alphaFeedback

WorksMathematics

Counterexamples to five conjectures of Graffiti.pc on domination

AI contributions
  • Anthropic Claude Code Opus 5.5 Ultracode
View details
Ancillary data
2 code links
View

Abstract

Written on the Wall II, the list of conjectures of the program Graffiti.pc, records Conjectures 319, 352, 358, 359 and 427 as open. We show that all five are false. The counterexamples are caterpillars for Conjectures 352, 358, 359 and 427, and graphs built from complete graphs for Conjecture 319; each comes in an infinite family, and for Conjectures 358, 359 and 427 the difference between the two sides is unbounded.

Introduction

Graffiti.pc is a program written by DeLaViña in 2001 [4] after Fajtlowicz’s Graffiti [9]; it proposes inequalities between graph invariants that hold on its database of graphs. Its conjectures are collected, with their status, in DeLaViña’s list Written on the Wall II [5]. Several hundred of them concern total domination, and some have been settled [7, 6, 8]. This note disproves five conjectures that the list still marks as open:1 four on total domination and one on independent domination. The proofs are by hand. As in [2], every finite statement is also re-checked by computer and in the Lean proof assistant; the computations are described in § 7 and Appendix § 8.

All graphs are finite, simple and connected. For a vertex vv of GG, N(v)N(v) is its set of neighbours and N[v]=N(v)∪{v}N[v] = N(v) \cup\{v\}; for X⊆V(G)X \subseteq V(G), N(X)=⋃v∈XN(v)N(X) = \bigcup_{v\in X} N(v) and ⟨X⟩\langle X\rangle is the subgraph induced by XX. A set DD is dominating if every vertex outside DD has a neighbour in DD, and totally dominating if every vertex of GG has a neighbour in DD. The domination number γ(G)\gamma(G) and the total domination number γt(G)\gamma_t(G) are the least sizes of such sets [3, 10]; the independent domination number i(G)i(G) is the least size of a dominating set that is independent. A total dominating set is minimal if no proper subset is a total dominating set; since every superset of a total dominating set is one, this holds as soon as no single vertex can be removed. GG is well totally dominated if all its minimal total dominating sets have the same size [1]. A pendant vertex or leaf has degree one, and a support vertex is a neighbour of a leaf. We write d(u,v)d(u,v) for the distance, ecc⁡(v)=max⁡ud(v,u)\operatorname{ecc}(v) = \max_u d(v,u) for the eccentricity, and C(G)C(G) for the center, the set of vertices of least eccentricity. Following [5], the eccentricity of a set XX is

ecc⁡(X)=max⁡v∉Xmin⁡x∈Xd(v,x).\operatorname{ecc}(X) = \max_{v\notin X}\min_{x\in X} d(v,x).

The number of vertices at even distance from vv, including vv itself, is denoted de⁡(v)\operatorname{de}(v); the list calls it disteven⁡(v)\operatorname{disteven}(v). In a tree TT, L(T)L(T) is the set of leaves, S(T)S(T) the set of support vertices, D2(T)D_2(T) the set of vertices of degree two, and M(T)M(T) the set of vertices of maximum degree. Finally iso⁡(H)\operatorname{iso}(H) and c(H)c(H) are the numbers of isolated vertices and of components of a graph HH.

The five conjectures read as follows in [5].

Corresponding author: [email protected]; ORCID 0000-0003-0466-750X.

Conjecture 352. If TT is a tree on n>2n>2 vertices, then γt(T)≥c⟨N(D2)∪D2⟩+⌈12ecc⁡avg⁡(M)⌉\gamma_t(T)\ge c\langle N(D_2)\cup D_2\rangle+\left\lceil\frac{1}{2}\operatorname{ecc}_{\operatorname{avg}}(M)\right\rceil, where ecc⁡avg⁡(M)\operatorname{ecc}_{\operatorname{avg}}(M) is the average eccentricity of the vertices of maximum degree.

Conjecture 358. If TT is a tree on n>2n>2 vertices, then γt(T)≥12ecc⁡(C)+iso⁡⟨S⟩\gamma_t(T)\ge\frac{1}{2}\operatorname{ecc}(C)+\operatorname{iso}\langle S\rangle.

