INCACIONES T
| MONTESAE
eINSTETUY
JUATERAT
1973 : vol, 4 No. 60°
Serie B: InvestigaciSn
“Investigador del CIMAS.
CRNTRO DE INVESTIGACION
EN MATEMATICAS APLICADAS:
VEN SISTEMAS £),EXTREMAL PROBLEMS CONCERNING N~PATH CONNECTED GRAPHS
BY F, ESCALANTE AND L. MONTEJANO
CIMAS, National University of Mexico
ABSTRACT
For given p23 and igngp~l wa determine all graphs which
are not n-path connected and with maximum number of edges.
AS 4 consequence, some results by @. Ore and G. Churtrané,
8.¥. Kapoor and H.V. Kronk concerning hamiltonian connected
gYaphs are extended.
il. DEFINITIONS
In this paper only finite, simple-graphs are considers!
and, if no otherwise stated, notation and definitions are
the same as in [8].
Bince Ore [12] introduced the term, considerable attencica
has been paid to hamiltonian connected graphs (or, shortly,
hrconnected graphs), i.e, graphs with the property that any
two different points are connected by a spanning path (see
2], [20], [41]}. ore himself found sufficient conditions
for a graph to be h-connected. These conditions are:
THEOREM 1 (Ore). If in 4 graph G with p points we have
deg uddeg.vzptl for every pair of non~adjacent soints u
and v, then G is h-connected.
7 iramrennererecemeeecr 1 Teen THD ermeanssan eeu epsnnunaycesdsnpecancnsnd: depanaetersaainacceunoneccauresceumcsemuecmssssemmsspatmapmaanaecscdcosiiasandaaanssarsemanssaspsspmsssseyTHEOREM 2 (Ore). If a graph G has pp4 poimts and
at 2) {e-2) 43 lines, thea G is h-connected.
THEOREM 3 (Ore). Let G have pz4 points an@ let deg un}
fox every vertex u of G. If G has at lease {271} B02)
edges, then G is h-connectad unlass G is asomorphic to
the graph Ge 5 of Fig. i.
Fig. 1 Ge 5
One sees readily that a necessary condition for a
& with pr4 points to be h-connected ig that the deg
any vertex be at least 3. Indeed, if we have a verter cf
degree 2, then there is no path between its two enéve:.
Hence the graph 4H, with p24 points consisting of a complene
graph on p=l points an a vertex of degree Nie, Lgigs, weA.B.
Observe also that if NycN,, then, according to
the Proposition, there exists precisely one Ry tin Ny)
such that NysWjewywj. Remark, finally, that if NiNyy
then we and Wy are both elements of Wye Such a pair
wy hy will be called a- complementary pair.
in order to prove first
lal gfE(Ge | (3:2)
an injective mapping f: ® -> E{G-x) will be construct
o
Case 2. Tf w,cW,, then define
case =. afta
E(w Jem, (3.33
Case 2. If w,eW,, then there exists a wy ely (and
only one) such that NyaN yrs Let us partition Ny
inte three sets Ny, Ber and Nyyy in the following way:
TE yycatYnca’
re yay) ya,, then Ny WyeNyy7
(Gd) .
TE var g?Y¥nrg Nar then Ny eM yeNyrye
a
#Y then Ny NseNy,
Observe that the partition is well defined because
ng4. Moreover, in case n=4 the sets By nd Ny, are st
ana yay Grey,
Let HRW U Wop Were be the corresponding partit
of Wye
REMARK 1. If w,,w, (cW,) is a complementary psiz, then
3
wibly, ant yOlu, ave both edges of FE.1d:
proof, TE, say, vfllw, vas an edge of G-x, then it
would shorten T, in one unit contradicting the fact
yeh (Gx) oye
Take now a complementary pair Wars ew, and lex
ay Dayil) o 1
WyeYanenYaraeWa: BY Remark 1, gyiaw,w, and gy teu,
are edges of SX. Further, at least one of yy is
not contained in "A Indeed, assume Oy Oye? then
QyeNy or Ry and O78, or Nye But Wye and hence
QyeR yr QSAN contradicting #,¥¥,, Without less of ge-
nerality let 0,¢N,, Then define
ElwJ=Qs,
vt (3.4)
E(w }*N =N_«
t 3) Psy
we further remark that a Q, has one endvertex from
Wy and one from W,7 moreover, the endvertex from W,
aniguely determines the Qs: Therefore £ is injective
on WAL Woy: Let @ be the set of all ay defined above.
It remains to extend f injectively to Wee Fase
tfake a complementary pair wy rye and let
wifey 3) wun ay. Hence, by Remark 1, posylly, and
ajyiny are edges of =X . Let Pa(P,}.then we have
REMARK 2. PA Qf.
Proof, No endvertex of a P, is in W,.ib
REMARK 3. PA\N, =f.
Proof. ‘Same reason as in Remark 2.
Wow take were, NyeNjew wy. According to Remark 1,
ti)
Mysreyghiwy and Ny gray Jha, are edges of EX. voreover,
also My sey (Uw. and nye, are both elements of
B(@X). For assume, say, 4, ,¢8(G-x). Then, since we),
the path yy(?y (4), verge ie, would Be an (nt) -
path between y and Wye Finally let M be the set of che
M's defined above,
REMARK 4. (PU QU MIN Naas.
Proof. No two elements of Ng Can have a vertex in common.
