Derek's sequence

Probability Level pending

Consider the sequence of complex numbers defined by the recurrence a k + 1 = a k i a k + i for k 1 , a_{k+1} = \frac{a_k - i}{a_k + i} \text{ for } k \geq 1, and let a 1 a_1 be any complex number such that a n a_n is defined for all positive integers n n . For each a 1 a_1 , let S ( a 1 ) S(a_1) be the set of all possible values of k = 1 p prime ( 1 ) k a k \displaystyle \sum_{k=1}^{p \text{ prime} } (-1)^ka_k , where p p ranges over the prime numbers. What is max a S ( a ) ? \max_{a} |S(a)| ?

This problem is posed by Derek K .

Details and assumptions

In this problem, i i is the imaginary unit, satisfying i 2 = 1 i^2=-1 .

S | S | denotes the number of distinct elements of set S S .


The answer is 4.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Aleksey Korobenko
Aug 12, 2013

Transition from a k a_k to a k + 1 a_{k+1} is given by the Mobius transformation with a matrix: M = ( 1 i 1 i ) M=\left(\begin{array}{cc} 1 & -i\\ 1 & i\end{array}\right) The composition of Mobius transformations is, again, Mobius transformation, with a matrix given by the matrix product of initial transformation matrices. Therefore, a k a_k to a k + 3 a_{k+3} transformation is given by: M 3 = M 3 = ( 1 i 1 i ) × ( 1 i 1 i ) × ( 1 i 1 i ) = ( 1 i 1 i 1 + i 1 i ) × ( 1 i 1 i ) = ( 2 2 i 0 0 2 2 i ) = ( 2 2 i ) ( 1 0 0 1 ) M_3=M^3=\left(\begin{array}{cc} 1 & -i\\ 1 & i\end{array}\right)\times\left(\begin{array}{cc} 1 & -i\\ 1 & i\end{array}\right)\times\left(\begin{array}{cc} 1 & -i\\ 1 & i\end{array}\right)=\left(\begin{array}{cc} 1-i & 1-i\\ 1+i & -1-i\end{array}\right)\times\left(\begin{array}{cc} 1 & -i\\ 1 & i\end{array}\right)=\left(\begin{array}{cc} 2-2i & 0\\ 0 & 2-2i\end{array}\right)=(2-2i)\left(\begin{array}{cc} 1 & 0\\ 0 & 1\end{array}\right) Since overall multiplier is not important for Mobius transforms, we see, that the composition of three transforms is an identity, and a k a_k is periodic with a period of a divisor of three, i.e. 1 or 3. The alternating series partial sums, used to compute members of S can therefore be written using just 3 terms: a 1 , a 1 a 2 , a 1 a 2 + a 3 , a 2 + a 3 , a 3 , 0 , a 1 , . . . a_1,a_1-a_2,a_1-a_2+a_3,-a_2+a_3,a_3,0,a_1,... i.e. it's periodic with a period of a divisor of 6. Quick check with a 1 = 2 i a_1=2i gives a 2 = 1 / 3 a_2=1/3 and a 3 = 0.8 0.6 i a_3=-0.8-0.6i , which gives partial sums: 2 i , 1 / 3 + 2 i , 17 / 15 + 7 / 5 i , 17 / 15 3 / 5 i , 4 / 5 3 / 5 i , 0 , 2 i , . . . 2i,-1/3+2i,-17/15+7/5i,-17/15-3/5i,-4/5-3/5i,0,2i,... i.e. for at least one a 1 a_1 these series of partial sums is periodic with a period of 6, and all the terms within a period are different.

For a given prime p p the partial sum would be determined by a remainder after division of p by 6. For p < 6 p<6 the reminders are: 2 , 3 , 5 2, 3, 5 . For p > 6 p>6 the remainders have to be coprime with 6, otherwise, according to the equality p = 6 q + r p=6q+r , p p has to divide by their common divisor, which is less then 6, and, therefore, not equal to p. Therefore the remainders 0, 2, 3 and 4 are impossible for p > 6 p>6 , or, there are no remainders 0 and 4 for any p p . Writing these remainders down for the first prime numbers 2 , 3 , 5 , 1 , . . . 2, 3, 5, 1, ... we see that other four possibilities are met for some primes. Answer: 4

Moderator note:

Yes, this problem is about a periodic Mobius transformation. Though one does not need to know anything about Mobius transformations to solve it.