Conjecture 359. If TT is a tree on n>2n>2 vertices, then γt(T)≥12ecc⁡(C)+c⟨S∪L⟩\gamma_t(T)\ge\frac{1}{2}\operatorname{ecc}(C)+c\langle S\cup L\rangle.

Conjecture 319. If GG has n>1n>1 vertices and max⁡vde⁡(v)=γ(G)\max_v\operatorname{de}(v)=\gamma(G), then GG is well totally dominated.

Conjecture 427. If GG has n>3n>3 vertices, then i(G)≤∣E(C,V−C)∣+⌊23∣E⟨V−N(P)⟩∣⌋i(G)\le|E(C,V-C)|+\left\lfloor\frac{2}{3}|E\langle V-N(P)\rangle|\right\rfloor, where PP is the set of pendant vertices, E⟨X⟩E\langle X\rangle is the edge set of ⟨X⟩\langle X\rangle and E(C,V−C)E(C,V-C) the set of edges with exactly one end in CC.

The printed statement of Conjecture 427 introduces the set MM of maximum-degree vertices and then uses the symbols CC and PP without defining them. In the neighbouring entries of the list, CC is the center (Conjectures 430a–c) and PP the set of pendant vertices (Conjecture 425d); we read the statement this way, and we show in Theorem 6.2 that it fails also if CC stands for MM.

Theorem 1.1. Conjectures 352, 358, 359, 319 and 427 are false. More precisely:

  1. for a≡3a\equiv3 and r≡0(mod4)r\equiv0\pmod{4} with r≥4r\ge4, the caterpillar T(a,r)T(a,r) on a+r+11a+r+11 vertices has γt=a+r+72\gamma_t=\frac{a+r+7}{2}, one less than the right side of Conjecture 352 (Theorem 3.2);

  2. for odd k≥3k\ge3, the caterpillar QkQ_k on 7k−27k-2 vertices has γt=3k\gamma_t=3k, while the right sides of Conjectures 358 and 359 both equal 3k+k−143k+\frac{k-1}{4} (Theorem 4.2);

  3. for k≥3k\ge3, the graph GkG_k on 3k+13k+1 vertices satisfies de⁡(v)=γ(Gk)=k+1\operatorname{de}(v)=\gamma(G_k)=k+1 for every vertex vv, and has minimal total dominating sets of sizes k+1k+1 and 2k2k (Theorem 5.2);

  4. for m≥2m\ge2, the caterpillar HmH_m on 4m+14m+1 vertices has i=2mi=2m, while the right side of Conjecture 427 equals 22 (Theorem 6.2).

The smallest counterexamples are T(3,4)T(3,4) with 18 vertices, Q3Q_3 with 19, G3G_3 with 10, and an 8-vertex tree for Conjecture 427. An exhaustive search (§ 7) shows that T(3,4)T(3,4) and Q3Q_3 are the only counterexamples of their orders and that there are none smaller. Likewise G3G_3 is the only counterexample to Conjecture 319 with at most 11 vertices, and the 8-vertex tree the only counterexample to Conjecture 427 with at most 8 vertices.

Two tools

Lemma 2.1. Every total dominating set of a tree contains all its support vertices, and every dominating set meets {s}∪(N(s)∩L)\{s\}\cup(N(s)\cap L) for each support vertex ss.

Proof. A leaf ℓ\ell has the single neighbour ss, so a total dominating set must contain ss, and a dominating set must contain ℓ\ell or ss. □\square

Lemma 2.2. Let DD be a total dominating set of a graph, RR a set of vertices, and QQ a set of vertices whose neighbours all lie in RR. If every vertex of RR has at most two neighbours in QQ, then ∣D∩R∣≥∣Q∣/2|D\cap R|\ge|Q|/2.

Proof. Each q∈Qq\in Q has a neighbour in DD, which lies in D∩RD\cap R. Counting the pairs (q,x)(q,x) with q∈Qq\in Q, x∈D∩Rx\in D\cap R and qxqx an edge gives ∣Q∣≤∑x∈D∩R∣N(x)∩Q∣≤2∣D∩R∣|Q|\le\sum_{x\in D\cap R}|N(x)\cap Q|\le2|D\cap R|. □\square

We also use Jordan’s theorem [11]: the center of a tree consists of the middle vertex, or the two middle vertices, of any longest path.

