This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

CONSTRUCTING CERTAIN FAMILIES OF 𝟑\mathbf{3}-POLYTOPAL GRAPHS

Riccardo W. Maffucci111EPFL, MA SB Batiment 8, Lausanne, Switzerland. [email protected].
Abstract

Let n3n\geq 3 and rnr_{n} be a 33-polytopal graph such that for every 3in3\leq i\leq n, rnr_{n} has at least one vertex of degree ii. We find the minimal vertex count for rnr_{n}. We then describe an algorithm to construct the graphs rnr_{n}. A dual statement may be formulated for faces of 33-polytopes. The ideas behind the algorithm generalise readily to solve related problems.

Moreover, given a 33-polytope tlt_{l} comprising a vertex of degree ii for all 3il3\leq i\leq l, ll fixed, we define an algorithm to output for n>ln>l a 33-polytope tnt_{n} comprising a vertex of degree ii, for all 3in3\leq i\leq n, and such that the initial tlt_{l} is a subgraph of tnt_{n}. The vertex count of tnt_{n} is asymptotically optimal, in the sense that it matches the aforementioned minimal vertex count up to order of magnitude, as nn gets large. In fact, we only lose a small quantity on the coefficient of the second highest term, and this quantity may be taken as small as we please, with the tradeoff of first constructing an accordingly large auxiliary graph.

Keywords: Algorithm, planar graph, degree sequence, valency, 33-polytope, polyhedron.
MSC(2010): 05C85, 52B05, 52B10, 05C07, 05C10, 05C75.

1   Introduction

1.1   The question

Graphs that are planar and 33-connected have the nice property of being 11-skeletons of 33-polytopes, as proven by Rademacher-Steinitz (see e.g. [11, Theorem 11.6]). We call these graphs 33-polytopal graphs, or 33-polytopes, or polyhedra interchangeably. These special planar graphs are uniquely embeddable in a sphere (as observed by Whitney, see e.g. [11, Theorem 11.5]). Their regions are also called ‘faces’, and are delimited by cycles (polygons) [6, Proposition 4.26].

Our starting point is the following question. Let n3n\geq 3 and GG^{\prime} be a 33-polytopal graph such that for every 3in3\leq i\leq n, GG^{\prime} has at least one ii-gonal face. What is the minimal number of faces for GG^{\prime}?

In what follows, we will work on the dual problem. Indeed, it is well-known that 33-polytopes have a unique dual graph, that is also 33-polytopal (see e.g. [11, Chapter 11]).

Definition 1.

A 33-polytope has the property 𝒫n\mathcal{P}_{n} if it has at least one vertex of degree ii, for each 3in3\leq i\leq n, and moreover it has minimal order (number of vertices) among 33-polytopes satisfying this condition.

The notation HGH\prec G indicates that HH is a subgraph of GG. Our first result is the following.

Theorem 2.

Let n3n\geq 3 and GG be a 33-polytopal graph with at least one vertex of degree ii, for every 3in3\leq i\leq n. Then the minimal number p(n)p(n) of vertices of GG is

p(n)=n211n+624,n14.p(n)=\left\lceil\frac{n^{2}-11n+62}{4}\right\rceil,\qquad\forall n\geq 14. (1.1)

For n13n\leq 13, we have the values in Table 1.

n3n78910111213p(n)n+1101114161923\begin{array}[]{|c||c|c|c|c|c|c|c|}\hline\cr n&3\leq n\leq 7&8&9&10&11&12&13\\ \hline\cr p(n)&n+1&10&11&14&16&19&23\\ \hline\cr\end{array}

Table 1: Values of p(n)p(n) for n13n\leq 13.

Moreover, starting from n=14n=14 and for every n16n\geq 16, Algorithm 8 constructs a 33-polytope rnr_{n} satisfying 𝒫n\mathcal{P}_{n} and the relations

rn{rn2 if n0(mod4),rn3 if n1(mod4),rn4 if n2(mod4),rn5 if n3(mod4)n16.r_{n}\succ\begin{cases}r_{n-2}&\text{ if }n\equiv 0\pmod{4},\\ r_{n-3}&\text{ if }n\equiv 1\pmod{4},\\ r_{n-4}&\text{ if }n\equiv 2\pmod{4},\\ r_{n-5}&\text{ if }n\equiv 3\pmod{4}\end{cases}\qquad n\geq 16. (1.2)

Graphs satisfying 𝒫n\mathcal{P}_{n} for 3n153\leq n\leq 15 are depicted in Figures 8, 9, 1, and 10.

Theorem 2 will be proven in sections 2, 3, and 4. Passing to the duals, we can answer the original question.

Theorem 2’.

If n3n\geq 3 and GG^{\prime} is a 33-polytopal graph with at least one ii-gonal face, for every 3in3\leq i\leq n, then the minimal number p(n)p(n) of faces for GG^{\prime} is given by (1.1) and Table 1.

1.2   A related problem

In our next result, given a 33-polytope HH containing vertices of valencies {3,4,,l}\{3,4,\dots,l\}, l5l\geq 5, and an integer n>ln>l, we aim to construct a 33-polytope GG containing a copy of HH as subgraph, and comprising vertices of degrees {3,4,,n}\{3,4,\dots,n\}. We start with the following definition.

Definition 3.

Let n5n\geq 5 be odd. We say that a 33-polytope satisfies property 𝒬n\mathcal{Q}_{n} if there is at least one vertex of degree 33, and moreover the polytope contains among its faces the triangles

Fn;j={vn;j;1,vn;j;2,vn;j;3},1j(n3)/2F_{n;j}=\{v_{n;j;1},v_{n;j;2},v_{n;j;3}\},\quad 1\leq j\leq(n-3)/2 (1.3)

where

{deg(vn;j;1):1j(n3)/2}{deg(vn;j;2):1j(n3)/2}={4,5,,n}\{\deg(v_{n;j;1}):1\leq j\leq(n-3)/2\}\cup\{\deg(v_{n;j;2}):1\leq j\leq(n-3)/2\}=\{4,5,\dots,n\} (1.4)

and

vn;j;3vn;i;1,vn;i;2,j,i.v_{n;j;3}\neq v_{n;i;1},v_{n;i;2},\quad\forall j,i.

Note that 𝒬n\mathcal{Q}_{n} together with minimality w.r.t. order is stronger than property 𝒫n\mathcal{P}_{n} of Definition 1.

Theorem 4.

