arXiv++ Combinatorics

We give a zero curvature proof of Dehn-Sommerville for finite simple graphs. It uses a parametrized Gauss-Bonnet formula telling that the curvature of the valuation G to f_G(t)=1+f0 t + ... + fd t^(d+1) defined by the f-vector of G is the anti-derivative F of f evaluated on the unit sphere S(x). Gauss Bonnet is then parametrized, f_G(t) = sum_x F_(S(x))(t), and holds for all simplicial complexes G. The Gauss-Bonnet formula chi(G)=sum_x K(x) for Euler characteristic chi(G) is the special case t=-1. Dehn-Sommerville is equivalent to the reflection symmetry f_G(t)+(-1)^d f_G(-1-t)=0 which is equivalent to the same symmetry for F. Gauss-Bonnet therefore relates Dehn-Sommerville for G with Dehn-Sommerville for the unit spheres S(x), where it is a zero curvature condition. A class X_d of complexes for which Dehn-Sommerville holds is defined inductively by requiring chi(G)=1+(-1)^d and S(x) in X_(d-1) for all x. It starts with X_(-1)={}. Examples are simplicial spheres, including homology spheres, any odd-dimensional discrete manifold, any even-dimensional discrete manifold with chi(G)=2. It also contains non-orientable ones for which Poincar'e-duality fails or stranger spaces like spaces where the unit spheres allow for two disjoint copies of manifolds with chi(G)=1. Dehn-Sommerville is present in the Barycentric limit. It is a symmetry for the Perron-Frobenius eigenvector of the Barycentric refinement operator A. The even eigenfunctions of A^T, the Barycentric Dehn-Sommerville functionals, vanish on X like 22 f1 - 33 f2 + 40 f3 - 45f4=0 for all 4-manifolds.

We study the average simplex cardinality Dim^+(G) = sum_x |x|/(|G|+1) of a finite abstract simplicial complex G. The functional is a homomorphism from the monoid of simplicial complexes to the rationals: the formula Dim^+(G + H) = Dim^+(G) + Dim^+(H) holds for the join + similarly as for the augmented inductive dimension dim^+(G) = dim(G)+1 where dim is the inductive dimension dim(G) = 1+ sum_x dim(S(x))/|G| with unit sphere S(x) (a recent theorem of Betre and Salinger). In terms of the generating function f(t) = 1+v_0 t + v_1 t^2 + ... +v_d t^(d+1) defined by the f-vector (v_0,v_1, \dots) of G for which f(-1) is the genus 1-X(G) with Euler characteristic X and f(1)=|G|+1 is the augmented number of simplices, the average cardinality is the logarithmic derivative Dim^+(f) = f'(1)/f(1) of f at 1. Beside introducing the average cardinality and establishing its compatibility with arithmetic, we prove two results: 1) the inequality dim^+(G)/2 <= Dim^+(G) with equality for complete complexes. 2) the limit C_d of Dim^+(G_n) for n to infinity is the same for any initial complex G_0 of maximal dimension d and the constant c_d is explicitly given in terms of the Perron-Frobenius eigenfunction of the universal Barycentric refinement operator and is for positive d always a rational number in the open interval ((d+1)/2,d+1).

We prove a Reeb sphere theorem for finite simple graphs. The result bridges two different definitions of spheres in graph theory. We also reformulate Morse conditions in terms of the center manifolds, the level surface graphs {f=f(x)} in the unit sphere S(x). In the Morse case these graphs are either spheres, the empty graph or the product of two spheres.

A finite simple graph is called a 2-graph if all of its unit spheres S(x) are cyclic graphs of length 4 or larger. A 2-graph G is Eulerian if all vertex degrees of G are even. An edge refinement of a graph splits an edge (a,b) to two edges (a,c),(c,b) and connects the newly added vertex c to the intersection of S(a) with S(b). Theorem I assures that every 2-graph can be rendered Eulerian by successive edge refinements. The construction is explicit using geodesic cutting. After the refinement, we have an Eulerian 2-graph that carries a natural geodesic flow. We construct some ergodic ones. A 2-graph with boundary is finite simple graph for which every unit sphere is either a path graph of length n larger than 1 or a cyclic graph of length larger than 3. 2-balls are special 2-graphs are simply connected with a circular boundary. Theorem II tells that every 2-ball can be edge refined using interior edges to become Eulerian if and only if its boundary has length divisible by 3. A billiard map is defined already if all interior vertices have even degree. We will construct some ergodic billiards in 2-balls, where the geodesics bouncing off at the boundary symmetrically and which visit every interior edge exactly once. A consequence of Theorem II is that an Eulerian billiard which is ergodic must have a boundary length that is divisible by 3. We also construct other 2-graphs like tori with ergodic geodesic flows. This clashes with experience in the continuum, where tori have periodic points minimizing the length in homology classes of paths. Ergodic Eulerian 2-graphs or billiards are exciting because they satisfy a Hopf-Rynov result: there exists a geodesic connection between any two vertices. We get so unique canonical metric associated to any ergodic Eulerian graph. It is non-local in the sense that two adjacent vertices can have large distance.