Conjecture 352

Definition 3.1. For a,r≥1a,r \ge1 let T(a,r)T(a,r) be the tree formed by a path

h, x1,…,xa, s1, s2, s3, y1,…,yr, zh,\ x_1,\ldots,x_a,\ s_1,\ s_2,\ s_3,\ y_1,\ldots,y_r,\ z

together with three leaves attached to hh and one leaf ℓi\ell_i attached to each sis_i (Figure 1). It has a+r+11a+r+11 vertices.

The tree $T(3,4)$ with shaded vertices forming a total dominating set

Figure 1. T(3,4)T(3,4), the smallest counterexample to Conjecture 352. The shaded vertices form a total dominating set of size 7; the right side of the conjecture is 8.

Theorem 3.2. Let a≡3(mod4)a \equiv3 \pmod{4} and r≡0(mod4)r \equiv0 \pmod{4}, r≥4r \ge4. Then

γt(T(a,r))=a+r+72<a+r+92=c⟨N(D2)∪D2⟩+⌈12ecc⁡avg(M)⌉.\gamma_t(T(a,r))=\frac{a+r+7}{2}<\frac{a+r+9}{2}=c\langle N(D_2)\cup D_2\rangle+\left\lceil\frac{1}{2}\operatorname{ecc}_{\mathrm{avg}}(M)\right\rceil.

Proof. Write T=T(a,r)T=T(a,r). The vertex hh has degree 4, the vertices s1,s2,s3s_1,s_2,s_3 have degree 3, the xix_i and yjy_j have degree 2, and all other vertices are leaves. Hence M={h}M=\{h\}, and ecc⁡(h)=d(h,z)=a+r+4\operatorname{ecc}(h)=d(h,z)=a+r+4, since every other vertex is closer to hh. As a+ra+r is odd, ⌈12ecc⁡(h)⌉=a+r+52\left\lceil\frac{1}{2}\operatorname{ecc}(h)\right\rceil=\frac{a+r+5}{2}.

Next, D2={x1,…,xa,y1,…,yr}D_2=\{x_1,\ldots,x_a,y_1,\ldots,y_r\} and

N(D2)∪D2={h,x1,…,xa,s1}∪{s3,y1,…,yr,z}.N(D_2)\cup D_2=\{h,x_1,\ldots,x_a,s_1\}\cup\{s_3,y_1,\ldots,y_r,z\}.

The two sets on the right induce paths, and no edge joins them because s2s_2 lies in neither and s1s3s_1s_3 is not an edge. So c⟨N(D2)∪D2⟩=2c\langle N(D_2)\cup D_2\rangle=2, and the right side of Conjecture 352 equals a+r+92\frac{a+r+9}{2}.

Upper bound. The set

D={h,x1,s1,s2,s3}∪{x4j,x4j+1:1≤j≤a−34}∪{y4j−1,y4j:1≤j≤r4}D=\{h,x_1,s_1,s_2,s_3\}\cup\{x_{4j},x_{4j+1}:1\le j\le\tfrac{a-3}{4}\}\cup\{y_{4j-1},y_{4j}:1\le j\le\tfrac{r}{4}\}

has 5+a−32+r2=a+r+725+\frac{a-3}{2}+\frac{r}{2}=\frac{a+r+7}{2} elements. It is a total dominating set: hh dominates its leaves and x1x_1; x1x_1 dominates hh and x2x_2; for j≥1j\ge1 the pair x4j,x4j+1x_{4j},x_{4j+1} dominates x4j−1,…,x4j+2x_{4j-1},\ldots,x_{4j+2}, so that x1,…,xa−1x_1,\ldots,x_{a-1} are all dominated; s1s_1 dominates xax_a, ℓ1\ell_1 and s2s_2; s2s_2 dominates s1s_1, s3s_3 and ℓ2\ell_2; s3s_3 dominates s2s_2, ℓ3\ell_3 and y1y_1; and the pair y4j−1,y4jy_{4j-1},y_{4j} dominates y4j−2,…,y4j+1y_{4j-2},\ldots,y_{4j+1}, where yr+1y_{r+1} means zz.

