum degree δ1. Then,(7)η112Δδ+δΔλ1,with equality holding if and only if G is regular.

Lemma 4.

(see ). Let G be a graph of order n. Then,(8)η1M1Gn,with equality holding if and only if G is regular.

Lemma 5.

(see ). A connected graph G of order n has only one positive eigenvalue in its adjacency spectrum if and only if G is a complete multipartite graph.

3. On AG Energy of a GraphLemma 6.

Let G be a graph of order n and size m, then(9)η12mn,equality holds if and only if G is a regular graph.

Proof.

Let us take any unit vector x=x1,x2,,xnT in Rn. By Lemma 1, we have(10)η1xTAagGx=vivjEGdi+djdidjxixj2vivjEGxixj.

Taking x=1/n,1/n,,1/nT into (10), we have(11)η1GxTAagGx2mn,then (9) holds.

For the equality in (9) to hold, all the inequalities in the above argument must be equalities. From (10), we have d1=d2==dn. Then, G is a regular graph.

Conversely, if G is regular, then d1=d2==dn. So x=1/n,1/n,,1/nT is an eigenvector of AagG corresponding to the eigenvalue η1. Then, the equality holds in (9).

Theorem 1.

Let G be a graph of order n and size m, then(12)agG4mn,with equality holding if and only if G is isomorphic to a regular complete multipartite graph or GK¯n.

Proof.

Applying Lemma 6, we have(13)agG=i=1nηi=2i=1,ηi0nηi2η14mn.

The first equality attained if and only if G has at most one positive AG eigenvalue, and the second equality holds if and only if G is regular. If G has no positive eigenvalue, then GK¯n. If G has only one positive eigenvalue, from Lemma 5, then G is isomorphic to a complete multipartite graph. And, if G is a regular graph, then its AG matrix is identical to its adjacency matrix. If so, we know that the equality in Theorem 1 holds if and only if G is isomorphic to a regular complete multipartite graph, or GK¯n.

Theorem 2.

Let G be a graph of order n and size m, and δ>0. Then,(14)agGt+n1m2Δδ+δΔ2t2,where t=max2m/n,Δ/δ+δ/Δm/2n. This bound is achieved if G is either n/2K2, Kn, or a noncomplete connected strongly regular graph with two nontrivial eigenvalues both with absolute value:(15)2m2m/n2n1.

Proof.

By Lemma 6, we have η12m/n. Moreover, since(16)i=1nηi2=12vivjEGdi+djdidj2=12vivjEGdidj+djdi212vivjEGΔδ+δΔ2=m2Δδ+δΔ2must hold, we have(17)i=2nηi2m2Δδ+δΔ2η12.

Combining the above inequality with the Cauchy–Schwartz inequality, we can see that the following inequality is obvious:(18)i=2nηin1m2Δδ+δΔ2η12.

Thus,(19)agGη1+n1m2Δδ+δΔ2η12.

Since the function(20)Fx=x+n1m2Δδ+δΔ2x2reaches the maximum for x=Δ/δ+δ/Δm/2n, we see that Fη1FΔ/δ+δ/Δm/2n must hold as well. In addition, if 2m/nΔ/δ+δ/Δm/2n, then Fη1F2m/n. From this fact and inequality (19), it immediately follows that inequality (14) holds when t=max2m/n,Δ/δ+δ/Δm/2n.

It is obvious that 2mn and we know that if G is either n/2K2, Kn, or a noncomplete connected strongly regular graph, the inequality 2m/nΔ/δ+δ/Δm/2n holds. As the graph n/2K2 is regular, its AG matrix is the same as its adjacency matrix. That is to say, the AG eigenvalues for n/2K2 are ±1 (both with multiplicity n/2). Similarly, the AG eigenvalues for Kn are n1 (multiplicity 1) and 1 (multiplicity n1) so that equality must hold in (14), if G is isomorphic to n/2K2 or Kn.

Moreover, if G is a noncomplete connected strongly regular graph with the trivial eigenvalue 2m/n and other two nontrivial eigenvalues both with absolute value 2m2m/n2/n1, the equality holds in (14) as well.

Remark 1.

The paper  proved that if G is a simple graph of order n and size m having no isolated vertices, then the following bound must hold:(21)agGΔδ2mn.

