Sei sulla pagina 1di 16

Abstract.

In this paper, we study a new variant of Fibonacci words,


called ”Watson-Crick” Fibonacci words and study various properties of
such words.

1 Preliminaries

An alphabet Σ is a finite non-empty set of symbols, and Σ ∗ denotes the set of all
words over Σ including the empty word λ, while Σ + is the set of all non-empty
words over Σ. The length of a word u ∈ Σ ∗ (i.e., the number of symbols in a
word) is denoted by |u|. We denote by |u|a the number of occurrences of a letter
a in u. By Σ m we denote the set of all words of length m > 0 over Σ. A language
L is a subset of Σ ∗ . The complement of a language L ⊆ Σ ∗ is Lc = Σ ∗ \L. For a
language L ⊆ Σ ∗ and i ≥ 2, let L(i) = {ui |u ∈ L} and L1 = L and Ln = Ln−1 L
for n ≥ 2.
A word x ∈ Σ ∗ is said to be a prefix of w ∈ Σ + (suffix, respectively) if
w = xα (w = βx, respectively) where α, β ∈ Σ ∗ . If α ∈ Σ + (β ∈ Σ + , respec-
tively), then x is said to be a proper prefix (proper suffix, respectively) of w.
By Pref(w) (Suff(w), respectively), let us denote the set of all prefixes of w.
(suffixes, respectively).
A word is called primitive if it cannot be expressed as a power of another
word. Let Q denote the set of all primitive words. For every word w ∈ Σ + ,
there exists a unique word ρ(w) ∈ Σ + , called the primitive root of w, such that
ρ(w) ∈ Q and w = ρ(w)n for some n ≥ 1.
For a word x = a1 a2 · · · ak with ai ∈ Σ, the reverse of x denoted by xR is
the word xR = ak ak−1 · · · a2 a1 . A word x is a palindrome if x = xR . We denote
by P the set of all palindromes over Σ ∗ . A function ϕ : Σ ∗ → Σ ∗ is said to be
a morphism if for all words u, v ∈ Σ ∗ we have that ϕ(uv) = ϕ(u)ϕ(v), and an
antimorphism if ϕ(uv) = ϕ(v)ϕ(u). The function ϕ is called an involution if ϕ2
is an identity on Σ ∗ .
Given a morphic or antimorphic involution ϕ : Σ ∗ → Σ ∗ – we sometimes
refer to such a function as an (anti)morphic involution) – and given a word
u ∈ Σ + , a ϕ-power of the word u is any word of the form w = u1 u2 · · · un , n ≥ 1
where u1 = u and ui ∈ {u, ϕ(u)}, for 2 ≤ i ≤ n [5]. Any word u such that w is a
ϕ-power of u is called a ϕ-period of w [5]. A word w ∈ Σ + is called ϕ-primitive
if w being a ϕ-power of u implies w = u and n = 1 [5]. By Qϕ we denote the set
of all ϕ-primitive words. As it was shown in [5], for every word w ∈ Σ + , there
uniquely exists a ϕ-primitive word ρϕ (w) ∈ Σ + , called the ϕ-root of w, such
that w is a ϕ-power of ρϕ (w).
A word w ∈ Σ + is called a weak ϕ-power of a word u if w ∈ {u, ϕ(u)}+ . Note
that, unlike ϕ-powers of u, a weak ϕ-power of a word u does not need to start
with u, it can also start with ϕ(u). Thus, if w is a ϕ-power of u, then w is also
a weak ϕ-power of u, but the converse does not hold. For example, for a given
(anti)morphic involution ϕ and word u ∈ Σ + , the words ui and ui ϕ(u)j , i, j ≥ 1
are both ϕ-powers and weak ϕ-powers of u, while ϕ(u)i and ϕ(u)i uj , i, j ≥ 1,
are weak ϕ-powers of u but not ϕ-powers of u.

1
A word x ∈ Σ ∗ is said to be a border of w ∈ Σ + if w = xα = βx where
α, β ∈ Σ ∗ , and a proper border of w if, in addition, x 6= w, see []1 . A word
w ∈ Σ + is called bordered if it has a proper border, and unbordered otherwise.
Similarly, for an (anti)morphic involution ϕ, a word x ∈ Σ ∗ is said to be a ϕ-
border of w ∈ Σ + if w = xα = βϕ(x) for some α, β ∈ Σ ∗ , and a proper ϕ-border
if, in addition, |x| 6= |w|, see [4]. Note that for any (anti)morphic involution ϕ
and any word w ∈ Σ + , the empty word λ is a ϕ-border of w. A nonempty word
is called ϕ-bordered if it has a proper ϕ-border, and is called ϕ-unbordered if it
is not ϕ-bordered, [4]. For w ∈ Σ + , let Lϕ d (w) denote the set of all proper ϕ-
borders of w, and Dϕ (i) denote the set of all words with exactly i ϕ-borders, [4].
In particular, Dϕ (1) denotes the set of all nonempty words that have λ as their
only ϕ-border, that is, they are ϕ-unbordered.
For convenience, in the remainder of this paper we use the convention that
ϕ denotes an (anti)morhic involution, θ denotes an antimorphic involution and
µ denotes a morphic involution. We recall the following results.