Extending a theorem of Whitney of 1931 we prove that all connected d-graphs are Hamiltonian for positive d. A d-graph is a type of combinatorial manifold which is inductively defined as a finite simple graph for which every unit sphere is a (d-1)-sphere. A d-sphere is d-graph such that removing one vertex renders the graph contractible. A graph is contractible if there exists a vertex for which the unit sphere and the graph without that vertex are both contractible. These inductive definitions are primed with the assumptions that the empty graph 0 is the (-1)-sphere and that the one-point graph 1 is the smallest contractible graph. The proof is constructive and shows that unlike for general graphs, the complexity of the construction of Hamiltonian cycles in d-graphs is polynomial in the number of vertices of the graph.

Defined by a single axiom, finite abstract simplicial complexes belong to the simplest constructs of mathematics. We look at a a few theorems.

While Euler characteristic X(G)=sum_x w(x) super counts simplices, Wu characteristics w_k(G) = sum_(x_1,x_2,...,x_k) w(x_1)...w(x_k) super counts simultaneously pairwise interacting k-tuples of simplices in a finite abstract simplicial complex G. More general is the k-intersection number w_k(G_1,...G_k), where x_i in G_i. We define interaction cohomology H^p(G_1,...,G_k) compatible with w_k and invariant under Barycentric subdivison. It allows to distinguish spaces which simplicial cohomology can not: it can identify algebraically the Moebius strip and the cylinder for example. The cohomology satisfies the Kuenneth formula: the Poincare polynomials p_k(t) are ring homomorphisms from the strong ring to the ring of polynomials in t. The Dirac operator D=d+d^* defines the block diagonal Hodge Laplacian L=D^2 which leads to the generalized Hodge correspondence b_p(G)=dim(H^p_k(G)) = dim(ker(L_p)) and Euler-Poincare w_k(G)=sum_p (-1)^p dim(H^p_k(G)) for Wu characteristic. Also, like for traditional simplicial cohomology, isospectral Lax deformation D' = [B(D),D], with B(t)=d(t)-d^*(t)-ib(t), D(t)=d(t)+d(t)^* + b(t) can deform the exterior derivative d. The Brouwer-Lefschetz fixed point theorem generalizes to all Wu characteristics: given an endomorphism T of G, the super trace of its induced map on k'th cohomology defines a Lefschetz number L_k(T). The Brouwer index i_T,k(x_1,...,x_k) = product_j=1^k w(x_j) sign(T|x_j) attached to simplex tuple which is invariant under T leads to the formula L_k(T) = sum_T(x)=x i_T,k(x). For T=Id, the Lefschetz number L_k(Id) is equal to the k'th Wu characteristic w_k(G) of the graph G and the Lefschetz formula reduces to the Euler-Poincare formula for Wu characteristic.

We prove that the spectrum of the Kirchhoff Laplacian H0 of a finite simple Barycentric refined graph and the spectrum of the connection Laplacian L of G determine each other: we prove that L-L^(-1) is similar to the Hodge Laplacian H of G which is in one dimensions the direct sum of the Kirchhoff Laplacian H0 and its 1-form analog H1. The spectrum of a single choice of H0,H1 or H alone determines the Betti numbers b0,b1 of G as well as the spectrum of the other matrices. It follows that b0 is the number of eigenvalues 1 of L and that b1 is the number of eigenvalues -1 of L. For a general abstract finite simplicial complex G, we express the matrix entries g(x,y) = w(x) w(y) X( St(x) cap St(y) ) of the inverse of L using stars St(x)= { z in G | x subset of z } of x and w(x)=(-1)^dim(x) and Euler characteristic X. One can see W+(x)=St(x) and W-(x)={ z in G | z subset x } as stable and unstable manifolds of a simplex x in G and g(x,y) =w(x) w(y) X(W+(x) cap W+(y)) as heteroclinic intersection numbers or curvatures and the identity L g=1 as a collection of Gauss-Bonnet formulas. The homoclinic energy w(x)=X(W+(x) cap W-(x)) by definition adds up to X(G). The matrix M(x,y)=w(x) w(y) X(W-(x) cap W-(y)) is similar to L(x,y)=X(W-(x) cap W-(y)). The sum of the matrix entries of M is the definition of Wu characteristic. For dimension 2 and higher we don't know yet how to recover the Betti numbers from the eigenvalues of the matrix H or from L. So far, it can only be obtained from a collection of block matrices, via the Hodge relations b_k = dim(H_k). A natural conjecture is that for a Barycentric refinement of a complex G, the spectrum of L determines the Betti vector. We know this now in one dimensions.

