A Cheeger inequality of a distance regular graph using the Green’s function Gil Chun Kim Department of Mathematics Dong-A University, Busan, Korea 1. Some definitions and facts 2. Green’s function 3. A Cheeger inequality of a distance regular graph 4. Explicit expression of the krawtchouk polynomial π be a subset of set of vertices in a graph Γ = (π, πΈ), where π is a set of vertices of Γ and πΈ is a set of edges of Γ. . Let (π) The edge boundary of π, denoted by ππ is defined as follows: ππ = π₯, π¦ ∈ πΈ Γ π₯ ∈ π πππ π¦ ∈ π − π}, where πΈ(Γ) is a edge set of Γ. (π) If π ≠ ∅, then the volume of π, denoted by π£ππ(π) is defined as follows: π£ππ π = π’∈π ππ’ where ππ’ is a valency of π’ in Γ. The volume of Γ is denoted by π£ππ Γ = π’ ππ’ (π) The πΆβπππππ πππ‘ππ of π, denoted by βπ is defined as |ππ| βπ = min{π£ππ π , π£ππ Γ − π£ππ π } (π) The πΆβπππππ ππππ π‘πππ‘ of Γ, denoted by βΓ is defined as βΓ = min βπ π⊂π In graph theory, the Cheeger constant is an important geometric meaning. The Cheeger constant give to us answer of the problem of separating the graph into two large components by making a small edge-cut. The Cheeger constant of a connected graph is strictly positive. If the Cheeger constant is “small” but postive, then there are two large sets of vertices with “few” edges between them. On the other hand if the Cheeger constant is “large”, then there are two sets of vertices with “many” edges between those two subsets. We are interested in finding bounds on the Cheeger constants of graphs. The Hamming graph is the graph that describes the distance – 1 relation in the Hamming scheme π»(π, π) The Hamming graph is a special class of graphs used in several branches of Mathematics and Computer Science. The Hamming graphs are interesting in connection with error-correcting codes and association schemes. The Cheeger constant of the Hamming graphs Γ(π, π) π+1 2 βΓ = π(π − 1) Hamming graph Γ(3,2) π = 000, 100, 010, 110 ππ = 4, π£ππ π = 12 4 1 βΓ = = 12 3 The Johnson graph is also an interesting class of graphs. A vertex set π of π€ π, π consists of all binary vectors of length π with Hamming weight π, such that π = ππ . The Johnson graph is the graph that describes the distance – 1 relation in the Johnson scheme π½(π, π). Two vertices are adjacent if they differ in two coordinates. The Cheeger constant of the Johnson graphs Γ(π, π) βΓ ≤ βΓ ≤ (π−π+1)2 π΄2 2π(π−π) π π π΄2 + 4 π π 2 π΄2 2 , if π is even and π is odd. , if π is even and π is even. βΓ ≤ βΓ ≤ (π+1)π΄2 +(π−1) 4π π π 2 π΄2 2 (π+1)π΄2 +(π−π−1) 4(π−π) π π 2 , if π is odd and π is odd. π΄2 2 , if π is odd and π is even. . π«πππππππ πππππππ πππππ Let Γ be a connected graph. For a vertex π₯ ∈ π , define Γπ π₯ to be the set of vertices which are at distance precisely π from π₯ (0 ≤ π ≤ π), where π = max π π₯, π¦ π₯, π¦ ∈ π}. A connected graph Γ with diameter π is called πππ π‘ππππ ππππ’πππ if there are integers ππ , ππ+1 0 ≤ π ≤ π − 1 such