Lemma 1. [4] Let θ be an antimorphic involution. Then for all x ∈ Σ + we


have that x is θ-bordered iff x = ayθ(a) for some a ∈ Σ and y ∈ Σ ∗ .

Definition 1. [3] Given an (anti)morphic involution ϕ on Σ ∗ and words u, v ∈


Σ ∗ , the binary operation ϕ-catenation is defined as

u ϕ v = {uv, uϕ(v)}.

This concept can be extended to cross-ϕ-catenation on words, as follows.

Definition 2. Given an (anti)morphic involution ϕ on Σ ∗ and words u, v ∈ Σ ∗ ,


the binary operation cross-θ-catenation is defined as

u ⊗ϕ v = {uv, ϕ(u)v, uϕ(v), ϕ(u)ϕ(v)},

and the operation can be extended to languages L1 , L2 ⊆ Σ ∗ by

L1 ⊗ϕ L2 = {u ⊗ v| u ∈ L1 , v ∈ L2 }.

In both the case of ϕ-catenation and cross-ϕ-catenation, of the (anti)morphic


involution ϕ is obvious from the context, the subscript ϕ will be omitted.

Definition 3. Let ϕ be an (anti)morphic involution over Σ ∗ . The n-th Fi-


bonacci set Wn is defined as

Wn = Wn−1 ⊗ Wn−2 .
1
Note that a border, respectively proper border of a word, is sometimes called d-
factor, respectively proper d-factor, see [?].

2
2 ϕ-Fibonacci words

Definition 4. [?] Let Σ be an alphabet with card(Σ) ≥ 2 and let u, v ∈ Σ +


with u 6= v, with u 6= v. The nth standard Fibonacci words are defined recursively
as:
f1 (u, v) = u, f2 (u, v) = v
fn (u, v) = fn−1 (u, v) fn−2 (u, v), n ≥ 3.
The sequence of standard Fibonacci words is defined as F (u, v) = {fn (u, v)}n≥1 ,
that is, F (u, v) = {u, v, vu, vuv, vuvvu, vuvvuvuv, vuvvuvuvvuvvu, . . .}. Simi-
larly, the nth reverse Fibonacci words are defined recursively as:

f10 (u, v) = u, f20 (u, v) = v

fn0 (u, v) = fn−2 (u, v) fn−1 (u, v), n ≥ 3


and the sequence of reverse Fibonacci words is defined as F 0 (u, v) = {fn0 (u, v)}n≥1 ,
that is, F 0 (u, v) = {u, v, uv, vuv, uvvuv, vuvuvvuv, uvvuvvuvuvvuv, . . .}.

In the following, if the first two Fibonacci words are obvious from the context,
the argument (u, v) will be ignored, and we will simply write the nth standard
Fibonacci word as fn , the nth reverse Fibonacci word as fn0 , the standard Fi-
bonacci sequence as F , and the reverse Fibonacci sequence as F 0 .
The notation F, F’ is from ”Properties of Fibonacci languages”,
Yu, Zhao, 2000; they denoted f by w and f ’ by w’
It was noted in [2], that every standard (reverse) Fibonacci sequence F (u, v)
(respectively F 0 (u, v)) is a homomorphic image of the standard (reverse) Fi-
bonacci sequence F (a, b) (respectively F 0 (a, b)), via the homorphism h(a) = u
and h(b) = v, where a 6= b. (check of this is true or we can have swaps F
to F’)
Thus, properties of the Fibonacci words and sequences over a two-letter al-
phabet, which are sometime called atom Fibonacci words and sequences [2] are
especially important. We recall the following.

Lemma 2. [1] Let Σ = {a, b}, with a 6= b, and let f1 = a and f2 = b. Then,
for all n ≥ 3, one has that fn = αn dn , where αn is palindrome, while dn = ab if
n is even and dn = ba if n is odd.

Lemma 3. [2] Let Σ = {a, b}, with a 6= b, and let f1 = a and f2 = b. For
all n ≥ 3, let fn = αn dn and fn0 = d0n αn0 where αn , dn , αn0 , d0n ∈ Σ ∗ , with
|dn | = |d0n | = 2. Then αn = αn0 .

Note that the proof Lemma 3 also entails that dn = ab and d0n = ba if n is
even, and dn = ba, d0n = ab if n is odd.
We now have the following observation.
Lemma 4. Let Σ = {a, b}, f1 = f10 = a, f2 = f20 = b, a 6= b. If fk and fk0 are
the kth standard and reverse Fibonacci words respectively, then fk0 = fkR .

3
Proof. We know that by Lemma 2 fk = αk dk . Then fkR = dR R
k αk . Moreover,
by Lemma 2, we have that αk is a palindrome, and by Lemma 3 we have that
αk = αk0 and d0k = dR R 0 0 0
k . Thus fk = dk αk = fk .

