Farey Sequences

advertisement
Farey Sequences
Definition
A Farey sequence F (n) is a complete sequence of reduced form rational fractions in
 0 a
a p 1
  with b j  n for j . The
the range 0..1 of the form F (n)    0  .....
b p 1
 1 b0
sequence is complete in the sense that, given a Farey sequence as defined above, we
aj
p a j 1
p
 
can find no rational fraction
with q  n such that
.
b j q b j 1
q
Theorem
Given a Farey sequence F (n) , the following equality holds for 0  j  p :
a j 1  a j 1
b j 1  b j 1

aj
bj
where a1  0 , a p 1  1 , b1  1 , b p 1  1 .
Proof
Proof is by induction. For n  2 , we have the trivial Farey sequence
 0 1 1
F (2)   , ,  where the above equality holds for j  0 . Given a Farey sequence
 1 2 1
F (n) , the sequence F (n  1) comprises a merge of F (n) and the sequence
kp 
 k
k
ki
F ' (n  1)   0 , 1 ..... n 1 , 0  k i  n  1 : i . If some
 F ' (n  1) is not
n 1
 n  1 n  1 n  1
k'
in reduced form then it can be represented as a reduced form rational fraction
with
n'
k '  n'  n . But this fraction will already be a member of F (n) since F (n) is
complete. Thus only reduced form fractions will be members of F ' (n  1) and will be
merged with F (n) . The merge is such that all elements of F ' (n  1) are inserted in
their correct position in the sequence F (n) in order to produce F (n  1) . (This
observation enables an algorithm for the construction of F (n) for any n to be
designed.)
The complete proof will use the following 4 lemmas.
Lemma 1
There exists integers a , b , k and n such that
k a k 1
 
for n  2, 0  k  n  1, a  b
n b
n
Proof
Choose a  k , b  n  1 . It is easy to verify that
k
k
k 1


.
n n 1
n
Lemma 2
After merging, no two consecutive members of F ' (n  1) will be consecutive
aj
members of F (n  1) . Thus we can always find some fraction
such that
bj
aj
ki
k

 i 1 and where a j  b j  n .
n 1 bj n 1
Proof
This follows directly from Lemma 1 since k i 1  k i  1 .
Lemma 3
Given a Farey sequence
 0 a
a p 1
F (n)    0  .....
  , for any 1  j  p ,
b p 1
 1 b0
a j 1b j  b j 1 a j  1 .
Proof
This follows directly from the theorem
a j 1  a j 1
b j 1  b j 1

aj
bj
a j 1  a j 1
b j 1  b j 1

aj
bj
.
 a j b j 1  a j b j 1  a j 1b j  a j 1b j  a j b j 1  a j 1b j  a j 1b j  a j b j 1
This applies for all 0  j  p and can be applied successively to create a sequence
a0b1  a1b0  a1b0  a0 b1  ......a j 1b j  a j b j 1  .......a p1b p  a p b p1 .
a1  0, a0  1, b1  1, b0  n  a0 b1  a1b0  1 and hence a j 1b j  a j b j 1  1
for 1  j  p .
But
aj
a j 1
k
ki
 i 
are
 F ' (n  1) such that
b j n  1 b j 1
n 1
aj
a j 1
consecutive members of F (n  1) and where
and
are consecutive members
bj
b j 1
From Lemma 2 we have some
of F (n) and
ki
is in reduced form.
n 1
Consider the rational fractions
k i  a j 1
,
ki  a j
n  1  b j 1 n  1  b j
. We can prove the following
lemma.
Lemma 4
k i  a j 1
n  1  b j 1

aj
bj
,
ki  a j
n 1 bj

a j 1
b j 1
Proof
a j 1
b j 1

k i  a j 1
ki
k
 k i b j 1  a j 1 (n  1)  k i (n  1)  a j 1 (n  1)  k i (n  1)  k i b j 1  i 
n 1
n  1 n  1  b j 1
Assume
k i  a j 1
n  1  b j 1
k i  a j 1
n  1  b j 1

aj
bj

aj
bj
 k i b j  a j 1b j  a j (n  1)  a j b j 1
 k i b j  a j (n  1)  a j 1b j  a j b j 1  k i b j  a j (n  1)  1
which follows from Lemma 3. But
Since k i b j  a j (n  1)
k i  a j 1
n  1  b j 1