Let l5l\geq 5 be odd, and tlt_{l} be a 33-polytope satisfying 𝒬l\mathcal{Q}_{l}. Fix an integer m14m\geq 14, m2(mod4)m\equiv 2\pmod{4}, and let n:=l+mkn:=l+mk, where kk is a non-negative integer. Then there exists a sequence of 33-polytopes

{tn}n=l+mk,k1,\{t_{n}\}_{n=l+mk,\ k\geq 1},

where each tnt_{n} satisfies 𝒬n\mathcal{Q}_{n}, such that along the sequence, for all ϵ>0\epsilon>0, one has

|V(tn)|n2411n4+(52m+ϵ)n|V(t_{n})|\leq\frac{n^{2}}{4}-\frac{11n}{4}+\left(\frac{5}{2m}+\epsilon\right)n (1.5)

as nn\to\infty. Moreover, for all k1k\geq 1, it holds that

tl+m(k1)tl+mk.t_{l+m(k-1)}\prec t_{l+mk}. (1.6)

For chosen tlt_{l} and N1N\geq 1, Algorithm 13 constructs tl+mkt_{l+mk} for 1kN1\leq k\leq N.

Theorem 4 will be proven in section 6.

Remark 5.

The order (1.5) of the sequence of polyhedra in Theorem 4 is asymptotically optimal, in the sense that the leading term is n2/4n^{2}/4 as in (1.1). The coefficient of the linear term is only slightly larger than the 11n/4-11n/4 of (1.1). This difference can be taken as small as we please if mm is chosen to be large, with the tradeoff of first constructing an accordingly large auxiliary graph sm+3s_{m+3}, as detailed in sections 5 and 6.

Remark 6.

In Definition 3, we could have supposed instead nn even. Then accordingly one would have taken j(n2)/2j\leq(n-2)/2 in (1.3), and the set {3,4,,n}\{3,4,\dots,n\} on the RHS of (1.4). This would have produced similar setup and ideas.

1.3   Related literature, notation, and plan of the paper

Related literature.

Necessary conditions for the degree sequence of a planar graph were given in [3, 5]. On the other hand, Eberhard [7] proved that any degree sequence where q=3p6q=3p-6 (p,qp,q being vertex and edge counts respectively) may be made planar by inserting a sufficiently large number of 66’s. There have been numerous generalisations and extensions since, see e.g. [1, 2, 9, 10, 13, 14]. In [12], the authors determine the sequences for regular, planar graphs. This was extended in [15] to sequences with highest and lowest valencies differing by one or two.

Notation.

All graphs that appear contain no loops and multiple edges. The vertex and edge sets of a graph GG are denoted by V(G)V(G) and E(G)E(G) respectively. The order and size of GG are the numbers |V(G)||V(G)| and |E(G)||E(G)|. The degree or valency degG(v)\deg_{G}(v) of a vertex vv counts the number of vertices adjacent to vv in GG. We use the shorthand deg(v)\deg(v) when GG is clear. The degree sequence of GG is the set of all vertex valencies.
We write GHG\cong H when G,HG,H are isomorphic graphs, and HGH\prec G when HH is (isomorphic to) a subgraph of GG.
A graph of order k+1k+1 or more is said to be kk-connected if removing any set of k1k-1 or fewer vertices produces a connected graph.
Regions of a 22-connected planar graph are cycles of length ii (ii-gons) [6, Proposition 4.26]. For these graphs, the terms ‘region’ and ‘face’ are interchangeable. The ii-gonal faces will be denoted by their sets of ii vertices. If {a,b,c}\{a,b,c\} is a triangle, we call splitting the operation of adding a vertex dd and edges da,db,dcda,db,dc.

Plan of the paper.

Theorem 2 is proven in sections 2 (lower bound (1.1)), 3 (cases n13n\leq 13) and 4 (Algorithm 8 for n14n\geq 14). Section 5 is about an application of a similar flavour, that we can tackle via a minor modification of Algorithm 8. The theory of section 5 will also be useful in section 6 to prove Theorem 4.
Appendix A presents another way to think about Algorithm 8. Appendix B contains figures of graphs mentioned in sections 3 and 4.

1.4   Acknowledgements

The author was supported by Swiss National Science Foundation project 200021_184927.

2   The lower bound

In this section we prove (1.1).

Lemma 7.

Let n3n\geq 3 and G(n)G(n) be a 33-polytopal graph with at least one vertex of degree ii, for every 3in3\leq i\leq n. Then its order is at least

p(n)n25n+306.p(n)\geq\left\lceil\frac{n^{2}-5n+30}{6}\right\rceil. (2.1)

Moreover, as soon as n8n\geq 8, we also have

p(n)n211n+624.p(n)\geq\left\lceil\frac{n^{2}-11n+62}{4}\right\rceil. (2.2)
Proof.

Let p=p(n)p=p(n), q=q(n)q=q(n) denote order and size of G(n)G(n), and dj:=deg(vj)d_{j}:=\deg(v_{j}). On one hand, via the handshaking lemma,

2q=i=3ni+j=n1pdjn(n+1)23+3(pn+2)=(n2)(n3)2+3p2q=\sum_{i=3}^{n}i+\sum_{j=n-1}^{p}d_{j}\geq\frac{n(n+1)}{2}-3+3(p-n+2)=\frac{(n-2)(n-3)}{2}+3p

where we used 33-connectivity. On the other hand, by planarity, q3p6q\leq 3p-6, so that altogether

pn25n+306p\geq\frac{n^{2}-5n+30}{6}

hence (2.1).
By [3, 5], for any 3k(p+4)/33\leq k\leq(p+4)/3, it holds that

i=1kdi2p+6k16.\sum_{i=1}^{k}d_{i}\leq 2p+6k-16. (2.3)

To optimise this lower bound for pp, the left hand side should contain as many numbers exceeding 55 as possible. We thus wish to take k=n5k=n-5, and we may do this as long as 3n5(p+4)/33\leq n-5\leq(p+4)/3, i.e.,

n8 and p3n19.n\geq 8\quad\text{ and }\quad p\geq 3n-19.