Lemma 4 thus justifies calling fn and Fn ”standard” Fibonacci words and


sequence, while fn0 and Fn0 are called ”reverse” Fibonacci words and sequence.
We can now define the nth standard and reverse ϕ-Fibonacci words in the fol-
lowing three ways corresponding to Definitions 3 and 4, where ϕ is an (anti)morphic
involution (See Fig 2).

fn = fn−1 fn−2
fn0 = fn−2
0 0
fn−1

gn = ϕ(gn−1 )gn−2 zn = zn−1 ϕ(zn−2 )


gn0 = gn−2
0 0
ϕ(gn−1 ) WK-Fibonacci Words zn0 = ϕ(zn−2
0 0
)zn−1

wn = ϕ(wn−1 )ϕ(wn−2 )
wn0 = ϕ(wn−2
0 0
)ϕ(wn−1 )

Fig. 1. Three possible ways of defining standard and reverse ϕ-Fibonacci Words.

Note that, as opposed to starting with an alphabet of cardinality at least


two, as required in Definition 4, in all these definitions we need at least a ternary
alphabet since even in the simplest case where we would require w1 = a, w2 = b
with a 6= b and a 6= ϕ(b).
We observe that the words fn , fn0 in Fig. 2, are the standard and reverse
Fibonacci words defined in Definition 4. One can also easily observe that the

4
Fibonacci words in the rightmost branch are nothing but subsequences of Fi-
bonacci words fn and fn0 . For example, consider the alphabet Σ = {A, C, G, T }
such that ϕ(A) = T , ϕ(G) = C, and vice versa. Without loss of generality, we
can choose g1 = g10 = A and g2 = g20 = C. Then we have the following,

gn = ϕ(gn−1 )gn−2 gn0 = gn−2


0 0
ϕ(gn−1 )
MI AMI MI AMI
A A A A
C C C C
GA GA AG AG
CT C T CC CT C CCT
GAGGA GGAGA AGGAG AGAGG
CT CCT CT C T CT CCT CC CT CT CCT C CCT CCT CT
.. .. .. ..
. . . .

Note that the words gn and gn0 are Fibonacci words fn and fn0 over binary
alphabet {A, G} when n is odd and are Fibonacci words fn and fn0 over the
binary alphabet {C, T } when n is even.
QN: May be we should prove this in general..without using an
example

Lemma 5. If µ is a morphic involution over an alphabet Σ, and a, b ∈ Σ,


a 6= b, then the standard Fibonacci words have the property that fn (a, b) =
µ(fn (µ(a), µ(b))) for n ≥ 1.

Proof. Strong induction on n. By definition, f1 (a, b) = a = µ(µ(a)) = µ(f1 (µ(a), µ(b))),


and f2 (a, b) = b = µ(µ(b)) = µ(f2 (µ(a), µ(b))), so the base case holds.
Assume now that fi (a, b) = µ(fi (µ(a), µ(b))), for all 1 ≤ i ≤ k. We have that:

fk+1 (a, b) = fk (a, b)fk−1 (a, b)


= µ(fk (µ(a), µ(b))) µ(fk−1 (µ(a), µ(b)))
= µ(fk (µ(a), µ(b)) fk−1 (µ(a), µ(b)))
= µ(fk+1 (µ(a), µ(b)))

and the inductive step, as well the proof, hold. t


u

Proposition 1. If µ is a morphic involution, over an alphabet Σ, and a, b ∈ Σ,


a 6= b, we have that:
(
fn (a, b), if n is odd
gn (a, b) =
fn (µ(a), µ(b)), if n is even.

Proof. Induction on n. By definition, g1 (a, b) = a = f1 (a, b), g2 (a, b) = µ(b) =


f2 (µ(a), µ(b)).
For 1 ≤ i ≤ n, we assume that g2∗i−1 (a, b) = f2∗i−1 (a, b), g2∗i (a, b) =
f2∗i (µ(a), µ(b)).

5
Consider the case that i = n + 1, we have that:

g2∗n+1 (a, b) = µ(g2∗n (a, b))g2∗n−1 (a, b)


= µ(f2∗n (µ(a), µ(b)))f2∗n−1 (a, b)
= f2∗n (a, b)f2∗n−1 (a, b)
= f2∗n+1 (a, b)
2∗n+2
ga,b = µ(g2∗n+1 (a, b))g2∗n (a, b)
= µ(f2∗n+1 (a, b))f2∗n (µ(a), µ(b))
= f2∗n+1 (µ(a), µ(b))f2∗n (µ(a), µ(b))
= f2∗n+2 (µ(a), µ(b))

t
u

Proposition 2. When ϕ is an antimorphic involution denoted by θ, the previous


relation does not hold.

Proof. The can be proven by a counterexample. g4 (A, T ) = T CC 6= CT C =


f4 (T, C).

