Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Bouweamp: Mathematics

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

MATHEMATICS

A CORRESPONDENCE

BETWEEN

TWO

SETS OF TREES

BY

DAVID
(Communicated

by

Prof.

C.

J.

A.

KLARNER

BOUWEAMP

at

the

meeting

of

March

29,

1969)

Synopsis
A planted, edge chromatic tree on k colors (V, E, v; f) is a tree with vertex
set V, edge set E, a root v E V with degree 1, and a mapping f sending E
into a I%-set of colors so that adjacent edges are mapped to different colors.
RIORDAN [5] found formulas for t(n, k), the number of classes of isomorphic,
planted, edge chromatic trees on le+ 1 colors having n + 1 vertices, and
every tree bearing the same color on its stem; however, Riordan overlooked the elegant formula
pointed

out by CARLITZ

t(n, E)=
[l].

It

was shown

in [4] that

p(n, k) =

(kn-n+
l),
where p(n, Ic) denotes ;the number of classes of isomorphic,
(h+ 1)-valent, planted plane trees having kn+ 2 vertices. Since t(n, k) =
=p(n, Ic), the question of constructing
a natural
one to one correspondence between the classes of isomorphic trees involved arises. Such
a correspondence is provided in this note ; also, precise definitions of the
trees under discussion are given in terms of binary relations on finite sets
which we hope will replace the topological definitions used formerly.
Definitions

and notation

A tree is a connected linear graph (V, E) having no cycles; V is the


vertex set, E is the edge set, and E is a set of 2-sets of V. (If (V, E) is a tree,
then 1VI = IEl+ 1.) A rooted tree (V, E, v) is a tree (V, E) with a distinguished vertex v E V called the root; a rooted tree is a planted tree if the degree
of the root is 1. Let K denote a k-set, then an edge chromatic tree (V, E; f)
on K (alternatively,
an edge chromatic tree on k cobrs) is a tree (V, E)
and a mapping f sending E into K such that adjacent edges are mapped
to distinct elements of K. Let T( V, K) denote the set of all planted, edge
chromatic trees on K having vertex set V, and every tree in T( V, K)
has the same color c E K assigned to the rooted edge. Two elements
(V, El, VI; /I), (V, Ea, ~12;f2) E T( V, K) are isomorphic if there exists a
permutation
pl of V such that (i) 212=q~(vr), (ii) Ez={(q~x, qy}: {x, y} E El),
and (iii) fl{x, y) = f&x, w}. I somorphism is an equivalence relation on
T( V, K) ; let T*( V, K) denote the set of equivalence classes defined in

293
T( V, K) by isomorphism. Suppose 1VI =n+ 1, jlij = k+ 1, and let t(n, k) : =
= [T*( I/, K)j. The elements of T*( V, K) are equivalent to what RIORDAN
[5] called chromatic planted trees with k + 1 line colors, n lines, and a given
color on the stem. CARLITZ [l] used Riordans equation involving
the
generating function
of {t(l, k), t(2, k), . ..} to deduce that

(1)

t(n, k) =

kn
?z /(kn--n+
( >

1).