By (2.1), these conditions certainly hold for all n8n\geq 8. In this case, equation (2.3) with k=n5k=n-5 reads

n(n+1)2152p+6(n5)16,\frac{n(n+1)}{2}-15\leq 2p+6(n-5)-16,

and rearranging this inequality we obtain (2.2). ∎

3   Proof of Theorem 2 for n13n\leq 13

The polyhedra up to 88 faces were tabulated in [4] and [8]. For 4n+184\leq n+1\leq 8, we are looking for polyhedra with at least one ii-gonal face for each 3in3\leq i\leq n. We consult [8, Table I], searching where the ‘Faces’ column has the maximal n2n-2 non-zero entries. It is straightforward to find the ten relevant cases (numbered 1, 2, 3, 4, 5, 13, 14, 15, 46, and 47 in [8]). Passing to the dual graphs, we obtain the ten 33-polytopes sketched in Figure 8. In particular, for 3n73\leq n\leq 7, we have p(n)=n+1p(n)=n+1 (cf. Table 1).

Next, we wish to find examples of polyhedra satisfying 𝒫n\mathcal{P}_{n} for 8n138\leq n\leq 13. We observe that each graph in Figure 8, save for the tetrahedron and square pyramid, may be obtained from a previous one via a splitting operation. Our strategy is then to apply repeated splitting on the faces of r7.1r_{7.1} from Figure 8, to obtain a new graph rnr_{n}. For 8n138\leq n\leq 13, the aim is to obtain a subset of vertices of valencies 4,5,,n4,5,\dots,n. In the effort to minimise the resulting graph’s order, we split triangles of r7.1r_{7.1} containing the maximal possible number of vertices of degree 44 or more, ideally all three of them. We thereby construct the graphs of Figure 9. Their orders match the largest of the lower bounds (2.1) and (2.2) proven in section 2. We thus complete Table 1.

In the next section, we will combine the above with other ideas to write Algorithm 8, proving the cases n14n\geq 14 of Theorem 2.

4   Proof of Theorem 2 for n14n\geq 14

4.1   Setup

Let r14r_{14} be the graph sketched in Figure 1.

Refer to caption
Figure 1: The 33-polytope r14r_{14}, satisfying 𝒫14\mathcal{P}_{14}.

It is straightforward to check that r14r_{14} is a polyhedron, and that the respective valencies of vjv_{j}, 1j111\leq j\leq 11, are

7,9,11,13,14,12,10,8,6,4,5.7,9,11,13,14,12,10,8,6,4,5.

The order of this graph is 2626, matching the lower bound (2.2) in the case n=14n=14, and there are vertices of degree 33 as well. Theorem 2 is hence proven in this case. In the following we recursively construct the rnr_{n}, n16n\geq 16, of Theorem 2. As for r15r_{15} (Figure 10), we obtain it from r14r_{14} via one edge deletion and again applying the ideas of section 3.

We need a preliminary definition, the operation of h\mathit{h}-splitting a triangle about a vertex, for some h1h\geq 1 (see Figure 2). To hh-split {a,b,c}\{a,b,c\} about cc, we begin by splitting it, introducing a new vertex c1c_{1}. Then we split {a,b,c1}\{a,b,c_{1}\} inserting c2c_{2}, and so on, until we have added the vertex chc_{h}. For instance referring to Figure 8, given the tetrahedron s3s_{3}, 11-splitting any face yields s4.2s_{4.2}, while 22-splitting any face produces s5.2s_{5.2}.

Refer to caption
Figure 2: Notation for hh-splitting a triangle {a,b,c}\{a,b,c\} about the vertex cc.

4.2   The case n2(mod4)n\equiv 2\pmod{4}

We now describe the algorithm producing the rnr_{n} of Theorem 2, starting from the case n2(mod4)n\equiv 2\pmod{4}. The remaining cases are covered in section 4.3.

Algorithm 8 (Part I).

Input. An integer N16N\geq 16.

Output. A set of graphs {rn:16nN}\{r_{n}:16\leq n\leq N\}, where each rnr_{n} has property 𝒫n\mathcal{P}_{n}.

Description. For all k1k\geq 1, we define the graph pck\text{pc}_{k} (‘kk-th piece’) of Figure 3. The half-lines and numbers in bold represent hh-splitting: for instance, face {ak,bk,ek}\{a_{k},b_{k},e_{k}\} is split h=4kh=4k times about the vertex eke_{k}. Letting pc0:=r14\text{pc}_{0}:=r_{14}, we label u0,v0,w0u_{0},v_{0},w_{0} its vertices of degrees 10,8,610,8,6 respectively (v7,v8,v9v_{7},v_{8},v_{9} is Figure 1). The vertex (of degree 33) adjacent to these three will be denoted by x0x_{0}. Note that also in each pck\text{pc}_{k}, k1k\geq 1, there are vertices uk,vk,wku_{k},v_{k},w_{k} of degrees 10,8,610,8,6, and xkx_{k} of degree 33 adjacent to them. We define

rn:=r14k=1(n14)/4pck,n14,n2(mod4),r_{n}:=r_{14}\cup\bigcup_{k=1}^{(n-14)/4}\text{pc}_{k},\qquad n\geq 14,\ n\equiv 2\pmod{4}, (4.1)

identifying in each union operation the vertices uk1,vk1,wk1,xk1u_{k-1},v_{k-1},w_{k-1},x_{k-1} with ak,bk,ck,dka_{k},b_{k},c_{k},d_{k} respectively.

Refer to caption
Figure 3: The graph pck\text{pc}_{k}. Dashed lines are not edges of pck\text{pc}_{k}. Half-lines and numbers in bold represent hh-splitting.
Proof of Theorem 2 for n14n\geq 14, n2(mod4)n\equiv 2\pmod{4}.

We claim that rnr_{n} (4.1) has order (1.1) and satisfies property 𝒫n\mathcal{P}_{n}. We argue by induction. The base case n=14n=14 has already been checked. The union (4.1) is still a 33-polytope by construction. By the inductive hypothesis, the graph

rn4=r14k=1(n18)/4pckr_{n-4}=r_{14}\cup\bigcup_{k=1}^{(n-18)/4}\text{pc}_{k}

satisfies 𝒫n4\mathcal{P}_{n-4}, and has order

p(n4)=n219n+1324.p(n-4)=\frac{n^{2}-19n+132}{4}.