Lower bound. Let DD be any total dominating set. By Lemma 2.1, h,s1,s2,s3∈Dh,s_1,s_2,s_3\in D. Apply Lemma 2.2 to R1=N(h)∪{x2,…,xa}R_1=N(h)\cup\{x_2,\ldots,x_a\} and Q1={h,x2,…,xa−1}Q_1=\{h,x_2,\ldots,x_{a-1}\}: the neighbours of hh and of each xix_i with 2≤i≤a−12\le i\le a-1 lie in R1R_1, and a vertex of R1R_1 is adjacent to at most two vertices of Q1Q_1. Hence ∣D∩R1∣≥a−12|D\cap R_1|\ge\frac{a-1}{2}. Likewise, with R2={y1,…,yr,z}R_2=\{y_1,\ldots,y_r,z\} and Q2={y2,…,yr,z}Q_2=\{y_2,\ldots,y_r,z\}, ∣D∩R2∣≥r2|D\cap R_2|\ge\frac{r}{2}. The sets {h,s1,s2,s3}\{h,s_1,s_2,s_3\}, R1R_1 and R2R_2 are disjoint, so ∣D∣≥4+a−12+r2=a+r+72|D|\ge4+\frac{a-1}{2}+\frac{r}{2}=\frac{a+r+7}{2}. □\square

Conjectures 358 and 359

Definition 4.1. For k≥1k\ge1 let QkQ_k be the caterpillar whose spine is the path

s1,c1,s1′,u1,w1,s2,c2,s2′,u2,w2,…,uk−1,wk−1,sk,ck,sk′s_1,c_1,s'_1,u_1,w_1,s_2,c_2,s'_2,u_2,w_2,\ldots,u_{k-1},w_{k-1},s_k,c_k,s'_k

with one leaf attached to each sjs_j and to each sj′s'_j (Figure 2). It has 7k−27k-2 vertices. We call {sj,cj,sj′}\{s_j,c_j,s'_j\} the jj-th block.

Q3, the smallest counterexample to Conjectures 358 and 359

Figure 2. Q3Q_3, the smallest counterexample to Conjectures 358 and 359. The center c2c_2 is circled; the shaded vertices form a total dominating set of size 9, while both conjectures demand 192\frac{19}{2}.

Theorem 4.2. Let k≥3k \ge3 be odd. Then γt(Qk)=3k\gamma_t(Q_k)=3k, ecc⁡(C)=5k−12\operatorname{ecc}(C)=\frac{5k-1}{2} and iso⁡⟨S⟩=c⁡⟨S∪L⟩=2k\operatorname{iso}\langle S\rangle=\operatorname{c}\langle S\cup L\rangle=2k. Hence the right side of both Conjecture 358 and Conjecture 359 equals 3k+k−14>γt(Qk)3k+\frac{k-1}{4}>\gamma_t(Q_k).

Proof. The leaves are the 2k2k attached vertices, so S={sj,sj′:1≤j≤k}S=\{s_j,s'_j: 1\le j\le k\}. Two support vertices are at distance at least 2 (they are separated by cjc_j or by uj,wju_j,w_j), so ⟨S⟩\langle S\rangle has no edges and iso⁡⟨S⟩=2k\operatorname{iso}\langle S\rangle=2k; in ⟨S∪L⟩\langle S\cup L\rangle each support vertex is joined only to its own leaf, so c⁡⟨S∪L⟩=2k\operatorname{c}\langle S\cup L\rangle=2k.

The spine has 2k+3(k−1)=5k−32k+3(k-1)=5k-3 edges, and every vertex lies on the spine or is a leaf attached to it; the spine ends s1s_1 and sk′s'_k carry leaves. So the path from the leaf at s1s_1 to the leaf at sk′s'_k, of length 5k−15k-1, is a longest path. Since kk is odd, 5k−15k-1 is even, and by Jordan’s theorem CC is the single vertex at distance 5k−12\frac{5k-1}{2} from both ends; this is cmc_m with m=k+12m=\frac{k+1}{2}, because the leaf at s1s_1 is at distance 1+5(m−1)+1=5k−121+5(m-1)+1=\frac{5k-1}{2} from cmc_m. Thus ecc⁡(C)=ecc⁡(cm)=5k−12\operatorname{ecc}(C)=\operatorname{ecc}(c_m)=\frac{5k-1}{2}.