Definition 5. Consider an alphabet Σ, and two characters a, b ∈ Σ. The Fi-


bonacci words tna,b is defined as:


 a, n = 1

b, n = 2
tn =
tn−1 tn−2 , n is odd, n ≥ 3


tn−2 tn−1 , n is even, n ≥ 3

Proposition 3. When ϕ is an antimorphic involution denoted by θ, we have


the following relation for n ≥ 1:
(
tn (A, G), n is odd
gn (A, G) =
(tn (T, C))R , n is even

Proposition 4. When ϕ is an antimorphic involution denoted by θ, we have


the following relation for n ≥ 4:
(
θ(g2 )g3 g5 ...gn−4 gn−2 gn−2 , n is odd
gn (A, G) =
θ(g1 )g2 g4 ...gn−4 gn−2 gn−2 , n is even

Corollary 1. For n ≥ 4, we have the following properties:


n−2
– gn has a square suffix gn−2∗i gn−2∗i for i ≤ 2 ;

6
– the longest square suffix of gn is gn−2 gn−2 ;

– For n is odd, n ≥ 6, gn has a square prefix θ(g2 )g3 g5 ...gn−4 θ(g2 )g3 g5 ...gn−4 ;

– For n is even, n ≥ 6, gn has a square prefix θ(g1 )g2 g4 ...gn−4 θ(g1 )g2 g4 ...gn−4 ;

– For n ≥ 6, gn has a square prefix (gn−2 (gn−4 )−1 )(gn−2 (gn−4 )−1 );

– For n ≥ 6, if we insert gn−4 to gn , it becomes cubic gn−2 gn−2 gn−2 , for


example, for n is odd:

gn (A, G) =θ(g2 )g3 g5 ...gn−4 gn−2 gn−2


θ(g2 )g3 g5 ...gn−4 gn−4 gn−2 gn−2 = gn−2 gn−2 gn−2

for n is even:

gn (A, G) =θ(g1 )g2 g4 ...gn−4 gn−2 gn−2


θ(g1 )g2 g4 ...gn−4 gn−4 gn−2 gn−2 = gn−2 gn−2 gn−2

Thus, we only concentrate on the middle two branches in Fig 2.

Definition 6. Consider an alphabet Σ such that |Σ| ≥ 3 and let ϕ be either a


morphic or an antimorphic involution on Σ extended to Σ ∗ . Then,

1. Let w1 = a, w2 = b such that a 6= b and a 6= ϕ(b) and the nth Standard


Palindromic-Fibonacci word wn is defined as wn = ϕ(wn−1 )ϕ(wn−2 ).
2. Let w10 = a, w20 = b such that a 6= b and a 6= ϕ(b) and the nth Reverse
0
Palindromic-Fibonacci word wn0 is defined as wn0 = ϕ(wn−2 0
)ϕ(wn−1 ).
th
3. Let z1 = a, z2 = b such that a 6= b and a 6= ϕ(b) and the n Standard
Hairpin-Fibonacci word zn is defined as wn = zn−1 ϕ(zn−2 ).
4. Let z10 = a, z20 = b such that a 6= b and a 6= ϕ(b) and the nth Reverse
0
Hairpin-Fibonacci word wn0 is defined as zn0 = ϕ(zn−2 0
)zn−1 .

We illustrate with the following example.

wn = ϕ(wn−1 )ϕ(wn−2 ) wn0 = ϕ(wn−2


0 0
)ϕ(wn−1 )
MI AMI MI AMI
A A A A
C C C C
GT GT TG TG
CAG ACG GAC GCA
GT CCA CGT AC ACCT G CAT GC
CAGGT GT C GT ACGCGT CT GT GGAC T GCGCAT G
.. .. .. ..
. . . .

7
zn = zn−1 ϕ(zn−2 ) zn0 = ϕ(zn−2 )zn−1
0

MI AMI MI AMI
A A A A
C C C C
CT CT TC TC
CT G CT G GT C GT C
CT GGA CT GAG AGGT C GAGT C
CT GGAGAC CT GAGCAG CAGAGGT C GACGAGT C
.. .. .. ..
. . . .

We have the following observations. Proof required


1. znR = zn0
2. wnR = wn0
3. For an antimorphic involution θ, wn can be classified as (similar to Lemma
9) and for k ≥ 1, we can show that

w3k = Aw3
w3k+1 = Bw4
w3k+2 = Cw2
where A, B, C are product of θ palindromes.
4. For an antimorphic involution θ, zn can be classified as (similar to Lemma
9) and for k ≥ 1we can show that

z3k = αz3 θ(α), for k odd


z3k = αθ(z3 )θ(α), for k even
z3k+1 = αθ(z1 )θ(α), for k odd
z3k+1 = αz1 θ(α), for k even
z3k+2 = αθ(z2 )θ(α), for k odd
z3k+2 = αz2 θ(α), for k even
It can also be noted that, above observations hold even in case of a ternary
alphabet, i.e., {a, b, c} and ϕ is an (anti)morphic involution such that ϕ(a) = b
and vice versa, and ϕ(c) = c.

3 Standard Palindromic-Fibonacci words


By F (k), let us denote the k th Fibonacci number defined by the recurrence
relation: F (0) = 0, F (1) = 1, F (2) = 1, F (k) = F (k − 1) + F (k − 2).
Let Σ = {A, C, G, T } and let θ be an (anti)morphic involution such that
ϕ(A) = T, ϕ(T ) = A, ϕ(C) = G and ϕ(G) = C. Furthermore, let

w1 = A, w2 = C, , and wn = ϕ(wn−1 )ϕ(wn−2 ).