that for any two vertices π₯, π¦ with π π₯, π¦ = π, there are precisely ππ neighbors of π¦ in Γπ−1 π₯ and ππ neighbors of π¦ in Γπ+1 π₯ . . π¨ππππππππππ ππππππ Let π be a nonempty finite set and π = {π 0 , π 1 , β― , π π } be a family of relations defined on π. Let π£ be the order of π. Then adjacency matrix π΄π of π π (π = 0,1, β― , π) is the π£ × π£ matrix defined by 1 , if (π₯, π¦) ∈ π π (π΄π )π₯,π¦ = { 0 , otherwise. We say that the pair π = (π, π ) is a symmetric association scheme with π classes if it satisfies the following conditions. 1 π΄π is symmetric, (2) ππ=0 π΄π = π½ (the all 1’s matrix), (3) π΄0 = πΌ, (4) π΄π π΄π = ππ=0 ππ,π π π΄π (π, π = 0,1,2, β― , π). Let π be the algebra spanned by the adjacency matrices π΄0 , π΄1 , β― , π΄π . Then π is called the Bose Mesner algebra of π and π has two distinguished bases π΄π and πΈπ , where the latter consist of primitive idempotent matrices. For π΄π and πΈπ , we express one in terms of the other: 1 π π΄π = ππ=0 ππ (π) πΈπ , πΈπ = π=0 ππ (π) π΄π for π = 0,1, β― , π. |π| The (π + 1) × π + 1 matrix π = (ππ π ) (respectively, Q= (ππ π ) ) is called the first eigenmatrix (respectively, the second eigenmatrix) of the association scheme. Then π = (ππ π ) and Q= (ππ π ) satisfy ππ π ππ = ππ π ππ , where ππ = ππππ(πΈπ ), ππ is the valency of π΄π ,and ππ π is the complex conjugate of ππ π . The adjacency matrices π΄π satisfy π΄π π΄π = ππ=0 ππ,π π π΄π for all π, π, where for (π₯, π¦) ∈ π π , ππ,π π is the number of π§ ∈ π such that (π₯, π§) ∈ π π and π§, π¦ ∈ π π . The non-negative integers ππ,π π are called the intersection numbers of π. Let π΅π be a matrix with (π, π)-entries ππ,π π ,and let β¬ be a algebra spanned by π΅0 , π΅1 , β― , π΅π . Then π΅π is called the π-th intersection matrix of π and β¬ is called the intersection algebra of π . In fact, the Bose-Mesner algebra π of π is isomorphic to β¬ by the map π΄π → π΅π . Let π = π, π π (π = 0,1, β― , π) be a symmetric association scheme, and let Γ be the graph whose set and edge set are π and π 1 respectively. Then, it is known that the following are equivalent. (π) Γ is distance regular graph. (π) π is a π–polynomial scheme with respect to π 0 , π 1 , β― , π π , that is π£π π΄1 (π = 0,1, β― , π) for some polynomial π£π π₯ of degree π. (π) The first eigenmatrix π = (ππ π ) satisfies ππ π = π£π (ππ ) for some polynomials π£π π₯ of degree π, where ππ = π1 π (π, π = 0,1, β― , π). (π) The first intersection matrix π΅1 is a tridiagonal matrix with non-zero off diagonal entries, π΄π = πͺππππππ ππππππππππ Let Γ be a graph of order π£, and let λ1 be the smallest positive eigenvalue of the Laplacian of Γ. Then there are Cheeger inequalities as follows: (π) [1] βΓ 2 2 ≤ π1 ≤ 2βΓ βΉ βΓ ≤ 2π1 ; [1] J. Dodziuk, W.S. Kendall, Combinatorial Laplacians and isoperimetric inequality, in: K.D. Elworthy (Ed.), From Local Times to Global Geometry, control and Physics, Research Notes in Mathematics Series, Vol. 