Previous: 3 Confrac expansion of
Up: Patterns Occurring During the
Next: 5 Relations between lengths
Definition 4.1
Let us define a pattern as a palindromic sequence

where

.
The associated matrix

is symmetrical and we define it's signature
and it's "whole fingerprint" as:
Notation 4.2
When convenient, signature and fingerprint
will be expanded using the corresponding digits of

.
This will be denoted as:
 |
(4.1) |
where, for example,

is the second binary digit of

and

. Occasionally,

will be used.
Definition 4.3
For a quadratic integer

, we say that it belongs to a pattern

when
Proof.
We have

.
Since

we must have

and similarly for

. The second part holds from
Notation 4.5
The notation

is efficient, and
not too cumbersome since

characterizes the sequence

.
For a given

, not a perfect square, the pattern associated
with

will be denoted by

(first kind), and notations
3.10,
3.12 completed
by:
Accordingly, when

, the pattern associated with

will be denoted by

(second kind), and following
notations will be used:
It should be noticed that some patterns

are of both kinds
(but not for the same

).
Corollary 4.6 (An Euler's result, proven by Lagrange)
. A method to solve Pell's
equation follows from proposition
4.4. Using

, we have
If

is even, then

is the
fundamental solution of

. Otherwise, this solution
is associated with

.
Theorem 4.7
There are

(over the

possibilities) signatures that actually occur for a pattern. Moreover,
these signatures can be grouped into three classes that depend only
of the existence and the parity of the middle quotient of the pattern
:
 |
(4.2) |
Proof.
It can be checked that, if

and

, then (using notations of definition
4.2)
we have:
where computations are done in

(and not in

!).
By recurrence, a pattern of even length is obtained when starting
from the empty pattern, i.e. from

and a pattern of odd length when starting from

where

is the middle quotient of the pattern. Starting from the
corresponding signatures (bold-faced in
4.2)
and using iteratively
4.3, we obtain the
three given classes.
Remark 4.8
The action of
4.3 over the odd middle class
can be summarized by F
IG. 4.1.
When lengthening a pattern by

,
one moves along radiuses when

is even and along circles when

is odd (clockwise along the inner circle and counter-clockwise
along the outer).
FIG. 4.1:
The odd middle class.
|
|
Proposition 4.9
There are

fingerprints that actually occur. The action induced
by

group them into three
transitive classes, depending on the existence and parity of the middle
quotient. The cardinal of the none, even and odd class are respectively

,

and

, leading to the following
rules (stated using notations
4.1):
 |
(4.4) |
Proof.
Exhaustion by formal computing, following the same guideline as in
the preceeding theorem. It should be noticed that the

are equalities in

while the

are only congruences.
Theorem 4.10
For a given pattern, the

such that equation

defines a quadratic integer

are exactly the positive elements
of the arithmetic progression:
 |
(4.5) |
except perhaps from the smallest. The discriminant of all the

take four values modulo

, except from the signature

where it takes only two values. According to the signature of

,
these values of

are given by:
 |
(4.6) |
Moreover, for a pattern whose signature is one of these four :

,

or

, all
the quotients

such that

verifies additionally

(and therefore such that

has an associated

) are given
by
4.5 where

has to be set according
to the last column of
4.6 (where

is as in
4.1).
Proof.
Since

, we have

(Bezout). Substituting

into equation

,
i.e. into

, we obtain

and therefore

since

,

are coprime and

is a quadratic integer.
For the remaining results, a computer aided proof is the easiest.
We consider the first five binary digits of
, i.e.
and refer these
,
,
(
) as
"the" digits. The signature fixes the value of
,
determining the value of
and therefore the exact
value of
.
For each signature, equation
can be iteratively
solved modulo the successive powers of
, fixing at each step the
value of another digit. When using the usual tools of formal computing,
these values must be carefully expressed since
is false
outside of
! It is therefore convenient to express a sum
in
by
in
.
We iterate this process as long as the equation remains linear in
at least a variable. It happens that
depends only
on
and that the set
collapses to the same set of cardinal
for each
-uple in the
remaining digits, except for signature
where it appears two sets of cardinal two :
and
.
The same formulae show that, for any first kind integer,
depend only on
(and not on the digit
)
and give the value of
leading to
(when possible, i.e. for adequate signatures).
Corollary 4.11
The patterns of second kind are characterized
equivalently by
(i) the signature is not
(ii) the middle element is odd or none
Proposition 4.13
Assume that

is not a perfect
square,

and

. When

is not empty, we have
Proof.
The condition on

follows directly from theorem
4.10
by substituting

into
4.5. By
definition,

for an

. Thus

and the conclusion on

follows. When

is empty, we
have

and conclusion remains when using

instead of

.
Proposition 4.15
A "nothing but ones"
pattern is a pattern of second kind for all length

. The corresponding

and

are given by:
 |
(4.7) |
where

and

is the

-th
Fibonacci number. Additionally, the parameter

must be
even when

.
Proof.
First part follows from theorem
4.10, and we have

. Thus the signatures are given by:
and criterion of corollary
4.11 is satisfied
(when

,

is also a pattern
of first kind and the parity condition over

is here to
select integers of second kind). The

and

values follow from :
Proposition 4.16
An "any, one, one" pattern
is a pattern of second kind if and only if its middle element is odd
or none. In such a case, all the

associated to a

belong to the same pattern of first kind

,
where

. And we have
Conversely, let be given a pattern

whose signature is

or

and whose quotients

are all even and at least

. Then
it exists infinitely many

such that

,
and all the associated

share the above described pattern
of second kind

.
Proof.
First assertion follows from theorem
4.10. The
second is obtained by factoring

according to :
The value of the

pattern follows from :
 |
(4.8) |
proving that

is a fixed point of

when

is a fixed point of

.
Conversely: if
is as described, the existence of the
such that
follows from theorem 4.10.
Then proposition 4.13 shows that
is a quadratic integer. And, by 4.8,
is a fixed point of
.
Corollary 4.17
When

is "nothing
but ones" and

then

is "nothing but fours" and

.
Corollary 4.18
It exists an infinite number of quadratic integers of each kind having
a given period length.
Remark
The values of

given in proposition
4.15
and corollary
4.17 for a given value of

or

are quickly huge, and give no hint about the smallest

leading to a given period length.
Previous: 3 Confrac expansion of
Up: Patterns Occurring During the
Next: 5 Relations between lengths
douillet@ensait.fr
2004-02-06