The set BB of all 3k3k block vertices is a total dominating set: each leaf is dominated by its support vertex, sjs_j and sj′s'_j by cjc_j, cjc_j by sjs_j, uju_j by sj′s'_j, and wjw_j by sj+1s_{j+1}. Conversely, let DD be a total dominating set. By Lemma 2.1, S⊆DS\subseteq D. For each jj the vertices sj,sj′s_j,s'_j need neighbours in DD, and these lie in Xj=N(sj)∪N(sj′)X_j=N(s_j)\cup N(s'_j), which contains no support vertex. The sets X1,…,XkX_1,\ldots,X_k are pairwise disjoint, because each vertex outside SS is adjacent to support vertices of only one block: cjc_j to sj,sj′s_j,s'_j; uju_j to sj′s'_j; wjw_j to sj+1s_{j+1}; a leaf to its support vertex. Hence ∣D∣≥2k+k=3k|D|\ge2k+k=3k, and γt(Qk)=3k\gamma_t(Q_k)=3k.

Finally 12ecc⁡(C)+2k=5k−14+2k=3k+k−14\frac{1}{2}\operatorname{ecc}(C)+2k=\frac{5k-1}{4}+2k=3k+\frac{k-1}{4}.

Conjecture 319

Remark 4.3. For even k≥4k\ge4 the center of QkQ_k is the edge uk/2wk/2u_{k/2}w_{k/2}, ecc⁡(C)=5k−22\operatorname{ecc}(C)=\frac{5k-2}{2}, and the same argument gives right sides 3k+k−24>3k=γt(Qk)3k+\frac{k-2}{4}>3k=\gamma_t(Q_k). For k=2k=2 the two sides are equal.

Definition 5.1. For k≥3k\ge3 let GkG_k consist of a complete graph on c1,…,ckc_1,\ldots,c_k, a vertex xx, and for each ii a path x,pi,qi,cix,p_i,q_i,c_i (Figure 3). It has 3k+13k+1 vertices.

G₃, the smallest counterexample to Conjecture 319

Figure 3. G3G_3, the smallest counterexample to Conjecture 319. Every vertex has four vertices at even distance (itself included) and γ(G3)=4\gamma(G_3)=4. The shaded sets are minimal total dominating sets of sizes 4 (a) and 6 (b).

Theorem 5.2. For k≥3k\ge3, every vertex vv of GkG_k has de⁡(v)=k+1\operatorname{de}(v)=k+1, and γ(Gk)=γt(Gk)=k+1\gamma(G_k)=\gamma_t(G_k)=k+1. The sets {x,p1,c2,…,ck}\{x,p_1,c_2,\ldots,c_k\} and {p1,…,pk,q1,…,qk}\{p_1,\ldots,p_k,q_1,\ldots,q_k\} are minimal total dominating sets of sizes k+1k+1 and 2k2k. Hence GkG_k satisfies the hypothesis of Conjecture 319 and is not well totally dominated.

Proof. Distances. Let i≠ji\ne j. The following lists, read off from the paths through xx and through the clique, give the vertices at distance 1, 2 and 3; no distance exceeds 3.

vertexdistance 1distance 2distance 3de
xxpip_iqiq_icic_i1+k1+k
pip_ixx, qiq_ipjp_j, cic_iqjq_j, cjc_j1+(k−1)+11+(k-1)+1
qiq_ipip_i, cic_ixx, cjc_jpjp_j, qjq_j1+1+(k−1)1+1+(k-1)
cic_icjc_j, qiq_ipip_i, qjq_jxx, pjp_j1+1+(k−1)1+1+(k-1)

Table 1.

For instance d(pi,cj)=3d(p_i,c_j)=3 along pi,qi,ci,cjp_i,q_i,c_i,c_j, while the route through xx has length 4; and d(qi,qj)=3d(q_i,q_j)=3 along qi,ci,cj,qjq_i,c_i,c_j,q_j. In every row de⁡=k+1\operatorname{de}=k+1.

Domination. The kk sets N[qi]={pi,qi,ci}N[q_i]=\{p_i,q_i,c_i\} are disjoint, so a dominating set DD has at least kk vertices. Suppose ∣D∣=k|D|=k. Then DD has exactly one vertex in each N[qi]N[q_i] and does not contain xx, so xx is dominated by some pi∈Dp_i\in D. Then ci∉Dc_i\notin D and qi∉Dq_i\notin D, so cic_i is dominated by some cj∈Dc_j\in D, j≠ij\ne i. But then pj,qj∉Dp_j,q_j\notin D and x∉Dx\notin D, so pjp_j is not dominated, a contradiction. Hence γ(Gk)≥k+1\gamma(G_k)\ge k+1.