150, Pitman, London, 1986, pp.68-74. π 2 For π£ ≥ 4, βΓ ≤ π1 2 − π1 . [2] J.Tan, On cheeger inequalities of a graph, Discrete Math. 269(2003) 315-323. Our Cheeger inequality provides an upper bound of the cheeger constant, which yields an improvement of the bound (π), (π). . [1] Let Γ be a distance regular graph. Then its edge-connectivity equals its valency π, and the only disconnecting sets of edges are the sets of π edges incident with a single vertex. [1] A.Brouwer, W.Haemers, Eigenvalues and perfect matchings, Linear Algebra and its Applications. 395 (2005). 155-162. . [2] Let Γ be a non-complete distance regular graph of valency π > 2. Then the vertex-connectivity π Γ equals π, and the only disconnecting sets of vertices of size not more than π are the point neighbourhoods. [2] A. Brouwer, J. Koolen, the vertex-connectivity of a distance regular graph, European Journal of Combinatorics. Vol. 30. No. 3. (2009). 668-673. . [3] Let Γ = (π, πΈ) be a simple graph with the vertex-connectivity π Γ and the edge-connectivity π(Γ). Then 2π Γ |π| ≤ 2π(Γ) |π| |ππ| |π| |π| . 2 ≤ πππ where π is a subset of π with |π| ≤ ≤ π Γ ≤ π(Γ), [3] G. Oshikiri, Cheeger constant and connectivity of graphs, Interdisciplinary Information Sciences, 8 (2) (2002), 147-150. From Propositions 1,2 and 3 we see that, for a distance regular graph, there are close connections between the Cheeger constant and vertex and edge connectivity. Also, we obtain an inequality 2 ≤ βΓ ≤ 1. |π| We are thus interested in finding optimal bounds of βΓ for distance regular graph. For our Cheeger inequality, we use the Green's function, which is defined as the inverse of the π½ -Laplacian. Green’s function Define a transition probability matrix π over π by π = 1 π΄ , where π1 1 π1 is the valency of π΄1 . For a function π βΆ π → β, we define a πΏππππππ ππππππ‘ππ β by βπ(π₯) = Then, we have β= πΌ − π = πΌ − 1 π1 π π₯ −π π¦ . π¦~π₯ 1 π΄ π1 1 and β is a matrix representation of the Laplacian β. For π = 0,1, β― , π and orthogonal eigenfunction ππ ∗ , we have β= where ππ is an eigenvalue of β. π π=0 ππ ππ ∗ ππ , Let βπ½ be the π½–normalized Laplacian by π½πΌ + β. For π½ > 0, let a Green’s function π’π½ denote the symmetric matrix satisfying βπ½ π’π½ = πΌ. Then we have π’π½ = 1 π π=0 π½+π π ππ ∗ ππ , For π½ > 0 we have π’π½ (π½πΌ + πΌ − π) = πΌ. Thus, this implies that βπ½ =(π½ + 1)πΌ − π = π π=0(π½ +1− 1 π π1 1 π )πΈπ . Hence, a Green’s function π’π½ can be expressed by π1 π π=0( π½+1 π −π π 1 1 π’π½ = Since πΈπ = 1 |π| ) πΈπ . ππ π π΄π , π’π½ is a linear combination of adjacency matrices π΄π as follows: π’π½ = π0 where ππ (π½) = 1 π 1 π½ (π½) ( + π1 π π΄0 + π1 1 π½+π1 (π½) π΄1 + β― + ππ + β― + ππ (π) 1 ) π½+ππ π½ π΄π , (π = 0,1, β― , π) Let πΏ be a ( π − 1) × |π| matrix obtained by the removal of the first row of 1 βπ½ =(π½ + 1)πΌ − π΄1 . π1 Since the Bose-Mesner algebra π is isomorphic to the intersection algebra π of 1 1 π then that isomorphism