Jimmy Kariznov
Aug 11, 2013

Let f ( z ) = z i z + i f(z) = \dfrac{z-i}{z+i} for all z i z \neq -i . So, a k + 1 = f ( a k ) a_{k+1} = f(a_k) for all k 1 k \ge 1 . Then, f ( f ( z ) ) = i z + 1 z 1 f(f(z)) = -i\dfrac{z+1}{z-1} and f ( f ( f ( z ) ) ) = z f(f(f(z))) = z hold for all z i , 1 z \neq -i,1 .

If a k = i , 1 a_k = -i, 1 then a k + 1 = f ( a k ) a_{k+1} = f(a_k) or a k + 2 = f ( f ( a k ) ) a_{k+2} = f(f(a_k)) will be undefined.

Thus, a k i , 1 a_k \neq -i, 1 , and so, a k + 3 = f ( f ( f ( a k ) ) ) = a k a_{k+3} = f(f(f(a_k))) = a_k for all k 1 k \ge 1 .

Let s n = k = 1 n ( 1 ) k a k s_n = \displaystyle\sum_{k = 1}^{n}(-1)^k a_k . So, S ( a 1 ) = { s p p is prime } S(a_1) = \{s_p \ | \ p \ \text{is prime}\} . Note that: s n + 6 s n = ( 1 ) n ( a n + 1 + a n + 2 a n + 3 + a n + 4 a n + 5 + a n + 6 ) s_{n+6} - s_n = (-1)^n(-a_{n+1} + a_{n+2} - a_{n+3} + a_{n+4} - a_{n+5} + a_{n+6})

= ( 1 ) n ( a n + 1 + a n + 2 a n + 3 + a n + 1 a n + 2 + a n + 3 ) = 0 = (-1)^n\left(-a_{n+1} + a_{n+2} - a_{n+3} + a_{n+1} - a_{n+2} + a_{n+3}\right) = 0 .

Hence, s n + 6 = s n s_{n+6} = s_n for all n 1 n \ge 1 . So, if n r ( m o d 6 ) n \equiv r \pmod{6} then s n = s r s_n = s_r .

For all primes p p we have p = 2 , 3 p = 2,3 or p 1 , 5 ( m o d 6 ) p \equiv 1,5 \pmod{6} .

Therefore, s p { s 1 , s 2 , s 3 , s 5 } s_p \in \{s_1,s_2,s_3,s_5\} for all primes p p .

Hence, S ( a 1 ) { s 1 , s 2 , s 3 , s 5 } S(a_1) \subseteq \{s_1,s_2,s_3,s_5\} , and thus, S ( a ) 4 |S(a)| \le 4 for all a a .

Now, we must check that S ( a ) = 4 |S(a)| = 4 is in fact attainable for some a a .

For a 1 = 2 a_1 = 2 we have a 2 = 3 5 4 5 i a_2 = \tfrac{3}{5} - \tfrac{4}{5}i , and a 3 = 3 i a_3 = -3i . Hence,

s 2 = a 1 + a 2 = 7 5 4 5 i s_2 = -a_1+a_2 = -\tfrac{7}{5} - \tfrac{4}{5}i , s 3 = a 1 + a 2 a 3 = 7 5 + 11 5 i s_3 = -a_1+a_2-a_3 = -\tfrac{7}{5} + \tfrac{11}{5}i ,

s 5 = a 3 = 3 i s_5 = -a_3 = 3i , and s 7 = s 1 = a 1 = 2 s_7 = s_1 = -a_1 = -2 . So, S ( 2 ) = 4 |S(2)| = 4 .

Since S ( a ) 4 |S(a)| \le 4 for all a a and S ( 2 ) = 4 |S(2)| = 4 , we have max a S ( a ) = 4 \displaystyle\max_{a} |S(a)| = \boxed{4} .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...