The reason is following: Fx defined in the proof of Theorem 2 reaches the maximum for x=Δ/δ+δ/Δm/2n, and the maximum is Fx=1/2Δ/δ+δ/Δ2mn. It follows that(22)agG12Δδ+δΔ2mnmust hold. Moreover, 1/2Δ/δ+δ/Δ2mnΔ/δ2mn. Hence, inequality (14) is an improvement on the bound in .

Theorem 3.

If 2mn and G is a graph on n vertices with m edges, then(23)agGmΔΓδΓ+δΓΔΓ,where Γ is the graph obtained from G by deleting all isolated vertices. Equality holds if and only if Γn/2K2.

Proof.

It is easy to know that G has at least n2m isolated vertices if 2mn. We can obtain a graph Γ with no isolated vertices by removing all isolated vertices. It follows that Γ has at most 2m vertices. Hence, by applying inequality (22) to Γ, we can immediately see that(24)agG12ΔΓδΓ+δΓΔΓ2mnmΔΓδΓ+δΓΔΓ.

Moreover, equality holds if and only if 2m=n; that is, G is the disjoint union of edges.

Theorem 4.

Let G be a graph of order n and size m and δ>0. Then,(25)agGn+nn8Δδ+δΔ2,holds. If G is a strongly regular graph with parameters n,n+n/2,n+2n/4,n+2n/4, the bound is achieved.

Proof.

It is obvious that 2mn. If 2m/nΔ/δ+δ/Δm/2n, then the inequality (14) is as follows:(26)agG2mn+n1m2Δδ+δΔ22mn2.

Using routine calculus, it is seen that the left-hand side of inequality (26)—considered as a function of m—is maximized when(27)m=n2+nn16Δδ+δΔ2,holds. Inequality (25) now follows by substituting this value of m into (26).

If Δ/δ+δ/Δm/2n2m/n, then by inequality (22) and 2mn, we know that(28)agGn4Δδ+δΔ2.

Since n+nn/8Δ/δ+δ/Δ2>n/4Δ/δ+δ/Δ2, then it is clear that (25) follows. Moreover, it follows by Theorem 2 that if G is a strongly regular graph with parameters n,n+n/2,n+2n/4,n+2n/4, equality holds in (25), which means that the upper bound is achieved. The theorem follows.

Let Δ¯, δ¯, and η¯1 be the maximum degree, minimum degree, and spectral radius of the complement graph G¯ of G, respectively. Here, we are presenting the lower and upper bounds on η1+η1¯.

Theorem 5.

Let G be a graph of order n. Then,(29)η1+η¯1M1Gn+M1G¯n,and the equality holds if and only if G is a regular graph.

Proof.

The proof of inequality in (29) follows directly from Lemma 4.

If the equality in (29) holds, that is, η1=M1G/n and η¯1=M1G¯/n, then by Lemma 4, G is regular.

If G is a regular graph, then(30)η1+η¯1=Δ+n1δ=n1=M1Gn+M1G¯n.

This finishes the proof.

Theorem 6.

Let G be a graph of order n. Then,(31)η1+η¯1n1,with equality holding if and only if G is a regular graph.

Proof.

Lemma 6 gives that(32)η12mn,η¯12EG¯n,where EG¯ denotes the number of edges in G¯. Then, we have(33)η1+η¯12mn+2EG¯n=2nn2=n1.

From Lemma 6, equality holds if and only if G is isomorphic to a regular graph.

Theorem 7.

Let G be a connected graph of order n and size m. And we write the number of vertices with di=Δ by t and the second largest degree of graph G by τ:

If Δ=n1 or Δ¯=n1, then(34)η1+η¯112n1δ+δn12mn+1+12nt1n1τ+n1τnt1n122m+t.

If Δn2 and Δ¯n2, then(35)η1+η¯112Δδ+δΔ2mδn1+δ1Δ+12n1δn1Δ+n1Δn1δn122mδn2Δ.

Moreover, the bounds are sharp.

Proof.

(1) Without loss of generality, let Δ=n1. Since δ1, by Lemmas 2 and 3, we have(36)η112n1δ+δn12mδn1+δ1n1=12n1δ+δn12mn+1.