of algebra takes (π½ + 1)πΌ − π΄1 to (π½ + 1)πΌ − π΅1 . Let πΏ′ = −π1 ( π½ + 1 πΌ − matrix. Let πΏπ π’π (π½) π1 − π1 π½ + 1 πΌ, which is a π + 1 × π + 1 be a matrix obtained by the removal of the first row of πΏ′ . Then, we obtain πΏπ π’π πΏπ π’π 1 π΅ )= π΅1 π1 1 π1 (π½) as follows: (π½) where π π = ππ − π1 (π½ + 1) for π = 1,2, β― , π and ππ , ππ , ππ are the same entries as in π΅1 . . For π½ > 0, let π’π½ be a Green’s function for a π–polynomial scheme. Then we have the following: (π) A Green’s function can be expressed as π’π½ = π‘π’0 (π½) π΄0 + π‘π’1 for some nonzero π‘ ∈ β, where (π’0 basis (π½) (π½) (π½) π΄1 + β― + π‘π’π , π’1 (π½) π½ , β― , π’π (π½) (π½) of the nullspace π©(πΏ(π½) ) of(π½) πΏπ π’π with π’ππ½ =1 1. π π’π (π) For π½ > 0, π0 π0 + π1 π1 + β― + ππ ππ = , where ππ is the valency of π΄π for π = 0, 1, β― , π. (π) π0 (π½) > π1 (π½) (π) As π½ → 0+ , π0 > β― > ππ (π½) ≈ π1 π½ (π½) > 0. ≈ β― ≈ ππ π½ . π½ π΄π (π½) ) is the unique For π½ > 0, let π’π½ be a Green’s function of the distance regular graph of order π£ We denote a subset πͺπ½ of 0,1,2, β― , π by πͺπ½ β{ π | 1 π½ − π£ππ π½ > 0}. It is clear that πͺπ½ is a non-empty set. When π½ is sufficiently close to 0+ , we consider a set πͺπ½ ′ = π Since π½π£ππ π½π£ππ π½ π½ π½ + π1 < π1 π½ π½+ππ = 1 + π1 π lim+ π½π£ππ π½→0 Thus, we have π½ 1 π½ ⇒ π½π£ππ π½π£ππ π½ π½ + π1 < π1 }. π½ + β― + ππ (π) 1 π½ < π1 ( − π£ππ π½ π½+ππ π½ ) , = 1. − π£ππ π½ > 0. That is, πͺπ½ ′ is a subset of πͺπ½ . and . For π½ > 0, let Γ be a distance regular graph of order π£ and let (π½) (π½) π’π½ = π0 π΄0 + π1 for π ∈ πͺπ½ ′ , we have (a) π½ 2 π£ππ 1−π½π£ππ π½ π΄1 + β― + ππ π½ π΄π be a Green’s function of Γ. Then, is decreasing in π ∈ πͺπ½ ′ π½ ′ (π) For π ∈ πͺπ½ , lim+ π½ 2 π£ππ π½→0 1−π½π£π π π½ π½ = πΌπ , where πΌπ = Moreover, for some π ∈ πͺπ½ ′ , πΌπ < π1 . (π) π½ 2 π£ππ 1−π½π£ππ π½ π½ is decreasing in π½ > 0. 1 −π1 π 1 − π1 β― −ππ π 1 ππ . - A Cheeger inequality of a distance regular graph . Let Γ be a distance regular graph of diameter π and let π1 be the smallest positive eigenvalue of the Laplacian. Then π 1 βΓ < πΌπ < πΌπ−1 < β― < π1 . . Let π be a subset of vertices of Γ with π£ππ(π) ≤ π£ππ(Γ)/2. Let 1−π½π£ππ π½π£ππ π½ = π½ 1 . π Then we have lim+ π½→0 number such that π΄ < π£ππ(π)π½π£ππ 1−π½π£ππ π½ π½ 1−π½π£ππ π½π£ππ π½ = 0+ . Let π΄ be a positive π½ . Then π½ π½ 1 − π½π£πππ½ π΄ < π£ππ(π)π½π£ππ ⇒ 1−π½π£ππ ′ + Let π½ = π½ . Since lim+ π½ = 0 , we have ′ < π½→0 π½π£ππ π΄ π£ππ(π) ⇒ π΄ π£ππ(π) < π΄ 1 π£ππ(π)( ′ ) π½ π½ ′ π£ππ π½′ 1−π½ ′ π£ππ < ⇒ π½′ 2 π½ ′ π£ππ 1−π½ ′ π£ππ π΄ π£ππ(π) π½′ π½′ ⇒ < π΄ π£ππ π π 2 π½ ′ π£ππ 1−π½ ′ π£ππ < π½′ π½′ 2 π½ ′ π£ππ 1−π½ ′ π£ππ π½π£ππ 1−π½π£ππ 1 π½′ π½′ π½′ π½ π½ . Since ππ π1 < π£ππ(π) and π΄ < ππ π1 π. Therefore 2 π½ ′ π£ππ Since πΌπ > 2 π½ ′ π£ππ 1−π½ ′ π£ππ π½′ 1−π½ ′ π£ππ π½′ π½′ π½′ > π£ππ(π)π½π£ππ 1−π½π£ππ ππ π1 π π£ππ π π = π½ π½ = π£ππ π π, we can choose π΄ = ππ π1 π£ππ(π) ≥ βΓ π1 . and πΌπ < πΌπ−1 < β― < π1 , we have π 1 βΓ < πΌπ < πΌπ−1 < β― < π1 . β . Let Γ be a distance regular graph and let βΓ be a Cheeger constant of Γ . Then we have πΌπ βΓ < ≤ π1 2 − π1 ≤ 2π1 , π1 where π1 is a smallest positive eigenvalue of the Laplacian. Let Γ be the Hamming graph π»(π, π). Then Γ is a distance regular graph with π vertices, valency π(π − 1) and π diameter. π We consider two cases: π π = 3, π = 2 and π π = 5, π = 2. Then 2 3 π π» 3,2 βΆ π£ = 8, π1 = 3, π1 = , πΌπ = βΓ < πΌπ π1 ≈ 0.545455 < 4 11 π1 2 − π1 = 0.942809 < 2 5 π π» 5,2 βΆ π£ = 32, π1 = 5, π1 = , πΌπ = βΓ < πΌπ π1 ≈ 0.437956 < . 24 137 π1 2 − π1 = 0.8 < 2π1 = 1.1547. . 2π1 = 0.894427. graph with π π Let Γ be the Johnson graph π½(π, π). Then Γ is a distance regular vertices, valency π(π − π) and π diameter. We consider two cases: π π = 10, π = 4 and π π = 11, π = 5. Then π π½ 10,4 βΆ π£ = 210, π1 = 24, π1 = βΓ < πΌπ π1 ≈ 0.429302 < πΌπ π1 ≈ 0.382344 < πΌπ = 11088 79091 π1 2 − π1 = 0.773879 < π π½ 11,5 βΆ π£ = 462, π1 = 30, π1 = βΓ < 11 , 30 10 , 24 πΌπ = 105 587 π1 2 − π1 = 0.812233 < . 2π1 = 0.856349. . 2π1 = 0.912871. In fact, πΌπ is 1 −π1 π 1 − π1 β― −ππ π 1 ππ , where π1 π , β― , ππ π are the π-numbers of a π-polynomial scheme. However, in general, the π -numbers are not easy to compute. We thus find an approximated value πΌπ of πΌπ with some error π½ > 0 in the following theorem. If π½ is sufficiently close to 0 then we obtain the estimate πΌπ which is close to πΌπ and easier to compute. π1 (π½) . Let Γ be a distance regular graph of order π£, and let π’π½ = π0 π΄1 + β― + ππ π½ (π½) π΄π be a Green’s function of Γ for π½ > 0. Then we have π½π£ππ πΌπ − 1 π½ −π£ππ π½ π½ < π½. π΄0 + πΏπ π’π (π½) For π½ > 0, let πΏ0 (π½) be the π × π matrix obtained by the removal of the first (π½) column of πΏπ π’π . Let πΏπ (π½) be the (π − π) × π − π matrix obtained by the removal from the first row(respectively, column) to the π–th row(respectively, column) of πΏ0 (π½) In the following proposition 5, we find an expression for πΌπ in terms of a basis (π’0 (π½) , π’1 (π½) , β― , π’π (π½) ) of π©(πΏπ π’π We also find an explicit expression for π’π πΏπ (π½) of πΏπ π’π (π½) . (π½) (π½) ) and the valencies ππ ’s. And by a determinant of a submatrix . (π½) Let Γ be a distance regular graph of order π£. For π½ > 0, let (π½) (π½) (π½) (π’0 , π’1 , β― , π’π ) be a basis of π©(πΏπ π’π ) with π’π π’π½ be a Green’s function of Γ. Then we have the following: π½π£ π πΌπ = lim+ , (π½) π½→0 π π’π (π½) where πΏπ π π π’ π=0 π π = (−1)π−π π½ (π½) = 1, and let −π£ det(πΏπ π½ ) ππ+1 ππ+2 β―ππ , π = 0,1, β― , π − 1, are submatrices of πΏπ π’π (π½) and det πΏπ π½ = 1. We consider a Johnson scheme π½(8,4) . Let Γ be a distance regular graph with respect to π΄1 . Then Γ is a distance regular graph with 70 vertices and valency 16 . Also, the valencies of π½(8,4) are 1, 16, 36, 16, 1 and πΏπ π’π (π½) Since πΌπ = 1 −π1 π 1 − π1 β― −ππ π 1 ππ , 8 14 18 20 , , ), 16 16 16 16 (π1 4 , β― ,π4 4 ) = (−7,20, −28,14) and π1 , β― , π4 = ( , have πΌπ = Let π½ = (π’0 315 1522 1 . 