Turning to rn=rn4pc(n14)/4r_{n}=r_{n-4}\cup\text{pc}_{(n-14)/4}, we record that for pckrn\text{pc}_{k}\prec r_{n}, k1k\geq 1, the vertices ak,bk,ck,dka_{k},b_{k},c_{k},d_{k} have respective valencies

degrn(ak)=10+1+4k+1+2=4k+14,\displaystyle\deg_{r_{n}}(a_{k})=10+1+4k+1+2=4k+14,
degrn(bk)=8+1+4k+3=4k+12,\displaystyle\deg_{r_{n}}(b_{k})=8+1+4k+3=4k+12,
degrn(ck)=6+1+(4k1)+2+5=4k+11,\displaystyle\deg_{r_{n}}(c_{k})=6+1+(4k-1)+2+5=4k+11,
degrn(dk)=3+1+1+(4k1)+7=4k+13.\displaystyle\deg_{r_{n}}(d_{k})=3+1+1+(4k-1)+7=4k+13.

In particular,

degrn(a(n14)/4)=n,\displaystyle\deg_{r_{n}}(a_{(n-14)/4})=n, degrn(b(n14)/4)=n2,\displaystyle\deg_{r_{n}}(b_{(n-14)/4})=n-2,
degrn(c(n14)/4)=n3,\displaystyle\deg_{r_{n}}(c_{(n-14)/4})=n-3, degrn(d(n14)/4)=n1.\displaystyle\deg_{r_{n}}(d_{(n-14)/4})=n-1.

The degree of the remaining vertices of rn4r_{n-4} has not changed in the union. Moreover, degrn(u(n14)/4)=10\deg_{r_{n}}(u_{(n-14)/4})=10, degrn(v(n14)/4)=8\deg_{r_{n}}(v_{(n-14)/4})=8, and degrn(w(n14)/4)=6\deg_{r_{n}}(w_{(n-14)/4})=6. As for order, pc(n14)/4\text{pc}_{(n-14)/4} introduces 1212 new vertices plus those given by the hh-splittings, namely 4(n14)/4+2+4(n14)/41=2n274(n-14)/4+2+4(n-14)/4-1=2n-27. It follows that

|V(rn)|=p(n4)+12+2n27=n211n+624.|V(r_{n})|=p(n-4)+12+2n-27=\frac{n^{2}-11n+62}{4}.

Therefore, rnr_{n} does indeed have property 𝒫n\mathcal{P}_{n}. Moreover, it is clear from the construction that rn4rnr_{n-4}\prec r_{n}. ∎

4.3   The cases n1,0,3(mod4)n\equiv 1,0,3\pmod{4}

Algorithm 8 (Part II).

Continuing Algorithm 8, for n16n\geq 16, n1n\equiv 1 (resp. 0\equiv 0) (mod4)\pmod{4}, we start by constructing rn3r_{n-3} (resp. rn2r_{n-2}) as above. We then define

rn:=rn3end1;nr_{n}:=r_{n-3}\cup\text{end}_{1;n}

(resp. rn:=rn2end0;nr_{n}:=r_{n-2}\cup\text{end}_{0;n}) with end1;n\text{end}_{1;n} (resp. end0;n\text{end}_{0;n}) given by Figure 4(a) (resp. 4(b)).
For n1n\equiv 1, in the union rn:=rn3end1;nr_{n}:=r_{n-3}\cup\text{end}_{1;n}, vertices u(n17)/4,v(n17)/4,w(n17)/4,x(n17)/4u_{(n-17)/4},v_{(n-17)/4},w_{(n-17)/4},x_{(n-17)/4} are identified with a,b,c,da,b,c,d respectively. For n0n\equiv 0, in the union rn:=rn2end0;nr_{n}:=r_{n-2}\cup\text{end}_{0;n}, vertices u(n16)/4,v(n16)/4,w(n16)/4,x(n16)/4u_{(n-16)/4},v_{(n-16)/4},w_{(n-16)/4},x_{(n-16)/4} are identified with a¯,b¯,c¯,d¯\bar{a},\bar{b},\bar{c},\bar{d} respectively.
Finally, if n3(mod4)n\equiv 3\pmod{4}, we take

rn:=rn2end0;n=rn5end1;n2end0;n,r_{n}:=r_{n-2}\cup\text{end}_{0;n}=r_{n-5}\cup\text{end}_{1;{n-2}}\cup\text{end}_{0;n},

identifying d,v,w,xd,v,w,x of end1;n2\text{end}_{1;{n-2}} with a¯,b¯,c¯,d¯\bar{a},\bar{b},\bar{c},\bar{d} of end0;n\text{end}_{0;n} respectively.

Refer to caption
(a) The graph end1;n\text{end}_{1;n}.
Refer to caption
(b) The graph end0;n\text{end}_{0;n}.
Figure 4: The graphs end1;n\text{end}_{1;n} and end0;n\text{end}_{0;n}. Half-lines and numbers in bold represent hh-splitting.
Proof of Theorem 2 for n16n\geq 16, n1,0,3(mod4)n\equiv 1,0,3\pmod{4}.

If n1n\equiv 1, in rnr_{n} one has deg(a)=n\deg(a)=n, deg(b)=n1\deg(b)=n-1, and deg(c)=n2\deg(c)=n-2. The remaining valencies 4\geq 4 of rn3r_{n-3} do not change in the union rnr_{n}. Moreover, deg(d)=10\deg(d)=10, deg(v)=8\deg(v)=8, and deg(w)=6\deg(w)=6. Lastly, by Theorem 2 in the already proven case n2n\equiv 2,

|V(rn)|=p(n3)+9+(n11)/2+(n13)/2+(n15)/2=n211n+624|V(r_{n})|=p(n-3)+9+(n-11)/2+(n-13)/2+(n-15)/2=\frac{n^{2}-11n+62}{4}

as required.

Similarly, if n0n\equiv 0, in rnr_{n} it holds that deg(a¯)=n\deg(\bar{a})=n, deg(b¯)=n1\deg(\bar{b})=n-1, deg(c¯)=10\deg(\bar{c})=10, deg(d¯)=8\deg(\bar{d})=8, and deg(w¯)=6\deg(\bar{w})=6. Further,

|V(rn)|=p(n2)+8+(n14)=n211n+624.|V(r_{n})|=p(n-2)+8+(n-14)=\left\lceil\frac{n^{2}-11n+62}{4}\right\rceil.

If n3n\equiv 3, in rnr_{n} we have deg(a)=n2\deg(a)=n-2, deg(b)=n3\deg(b)=n-3, deg(c)=n4\deg(c)=n-4, deg(a¯)=n\deg(\bar{a})=n, deg(b¯)=n1\deg(\bar{b})=n-1, deg(c¯)=10\deg(\bar{c})=10, deg(d¯)=8\deg(\bar{d})=8, and deg(w¯)=6\deg(\bar{w})=6. Via the already proved case n1n\equiv 1,

|V(rn)|=p(n2)+8+(n14)=n211n+624.|V(r_{n})|=p(n-2)+8+(n-14)=\left\lceil\frac{n^{2}-11n+62}{4}\right\rceil.