8
wn = ϕ(wn−1 )ϕ(wn−2 )
MI AMI
A A
C C
GT GT
CAG ACG
GT CCA CGT AC
CAGGT GT C GT ACGCGT
.. ..
. .

Lemma 6. New! For an antimorphic involution θ, wn can be classified as

w3k = Aw3 , k ≥ 2

w3k+1 = Bw4 , k ≥ 2

w3k+2 = Cw2 , k ≥ 1

where

A = w3k−3 θ(w3k−3 )θ(w3k−4 )w3k−4 w3k−6 θ(w3k−6 )θ(w3k−7 )w3k−7 · · · w3 θ(w3 )θ(w2 )w2

B = w3k−2 θ(w3k−2 )θ(w3k−3 )w3k−3 w3k−5 θ(w3k−5 )θ(w3k−6 )w3k−6 · · · w4 θ(w4 )θ(w3 )w3

C = w3k−1 θ(w3k−1 )θ(w3k−2 )w3k−2 w3k−4 θ(w3k−4 )θ(w3k−5 )w3k−5 · · · w2 θ(w2 )θ(w1 )w1

Proof. We prove only the case when n mod 3 ≡ 0 as other cases can be proved
similarly. Let n mod 3 ≡ 0, then n = 3k where k > 1. We further prove the
result by induction on k.
Base case: Let k = 2, then w6 = θ(w5 )θ(w4 ) = w3 w4 w2 w3 = w3 θ(w3 )θ(w2 )w2 w3 .
For k = 3,

w9 = θ(w8 )θ(w7 )
= w6 w7 w5 w6
= w6 θ(w6 )θ(w5 )w5 w3 θ(w3 )θ(w2 )w2 w3

Induction hypothesis: Assume that w3N = w3N −3 θ(w3N −3 )θ(w3N −4 )w3N −4


w3N −6 θ(w3N −6 )θ(w3N −7 )w3N −7 · · · w3 θ(w3 )θ(w2 )w2 w3 for all 2 ≤ k ≤ N .
Induction step: Consider

w3(N +1) = w3N +3 = θ(w3N +2 )θ(w3N +1 )


= w3N w3N +1 w3N −1 w3N
= w3N θ(w3N )θ(w3N −1 )w3N −1 w3N

Hence the result. t


u

9
Lemma 7. New! For an antimorphic involution θ, zn can be classified as

z3k = αz3 θ(α), for k odd


z3k = αθ(z3 )θ(α), for k even
z3k+1 = αθ(z1 )θ(α), for k odd
z3k+1 = αz1 θ(α), for k even
z3k+2 = αθ(z2 )θ(α), for k odd
z3k+2 = αz2 θ(α), for k even
Proof. We here prove a more general result, i.e., zn = xn yn θ(xn ) where |yn | = 1
if |zn | is odd and |yn | = 2 if |zn | is even. We prove by a string induction on n.
Base case: We know that z1 = A, z2 = C and z3 = CT . Thus we get
z4 = C(T )G, z5 = CT (G)AG and z6 = CT G(AG)CAG.
Induction hypothesis: Suppose the result holds for all zn such that 4 ≤ n ≤ k.
Induction step: We prove that the result holds for zk+1 . We have following
two cases
Case 1: Let |zk+1 | be even. Then |zk | and |zk−1 | is odd, and |zk−2 | is even.
Furthermore,

zk+1 = zk θ(zk−1 )
= zk−1 θ(zk−2 )θ(zk−1 )
= zk−1 θ(xk−2 yk−2 θ(xk−2 ))θ(zk−1 )
= zk−1 xk−2 θ(yk−2 )θ(xk−2 )θ(zk−1 )

By induction hypothesis |yk−2 | = 2.


Case 2: Let |zk+1 | be odd. Then one of |zk | or |zk−1 | is odd. Without loss of
generality, let |zk | be odd. Then |zk−1 | is even and |zk−2 | is odd. Furthermore,
since zk+1 = zk θ(zk−1 ) = zk−1 θ(zk−2 )θ(zk−1 ), induction hypothesis can be ap-
plied to zk−2 and hence we get zk+1 = zk−1 xk−2 θ(yk−2 )θ(xk−2 )θ(zk−1 ) with
|yk−2 | = 1.
Furthermore, we have that z4 = Cθ(z1 )G, z5 = CT θ(z2 )AG, z6 = CT Gθ(z3 )CAG
and for any k 0 , zk0 = xk0 θ(yk0 −3 )θ(xk0 ), we get the required result. t
u
Corollary 2. New! For an antimorphic involution θ, zi ∈ Dθ (j) where j = d k2 e
and k = |zi |.
Lemma 8. New! Let θ be an antimorphic involution and {wn0 }n≥1 be the se-
quence of Palindromic-Fibonacci words. Then,
0 0 0 0
w3k+1 = θ(w3k−1 )w3k−1 θ(w3k−4 )w3k−4 · · · θ(w50 )w50 θ(w20 )w10 , k ≥ 1
0 0 0 0
w3k+2 = θ(w3k )w3k θ(w3k−3 )w3k−3 · · · θ(w60 )w60 θ(w30 )w30 w20 , k ≥ 1
and
0 0 0 0
w3k = θ(w3k−2 )w3k−2 θ(w3k−5 )w3k−5 · · · θ(w70 )w70 θ(w40 )w40 w30 , k ≥ 2