1000 (π½) , π’1 we ≈ 0.206965. By proposition 5, a basis for π©(πΏπ π’π (π½) ,π’2 (π½) , π’3 (π½) , π’4 (π½) )= (π½) ) is 263196691 15762389 94021 1001 , , , ,1 244140625 15625000 93750 1000 . Thus, we have (70) 1 π’0 (π½) + 16 π’1 (π½) + 36 π’2 1 1000 (π½) + 16 π’3 (π½) + 1 π’4 (π½) − 70 ≈ 0.206609. Also, by Theorem 4, we have πΌπ − πΌπ < 0.001 and πΌπ < πΌπ ≈ 0.206609 + 0.001 = 0.207609. Let π be the set of π × π matrices over πΊπΉ(ππ‘ ) (π ≤ π). Define the π–th relation on π by π₯, π¦ ∈ π π ⇔ ππππ π₯ − π¦ = π. Then π = π, π π (0 ≤ π ≤ π) is a π-polynomial scheme with respect to the ordering π 0 , π 1 , β― , π π . Let π = 2, π‘ = 1, π = 4, π = 5. Then, πΏπ π’π (π½) is as follows: Also, we have π = π£ = 1048576, π0 = 1, π1 = 465, π2 = 32550, π3 = 390600 π π π β―π and π4 = 624960 by using ππ = 1 1 2 π−1 π = 2,3, β― , π . π2 π3 β―ππ 1 Let π½ = . Then by Proposition 5, we obtain the unique basis ofπ©(πΏπ π’π 100 as follows: (π’0 (π½) , π’1 (π½) ,π’2 (π½) , π’3 (π½) , π’4 (π½) )= 3921317781669 486743013 661683 831 , , , ,1 358400000 17920000 448000 800 . (π½) ) Thus, we have (1048576) 1 π’0 (π½) + 465 π’1 (π½) + 32550 π’2 (π½) 1 100 + 390600 π’3 (π½) + 624960 π’4 (π½) − 1048576 ≈ 0.195023. Also, by Theorem 4, we have πΌπ < πΌπ ≈ 0.195023 + 0.01 = 0.205023. Let Γ be a distance regular graph with respect to π΄1 . Since an eigenvalue π1 of the Laplacian is βΓ < 256 , 465 we obtain βΓ < πΌπ ≈ 0.372405 < π1 πΌπ π1 < πΌπ π1 ≈ 0.372405. Moreover, π1 2 − π1 = 0.893299 < 2π1 = 1.04932. - Explicit expression of the krawtchouk polynomial For π½ > 0, let π’π½ be a Green’s function for a π–polynomial scheme. Then we have A Green’s function can be expressed as π’π½ = π‘π’0 (π½) π΄0 + π‘π’1 (π½) π΄1 + β― + π‘π’π for some nonzero π‘ ∈ β, where is the (π’0 the nullspace π©(πΏπ π’π πΏπ π’π (π½) ) of πΏπ π’π (π½) (π½) with π’π , π’1 (π½) (π½) π½ π΄π , β― , π’π (π½) ) unique basis of = 1. (π½) where π π = ππ − π1 (π½ + 1) for π = 1,2, β― , π and ππ , ππ , ππ are the same entries as in π΅1 . A Green's function π’π½ is defined only for π½ > 0, and π’π½ is expressed as a linear combination of adjacency matrices π΄π . But, for π½ ≤ 0, π’π½ may be a singular matrix, so there is no Green's function notion for this case. We however still have ππππ(πΏπ π’π (π’0 (π½) , π’1 (π½) , β― , π’π (π½) (π½) ) = π for π½ ∈ β, so we can obtain a unique basis ) of π©(πΏπ π’π (π½) ) with π’π (π½) = 1. We extend a notion of a Green's function associated with any real number π½ as follows. The following definition plays an important role for computation of the πnumber and the π-number. with π’π (π½) . For π½ ∈ β , let (π’0 (π½) , π’1 (π½) = 1 and let (π½) (π½) , β― , π’π (π½) )∈ π©(πΏπ π’π π½ (π½) ) π’π½,π© = π‘π’0 π΄0 + π‘π’1 π΄1 + β― + π‘π’π π΄π , where π‘ is some nonzero ∈ β if π½ > 0 and π‘ = 1 if π½ ≤ 0. Then π’π½,π© is called The normalized Green’s function. Let π»(5,3) be a Hamming scheme . Choosing π½ = − obtain a 5 × 6 matrix πΏπ π’π (π½) as follows: 40 (π½) 1 , 10 26 1 1 4 Also, a basis of π©(πΏπ π’π ) is (− , − , − , , , 1). 