The connection zeta function of a finite abstract simplicial complex G is defined as zeta_L(s)=sum_x 1/lambda_x^s, where lambda_x are the eigenvalues of the connection Laplacian L defined by L(x,y)=1 if x and y intersect and 0 else. (I) As a consequence of the spectral formula chi(G)=sum_x (-1)^dim(x) = p(G)-n(G), where p(G) is the number of positive eigenvalues and n(G) is the number of negative eigenvalues of L, both the Euler characteristic chi(G)=zeta(0)-2 i zeta'(0)/pi as well as determinant det(L)=e^zeta'(0)/pi can be written in terms of zeta. (II) As a consequence of the generalized Cauchy-Binet formula for the coefficients of the characteristic polynomials of a product of matrices we show that for every one-dimensional simplicial complex G, the functional equation zeta(s)=zeta(-s) holds, where zeta(s) is the Zeta function of the positive definite squared connection operator L^2 of G. Equivalently, the spectrum sigma of the integer matrix L^2 for a 1-dimensional complex always satisfies the symmetry sigma = 1/sigma and the characteristic polynomial of L^2 is palindromic. The functional equation extends to products of one-dimensional complexes. (III) Explicit expressions for the spectrum of circular connection Laplacian lead to an explicit entire zeta function in the Barycentric limit. The situation is simpler than in the Hodge Laplacian H=D^2 case where no functional equation was available. In the connection Laplacian case, the limiting zeta function is a generalized hypergeometric function which for an integer s is given by an elliptic integral over the real elliptic curve w^2=(1+z)(1-z)(z^2-4z-1), which has the analytic involutive symmetry (z,w) to (1/z,w/z^2).

We prove that that the number p of positive eigenvalues of the connection Laplacian L of a finite abstract simplicial complex G matches the number b of even dimensional simplices in G and that the number n of negative eigenvalues matches the number f of odd-dimensional simplices in G. The Euler characteristic X(G) of G therefore can be spectrally described as X(G)=p-n. This is in contrast to the more classical Hodge Laplacian H which acts on the same Hilbert space, where X(G) is not yet known to be accessible from the spectrum of H. Given an ordering of G coming from a build-up as a CW complex, every simplex x in G is now associated to a unique eigenvector of L and the correspondence is computable. The Euler characteristic is now not only the potential energy summing over all g(x,y) with g=L^{-1} but also agrees with a logarithmic energy tr(log(i L)) 2/(i pi) of the spectrum of L. We also give here examples of L-isospectral but non-isomorphic abstract finite simplicial complexes. One example shows that we can not hear the cohomology of the complex.

We define a ring R of geometric objects G generated by finite abstract simplicial complexes. To every G belongs Hodge Laplacian H as the square of the Dirac operator determining its cohomology and a unimodular connection matrix L). The sum of the matrix entries of the inverse of L is the Euler characteristic. The spectra of H as well as inductive dimension add under multiplication while the spectra of L multiply. The nullity of the Hodge of H are the Betti numbers which can now be signed. The map assigning to G its Poincare polynomial is a ring homomorphism from R the polynomials. Especially the Euler characteristic is a ring homomorphism. Also Wu characteristic produces a ring homomorphism. The Kuenneth correspondence between cohomology groups is explicit as a basis for the product can be obtained from a basis of the factors. The product in R produces the strong product for the connection graphs and leads to tensor products of connection Laplacians. The strong ring R is also a subring of the full Stanley-Reisner ring S Every element G can be visualized by its Barycentric refinement graph G1 and its connection graph G'. Gauss-Bonnet, Poincare-Hopf or the Brouwer-Lefschetz extend to the strong ring. The isomorphism of R with a subring of the strong Sabidussi ring shows that the multiplicative primes in R are the simplicial complexes and that every connected element in the strong ring has a unique prime factorization. The Sabidussi ring is dual to the Zykov ring, in which the Zykov join is the addition. The connection Laplacian of the d-dimensional lattice remains invertible in the infinite volume limit: there is a mass gap in any dimension.