The relations (1.2) are clear by construction. This concludes the proof of Theorem 2. ∎

Remark 9.

The time to implement Algorithm 8 is quadratic in nn, and this is optimal in the sense that the order of rnr_{n} is itself quadratic (1.1). Moreover, constructing rNr_{N} takes no more time than obtaining all of

r14,r18,,rN(Nmod4+2),rNr_{14},r_{18},\dots,r_{N-(N\mod 4+2)},r_{N}

(cf. (1.2)).

5   Another application

The ideas of the preceding sections have solved the problem of finding for all nn a graph satisfying 𝒫n\mathcal{P}_{n} of Definition 1. The following lemma constitutes an application of the same ideas, and it illustrates how a minor modification of Algorithm 8 allows to answer similar questions. Moreover, the result of the lemma will be needed in section 6.

Lemma 10.

Let n17n\geq 17, n1(mod4)n\equiv 1\pmod{4}, and HH be an order pp polyhedral graph with at least one vertex of degree ii, 3in3\leq i\leq n, and at least three vertices of degree n1n-1. Then its minimal order is

n27n+344.\frac{n^{2}-7n+34}{4}. (5.1)
Proof.

Let us show the lower bound first. Similarly to Lemma 7, we begin by imposing

6p12n+3(n1)+(n2)(n1)23+j=npdjn2n102+3p,6p-12\geq n+3(n-1)+\frac{(n-2)(n-1)}{2}-3+\sum_{j=n}^{p}d_{j}\geq\frac{n^{2}-n-10}{2}+3p,

leading to

pn2n+146.p\geq\frac{n^{2}-n+14}{6}.

Thereby, for n17n\geq 17, we certainly have 3n4(p+4)/33\leq n-4\leq(p+4)/3. Applying (2.3) with k=n4k=n-4 yields the inequality

n+3(n1)+(n2)(n1)2212p+6(n4)16,n+3(n-1)+\frac{(n-2)(n-1)}{2}-21\leq 2p+6(n-4)-16,

and rearranging this inequality and imposing n1(mod4)n\equiv 1\pmod{4} we obtain (5.1) as a lower bound.

To show the upper bound, we will actually construct a graph sns_{n} satisfying the assumptions of the present lemma, of order (5.1). We set

sn:=rnendnn17,n1(mod4)s_{n}:=r_{n}\cup\text{end}^{\prime}_{n}\qquad n\geq 17,\ n\equiv 1\pmod{4} (5.2)

where rnr_{n} was constructed in Algorithm 8 and endn\text{end}^{\prime}_{n} is depicted in Figure 5.

Refer to caption
Figure 5: The graph endn\text{end}^{\prime}_{n}. Dashed lines are not edges of endn\text{end}^{\prime}_{n}.

Since n1(mod4)n\equiv 1\pmod{4},

sn=rn3end1;nendn.s_{n}=r_{n-3}\cup\text{end}_{1;n}\cup\text{end}^{\prime}_{n}.

While performing the union, we identify vertices d,v,w,xd,v,w,x of end1;n\text{end}_{1;n} (Figure 4(a)) with a,b,c,da^{\prime},b^{\prime},c^{\prime},d^{\prime} of endn\text{end}^{\prime}_{n} in this order. Then sns_{n} is clearly still a polyhedron. Further, degsn(a)=10+2+n13=n1\deg_{s_{n}}(a^{\prime})=10+2+n-13=n-1, degsn(b)=8+4+n13=n1\deg_{s_{n}}(b^{\prime})=8+4+n-13=n-1, degsn(c)=6+4=10\deg_{s_{n}}(c^{\prime})=6+4=10, degsn(d)=3+5=8\deg_{s_{n}}(d^{\prime})=3+5=8, and degsn(w)=6\deg_{s_{n}}(w^{\prime})=6. Since rnr_{n} has the property 𝒫n\mathcal{P}_{n}, then sns_{n} has vertices of each degree ii, 3in3\leq i\leq n. The union with endn\text{end}^{\prime}_{n} has inserted two more vertices of valency n1n-1. Lastly,

|V(sn)|=|V(rn)|+6+(n13)=n211n+62+4n284=n27n+344.|V(s_{n})|=|V(r_{n})|+6+(n-13)=\frac{n^{2}-11n+62+4n-28}{4}=\frac{n^{2}-7n+34}{4}. (5.3)

The proof of Lemma 10 is complete. ∎

Remark 11.

We record the following property of the graph sns_{n}. It plainly has degree 33 vertices, and contains among its faces the triangles

Fn;j={vn;j;1,vn;j;2,vn;j;3},1j(n1)/2F_{n;j}=\{v_{n;j;1},v_{n;j;2},v_{n;j;3}\},\quad 1\leq j\leq(n-1)/2 (5.4)

where deg(vn;1;1)=deg(vn;1;2)=n1\deg(v_{n;1;1})=\deg(v_{n;1;2})=n-1,

{deg(vn;j;1):2j(n1)/2}{deg(vn;j;2):2j(n1)/2}={4,5,,n}\{\deg(v_{n;j;1}):2\leq j\leq(n-1)/2\}\cup\{\deg(v_{n;j;2}):2\leq j\leq(n-1)/2\}=\{4,5,\dots,n\}

and

vn;j;3vn;i;1,vn;i;2,j,i.v_{n;j;3}\neq v_{n;i;1},v_{n;i;2},\quad\forall j,i.

This property, stronger than 𝒬n\mathcal{Q}_{n} of Definition 3, shall be denoted by n\mathcal{R}_{n}. Indeed for sns_{n}, n\mathcal{R}_{n} is easily observed by construction. For instance, we may take the pairs

(7,4),(13,12),(9,5),(14,11),\displaystyle(7,4),(13,12),(9,5),(14,11),
(4k+14,4k+12),(4k+11,4k+13) for 1k(n17)/4,\displaystyle(4k+14,4k+12),(4k+11,4k+13)\quad\text{ for }1\leq k\leq(n-17)/4,
(n,n1),(n2,10),(n1,n1),(8,6)\displaystyle(n,n-1),(n-2,10),(n-1,n-1),(8,6)

where the notation (a,b)(a,b) means that if u,vu,v denote two vertices of one of the triangles Fn;jF_{n;j}, then deg(u)=a\deg(u)=a and deg(v)=b\deg(v)=b.