The set A={x,p1,c2,…,ck}A=\{x,p_1,c_2,\ldots,c_k\} is a total dominating set: xx is dominated by p1p_1; every pip_i by xx; q1q_1 by p1p_1; qiq_i by cic_i for i≥2i\ge2; c1c_1 by c2c_2; and cic_i, i≥2i\ge2, by another cjc_j with j≥2j\ge2, which exists since k≥3k\ge3. As γ≤γt≤∣A∣=k+1\gamma\le\gamma_t\le|A|=k+1, we get γ(Gk)=γt(Gk)=k+1\gamma(G_k)=\gamma_t(G_k)=k+1, and AA is minimal because it has the least possible size.

The set B={p1,…,pk,q1,…,qk}B=\{p_1,\ldots,p_k,q_1,\ldots,q_k\} is a total dominating set: xx is dominated by p1p_1, pip_i and qiq_i by each other, and cic_i by qiq_i. It is minimal: without pip_i the vertex qiq_i has no neighbour in BB, since N(qi)={pi,ci}N(q_i)=\{p_i,c_i\}; without qiq_i the vertex cic_i has none, since N(ci)={qi}∪{cj:j≠i}N(c_i)=\{q_i\}\cup\{c_j:j\ne i\}. As 2k>k+12k>k+1, GkG_k is not well totally dominated, while max⁡vde⁡(v)=k+1=γ(Gk)\max_v\operatorname{de}(v)=k+1=\gamma(G_k). □

Remark 5.3. If vv itself is not counted in dist⁡even(v)\operatorname{dist}_{\mathrm{even}}(v), Conjecture 319 already fails for the path v1v2v3v4v5v_1v_2v_3v_4v_5: the largest count is 2=γ(P5)2=\gamma(P_5), while {v2,v3,v4}\{v_2,v_3,v_4\} and {v1,v2,v4,v5}\{v_1,v_2,v_4,v_5\} are minimal total dominating sets. So the conjecture is false under either reading; the definition in [5] counts all vertices at even distance, which includes vv.

Conjecture 427

Definition 6.1. For m≥2m \ge2 let HmH_m be the path v−m,…,v0,…,vmv_{-m},\ldots,v_0,\ldots,v_m with one leaf ℓj\ell_j attached to each vjv_j, j≠0j \ne0; it has 4m+14m+1 vertices. For m≥3m \ge3 let FmF_m be the path v1,…,vmv_1,\ldots,v_m with one leaf attached to each vjv_j and a second leaf attached to v2v_2; it has 2m+12m+1 vertices.

Theorem 6.2. For m≥2m \ge2, the right side of Conjecture 427 for HmH_m equals 2, while i(Hm)=2mi(H_m)=2m. If CC is replaced by the set MM of maximum-degree vertices, the right side for FmF_m equals 4, while i(Fm)=mi(F_m)=m; so for m≥5m \ge5 the conjecture fails in this reading too.

Proof. In HmH_m the path ℓ−m,v−m,…,vm,ℓm\ell_{-m},v_{-m},\ldots,v_m,\ell_m is a longest path, so C={v0}C=\{v_0\} by Jordan’s theorem. As v0v_0 has degree 2, ∣E(C,V−C)∣=2|E(C,V-C)|=2. The pendant vertices are the ℓj\ell_j, so N(P)={vj:j≠0}N(P)=\{v_j:j\ne0\} and V−N(P)={v0}∪{ℓj}V-N(P)=\{v_0\}\cup\{\ell_j\}, an independent set; the second term vanishes and the right side is 2. By Lemma 2.1 an independent dominating set meets each of the 2m2m disjoint sets {vj,ℓj}\{v_j,\ell_j\}, j≠0j\ne0, so i(Hm)≥2mi(H_m)\ge2m; and {vj:j odd}∪{ℓj:j≠0 even}\{v_j:j\text{ odd}\}\cup\{\ell_j:j\ne0\text{ even}\} is an independent dominating set of size 2m2m (it dominates v0v_0 through v±1v_{\pm1}).