The Zykov ring of signed finite simple graphs with topological join as addition and compatible multiplication is an integral domain but not a unique factorization domain. We know that because by taking graph complements, it becomes isomorphic to the strong Sabidussi ring with disjoint union as addition. We prove that the Euler characteristic is a ring homomorphism from the strong ring to the integers by demonstrating that the strong ring is homotopic to a Stanley-Reisner Cartesian ring. More generally, the Kuenneth formula holds on the strong ring so that the Poincare polynomial is compatible with the ring structure. The Zykov ring has the clique number as a ring homomorphism. Furthermore, the Cartesian ring has the property that the functor which attaches to a graph the spectrum of its connection Laplacian is multiplicative. The reason is that the connection Laplacians do tensor under multiplication, similarly to what the adjacency matrix does for the weak ring. The strong ring product of two graphs contains both the weak and direct product graphs as subgraphs. The Zykov, Sabidussi or Stanley-Reisner rings are so manifestations of a network arithmetic which has remarkable cohomological properties, dimension and spectral compatibility but where arithmetic questions like the complexity of detecting primes or factoring are not yet studied well. We illustrate the Zykov arithmetic with examples, especially from the subring generated by point graphs which contains spheres, stars or complete bipartite graphs. While things are formulated in the language of graph theory, all constructions generalize to the larger category of finite abstract simplicial complexes.

Assume G is a finite abstract simplicial complex with f-vector (v0,v1, ...), and generating function f(x) = sum(k=1 v(k-1) x^k = v0 x + v1 x^2+ v2 x^3 + ..., the Euler characteristic of G can be written as chi(G)=f(0)-f(-1). We study here the functional f1'(0)-f1'(-1), where f1' is the derivative of the generating function f1 of G1. The Barycentric refinement G1 of G is the Whitney complex of the finite simple graph for which the faces of G are the vertices and where two faces are connected if one is a subset of the other. Let L is the connection Laplacian of G, which is L=1+A, where A is the adjacency matrix of the connection graph G', which has the same vertex set than G1 but where two faces are connected they intersect. We have f1'(0)=tr(L) and for the Green function g L^(-1) also f1'(-1)=tr(g) so that eta1(G) = f1'(0)-f1'(-1) is equal to eta(G)=tr(L-L^(-1). The established formula tr(g)=f1'(-1) for the generating function of G1 complements the determinant expression det(L)=det(g)=zeta(-1) for the Bowen-Lanford zeta function zeta(z)=1/det(1-z A) of the connection graph G' of G. We also establish a Gauss-Bonnet formula eta1(G) = sum(x in V(G1) chi(S(x)), where S(x) is the unit sphere of x the graph generated by all vertices in G1 directly connected to x. Finally, we point out that the functional eta0(G) = sum(x in V(G) chi(S(x)) on graphs takes arbitrary small and arbitrary large values on every homotopy type of graphs.

We prove a Gauss-Bonnet formula X(G) = sum_x K(x), where K(x)=(-1)^dim(x) (1-X(S(x))) is a curvature of a vertex x with unit sphere S(x) in the Barycentric refinement G1 of a simplicial complex G. K(x) is dual to (-1)^dim(x) for which Gauss-Bonnet is the definition of Euler characteristic X. Because the connection Laplacian L'=1+A' of G is unimodular, where A' is the adjacency matrix of of the connection graph G', the Green function values g(x,y) = (1+A')^-1_xy are integers and 1-X(S(x))=g(x,x). Gauss-Bonnet for K^+ reads therefore as str(g)=X(G), where str is the super trace. As g is a time-discrete heat kernel, this is a cousin to McKean-Singer str(exp(-Lt)) = X(G) for the Hodge Laplacian L=dd^* +d^*d which lives on the same Hilbert space than L'. Both formulas hold for an arbitrary finite abstract simplicial complex G. Writing V_x(y)= g(x,y) for the Newtonian potential of the connection Laplacian, we prove sum_y V_x(y) = K(x), so that by the new Gauss-Bonnet formula, the Euler characteristic of G agrees with the total potential theoretic energy sum_x,y g(x,y)=X(G) of G. The curvature K now relates to the probability measure p minimizing the internal energy U(p)=sum_x,y g(x,y) p(x) p(y) of the complex. Since both the internal energy (here linked to topology) and Shannon entropy are natural and unique in classes of functionals, we then look at critical points p the Helmholtz free energy F(p)=(1-T) U(p)-T S(p) which combines the energy functional U and the entropy functional S(p)=-sum_x p(x) log(p(x)). As the temperature T changes, we observe bifurcation phenomena. Already for G=K_3 both a saddle node bifurcation and a pitchfork bifurcation occurs. The saddle node bifurcation leads to a catastrophe: the function T -> F(p(T),T) is discontinuous if p(T) is a free energy minimizer.

