SIMPLICES AND SPECTRA OF GRAPHS, CONTINUED BOJANMOHARANDIGORRIVIN 9 0 Abstract. Inthisnoteweshowthatthe(n−2)-dimensionalvolumesofcodimen- 0 sion2facesofann-dimensionalsimplexarealgebraicallyindependentquantities 2 ofthevolumesofitsedge-lengths. Theproofinvolvescomputationoftheeigen- n valuesofKnesergraphs. a J 0 2 ntroduction I ] G Let T be the set of congruence classes of n-simplices in Euclidean space En. n M The set T is an open manifold (also a semi-algebraic set) of dimension n+1 . n 2 h. Coincidentally, a simplex T ∈ Tn is determined by the n+21 lengths of its ed(cid:0)ges(cid:1). t Furthermore, the square ofthevolume ofT ∈ Tn isapol(cid:0)yno(cid:1)mial inthe squaresof ma theedge-lengthsℓij = kvi−vjk2 (1 ≤ i < j ≤ n+1),wherev1,...vn+1 arethevertices of T. This polynomial is given by the Cayley-Menger determinant formula (cf., [ e.g., [5]or[2]): 1 v (−1)n+1 4 (1) V2(T) = detC, 8 2n(n!)2 2 3 whereCistheCayley-Mengermatrixofdimension(n+2)×(n+2),whoserowsand . columns are indexedby {0,1,...,n+1}andwhose entriesare definedasfollows: 1 0 9 = 0, i j 0 : C = 1, ifi = 0 or j = 0, andi , j v ij i ℓ2 , otherwise. X ij ar Notethatann-simplexhasn+21 edgesandthesamenumberof(n−2)-dimensional faces, and sothe following question isnatural: (cid:0) (cid:1) Date:January21,2009. SupportedinpartbytheARRSResearchProgramP1-0297,byanNSERCDiscoveryGrantand by the Canada Research Chair program. On leave from Department of Mathematics, IMFM & FMF,UniversityofLjubljana,Ljubljana,Slovenia. TheauthorwouldliketothanktheAmericanInstitute ofMathematicsforaninvitationtothe workshopon“RigidityandPolyhedralCombinatorics”,wherethisworkwasstarted. Theauthor hasprofitedfromdiscussionswithIgorPak,EzraMiller,andBobConnelly. 1 2 BOJANMOHARANDIGORRIVIN Question 1. Is the congruence class of every n-simplex determined by the (n−2)- dimensional volumesof its (n−2)-faces? Question 1 must be classical, but the earliest reference stating it that we are aware ofis WarrenSmith’s PhDthesis [10]. AttheAIMworkshoponRigidityandPolyhedralCombinatorics,BobConnelly (who wasunaware ofthe reference [10])raised the following: Question2. Isthevolumeofeveryn-simplexdeterminedbythe(n−2)-dimensional volumes ofits (n−2)-faces? = In fact, Connelly stated Question 2 for n 4, which is the first case where the question is open. For n = 3 the answer is trivially ”Yes”, since 3 −2 = 1, and we are simply asking if the volume of the simplex is determined by its edge-lengths. In dimension 2, the answer is trivially ”No”, since 2 − 2 = 0, and the volume of codimension-2 facesofa triangle carriesnoinformation. ffi ffi Clearly,thea rmativeanswertoQuestion1wouldimplyana rmativeanswer toQuestion2. InthispaperwefirstshowthattheanswertoQuestion2,andhence also to Question 1 is negative for every n ≥ 4. We actually found out that this = has been answered previously for n 4 in [1], where an example is given and attributed to Philip Tuckey; seealso [3]. Our examples are given in a separate section. Several reasons suggest that the ffi following question maystill havean a rmative answer: Question 3. Is it true that for every choice of n+1 positive real numbers, there are 2 only finitely many congruence classes of n-simplices whose (n − 2)-dimensional (cid:0) (cid:1) volumes ofthe (n−2)-facesare equalto these numbers? Inthis note we show thata weakerstatement holds: Theorem 4. The n+1 (n− 2)-dimensional volumes of the (n−2)-faces of an n-simplex 2 are algebraicallyindependentoverC[ℓ ; 1 ≤ i < j ≤ n+1]. (cid:0) (cid:1) ij Theorem4isclearlyanecessarystepinthedirectionofresolvingQuestion3,but is far from sufficient. To show it, consider the map of R(n+1)n/2 to R(n+1)n/2, which sends the vector ℓ of edge-lengths of an n-simplex to the vector Y of volumes of (n − 2)-dimensional faces. To show Theorem 4, it is enough to check that the = Jacobian J(ℓ) ∂Y/∂ℓ is non-singular at one point. We will use the most obvious point p , the one corresponding to a regular simplex with all edge-lengths equal 1 = to 1. By symmetry considerations, the Jacobian J(p ) can be written as J(p ) cM, 1 1 where cis aconstant andMis 1, ifthe edgee isincidentwith the (n−2)-face F = M e,F (0, otherwise. SIMPLICESANDSPECTRAOFGRAPHS, CONTINUED 3 The first observation isthatthe constant cabove isnot equalto 0: 1 = Lemma5. J(p ) M. 1 (n−2)!