aj
bj
rational fraction
. Since
aj
ki

 k i b j  a j (n  1)  0
n 1 bj
is an integer, our original assumption is false and
k i  a j 1
n  1  b j 1
 F ( n) ,
k i  a j 1
n  1  b j 1

aj
bj
because there is no
aj
k
p
p
  i as F (n) is complete.
 F (n) such that
bj q n 1
q
The second equality can be proved using a similar technique
aj
bj

ki  a j
ki
k
 k i b j  a j (n  1)  k i (n  1)  k i b j  k i (n  1)  a j (n  1)  i 
n 1
n 1 n 1 bj
Assume
ki  a j
n 1 bj
ki  a j
n 1 bj

a j 1
b j 1

a j 1
b j 1
 k i b j 1  a j b j 1  a j 1 (n  1)  a j 1b j
 a j 1b j  a j b j 1  a j 1 (n  1)  k i b j 1  a j 1 (n  1)  k i b j 1  1
a j 1
ki

 a j 1 (n  1)  k i b j 1  0
n  1 b j 1
which follows from Lemma 3. But
Since a j 1 (n  1)  k i b j 1 is an integer, our original assumption is false and
ki  a j
n 1 bj

a j 1
b j 1
rational fraction
. Since
ki  a j
n 1 bj
 F ( n) ,
ki  a j
n 1 bj

a j 1
b j 1
because there is no
k
p a j 1
p
as F (n) is complete.
 F (n) such that i  
n  1 q b j 1
q
This proves Lemma 4.
From Lemmas 3 and 4, the proof of the theorem follows directly.
ki  a j
n 1 bj

k i  a j 1
n  1  b j 1
a j 1
b j 1

aj
bj
 k i b j 1  a j b j 1  a j 1 (n  1)  b j a j 1  a j 1 (n  1)  k i b j 1  1
 k i b j  a j 1b j  a j (n  1)  a j b j 1  k i b j  a j (n  1)  1
k i b j  a j (n  1)  a j 1 (n  1)  k i b j 1 
a j  a j 1
ki

n  1 b j  b j 1
ki
inserted into F (n) obeys the equality
n 1
given in the Theorem. However, we still have to consider preceding values to check
whether the Theorem is generally applicable to F (n  1) .
This shows that the rational fraction
There are 2 possibilities:
where
a j 1
b j 1

aj
bj

a j 1
a j 1
aj
a j 1
ki
k
k

 i 1 
 i 
and
n  1 b j 1
b j 1 n  1 b j n  1 b j 1
a j 1 a j a j 1
, ,
are consecutive terms in sequence F (n) .
b j 1 b j b j 1
Taking the first possibility,
a j 1  a j 1
b j 1  b j 1
a j  a j 1
b j  b j 1


aj
bj
a j 1
b j 1

aj

bj
a j 1
ki

n  1 b j 1
 a j 1  a j 1  pa j , b j 1  b j 1  pb j for some integer p .
ki
 a j  a j 1  k i , b j  b j 1  (n  1) because b j  b j 1  2(n  1)
n 1
 a j 1  a j  pa j  k i , b j 1  b j  pb j  (n  1)
 a j 1  k i  a j (1  p), b j 1  (n  1)  b j (1  p)

a j 1  k i
b j 1  (n  1)

aj
bj
ki
is the third term in the triplet. An identical
n 1
ki
proof shows that the equality holds when
is the first term of the triplet and so
n 1
ki  a j 2
a j 1

(n  1)  b j  2 b j 1
Thus the equality holds for when
Taking the second possibility
a j 1  a j 1
b j 1  b j 1
a j 1  a j
b j 1  b j
a j  a j 1
b j  b j 1

aj
bj
a j 1
b j 1

aj
a j 1
k i 1
k

 i 
,
n  1 b j n  1 b j 1
 a j 1  a j 1  pa j , b j 1  b j 1  pb j for some integer p .

k i 1
 a j 1  a j  k i 1 , b j 1  b j  (n  1)
n 1

ki
 a j  a j 1  k i , b j  b j 1  (n  1)
n 1
k i  k i 1  a j 1  2a j  a j 1  k i , n  1 
 k i  k i 1  a j (2  p ), n  1 

b j 1  2b j  b j 1
2
b j (2  p)
2
k i  k i 1 a j

2(n  1) b j
Thus the equality holds in this case also.
We have thus proved by induction that if the equality holds for F (n) , then it must
hold for F (n  1) which completes the proof of the Theorem.
Download