In FmF_m the vertex v2v_2 has degree 4 and every other vertex has degree at most 3, so M={v2}M=\{v_2\} and ∣E(M,V−M)∣=4|E(M,V-M)|=4. Every spine vertex is a support vertex, so V−N(P)V-N(P) is the set of leaves, which is independent; the right side is 4. By Lemma 2.1, i(Fm)≥mi(F_m)\ge m, and the vertices vjv_j with jj even together with the leaves at the vjv_j with jj odd form an independent dominating set of size $m. □

The smallest counterexample has 8 vertices: the path a1a2⋯a7a_1a_2\cdots a_7 with a leaf bb attached to a5a_5 (Figure 4(a)). The path a1⋯a7a_1\cdots a_7 is a longest path, so the center is {a4}\{a_4\}, which has degree 2. The pendant vertices are a1a_1, a7a_7 and bb, so N(P)={a2,a5,a6}N(P)=\{a_2,a_5,a_6\}, and ⟨V−N(P)⟩\langle V-N(P)\rangle has the single edge a3a4a_3a_4. The right side is 2+⌊23⌋=22+\lfloor\frac{2}{3}\rfloor=2. On the other hand N[a5]N[a_5] has four vertices and every other closed neighbourhood at most three, so two vertices dominate at most seven of the eight vertices; and {a2,a5,a7}\{a_2,a_5,a_7\} is an independent dominating set. So i=3i=3.

Smallest counterexample and graph $F_5$

Figure 4. (a) The smallest counterexample to Conjecture 427: the right side is 2 (the center is circled) and i=3i=3, attained by the shaded set. (b) F5F_5: with CC read as the set of maximum-degree vertices the right side is 4, while i=5i=5.

Exhaustive search

The trees with at most 24 vertices (63,242,254 trees with 3 to 24 vertices) were generated with gentreeg from nauty [12], which implements the algorithm of Wright, Richmond, Odlyzko and McKay [13], and tested against Conjectures 352, 358 and 359; γt\gamma_t was computed by a dynamic programme over a rooted tree, and compared with a brute-force computation for all trees with at most 16 vertices. Table 1 gives the numbers of counterexamples. The connected graphs were generated with geng. Among the 1,018,690,328 connected graphs with 2 to 11 vertices, 46 satisfy the hypothesis of Conjecture 319, and G3G_3 is the only one of them that is not well totally dominated. Among the 11,989,760 connected graphs with 4 to 10 vertices, Conjecture 427 has 1, 12 and 72 counterexamples with 8, 9 and 10 vertices and none smaller; with CC read as MM it has 5, all with 10 vertices.

nntreesConjecture 352Conjectures 358, 359
3–1781,13500
18123,86710
19317,95541
20823,065133
212,144,5053114
225,623,7567338
2314,828,074157108
2439,299,897318255

Table 1 (PDF p. 7). Counterexamples among all trees with nn vertices.

Verification

The proofs above use no computation. The computer checks the finite claims about the smallest members of each family and the searches of § 7, independently in four languages; all files are in the repository listed under Data and code availability.

  1. C. Exhaustive searches: search352.c and search358.c over trees in graph6 format (total domination number by dynamic programming, optionally compared with brute force over all vertex subsets), c319.c and c427.c over connected graphs (all parameters by exhaustion over vertex subsets, in integer arithmetic). The search for Conjecture 319 on the 1,006,700,565 connected graphs with 11 vertices was split into three parts with geng’s res/mod option.

  1. Python (exact integer and rational arithmetic): for T(a,r)T(a,r) with a∈{3,7,11}a \in\{3,7,11\}, r∈{4,8,12,16}r \in\{4,8,12,16\}, for QkQ_k with odd k≤11k \le11, for GkG_k with k≤5k \le5, for HmH_m with m≤5m \le5 and for FmF_m with 5≤m≤85 \le m \le8, recomputes every quantity in Theorems 3.2–6.2 from the edge lists, and checks the stated values of γt\gamma_t, γ\gamma and ii by exhaustion for the smaller members.

  1. Julia: the same quantities, with γt\gamma_t, γ\gamma, ii and the minimal total dominating sets found by brute force over all vertex subsets (up to 2262^{26} subsets), except that γt(Qk)\gamma_t(Q_k) for k≥5k \ge5 is computed by a tree dynamic programme.