It clearly follows that G¯ has t isolated vertices. Let Γ1 be the graph obtained from G¯ by removing the isolated vertices. It is clear that η¯1=η1Γ1, Δ¯=ΔΓ1nt1, and δ¯=δΓ1=n1τ, so(37)η¯112ΔΓ1δΓ1+δΓ1ΔΓ12eΓ1δΓ1VΓ11+δΓ11ΔΓ112nt1n1τ+n1τnt12n22mnt1=12nt1n1τ+n1τnt1n122m+t.

The proof of the inequality (34) follows.

(2) Let Δn2 and Δ¯n2. It follows that δ1 and δ¯1. From Lemmas 2 and 3, we have(38)η112Δδ+δΔ2mδn1+δ1Δ,η¯112Δ¯δ¯+δ¯Δ¯2n22mδ¯n1+δ¯1Δ¯=12n1δn1Δ+n1Δn1δ2n22mn1Δn1+n2Δn1δ=12n1δn1Δ+n1Δn1δn122mδn2Δ.

Hence, the inequality (35) holds.

To show the sharpness of upper bounds in Theorem 7, we consider the following examples.

Example 1.

Let G be a star of order n. That is to say, t=1 and τ=1. Then, G¯ is the union of an isolated vertex and a complete graph Kn1. We have(39)η1=n2=12n1+1n1n1=12n1δ+δn12mn+1,η¯1=n2=12nt1n1τ+n1τnt1n122m+t.

Example 2.

Let G be a regular graph. It is clear that G¯ is still a regular graph. We have(40)η1+η¯1=Δ+n1Δ=n1=12Δδ+δΔ2mδn1+δ1Δ+12n1δn1Δ+n1Δn1δn122mδn2Δ.

5. Nordhaus–Gaddum-Type Relation for AG Energy

In the following theorems, some lower and upper bounds for agG+agG¯ are obtained.

Theorem 8.

Let G be a connected graph of order n and size m and Γ1,Γ2,,Γr be all connected components of G¯. Then,(41)agG+agG¯4mn+i=1r4EΓiVΓi,where VΓi and EΓi are the number of vertices and edges in Γii=1,,r, respectively. Moreover, the equality holds if and only if G is isomorphic to a regular complete multipartite graph, or GK¯n.

Proof.

Applying Theorem 1, we have(42)agG4mn,agΓi4EΓiVΓi.

Hence, we can obtain(43)agG+agG¯=agG+i=1ragΓi4mn+i=1r4EΓiVΓi.

From Theorem 1, we know that equality is possible if and only if G is isomorphic to regular complete multipartite graph, or GK¯n. The complement of regular complete multipartite graph is the disjoint union of k complete graphs of order n/k, and Kn is the complement graph of K¯n. If so, agΓi=4EΓi/VΓi holds. Hence, the equality holds if and only if G is isomorphic to a regular complete multipartite graph, or GK¯n.

Theorem 9.

Let G be a graph of order n and size m, then(44)agG+agG¯2n1,with equality holding if and only if either GKn or GK¯n.

Proof.

Note that(45)agG+agG¯4mn+4EG¯n=4nn2=2n1.

Similar to the analysis of equality in Theorem 8, we consider the following two cases:

GKn, which implies G¯K¯n.

Both G and G¯ are isomorphic to regular complete multipartite graphs. This is impossible because complete multipartite graphs are connected, whereas their complements are disconnected.

Theorem 10.

Let G be a graph of order n and G¯ is the complement graph of G. If both G and G¯ have no isolated vertices, then(46)agG+agG¯n+nn4Δn1δδn1Δ+3.

If G is a strongly regular graph with parameters n,n+n/2,n+2n/4,n+2n/4, the equality holds.

Proof.

From Theorem 4, we have(47)agG+agG¯n+nn8Δδ+δΔ2+n+nn8Δ¯δ¯+δ¯Δ¯2=n+nn8Δδ+δΔ2+n1δn1Δ+n1Δn1δ2n+nn4Δn1δδn1Δ+3.

If G is a strongly regular graph, we know that G¯ is also a strongly regular graph. Moreover, all inequalities in the above argument must be equalities. The first equality holds if G is isomorphic to a strongly regular graph with parameters n,n+n/2,n+2n/4,n+2n/4. And, the last equality holds if G is regular. The theorem now follows immediately.