6   Proof of Theorem 4

6.1   Premise

We first introduce the main ideas of the proof, via the following lemma.

Lemma 12.

Let tlt_{l} be as in Theorem 4. Then we may construct a sequence of 33-polytopes

{tn}n=l+2k,k1,\{t_{n}\}_{n=l+2k,\ k\geq 1},

where tl+m(k1)tl+mkt_{l+m(k-1)}\prec t_{l+mk} for k1k\geq 1, and each tnt_{n} satisfies 𝒬n\mathcal{Q}_{n}. Moreover along the sequence, for all ϵ>0\epsilon>0, one has

|V(tn)|n24+(1+ϵ)n|V(t_{n})|\leq\frac{n^{2}}{4}+\left(-1+\epsilon\right)n (6.1)

as nn\to\infty.

Here we prove the first statement, relegating the proof of the second one (6.1) to section 6.3. The base case is just the assumption of the lemma. We take the inductive hypothesis that tl+2(k1)t_{l+2(k-1)} verifies property 𝒬l+2(k1)\mathcal{Q}_{l+2(k-1)} of Definition 3. Suppose that at step kk, 1kN1\leq k\leq N, we were to perform 22-splitting on each triangle

Fl+2(k1);j,1j(l+2(k1)3)/2F_{l+2(k-1);j},\quad 1\leq j\leq(l+2(k-1)-3)/2

(1.3) about vertex vl+2(k1);j;3v_{l+2(k-1);j;3}. That would raise by 22 the degrees of a set of vertices of valencies 4,5,,l+2(k1)4,5,\dots,l+2(k-1). However, in the resulting graph, we would not be guaranteed vertices of degrees 44 and 55. Therefore, the 22-splitting is taken only for

2j(l+2(k1)3)/22\leq j\leq(l+2(k-1)-3)/2

(i.e., all these faces save Fl+2(k1);1F_{l+2(k-1);1}). We replace Fl+2(k1);1F_{l+2(k-1);1} with the graph SS of Figure 6, identifying vl+2(k1);1;1,vl+2(k1);1;2,vl+2(k1);1;3v_{l+2(k-1);1;1},v_{l+2(k-1);1;2},v_{l+2(k-1);1;3} with a,b,ca,b,c respectively.

Refer to caption
Figure 6: The graph SS.

In this way, the valencies of vl+2(k1);1;1v_{l+2(k-1);1;1} and vl+2(k1);1;2v_{l+2(k-1);1;2} increase by 22, and they belong to the new triangle Fl+2k;(l+2k3)/2:={a,b,g}F_{l+2k;(l+2k-3)/2}:=\{a,b,g\}. Two vertices of valencies 44 and 55 are introduced, namely d,ed,e. Moreover, these two belong to a triangle Fl+2k;1:={d,e,f}F_{l+2k;1}:=\{d,e,f\}. We also set

Fl+2k;j:={vl+2(k1);j;1,vl+2(k1);j;2,vl+2(k1);j;3′′},2j(l+2k3)/21F_{l+2k;j}:=\{v_{l+2(k-1);j;1},v_{l+2(k-1);j;2},v_{l+2(k-1);j;3}^{\prime\prime}\},\qquad 2\leq j\leq(l+2k-3)/2-1

where vl+2(k1);j;3′′v_{l+2(k-1);j;3}^{\prime\prime} is the second of the two vertices introduced in the 22-splitting of Fl+2(k1);jF_{l+2(k-1);j}. We have thus constructed tl+2kt_{l+2k} satisfying 𝒬l+2k\mathcal{Q}_{l+2k} as claimed.

In section 6.3, it will be shown that the above produces a sequence of graphs tnt_{n} verifying (6.1). Our goal is to optimise this method to asymptotically improve this upper bound on |V(tn)||V(t_{n})|.

6.2   The algorithm

In section 6.1 we have used the polyhedron SS as it has the 22 triangles {a,b,g}\{a,b,g\} and {d,e,f}\{d,e,f\}, with vertices of appropriate valencies deg(d)=5\deg(d)=5 and deg(e)=deg(a)=deg(b)=4\deg(e)=\deg(a)=\deg(b)=4. A refinement of this idea is then to pick mm even and use in place of SS a polyhedron containing (m+2)/2(m+2)/2 triangles, where two vertices from each form a set of vertices of degrees

m+3,m+2,m+2,m+2,m+1,m,m1,,5,4.m+3,m+2,m+2,m+2,m+1,m,m-1,\dots,5,4.

We have seen in Remark 11 that for m14m\geq 14, sm+3s_{m+3} has the desired property m+3\mathcal{R}_{m+3}.

Algorithm 13.

Input. A 33-polytopal graph tlt_{l} satisfying 𝒬l\mathcal{Q}_{l}, an integer m14m\geq 14, m2(mod4)m\equiv 2\pmod{4}, and a positive integer NN.

Output. A set of graphs {tn=tl+mk:1kN}\{t_{n}=t_{l+mk}:1\leq k\leq N\}, each satisfying property 𝒬n\mathcal{Q}_{n}, and tl+m(k1)tl+mkt_{l+m(k-1)}\prec t_{l+mk}. These are the first NN entries of a sequence verifying (1.5).

Description. Starting from tlt_{l}, we perform steps 1kN1\leq k\leq N as follows. The graph tl+m(k1)t_{l+m(k-1)} verifies 𝒬l+m(k1)\mathcal{Q}_{l+m(k-1)}, i.e. it has (l+m(k1)3)/2(l+m(k-1)-3)/2 triangular faces

Fl+m(k1);j={vl+m(k1);j;1,vl+m(k1);j;2,vl+m(k1);j;3},1j(l+m(k1)3)/2F_{l+m(k-1);j}=\{v_{l+m(k-1);j;1},v_{l+m(k-1);j;2},v_{l+m(k-1);j;3}\},\quad 1\leq j\leq(l+m(k-1)-3)/2

where

{deg(vl+m(k1);j;1):1j(l+m(k1)3)/2}{deg(vl+m(k1);j;2):1j(l+m(k1)3)/2}={4,5,,l+m(k1)}.\{\deg(v_{l+m(k-1);j;1}):1\leq j\leq(l+m(k-1)-3)/2\}\ \\ \cup\ \{\deg(v_{l+m(k-1);j;2}):1\leq j\leq(l+m(k-1)-3)/2\}=\{4,5,\dots,l+m(k-1)\}.