Given an abstract simplicial complex G, the connection graph G' of G has as vertex set the faces of the complex and connects two if they intersect. If A is the adjacency matrix of that connection graph, we prove that the Fredholm characteristic det(1+A) takes values in {-1,1} and is equal to the Fermi characteristic, which is the product of the w(x), where w(x)=(-1)^dim(x). The Fredholm characteristic is a special value of the Bowen-Lanford zeta function and has various combinatorial interpretations. The unimodularity theorem proven here shows that it is a cousin of the Euler characteristic as the later is the sum of the w(x). Unimodularity implies that the matrix 1+A has an inverse which takes integer values. Experiments suggest the conjecture that the range of the Green function values, the union of the entries of the inverse of 1+A form a combinatorial invariant of the simplicial complex and do not change under Barycentric or edge refinements.

The counting function on the natural numbers defines a discrete Morse-Smale complex with a cohomology for which topological quantities like Morse indices, Betti numbers or counting functions for critical points of Morse index are explicitly given in number theoretical terms. The Euler characteristic of the Morse filtration is related to the Mertens function, the Poincaré-Hopf indices at critical points correspond to the values of the Moebius function. The Morse inequalities link number theoretical quantities like the prime counting functions relevant for the distribution of primes with cohomological properties of the graphs. The just given picture is a special case of a discrete Morse cohomology equivalent to simplicial cohomology. The special example considered here is a case where the graph is the Barycentric refinement of a finite simple graph.

We prove Gauss-Bonnet and Poincare-Hopf formulas for multi-linear valuations on finite simple graphs G=(V,E) and answer affirmatively a conjecture of Gruenbaum from 1970 by constructing higher order Dehn-Sommerville valuations which vanish for all d-graphs without boundary. An example of a quadratic valuation is the Wu characteristic w(G) which sums (-1)^(dim(x)+dim(y)) over all intersecting pairs of complete subgraphs x,y of a G. More generally, an intersection number w(A,B) sums (-1)^(dim(x)+dim(y)) over pairs x,y, where x is in A and y is in B and x,y intersect. w(G) is a quadratic Euler characteristic X(G), where X sums (-1)^dim(x) over all complete subgraphs x of G. We prove that w is multiplicative, like Euler characteristic: w(G x H) = w(G) w(H) for any two graphs and that w is invariant under Barycentric refinements. We construct a curvature K satisfying Gauss-Bonnet w(G) = sum K(a). We also prove w(G) = X(G)-X(dG) for Euler characteristic X which holds for any d-graph G with boundary dG. We also show higher order Poincare-Hopf formulas: there is for every multi-linear valuation X and function f an index i(a) such that sum i(a)=X(G). For d-graphs G and X=w it agrees with the Euler curvature. For the vanishing multi-valuations which were conjectured to exist, like for the quadratic valuation X(G) = (V X) Y with X=(1,-1,1,-1,1),Y=(0,-2,3,-4,5) on 4-graphs, discrete 4 manifolds, where V_{ij}(G) is the f-matrix counting the number of i and j-simplices in G intersecting, the curvature is constant zero. For all graphs and multi-linear Dehn-Sommerville relations, the Dehn-Sommerville curvature K(v) at a vertex is a Dehn-Sommerville valuation on the unit sphere S(v). We show X V(G) Y = v(G) Y for any linear valuation Y of a d-graph G with f-vector v(G). This provides examples for the Gruenbaum conjecture.