On the other side all elements of P,Q and R have an end=-
vertex in Wa
REMARK 5. Let. yw, be a complementary paix in W,. Then
at most two elements of the set Mg eMysiHy gr By belong
te Nae
Proof. If one of the elements belongs to Nye then it
necessarily has an endvertex from W,. But only two end-
‘a
vertices are available which possibly belong to %
ff
namely yi) and 7S). soreover, a vertex wt
uniquely determines its N,, and then w, is endvertex
of Hy. Hence our claim follows.42
REMARK 6. MAQ=f.
proof. Let Q,2Q correspond to the complementary pair
(4) ov tliws
wy rye i.e. QyAY yr 4™ 4 . Hence wieWygge Suppose
QyeN. Since yGhew, and one of the ‘enavertices of the
elements cf M is always in W,, we conclude wyc¥,, what
is impossible.
REMARK 7. Given three elements among the M's and the
pig at least two of them are different.
proof. Since complementary pairs are disjoint and
since every element of MUP has precisely one endvertex
in the complementary paix which defines it, its main
vertex, then, if two elements coincide, they necessarily
have different main vertices. Obviously this implies
that no three elements of MUP can be equal.
Let M' denote the set M-N,. Then
iw tlh!
x Ii
is leisy l=
and by Remarks 3,5 and 7
ainyl+2iuy.| fw it
[mtu Pi ing = Z 3E
Hence we have [N. ght thy thet PighMy iti ¥ay) different
edges in @°X and therefore we can map Hj, V8,, injeccively
inte BUN MUP. Now, by definition of M? and Remarks
2,3,4 and 5, (Ny Ny NY BIA (Nt Q)=8, so we can13
extend the £ defined in (3.4) to an injective mapping
from W into E({G=X) proving (3.2).
tn order to prove the strict inequality (3.1) we have
to Find an edge of G-x which is not an image under £,
Let first n2$. Both endvertices of any edge of fi}
belong to Ser.. “Moreover, the vertices used in this
way and wien are “closest” to y are the y(t). re nos
nea”
qi) ci}
we have yUl¢y for all i and hence yy$*) ent@=H)
w)
for all 4 (the selection of yy!!! ie possible because
WS). Our lemma is thus proved for, nz5.
Finally let n=4. In this case, as ckserved before,
My and Nyyzare empty, i.e. NgeNey: Our mapping £ res-
tricts then to ,
fy: Wa? ry and
£,? 4, -> P, Now we show
REMARK 8. For n=4 all P's are distinct.
Proof. Let Py =P =yw,. Since yak, wy, ds the main vertex
il 3
of both Py and Pye But this is a contradiction since
complementary pairs are disjoint.
Now, if Wy=#, then select any yyi*). this edge of
G-x is not contained in £(w)=£(W,).
If, on the contrary, Wa#¥, then there is at least
one complementary pair Wary: Set then14
£0 )8Py, Fw 5)=P AE ww yclig and
Ew =Ny, if wey.
Remark 8 ensures that the mapping is injective. Yore~
ne
over, since W,¥%, there is at least the edge wow, 6X.
3 a 4e%B
his implies the strict imequality (3.4).$4, PROOF OF THE LEMMA FOR THE CASE n=3
As in §3 assume G-x is not complete, i.e. H¥9,
and let W be partitioned into W, and W, as before.
if WyeB, then set Fiv,dey{w, for all £. Then
there exists an.edge in E(G"k}-f£(W), namely yi) .
because az3. This proves (3.1) in case Wy78-
REMARK 9. If n=3 and wyeWye then yejeE(G).
Proof. Let y be the complementary vertex to w,.
Then N,=N, eee whew, vhewe But n*3, hence
itt =
yidleyt rawr, and analogously for w,. Therefore v,
and ¥y are adjacent to y.
REMARK 10. Let n=3, If at least one of two arbitrary
vertices Wye Bp of W belongs to Wer then wyWyeE(S“k) .
Proof. Assume w,cW,. By Remark 7, yw ,eE(G) and, by
hypothesis, YWo#E( {G-x) 2) . Therefore wy Wek {G-%) .
Let us consider now the case W.#9. Recording to
Remark 20, .
|B oeiay) layl
ve Pe
where the term +1 appears due to yy) ch(G-e). Hence
{B(G=x) eh,
[eS [> fw, i+] ey! 2 W,¥8 or if lwyle2s and
[et@=x) [=|w,i+iw,| if Wy=8 and iw, ]=2.
asThis proves the lemma in the remaining ease n=3
except if Waa and iw}=2- But then G consists of
exactly 5 vertices, namely x,¥,W,/¥ and yPay 3?)
and the graph is the exceptional graph G; 4 of Fig. 2.
,
Now the lemma is proved for all nz3.A?
65, CHARACTERIZATION OF THE SETS 2a
“wet FL, pr P23) Ograp, Pe the graph on p points
. : =
consisting of ;the complete graph Koo? together with
two adjacent points u and v with the property that
x vertices of Kine are adjacent to u ané prg-roote
v. Then all graphs F, , have (enb} ter?) 3 eaves.
Let 0, be the set of the fe 41 different graphs
L
Ber
THEOREM 8. 9. if o3.
pe2*p
Proof. Take Gee, 2" pg3. Let u,v be two vertices for
which there is no 2-path joining them. Then, because
of the maximality of G, we have uveE(S). Also Gesev
must be complete, since otherwise we gould ad& rou edges
to Gru-v without creating @ Qe-path between u ane 7.
Finally, since u and v have no common neighbour, ve
must have deggurieg,VePs But this means SF is
2
an x, ogrs 2]. Moreover, since any F, has
property that the-vertices u and v have no common
neighbour, it follows that (F,, ,),#K,. This proves
the theorem.
THEOREM 9. For 5sn=p-2, 8 a7 Fp) where 35 is the
graph described in page 6.18
Proof. It is a routine matter to check the theortr tor
ne5 and p»7. Assume then 6inep-2 and take GeO, eat
We claim that in G there exists a vertex of degree