10
Proof. Same as the proof for wn .
Proposition 5. New! For an antimorphic involution θ, let {wn }n≥1 and {wn0 }n≥1
be the sequence of Palindromic-Fibonacci words. If wk and wk0 are the k-th
Palindromic-Fibonacci words of the same length, then wkR = wk0 .
Proposition 6. New! For morphic involution µ, wn wn+1 almost µ-commutes
with wn+1 wn , i.e., wn wn+1 and µ(wn+1 )wn shares a common prefix of length
2F (n).
Proof. We know that wn wn+1 = wn µ(wn µ(wn−1 ) = wn wn−1 wn−2 wn−2 wn−3 .
Similarly, µ(wn+1 )wn = wn wn−1 wn−2 wn−3 µ(wn−2 ). It is clear from above two
equations that wn wn+1 and µ(wn+1 )wn shares a common prefix of length 2F (n).
However, since wn−2 and wn−3 begins with a different letter, wn−2 wn−3 6=
wn−3 µ(wn−2 ). t
u
Lemma 9. Let θ be an antimorphic involution and {wn }n≥1 be the sequence of
Palindromic-Fibonacci words. Then for k > 1,

w3k+1 = w1 w2 θ(w2 )w5 θ(w5 ) · · · w3k−4 θ(w3k−4 )w3k−1 θ(w3k−1 )

w3k+2 = w2 w3 θ(w3 )w6 θ(w6 ) · · · w3k−3 θ(w3k−3 )w3k θ(w3k )


and
w3k = w3 w4 θ(w4 )w7 θ(w7 ) · · · w3k−5 θ(w2k−5 )w3k−2 θ(w3k−2 )
Proof. We prove only the case when n mod 3 ≡ 0 as other cases can be proved
similarly. Let n mod 3 ≡ 0, then n = 3k where k > 1. We further prove the
result by induction on k.
Base case: Let k = 2, then w6 = θ(w5 )θ(w4 ) = w3 w4 θ(w4 ). For k = 3,

w9 = θ(w8 )θ(w7 )
= w6 w7 θ(w7 )
= θ(w5 )θ(w4 )w7 θ(w7 )
= w3 w4 θ(w4 )w7 θ(w7 )

Induction hypothesis: Assume that w3N = w3 w4 θ(w4 )w7 θ(w7 ) · · · w3n−2 θ(w3n−2 )
for all 2 ≤ k ≤ N .
Induction step: Consider

w3(N +1) = w3N +3 = θ(w3N +2 )θ(w3N +1 )


= w3N w3N +1 θ(w3N +1 )
= w3 w4 θ(w4 )w7 θ(w7 ) · · · w3N −2 θ(w3N −2 )w3N +1 θ(w3N +1 )

Hence the result. t


u
Corollary 3. Let θ be an antimorphic involution and {wn }n≥1 be the sequence
of Palindromic-Fibonacci words. Then for n > 3, wn = βα1 α2 · · · αk where
αi ∈ Pθ for 1 ≤ i ≤ k and β ∈ {w1 , w2 , w3 }.

11
Definition 7. An infinite language L is a regular-free language if every infinite
subset of L is not a regular language.

Theorem 1. The language L of Palindromic-Fibonacci words is regular-free for


an (anti)morphic involtion ϕ.

Proof. Let L1 = {u1 , u2 , u3 , . . .} ⊆ {w1 , w2 , w3 , . . .} be an infinite subset of L.


For the sake of contradiction let us assume that L1 is regular. By pumping lemma
for regular languages, there exists a positive integer m such that for any ui ∈ L1
with |ui | ≥ m, ui can be decomposed as ui = xyz with |xy| ≤ m, |y| ≥ 1 and
uij = xy j z ∈ L1 for all j ≥ 0. However, due the fact that the length of every
Palindromic-Fibonacci word is a Fibonacci number, there certainly exists some
j ≥ 0 such that uij ∈/ L1 , a contradiction. t
u

In the following we show that the Palindromic-Fibonacci words are indeed θ-


unbordered for an antimorphic involution θ.

Proposition 7. For an antimorphic involution θ, wi ∈ Dθ (1) for all i ≥ 1.

Proof. We prove the result by the induction on i.


Base case: We have w1 = A, w2 = C ∈ Dθ(1) . Also, w3 = GT, w4 =
ACG, w5 = CGT AC ∈ Dθ (1).
Induction hypothesis: Suppose the result holds for 6 ≤ i ≤ n.
Induction step: For the sake of contradiction, assume that wn+1 ∈
/ Dθ(1) .
Then by Lemma 1 wn+1 = axθ(a) where x ∈ Σ + and a ∈ Σ. This implies

wn+1 = θ(wn )θ(wn−1 )


= wn−2 wn−1 wn−3 wn−2
= axθ(a).