Suppose (8, E, V) is a rooted tree and let e(x) denote the length of the
path from v to x for each x E V; in particular,
e(v) = 0. A planted plane
tree (8, E, V, R) is a planted tree (V, E, w) with a linear order relation R
on V possessing two properties : (i) If 2, y E V and Q(X)<e(y), then (x, y) E R.
(ii) If (r, s>, {x, y} E E, e(r)=~(x)=e(s)-1
=~(y)-1,
and (Y; x) E R, then
(s, y) E R. Let P(V) d enote the set of planted plane trees with vertex
set V. Planted plane trees as we have defined them are equivalent to the
planted plane trees defined by HARARY, PRIM, and TUTTE [S]; our definition is in terms of binary relations on finite sets while the latter definition
is given in topological terms.
There are nice diagrams corresponding
to planted plane trees ; the
vertices of the tree can be arranged in consecutive levels 0, 1, . . . so that
x E V is in level Q(X), and then the vertices in each level can be put in
a linear order from left to right according to the order relation R. The
conditions on R imply that the edges of the tree will occur in the diagram
between consecutive levels, and if the edges of the tree are indicated by
straight lines in the diagram, then no lines intersect in the diagram.
Thus, planted plane trees are equivalent
to topological trees that have
been embedded in the plane. The diagrams of the elements of P{a, b, c,
d, e, j} corresponding
to a given planted tree are shown in figure 1.
Two elements (V, El, ~1, RI), (V, Ez, ~2, R,) E P( V) are isomorphic if
there exists a permutation
p of V such that (i) v2 = T(Q), (ii) Ez = {{qz, qy) :
{x, y> E El}, and (iii) (x, y) E RI if and only if (CJX,CJZJ)
E R2. Isomorphism
is an equivalence relation on P(V) ; hence, let P*(V) denote the classes
of isomorphic, planted plane trees with vertex set V. (Elements of P*(V)
are equivalent
to what HARARY,
PRINS and TUTTE
[3] called mapisomorphic,
planted plane trees.)
A tree is k-vabent if every vertex of the tree has degree 1 or k. Let P( V, k)

Fig.

1.

Some planted

plane

trees.

294
denote the set of all (k+ 1)-valent, planted plane trees with vertex set V,
and let P*( P, k) denote the set of classes of isomorphic planted plane
trees in P( 8, lc). For a fixed le it is easy to prove by induction on n that
P( I, k) # + implies 1VI = kn + 2 for some n. Suppose j VI = kn + 2, and let
p(n, k): = IP*( V, k)j. It was shown in [4] that

(2)

p(n, 4 =

The diagrams
P*( V, 2) with

010101

corresponding
/VI =2(4)+2=

010110

011001

YY

2.

Diagrams

101100

representing

1).

to the ~(4, 2)=

z /5= 14 elements
0
10 are shown in figure 2.

011010

011100

Fig.

lo2 /(i&z-n+
( n >

of

c
100101

v
111000

the elements
sponding
code

110100

of P*(
words.

V, 2),

1VI = 10, with

the

corre-

Let B(n, k) denote the set of all binary sequences (bi, . . . . &k-k) containing exactly n-1
units such that bl+ . . . $ bjk>j for j= 1, . . . . n-l ;
also, let b(n, k) : = IB(n, k)l. W e o b served in [4] that there is a one to one
correspondence
[ between the elements of B(n, k) and P*( V, k) when
/VI =kn+2;
thus, b(n, k)=p(n, k).
Now we are going to construct a one to one correspondence E between
the elements of B(n, k) and the elements of T*( V, K) with 1V( = n + 1 and
/li/ = k + 1. Once this has been accomplished,
5 and E can be combined
to obtain a one to one correspondence between the elements of P*(U, k)
with j U] = kn + 2, and the elements of T*( V, K) with j VI = n+ 1 and
/K/=/%+1.
The correspondence

Let K denote a (k+ 1)-set; we will require


so we assume K={l,
. . . . k+l>
in the first
ordering of the integers. Let V denote an
element T:=(V,
E, v; f) E T(V, K). N ow we
G on V in terms of E, v, and f ; however, we

a linear order relation on K,


place and use the natural
(n+ l)-set, and consider an
define a linear order relation
plan to use the symbol G