(n−1)1/22(n−4)/2 (n−1)1/2 Proof. Letν = denotethe(n−2)-dimensionalvolumeoftheregular (n−2)!2(n−2)/2 (n − 2)-simplex with all edge-lengths 1. Let us observe that the volume of a k- dimensionalsimplexisahomogeneousfunctionofdegreekoftheedge-lengths. An application of Euler’sHomogeneous Function Theorem shows thatat p , 1 ∂Y 2 ν, ifthe edge eisincident with the (n−2)-face F F = n−1 ∂ℓ (0, otherwise. e Thisimpliesthat c = 2 ν andcompletes the proof. (cid:3) n−1 heeigenvalues of T M Asshownabove,Theorem4reducestotheassertionthatthedeterminantofthe matrix M is not zero. We will actually be able to compute all eigenvalues of M, which isof interest in itsown right. Theorem6. Eigenvaluesof Mareλ = n−1 (simpleeigenvalue),λ = 1withmultiplic- 1 2 2 ity 12(n+1)(n−2), and λ3 = 2−nwith(cid:0)mul(cid:1)tiplicityn. Corollary 7. Theabsolute valueof the determinantof Mequals 1(n−2)n+1(n−1) , 0, 2 for n > 2. To prove Theorem 6, let us first observe that the n+1 rows of M are indexed by 2 the2-elementsubsetsofthesetR = {1,...,n+1},anditscolumnsareindexedbythe (cid:0) (cid:1) (n−1)-subsetsFofR. ByreplacingeachcolumnindexFwithitscomplementR\F, then the columns are indexed by the same set as the rows. After this convention, = thematrixMbecomesasymmetricmatrixwithzerodiagonalsinceM 1ifand e,f only if e ⊆ R\ f, which is equivalent to f ⊆ R \ e. Therefore, M is the adjacency matrix of a graph G whose vertices are the 2-element subsets of R, and two of n themareadjacentifandonlyiftheyaredisjoint. Thus,thecomplementG ofG is n n + isomorphic to the line graph L(Kn+1)ofthe complete graph Kn+1 on n 1 vertices. TheeigenvaluesofL(Kn+1)are(see[4,p.19]): t1 = 2n−2,t2 = −2,andt3 = n−3, withthesamemultiplicities(respectively)asclaimedabovefortheeigenvaluesof M. SincethegraphL(Kn+1)isregular,itisaneasyexercisetoseethatitsadjacency matrix A and the adjacency matrix M of its complement have the same set of 4 BOJANMOHARANDIGORRIVIN eigenvectors. By using the fact that A + M + I = xt · x, where x = (1,...,1)t is the eigenvector of A and M corresponding to the dominant eigenvalues of these matrices,weconcludethattheeigenvaluesofMareλ = n+1 −t −1andλ = −t −1 1 2 1 i i fori = 2,3(preserving multiplicities). Thus, λ1 = n+21 −(cid:0)2n+(cid:1) 1 = n−21 ,λ2 = 1,and λ = 2−n, respectively. 3 (cid:0) (cid:1) (cid:0) (cid:1) ingular examples S Let us consider the n-simplex in Rn with vertices v ,v ,...,v given as follows. 0 1 n The vertex v has the first n − 2 coordinates equal to ((n − 1)1/2 + 1)/(21/2(n − 2)), 0 while its last two coordinates are 0. For i = 1,2,...,n − 2, the vertex v has ith i coordinate equal to 2−1/2 and all other coordinates 0. These n − 1 vertices form a regular (n − 2)-simplex contained in Rn−2 ⊂ Rn with all side lengths 1. Let a = 1 n−2v be its barycenter, and let c := kv − ak denote the distance from a n−1 i=0 i 0 2 to the vPertices vi. A short calculation shows that c2 = 12 − 2(n1−1). Now, let vn−1 be obtained from a by changing its last two coordinates to be real numbers p and q satisfyingp2+q2 = 1−c2. Similarly,letv beobtainedinthesamewaybychoosing n anotherpairr,sofnumberssatisfyingr2+s2 = 1−c2. Thisgivesrisetoann-simplex whose all sides are equal to 1 except for the side v v whose square length is n−1 n t := (p−r)2+(q−s)2. Byfixingt,thissimplexisdetermineduptocongruence,and we denote it by T(t). Observe that t may take any value between 0 and 4(1−c2), byselecting p,q,r,sappropriately. Nextweobservethatthevolumesofthe(n−2)-facesofT(t)takeonlytwovalues. If an (n − 2)-face does not contain both v and v , then it is a regular simplex, n−1 n whosevolumeisindependentoft. Ontheotherhandifan(n−2)-simplexcontains = v and v ,its volume w w(t)isuniquelydeterminedby