This further implies a <p wn−2 and θ(a) <s wn−2 , and hence wn−2 ∈ / Dθ (1), a
contradiction to the induction hypothesis. Thus wi ∈ Dθ (1) for all i ≥ 1. t
u

Lemma 10. Let Bµ [n] denote the length of the longest µ-border of wn for a
morphic involution µ. Then for every n ≥ 4, Bθ [n] = F (n − 2) where F (n − 2)
is the length of the n − 2nd Palindromic-Fibonacci word and is also n − 2nd
Fibonacci number.

Proof. We prove by induction on n.


Base case: Let n = 4. Then w4 = µ(w3 )µ(w2 ) = w2 w1 µ(w2 ). Thus w2 = C
is a µ-border of w4 , and the length of w2 is F (2) = 1, thus Bµ [4] = 1.
Induction hypothesis: Let us assume that the result holds for 5 ≤ n < k − 1.
Induction step: Assume that the result does not hold for k. Since wk =
µ(wk−1 )µ(wk−2 ) = wk−2 wk−3 µ(wk−2 ), Bµ [k] > F (k − 2). Furthermore, by
induction hypothesis, Bµ [k − 1] = F [k − 3] and for every integer N ≥ 4,
Bµ [N ] ≤ Bµ [N − 1] + F (N − 2), Bµ [k] ≤ F (k − 2) + F (k − 3) = F (k − 1).

12
Let Bµ [k] = F (k − 2) + i where 1 ≤ i ≤ F (k − 3). Then since µ(wn−3 ) is a
prefix of wn−2 ,

wk [1..i] = µ(wn [F (k − 2) + 1..F (k − 2) + i]) (1)


= µ(wk [F (k − 1) − i + 1..F (k − 1)]) (2)
= µ(wk [F (k − 1) + 1..F (k − 1) + i]) (3)
= wk [F (k) − i..F (k)] (4)

Thus by (3) and (4), w[1 · · · i] is a µ-border of wk−2 , and hence by induction
hypothesis i ≤ F (k − 4). However, from (1) and (2),it is clear that wn [1 · · · i]
is also a border of wk−3 . This implies w[1 · · · i] ∈ Suff(wk−2 ) ∩ Suff(wk−3 ), a
contradiction. t
u

Explanation for equations from 1 to 4:


We know that wk = µ(wk−3 )µ(wk−4 )wk−3 µ(wk−2 ). If we consider first i
letters of wk , i.e., first i letters of θ(wk−3 ) (this is possible because i ≤ F (k − 3)),
the next wk−3 occurs at F (k − 2) + 1th position, hence the first equation. Also,
since wk = µ(wk−1 )µ(wk−2 ) = wk−2 wk−3 µ(wk−3 ), the next wk−3 occurs at
F (k − 1) − i + 1th position, hence the second equation. Similarly, since wk =
µ(wk−1 )wk−3 wk−4 , wk−3 occurs at F (k − 1) + 1th position, hence the third
equation. Since, wk = wk−2 wk−3 µ(wk−2 ), the same prefix wk [1 · · · i] occurs as a
suffix of wk , hence the fourth equation.

Proposition 8. Let µ be a morphic involution and let |wn | = F (n). Then µ-


borders of wn are

{wn−2 , wn−6 , . . . w2 }


if n mod 4 ≡ 0
{w
n−2 , wn−6 , . . . w3 } if n mod 4 ≡ 1
Lµd (wn ) =
{wn−2 , wn−6 , . . . w4 }
 if n mod 4 ≡ 2

{wn−2 , wn−6 , . . . w5 } if n mod 4 ≡ 3

Proof. We prove the result only when n mod 4 ≡ 0 as other cases can be proved
similarly. We split the proof into following three steps:

1. Show that {wn−2 , wn−6 , . . . , w2 } ⊆ Lµd (wn ).


2. Show that there does not exists i other than one belonging to {n − 2, n −
6, . . . , 2} such that wi is a µ-border of wn .
3. Show that there does exists any x such that x ∈ / {wi }i≥2 is a µ-border of
wn .

1. Let n mod 4 ≡ 0. It is easy to see that each Palindromic-Fibonacci word


begins either with w2 or µ(w2 ). Moreover, if i is even then wi begins with w2 , and
if i is odd then wi begins with µ(w2 ). Now, we show that {wn−2 , wn−6 , . . . w2 }

13
is a µ-border of wn . Let

wn = µ(wn−1 )µ(wn−2 ) (5)


= wn−2 wn−3 µ(wn−2 ) (6)
= µ(wn−3 )µ(wn−4 )wn−3 wn−3 wn−4 (7)
= wn−4 αwn−4 (8)
= µ(wn−5 )µ(wn−6 )αµ(wn−5 )θ(wn−6 ) (9)
= wn−6 wn−7 αµ(wn−5 )µ(wn−6 ) (10)
.
= .. (11)
= w2 βµ(w2 ) (12)

where α, β ∈ Σ + . Thus {wn−2 , wn−6 , . . . , w2 } ⊆ Lµd (wn ).