ambiguously
and depend on the context in which it is used to make its
meaning clear. (Recall that e(x) denotes the length of a path from x E I
to the root v in T.) First, define x<y in T if Q(X)C&J); next, vertices
belonging to the same level in T can be linearly ordered by induction
on the levels of the vertices by using the following rule : If {Y, 81, {z, y} E E,
e(r) = e(x) = e(s)-1 =e(y)-1,
then define s < y in T if and only if f(r, s><
c/(x, y}. (It is easy to check that if p is an isomorphism sending TI onto
Tz where TI, T2 E T( V, K), then x G y in TI if and only if vx<vy in Tz;
just use the fact that Q(X)=Q(~x), and fi{x, y)=fz{~?x, vy).)
Now we assign a binary sequence of length n&---r% to each element of
T(V, K). Suppose T:=(V,
E, V; f) ET(V, K), and suppose ~=wa<~r<
< . . . <v, is the linear order on V in T defined in the last paragraph,
then E(T)= (by, . . . . bnkPn) is defined as follows : For i = 1, . . . , n let aj denote
the color assigned to the edge leading from vj to the root v; also, suppose
the elements of K\{aj> are cl < . . . < ck ; now, define bkj+g= 1 if some edge
joined to vj is assigned the color ci, and define bY+t= 0 otherwise for
i=l
to check that l(Tl)=[(Tz)
if and only
> ***, Ic. It is a routine matter
if T1 and Tz are isomorphic. Thus, indulging in a slight abuse of notation,
6 can be considered a mapping sending T*( V, K) into B(n, /%).To see that
l(T) E B(n, k) for every T E T( V, K), note that bl+ . . . + bkf is the sum
of the degrees of vertices vi, . . . , vj, and T with its rooted edge deleted
is a tree. Also, T with its rooted edge deleted is a tree with n vertices,
units, each unit corresponding
to an
so (bl, . . . . bkn-k) has exactly n-l
edge in the tree. Conversely, given (bl, . . . . bk+k) E B(n, Ic), a representative
element of a class in T*( V, K) can be constructed. The diagrams corresponding to the elements of T*( V, K) with 1V I= n + 1= 5 and 1K I= k + 1 = 3
are shown in figure 3; in the diagrams an edge assigned i E K has been
indicated with i parallel lines. The binary sequence of length (4-l) x
(3-l)=
6 assigned to each tree by [ is also indicated in the figure.

011100

Fig.

3.

Diagrams

101100

representing

111000

110100

the elements
of P( V, K),
corresponding
code
words.

110010

/ VI =5,

110001

IKK/ = 3 with

the

N. G. de Bruijn has observed that the correspondence E is the same as


the one suggested by figure 4. Suppose X E P*( V, k), 1VI =kn+ 2, and
select a representative
element T E X. There is a unique subtree T' of
T which has as its vertex set the root and all vertices of degree k+ 1
in T; since T has exactly n vertices of degree k+ 1, T has exactly n+ 1

Fig.

4.

vertices. We are going to color the edges of T so that T' without the linear
order relation on its vertices becomes an element of T*( 'CT',
ii), / V'I =n+ 1.
Let K={l,
. . . . Ic+ l> denote a set of colors, and assign 1 to the rooted
edge of T. Begin with the edges adjacent to the rooted edge of T: Suppose
the edge below a vertex x of degree ES 1 (the edge going toward the root
from x) has been colored i E K, and suppose cl < . . . <ck are the colors in
K\(i),
then the k edges above and adjacent to x are colored cl, . . . . ck
from left to right. This coloring of T is also a coloring of T', and T' E
E T( I, K). It is easy to check that this mapping of P( V, k) into T( V', K)
defines a one to one correspondence
between the elements of P*( V, k)
and T*( v, k) which is equivalent
to E.
Technological
Eindhoven,

The

University,
Netherlands

REFERENCES
1.

2.
3.
4.

5.

CARLITZ, L.,

A note
on the enumeration
of line chromatic
trees,
Journal
of
Combinatorial
Theory
6, 99-101
(1969).
BRIJIJN, N. G. DE and B. J. M. MORSELT, A note on plane trees,
Journal
of
Combinatorial
Theory,
2, 27-34
(1967).
HARARY, I?., G. PRINS and W. T. TUTTE, The number
of plane
trees,
Nederl.
Akad.
Wetensch.
Proc.
Ser. A 67, (=Indag.
Math.,
26, 319-329
(1964).
KLARNER,
D. A., Correspondences
between
plane
trees
and binary
sequences,
to appear
in Journal
of Combinatorial
Theory.
RIORDAN, J., The number
of labeled,
colored
and chromatic
trees,
Acta
Math.
97, 211-225
(1957).

You might also like