6. Conclusions

In this paper, we obtain an upper bound on AG spectral radius, and furthermore, the Nordhaus–Gaddum-type relation of AG spectral radius is derived. Some new upper and lower bounds on arithmetic-geometric energy are obtained. Moreover, we illustrate one of the upper bounds is better than the result in paper . In the end, we state the Nordhaus–Gaddum-type relation of AG energy.

Data Availability

No data were used to support this study.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

Acknowledgments

This research was funded by the Shanxi Scholarship Council of China (no. 201901D211227).

ShegehalliV.KanaburR.Arithmetic-geometric indices of path graphJournal of Computer and Mathematical Sciences2015611924GutmanI.TrinajstićN.Graph theory and molecular orbitals. Total φ-electron energy of alternant hydrocarbonsChemical Physics Letters197217453553810.1016/0009-2614(72)85099-12-s2.0-33845352210GodsilC.RoyleG. F.Algebraic Graph Theory2001New York, NY, USASpringerGuoX.GaoY. B.Arithmetic-geometric spectral radius and energy of graphsMATCH Communications in Mathematical and in Computer Chemistry2020833651660ZhengL.TianG. X.CuiS. Y.On spectral radius and energy of arithmetic-geometric matrix of graphsMATCH Communications in Mathematical and in Computer Chemistry2020833635650HafeezS.FarooqR.Inverse sum indeg energy of graphsIEEE Access2019710086010086610.1109/access.2019.2929528ZhangX.Inertia and distance energy of line graphs of unicyclic graphsDiscrete Applied Mathematics201925422223310.1016/j.dam.2018.06.0232-s2.0-85049885924PirzadaS.GanieH. A.RatherB. A.Ul ShabanR.On generalized distance energy of graphsLinear Algebra and Its Applications202060311910.1016/j.laa.2020.05.022ChenX.On ABC eigenvalues and ABC energyLinear Algebra and Its Applications201854414115710.1016/j.laa.2018.01.0112-s2.0-85041444974GaoY.ShaoY.The minimum ABC energy of treesLinear Algebra and Its Applications201957718620310.1016/j.laa.2019.04.0322-s2.0-85065101561ZhuJ.YangJ.On the minimal matching energies of unicyclic graphsDiscrete Applied Mathematics201925424625510.1016/j.dam.2018.06.0132-s2.0-85049356026AshrafF.Energy, matching number and odd cycles of graphsLinear Algebra and Its Applications201957715916710.1016/j.laa.2019.04.0292-s2.0-85065066849DasK. C.SunS.Extremal graphs for Randić energyMATCH Communications in Mathematical and in Computer Chemistry20177717784AltındagI.Some statistical results on Randić energy of graphsMATCH Communications in Mathematical and in Computer Chemistry2018792331339BhatM. A.PirzadaS.On unicyclic graphs with given number of pendent vertices and minimal energyLinear Algebra and Its Applications2020597869310.1016/j.laa.2020.03.024GanieH. A.SameeU.PirzadaS.AlghamdiA. M.Bounds for graph energy in terms of vertex covering and clique numbersElectronic Journal of Graph Theory and Applications20197231532810.5614/ejgta.2019.7.2.9GanieH. A.SameeU.PirzadaS.On graph energy, maximum degree and vertex cover numberLe Matematiche2019741163172NordhausE. A.GaddumJ. W.On complementary graphsThe American Mathematical Monthly195663317517710.2307/2306658AouchicheM.HansenP.A survey of Nordhaus-Gaddum type relationsDiscrete Applied Mathematics20131614-546654610.1016/j.dam.2011.12.0182-s2.0-84873411649BondyJ. A.MurtyU. S. R.Graph Theory with Applications1976London, UKMacmillanLiX.ShiY.GutmanI.Graph Energy2012New York, NY, USASpringerZhangF.Basic results and techniquesMatrix Theory2011New York, NY, USASpringerCaoD.Bounds on eigenvalues and chromatic numbersLinear Algebra and Its Applications19982701–311310.1016/s0024-3795(96)00199-12-s2.0-0041853219SmithJ. H.Some properties of the spectrum of a graphCombinatorial Structures and Their Applications1970New York, NY, USAGordon and Breach403406