At step kk, we mm-split the Fl+m(k1);jF_{l+m(k-1);j} about vl+m(k1);j;3v_{l+m(k-1);j;3}, for j=2,,(l+m(k1)3)/2j=2,\dots,(l+m(k-1)-3)/2. Next, we replace the remaining triangle Fl+m(k1);1F_{l+m(k-1);1} with a copy of sm+3s_{m+3} from Lemma 10, identifying vl+m(k1);1;1v_{l+m(k-1);1;1} and vl+m(k1);1;2v_{l+m(k-1);1;2} with two adjacent vertices of degree m+2m+2 in sm+3s_{m+3}. This is well defined, since m+317m+3\geq 17, m+31(mod4)m+3\equiv 1\pmod{4}, and sm+3s_{m+3} has property m+3\mathcal{R}_{m+3} (Remark 11). In this way, we have increased by mm the degrees of a set of vertices of valencies 4,5,,l+m(k1)4,5,\dots,l+m(k-1), and we have introduced mm new vertices of degrees {4,5,,m+3}\{4,5,\dots,m+3\} (applying Lemma 10). Moreover, these new vertices, together with vl+m(k1);1;1v_{l+m(k-1);1;1} and vl+m(k1);1;2v_{l+m(k-1);1;2}, belong pairwise to (m+2)/2(m+2)/2 triangles, by the construction of sm+3s_{m+3} (Remark 11). We have thus obtained tl+mkt_{l+mk} satisfying 𝒬l+mk\mathcal{Q}_{l+mk}. Relation (1.6) follows by construction.

6.3   Concluding the proofs of Theorem 4 and Lemma 12

It remains to show (1.5). In Algorithm 13, starting with |V(tl)||V(t_{l})| vertices, at step k1k\geq 1 we have inserted mm of them for each of (l3+m(k1))/21(l-3+m(k-1))/2-1 mm-splittings, plus |V(sm+3)|3|V(s_{m+3})|-3 for the operation on Fl+m(k1);1F_{l+m(k-1);1} (i.e. replacing this triangle with a copy of sm+3s_{m+3}). Therefore,

|V(tn)|\displaystyle|V(t_{n})| =|V(tl)|+k=1(nl)/m[ml5+m(k1)2+|V(sm+3)|3]\displaystyle=|V(t_{l})|+\sum_{k=1}^{(n-l)/m}\left[m\cdot\frac{l-5+m(k-1)}{2}+|V(s_{m+3})|-3\right]
=|V(tl)|+m22k=1(nl)/m(k1)+m(l5)2nlm+(|V(sm+3)|3)nlm\displaystyle=|V(t_{l})|+\frac{m^{2}}{2}\sum_{k=1}^{(n-l)/m}(k-1)+\frac{m(l-5)}{2}\cdot\frac{n-l}{m}+(|V(s_{m+3})|-3)\cdot\frac{n-l}{m}
n24+(4|V(sm+3)|m210m124m+ϵ)n,\displaystyle\leq\frac{n^{2}}{4}+\left(\frac{4|V(s_{m+3})|-m^{2}-10m-12}{4m}+\epsilon\right)n,

where as nn\to\infty we have bounded the constant terms via ϵn\epsilon n, for all ϵ>0\epsilon>0. Substituting the value (5.1), we have as nn\to\infty

|V(tn)|n24+(11m+104m+ϵ)n|V(t_{n})|\leq\frac{n^{2}}{4}+\left(\frac{-11m+10}{4m}+\epsilon\right)n

as required. The proof of Theorem 4 is complete.

Note that mm was chosen so that m+317m+3\geq 17 and m+31(mod4)m+3\equiv 1\pmod{4} hold, in order to minimise the quantity

4|V(sm+3)|m210m124m\frac{4|V(s_{m+3})|-m^{2}-10m-12}{4m}

(Lemma 10) so that ultimately the coefficient of nn in (1.5) is as small as this method allows.

In Lemma 12, we had fixed instead m=2m=2, and used the graph SS (Figure 6) in place of sm+3s_{m+3}. Since |V(S)|=7|V(S)|=7, we get (6.1). This concludes the proof of Lemma 12.

Future directions.

The ideas behind Algorithms 8 and 13 are readily generalisable to tackle problems of a similar flavour, as shown for instance in section 5. The constructions, or a slight modification thereof, allow to minimise the total number of vertices of a graph, where certain valencies have been fixed.

Appendix A Another way to present Algorithm 8

The following construction of rnr_{n} may be more intuitive than Algorithm 8, albeit less apt for implementation. We begin by fixing n9n\geq 9 and defining a polyhedron A(n)A(n) of order n5n-5 as follows. Given an initial triangle {v1,v2,v3}\{v_{1},v_{2},v_{3}\}, we add in order v4,v5,,vn5v_{4},v_{5},\dots,v_{n-5} together with edges

vivi1,vivi2,vivi3,i=4,5,,n5v_{i}v_{i-1},\quad v_{i}v_{i-2},\quad v_{i}v_{i-3},\qquad i=4,5,\dots,n-5

(splitting operations). The resulting A(n)A(n) for n=14,21n=14,21 are illustrated in Figure 7.

Refer to caption
(a) A(14)A(14)
Refer to caption
(b) A(21)A(21)
Figure 7: Illustration of the construction A(n)A(n).

We note that A(9)r3A(9)\cong r_{3}, A(10)r4.2A(10)\cong r_{4.2}, A(11)r5.2A(11)\cong r_{5.2}, and A(12)r6.3A(12)\cong r_{6.3} from Figure 8. For all n11n\geq 11, the degree sequence of these graphs is

3,4,5,6n11,5,4,33,4,5,6^{n-11},5,4,3

where the superscript is a shorthand indicating quantities of repeated numbers, e.g. 6n116^{n-11} means n11n-11 vertices of degree 66.

Assuming n14n\geq 14, we pass from A(n)A(n) to another polyhedron B(n)B(n) in the following way. Firstly, we apply the splitting operation to every face of A(n)A(n). This has the effect of doubling all previous vertex degrees, and introducing 2(n5)42(n-5)-4 new ones (A(n)A(n) is a triangulation – it is maximal planar) so that the sequence is now

6,8,10,12n11,10,8,6,32n14.6,8,10,12^{n-11},10,8,6,3^{2n-14}.