3 15 15 2 5 Thus we obtain a normalized Green’s function π’π½,π© as follows: π’π½,π© = − 40 π΄ 3 0 − 26 π΄ 15 1 − 1 π΄ 15 2 1 2 4 5 + π΄3 + π΄4 + π΄5 , where π΄π (π = 0,1,2, β― , 5) are the adjacency matrices of π» 5,3 . then we For some π½π ∈ β, let (π’0 (π) , π’1 (π) , β― , π’π (π) ) be a basis of πΏπ π’π (π½π ) Let π = π, π π (π = 0,1, β― , π) be a π-polynomial scheme, and let π = (ππ π ) (respectively, Q= (ππ π ) ) be the first eigenmatrix (respectively, the second eigenmatrix) of π. In the following proposition 6, we show that the π-th column vector of the second eigenmatrix Q belongs to π©(πΏπ π’π (π½π ) π1 π π1 (π) ) for π½π = − 1 (j = 0,1,2, β― , π) . That is, we show the relationship between π’π and component ππ π of the second eigenmatrix Q= (ππ π ) over the π - polynomial scheme. Let π = π, π π π½π = π1 π π1 (π = 0,1, β― , π) be a π-polynomial scheme. For − 1 j = 0,1,2, β― , π , let π’π½π,π© = π’0 π a normalized Green’s function with π’π (π) π΄0 + π’1 (π) π΄1 + β― + π’π π = 1. Then ππ π satisfy ππ π = ππ π΄π be π’π π’0 (π) (π) (π = 0,1, β― , π). That is, the π-th column of the second eigenmatrix Q= (ππ π ) is equal to ππ π’0 (π) (π’0 (π) , π’1 (π) , β― , π’π Moreover, we have ππ π = ππ π π’π (π) )π , where ππ is the π-th multiplicity of π. (π) . By Proposition 5 (b), π’π (π) = (−1)π−π where πΏπ π½π det(πΏπ π½π ) ππ+1 ππ+2 β―ππ , π = 0,1, β― , π − 1, are submatrices of πΏπ π’π (π½π ) and det πΏπ π½π = 1. Let π»(π, π) be a Hamming scheme. Since the Hamming scheme is a self-dual scheme, π = (ππ π ) is equal to Q= (ππ π ) . The π-number ππ π of a Hamming scheme π»(π, π) is defined by the Krawtchouk polynomial. Thus, we have π1 π = π π − 1 − ππ. And, π1 = π π − 1 . Thus π½π = π1 π π1 −1= π π−1 −ππ−π π−1 π π−1 =− ππ π(π−1) . For π½π = − ππ , π(π−1) we have a matrix πΏπ π’π (π½π ) as follows: where π π = π π − 2 − π(π − 1)(π½π + 1) for π = 1,2, β― , π, and π‘π = (π − π)(π − 1) for π = 1,2, β― , π − 1. . Let π» π, π a Hamming scheme. Let πΏπ submatrix of πΏπ π’π (π½π ) for π½π = − ππ π π−1 ππ π = ππ π π’π (π) π½π be an (π − π) × (π − π) (j = 0,1,2, β― , π). Then we have = π!det(πΏπ (−1)π+π−π ππ π! π½π ) . . Let π»(5,3) be a Hamming scheme. Then, the first eigenmatrix π = (ππ π ) is as follows: Let π = 1, then we obtain the π-number π1 π as the entries of the second column of the first eigenmatrix π. Let πΏπ π’π (π½1 ) be a 5 × 6 matrix for π½1 = − 3 1 5 3−1 =− 3 10 as follows: Then, the matrices πΏ0 Also, det πΏ0 − 4, det πΏ4 π½1 π½1 π½1 , πΏ1 π½1 , πΏ2 = 240 , det πΏ1 π½1 π½1 , πΏ3 π½1 and πΏ4 π½1 = −168, det πΏ2 are π½1 = −2. Therefore, π1 π (π = 0,1,2,3,4) are = 48, det πΏ3 π½1 = respectively. Thus, (10,7,4,1, −2, −5)π is the first column vector of the first eigenmatrix π. Thank you