t. In fact, ifwe putthe n−1 n square distances in the Cayley-Menger determinant, we conclude that w(t)2 is a quadratic polynomial in t, w(t)2 = αt2 +βt+γ. If t = 0, the volume is 0, so γ = 0. For t = 1 we have the regular (n−2)-simplex, so α+β = n−1 . Finally, using 2n−2((n−2)!)2 (1)(withthevalueofnbeingreplacedbyn−2)andlookingattheCayley-Menger determinantexpansion term with t2, we conclude that (−1)n−1 α = − det(J −I ), 2n−2((n−2)!)2 n−2 n−2 where J is the all-1-matrix and I is the identity matrix of order n− 2. Since n−2 n−2 det(J −I ) = (−1)n−3(n−3), we conclude that α = −(n−3)22−n/((n −2)!)2 and n−2 n−2 β = (n−2)21−n/((n−2)!)2. Inparticular, 1 w(t)2 = ((3−n)t2 +(2n−4)t). 2n−2((n−2)!)2 SIMPLICESANDSPECTRAOFGRAPHS, CONTINUED 5 This function is symmetric around the point t = n−2. Consequently, the non- 0 n−3 congruent n-simplices T(t − x) and T(t + x) have the same (n − 2)-volumes of 0 0 their (n−2)-faces for each admissible value of x, i.e. for 0 < x < n−2−4/(n−1). These n−3 examplesthusshow that Questions 1 and2 have negative answers. oncluding remarks C One can ask the same question as above for other dimension-complementary volumes, i.e. about the volumes of the (k − 1)-faces and the (n − k)-faces of an n-simplex,where2 ≤ k ≤ n/2. Ifonewouldcompare,similarlyasinthecasek = 2 above,thedependenceof(n−k)-volumesofan(n−k)-faceQonthe(k−1)-volumes ofthe(k−1)-facesF ⊂ Q,thecorresponding“Jacobian”wouldagainbeaconstant multiple of a symmetric matrix M, whose entries are indexed by the k-subsets of the set R = {1,...,n + 1} (after the column indices pass to the complementary subsets), and 1, ifthe E∩F = ∅ = (2) M E,F (0, otherwise. + The graph whose adjacency matrix is M is known as the Kneser graph K(n 1,k). Its eigenvalues can be comuted using the methods from the theory of association schemesand can befound, forexample, in [8,Section 9.4]. Theorem8. Letnandkbeintegers,where2 ≤ k ≤ n/2,and letMbethematrixoforder n+1 × n+1 whoseentriesare determinedby(2). The eigenvaluesof Mare theintegers k k (cid:0) (cid:1) (cid:0) (cid:1) n−k−i+1 λ = (−1)i , i = 0,1,...,k. i k−i ! Since 2 ≤ k ≤ n/2, none of the eigenvalues in Theorem 8 is zero. This raises the question whether there is an analogy with Theorem 4 for 2 ≤ k ≤ n/2, between the collection of the n+1 (n − k)-dimensional volumes of the (n − k)-faces of an k n-simplexand thecollection ofall (k−1)-dimensional volumesofits(k−1)-faces. (cid:0) (cid:1) Asa final remark, we would like to point out that our original approach to this problem [9] used results about divisors [6] (also known as equitable partitions [8]) combined with the representation theory of the symmetric group and the notion ofGelfand pairsasdevelopedin [7]. eferences R [1] JohnW.Barrett.FirstorderReggecalculus.Class.QuantumGrav.11(1994)2723–2730. [2] MarcelBerger.GeometryI.Springer,Berlin,1987. [3] EugenioBianchiandLeonardoModesto.TheperturbativeRegge-calculusregimeofloopquantum gravity.NuclearPhysicsB796[FS](2008)581–621. 6 BOJANMOHARANDIGORRIVIN [4] NormanBiggs.Algebraicgraphtheory,secondedition.CambridgeUniversityPress,2001. [5] LeonardM.Blumenthal.Theoryandapplicationsofdistancegeometry.ClarendonPress,Oxford, 1953. [6] Dragosˇ M. Cvetkovic´, Michael Doob, and Horst Sachs. Spectra of graphs. Johann Ambrosius Barth,Heidelberg,thirdedition,1995. [7] PersiDiaconis.Grouprepresentationsinprobabilityandstatistics.InstituteofMathematicalStatis- ticsLectureNotes—MonographSeries,11.InstituteofMathematicalStatistics,Hayward,CA, 1988. [8] Chris Godsil and Gordon Royle. Algebraic graph theory. Graduate Texts in Mathematics, 207. Springer-Verlag,NewYork,2001. [9] IgorRivin,Simplicesandspectraofgraphs.arXiv:0803.1317v1. [10] Warren Douglas Smith. Studies in computational geometry motivated by mesh generation. PhD thesis,PrincetonUniversity,1989. DepartmentofMathematics,SimonFraserUniversity,Burnaby,B.C.V5A1S6 E-mailaddress: [email protected] DepartmentofMathematics,TempleUniversity,Philadelphia E-mailaddress: [email protected]