2 We need to show that wi ∈ / Lθd (wn ) where i ∈
/ {n − 2, n − 6, . . . , w2 }. By
equation (8), wn−4 is a border of wn . Similarly we can prove that if i is even
such that i ∈ / {n − 2, n − 6, . . . , w2 }, wi is a border of wn . Since µ(wn−2 ) is a
suffix of wn , every suffix of wn is either wj or µ(wj ) where j is even. Thus wk
where k is odd can never be a µ-border of wn . Hence the result.
3 We need to show that no x ∈ / {wi }i≥2 can be a µ-border of wn . For
the sake of contradiction, assume that x ∈ Lµd (wn ). Now, since wn−2 is the
longest µ-border of wn , we must have |x| < |wn−2 |, and moreover x <p wn−2 .
If x ∈ Lθd (wn ), then wn = xα = βθ(x) where α, β ∈ Σ + . Now, we show that
by induction that if wn ends with a ∈ Σ, then in the prefix of length n − 2 of
wn , µ(a) occurs only at |wn−2 |, |wn−6 |, . . . , |w2 | positions. Since n mod 4 ≡ 0,
let n = 4k where k ≥ 1.
Base case: Let k = 1. Then w4 = w2 w1 µ(w2 ). The longest θ-border of w4 is
w2 , and w4 ends with µ(w2 ), the result holds.
Induction hypothesis: Assume that the result holds for all 1 ≤ k ≤ N .
Induction step: We know that w4(N +1) = µ(w4N +4 ) = µ(w4N +3 )θ(w4N +2 ) =
w4N +2 w4N +1 µ(w4N +2 ). By Lemma 10, the longest µ-border is w4N +2 which in
turn has w4N as the longest µ-border. Now, if a ∈ Suff(w4N +4 ) for a ∈ Σ, then
a ∈ Suff(w4N +2 ) and hence a ∈ Suff(w4N ). By induction hypothesis, in the
prefix of length less that |w4N +2 |, µ(a) occurs only at |w4N |, |w4N −4 |, . . . , |w2 |
positions. Hence the result.
Thus Lµd (wn ) = {wn−2 , wn−6 , . . . w2 } if n mod 4 ≡ 0. t
u

Corollary 4. Let µ be a morphic involution. Then wj ∈ Dµ (i + 1) where 4i ≤


j ≤ 4i + 3 for all i ≥ 1.

Conjecture 1. Let θ be an antimorphic involution, then the number distinct θ-


palindromic factors in ith Palindromic-Fibonacci word wi is F (i) − 2.

Conjecture 2. Let θ be an antimorphic involution. Then the number of distinct


θ-squares in ith Palindromic-Fibonacci word wi for i ≥ 7 is

S(wi ) = S1 + S2 − (2 × (i − 6))

14
where S1 = F (i)−2 are the θ-squares of the form xθ(x), and S2 = 2(F (i−2)−3)
are θ-squares of the form xx where x ∈ Σ + .
Squares in Palindromic-Fibonacci word for AMI

wn = θ(wn−1 )θ(wn−2 ) Length F (n) Squares(xx) Squares (xθ(x)) Total Common


w1 1 0 0 0 0
w2 1 0 0 0 0
w3 2 0 0 0 0
w4 3 0 1 1 0
w5 5 0 3 3 0
w6 8 1 6 6 1
w7 13 4 11 13 2
w8 21 10 19 25 4
w9 34 20 32 46 6
w10 55 36 53 81 8
w11 89 162 87 139 10
w12 144 104 142 234 12
w13 233 172 231 389 14
Squares in Palindromic-Fibonacci word for MI

wn = µ(wn−1 )µ(wn−2 ) Length F (n) Squares(xx) Squares (xµ(x)) Total squares


w1 1 0 0 0
w2 1 0 0 0
w3 2 0 0 0
w4 3 0 0 0
w5 5 1 0 1
w6 8 2 0 2
w7 13 4 1 5
w8 21 6 4 10
w9 34 8 9 17
w10 55 10 16 26
w11 89 12 25 37
w12 144 14 38 52
w13 233 16 58 74
w14 377 18 89 107
w15 610 20 138 158
w16 987 22 216 238
I am unable to derive the formula for the number of squares

References
1. A. de Luca. A combinatorial property of the fibonacci words. Information Processing
Letters, 12(4):193–195, 1981.

15
2. C.-M. Fan and H. J. Shyr. Some properties of Fibonacci languages. Tamkang
Journal of Mathematics, 27(2):165–182, 1996.
3. L. Kari and M. S. Kulkarni. Pseudo-identities and bordered words. In G. Păun,
G. Rozenberg, and A. Salomaa, editors, Discrete Mathematics and Computer Sci-
ence, pages 207–222. Editura Academiei Române, 2014.
4. L. Kari and K. Mahalingam. Involutively bordered words. International Journal of
Foundations of Computer Science, 18(05):1089–1106, 2007.
5. E. Czeizler, L. Kari, and S. Seki. On a special class of primitive words. Theoretical
Computer Science, 411:617–630, 2010.

16

Potrebbero piacerti anche