Given a finite simple graph G, let G' be its barycentric refinement: it is the graph in which the vertices are the complete subgraphs of G and in which two such subgraphs are connected, if one is contained into the other. If L(0)=0<L(1) <= L(2) ... <= L(n) are the eigenvalues of the Laplacian of G, define the spectral function F(x) as the function F(x) = L([n x]) on the interval [0,1], where [r] is the floor function giving the largest integer smaller or equal than r. The graph G' is known to be homotopic to G with Euler characteristic chi(G')=chi(G) and dim(G') >= dim(G). Let G(m) be the sequence of barycentric refinements of G=G(0). We prove that for any finite simple graph G, the spectral functions F(G(m)) of successive refinements converge for m to infinity uniformly on compact subsets of (0,1) and exponentially fast to a universal limiting eigenvalue distribution function F which only depends on the clique number respectively the dimension d of the largest complete subgraph of G and not on the starting graph G. In the case d=1, where we deal with graphs without triangles, the limiting distribution is the smooth function F(x) = 4 sin^2(pi x/2). This is related to the Julia set of the quadratic map T(z) = 4z-z^2 which has the one dimensional Julia set [0,4] and F satisfies T(F(k/n))=F(2k/n) as the Laplacians satisfy such a renormalization recursion. The spectral density in the d=1 case is then the arc-sin distribution which is the equilibrium measure on the Julia set. In higher dimensions, where the limiting function F still remains unidentified, F' appears to have a discrete or singular component.

We construct a Cartesian product G x H for finite simple graphs. It satisfies the Kuenneth formula: H^k(G x H) is a direct sum of tensor products H^i(G) x H^j(G) with i+j=k and so p(G x H,x) = p(G,x) p(H,y) for the Poincare polynomial p(G,x) and X(G x H) = X(G) X(H) for the Euler characteristic X(G)=p(G,-1). G1=G x K1 has as vertices the simplices of G and a natural digraph structure. We show that dim(G1) is larger or equal than dim(G) and G1 is homotopic to G. The Kuenneth identity is proven using Hodge describing the harmonic forms by the product f g of harmonic forms of G and H and uses a discrete de Rham theorem given by a combinatorial chain homotopy between simplicial and de Rham cohomology. We show dim(G x H) = dim(G1) + dim(H1) implying that dim(G x H) is larger or equal than dim(G) + dim(H) as for Hausdorff dimension in the continuum. The chromatic number c(G1) is smaller or equal than c(G) and c(G x H) is bounded above by c(G)+c(H)-1. The automorphism group of G x H contains Aut(G) x Aut(H). If G~H and U~V then (G x U) ~ (H x V) if ~ means homotopic: homotopy classes can be multiplided. If G is k-dimensional geometric meaning that all unit spheres S(x) in G are (k-1)-discrete homotopy spheres, then G1 is k-dimensional geometric. If G is k-dimensional geometric and H is l-dimensional geometric, then G x H is geometric of dimension (l+k). The product extends to a ring of chains which unlike the category of graphs is closed under boundary operation taking quotients G/A with A subset Aut(G). As we can glue graphs or chains, joins or fibre bundles can be defined with the same features as in the continuum, allowing to build isomorphism classes of bundles.

d-spheres in graph theory are inductively defined as graphs for which all unit spheres S(x) are (d-1)-spheres and that the removal of one vertex renders the graph contractible. Eulerian d-spheres are geometric d-spheres which are d+1 colorable. We prove here that G is an Eulerian sphere if and only if the degrees of all the (d-2)-dimensional sub-simplices in G are even. This generalizes a Kempe-Heawood result for d=2 and is work related to the conjecture that all d-spheres have chromatic number d+1 or d+2 which is based on the geometric conjecture that every d-sphere can be embedded in an Eulerian (d+1)-sphere. For d=2, such an embedding into an Eulerian 3-sphere would lead to a geometric proof of the 4 color theorem, allowing to see "why 4 colors suffice". To achieve the goal of coloring a d-sphere G with d+2 colors, we hope to embed it into a (d+1)-sphere and refine or thin out the later using special homotopy deformations without touching the embedded sphere. Once rendered Eulerian and so (d+2)-colorable, it colors the embedded graph G. In order to define the degree of a simplex, we introduce a notion of dual graph H' of a subgraph H in a general finite simple graph G. This leads to a natural sphere bundle over the simplex graph. We look at geometric graphs which admit a unique geodesic flow: their unit spheres must be Eulerian. We define Platonic spheres graph theoretically as d-spheres for which all unit spheres S(x) are graph isomorphic Platonic (d-1)-spheres. Gauss-Bonnet allows a classification within graph theory: all spheres are Platonic for d=1, the octahedron and icosahedron are the Platonic 2-spheres, the sixteen and six-hundred cells are the Platonic 3-spheres. The cross polytop is the unique Platonic d-sphere for d>3. It is Eulerian.