Secondly, we split either of the two faces containing v1,v4v_{1},v_{4}. This yields in particular a vertex of degree 44. To obtain one of degree 55, we split the two faces that are adjacent to one another and that contain v2,v5v_{2},v_{5} and v3,v5v_{3},v_{5} respectively. For instance, in Figure 1, inserting v24v_{24} raises the valency of v10v_{10} to 44, and inserting v25,v26v_{25},v_{26} raises the valency of v11v_{11} to 55. The constructed polyhedron shall we denoted by B(n)B(n). Its order is

|V(B(n))|=(n5)+(2(n5)4)+3=3n16,|V(B(n))|=(n-5)+(2(n-5)-4)+3=3n-16, (A.1)

and its sequence

7,9,11,13,14,12,12n14,10,8,6,5,4,32n13.7,9,11,13,14,12,12^{n-14},10,8,6,5,4,3^{2n-13}. (A.2)

In (A.2), we have purposefully isolated a subset of vertices of degree 1212

V:=V(B(n))V(B(n)),V^{\prime}:=V^{\prime}(B(n))\subset V(B(n)), (A.3)

with cardinality n14n-14, keeping the remaining one aside.

We have B(14)r14B(14)\cong r_{14} (Figure 1). For n16n\geq 16 our strategy is outlined as follows. The vertices in A(n)A(n) have been designated to eventually correspond to ones of degree 66 or higher in rnr_{n}. Following the ideas of section 3, B(n)B(n) has been constructed by splitting faces of A(n)A(n) that contain three of these vertices of high degree. Next, starting from B(n)B(n), we split faces containing two of them. We take four vertices from VV^{\prime} of (A.3). Via nine repeated splitting operations, we aim to raise their degrees to 15,16,17,1815,16,17,18. Similarly, the next four shall become of degrees 19,20,21,2219,20,21,22, and so forth 4k+11,4k+12,4k+13,4k+144k+11,4k+12,4k+13,4k+14. This procedure ends when there remain either 2,3,02,3,0, or 55 vertices in VV^{\prime}, depending on whether n0,1,2n\equiv 0,1,2, or 3(mod4)3\pmod{4}. For n2n\equiv 2 the algorithm stops here. In the other cases n0,1,3n\equiv 0,1,3, we look to apply further triangle splittings, to obtain vertices of degrees

n1,n,n2,n1,n, or n4,n3,n2,n1,nn-1,n,\qquad n-2,n-1,n,\qquad\text{ or }\ n-4,n-3,n-2,n-1,n

respectively. The details have already been presented in section 4.

Appendix B Illustrations for small cases of Theorem 2

Here we sketch certain 33-polytopes mentioned in sections 3 and 4.

Refer to caption
(a) r3r_{3}
Refer to caption
(b) r4.1r_{4.1}
Refer to caption
(c) r4.2r_{4.2}
Refer to caption
(d) r5.1r_{5.1}
Refer to caption
(e) r5.2r_{5.2}
Refer to caption
(f) r6.1r_{6.1}
Refer to caption
(g) r6.2r_{6.2}
Refer to caption
(h) r6.3r_{6.3}
Refer to caption
(i) r7.1r_{7.1}
Refer to caption
(j) r7.2r_{7.2}
Figure 8: The 1010 polyhedra with p(n)=n+1p(n)=n+1.
Refer to caption
(a) r8r_{8}
Refer to caption
(b) r9r_{9}
Refer to caption
(c) r10r_{10}
Refer to caption
(d) r11r_{11}
Refer to caption
(e) r12r_{12}
Refer to caption
(f) r13r_{13}
Figure 9: Examples of 33-polytopes rnr_{n} satisfying 𝒫n\mathcal{P}_{n}, for 8n138\leq n\leq 13.
Refer to caption
Figure 10: A graph r15r_{15} satisfying 𝒫15\mathcal{P}_{15}, obtained from r14r_{14} (Figure 1) by deleting the dashed edge and inserting vjv_{j}, 27j3127\leq j\leq 31, and their incident edges.

References

  • [1] David Barnette. On pp-vectors of 33-polytopes. Journal of Combinatorial Theory, 7(2):99–103, 1969.
  • [2] David Barnette, Peter Gritzmann, and Rainer Höhne. On valences of polyhedra. Journal of Combinatorial Theory, Series A, 58(2):279–300, 1991.
  • [3] Robert Bowen. On sums of valencies in planar graphs. Canadian Mathematical Bulletin, 9(1):111–114, 1966.
  • [4] Doyle Britton and Jack David Dunitz. A complete catalogue of polyhedra with eight or fewer vertices. Acta Crystallographica Section A: Crystal Physics, Diffraction, Theoretical and General Crystallography, 29(4):362–371, 1973.
  • [5] Vašek Chvátal. Planarity of graphs with given degrees of vertices. Nieuw Arch. Wisk, 3(17):47–60, 1969.
  • [6] Reinhard Diestel. Graph theory 3rd ed. Graduate texts in mathematics, 173, 2005.
  • [7] Victor Eberhard. Zur morphologie der polyeder: mit vielen figuren im text. BG Teubner, 1891.
  • [8] PJ Federico. Polyhedra with 4 to 8 faces. Geometriae Dedicata, 3(4):469–481, 1975.
  • [9] JC Fisher. An existence theorem for simple convex polyhedra. Discrete Mathematics, 7(1-2):75–97, 1974.
  • [10] Branko Grünbaum. Some analogues of Eberhard’s theorem on convex polytopes. Israel journal of mathematics, 6(4):398–411, 1968.
  • [11] Frank Harary. Graph theory. Addison-Wesley, 1991.
  • [12] AF Hawkins, AC Hill, JE Reeve, and JA Tyrrell. On certain polyhedra. The Mathematical Gazette, 50(372):140–144, 1966.
  • [13] Stanislav Jendrol. On face vectors and vertex vectors of convex polyhedra. Discrete mathematics, 118(1-3):119–144, 1993.
  • [14] Stanislav Jendrol. On face-vectors and vertex-vectors of polyhedral maps on orientable 22-manifolds. Mathematica Slovaca, 43(4):393–416, 1993.
  • [15] EF Schmeichel and SL Hakimi. On planar graphical degree sequences. SIAM Journal on Applied Mathematics, 32(3):598–609, 1977.