  1. Lean 4: self-contained files without Mathlib. For T(3,4)T(3,4), Q3Q_3, G3G_3, the 8-vertex tree and F5F_5 they compute the right sides of the conjectures from the edge lists and prove that the conjectures fail. The refutations of Conjectures 352, 358 and 359 exhibit a total dominating set that is too small and depend on no axioms; those of Conjectures 319 and 427 evaluate γ\gamma and ii over all vertex subsets in the kernel and depend only on propext. The exact values γt(T(3,4))=7\gamma_t(T(3,4)) = 7 and γt(Q3)=9\gamma_t(Q_3) = 9 are also checked, by compiled evaluation (native_decide); the refutations do not use them.

Statements and Declarations

Funding

Funding. No funding was received for conducting this study.

Competing interests

Competing interests. The author has no competing interests to declare that are relevant to the content of this article.

Data and code availability

The programs, their outputs and the Lean files are available at https://github.com/creelie/eCY and archived on Zenodo, https://doi.org/10.5281/zenodo.23252219.

Use of generative AI

The author used Claude (Anthropic) for programming, computation and drafting; the author checked the mathematics and takes full responsibility for the content.

Formerly, Electro-Gravitational Space Propulsion Laboratory (EGSPL), Bhubaneswar, Odisha 751030, India

Email address: [email protected]

Email address: [email protected]

URL: https://orcid.org/0000-0003-0466-750X

References

  1. [1]S. Bahadır, T. Ekim and D. Gözüpek, Well-totally-dominated graphs, Ars Math. Contemp. 20 (2021), 209–222. https://doi.org/10.26493/1855-3974.2465.571
  2. [2]D. Bhattacharjee, P. Mandal and U. Bhattacharya, Resolving Erdős–Ulam monochromatic union-closed family conjectures, preprint, 2026. arXiv:2610.02833
  3. [3]E. J. Cockayne, R. M. Dawes and S. T. Hedetniemi, Total domination in graphs, Networks 10 (1980), 211–219. https://doi.org/10.1002/net.3230100304
  4. [4]E. DeLaViña, Graffiti.pc: a variant of Graffiti, in: Graphs and Discovery, DIMACS Ser. Discrete Math. Theoret. Comput. Sci. 69, Amer. Math. Soc., Providence, RI, 2005, 71–79. https://doi.org/10.1090/dimacs/069/05
  5. [5]E. DeLaViña, Written on the Wall II (conjectures of Graffiti.pc), http://cms.uhd.edu/faculty/delavinae/research/wowII/; version of 26 July 2026 archived at https://web.archive.org/web/20260726061534/http://cms.uhd.edu/faculty/delavinae/research/wowII/all.html
  6. [6]E. DeLaViña, C. E. Larson, R. Pepper and B. Waller, Graffiti.pc on the total domination number of a tree, Congr. Numer. 195 (2009), 5–18.
  7. [7]E. DeLaViña, Q. Liu, R. Pepper, B. Waller and D. B. West, Some conjectures of Graffiti.pc on total domination, Congr. Numer. 185 (2007), 81–95.
  8. [8]W. J. Desormeaux and M. A. Henning, Lower bounds on the total domination number of a graph, J. Comb. Optim. 31 (2016), 52–66. https://doi.org/10.1007/s10878-014-9708-2
  9. [9]S. Fajtlowicz, On conjectures of Graffiti, Discrete Math. 72 (1988), 113–118. https://doi.org/10.1016/0012-365X(88)90199-9
  10. [10]M. A. Henning and A. Yeo, Total Domination in Graphs, Springer Monographs in Mathematics, Springer, New York, 2013. https://doi.org/10.1007/978-1-4614-6525-6
  11. [11]C. Jordan, Sur les assemblages de lignes, J. Reine Angew. Math. 70 (1869), 185–190. https://doi.org/10.1515/crll.1869.70.185
  12. [12]B. D. McKay and A. Piperno, Practical graph isomorphism, II, J. Symbolic Comput. 60 (2014), 94–112. https://doi.org/10.1016/j.jsc.2013.09.003
  13. [13]R. A. Wright, B. Richmond, A. Odlyzko and B. D. McKay, Constant time generation of free trees, SIAM J. Comput. 15 (1986), 540–548. https://doi.org/10.1137/0215039

Paper details

vlnc:W_ypw5aj54
v1 · latestHistory (1)

published · Fresh — added to the Hub within the last 30 days

added ·

Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

CC BY 4.0

Original source

  • Mathematics

Anthropic Claude Code Opus 5.5 Ultracode · View statement →

Looking for related papers…
Contents