MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The system has the form

$$ (n-2)f_n^{(1)}=n(f_{n-1}^{(1)}+1), $$ $$ (n-2\cdot 2) f_n^{(2)}=n(f_{n-1}^{(2)}+f_{n-1}^{(1)}), $$ $$ \ldots $$ $$ (n-2k)f_n^{(k)}=n(f_{n-1}^{(k)}+f_{n-1}^{(k-1)}), $$ for the unknown sequences $f_n^{(1)},f_n^{(2)},\ldots,f_n^{(k)}$ with the initial conditions $f^{(i)}_k=0,$ for all $k=0,1,\ldots,2i.$

By direct calculation I have got

$$ f_n^{(1)}=n(n-2), $$ $$ f_n^{(2)}=\frac{1}{2!} (n-4) { n \choose 2} (3n-7), $$ $$ f^{(3)}_n=\frac{1}{3!} (n-6) { n \choose 3} \left( 19{n}^{2}-141n+254 \right), $$ $$ f^{(4)}_n=\frac{1}{4!} (n-8) { n \choose 4} \left( 211{n}^{3}-3258{n}^{2}+16481n-27306 \right), $$ $$ f^{(5)}_n=\frac{1}{5!} (n-10) { n \choose 5} \left( 3651{n}^{4}-96550{n}^{3}+946185{n}^{2}-4071950n+ 6492024 \right) $$

Question. What is a general expression for $f_n^{(i)}$?

The ordinary generating function for the above sequences has the form $$ G(f_n^{(1)},z)={\frac {{z}^{3} \left( 3-z \right) }{ \left( 1-z \right) ^{3}}}=3{z}^{3}+8{z}^{4}+15{z}^{5}+\cdots $$ $$ G(f_n^{(2)},z)={\frac {z^5(-3{z}^{3}+16{z}^{2}-35z+40)}{ \left( 1-z \right) ^{5}}}, $$ $$ G(f_n^{(3)},z)={\frac {{z}^{7} \left( -40{z}^{5}+288{z}^{4}-897{z}^{3}+1575{z} ^{2}-1701z+1155 \right) }{ \left( 1-z \right) ^{7}}} $$

share|cite|improve this question
Where this system comes from? – Daniele Zuddas Apr 27 '12 at 18:57
You might want to restate your question without typos. As it reads, I get $f_n^{(1)}=-n/2$, which is clearly not what you intend. – Michael Renardy Apr 27 '12 at 19:00
to Michael. Sorry, corrected – Melania Apr 27 '12 at 19:07
I don't know that this'll help, but the sequence 3,19,211,3651 shows up at with next term 90921. – Barry Cipra Apr 27 '12 at 19:11
to Barry. Thanks. I already saw it but it doesnt help. – Melania Apr 27 '12 at 19:20
up vote 4 down vote accepted

This is not a full solution, but it reduces the system to another recurrence relation which involves only coefficients. Let $$f_n^{(k)}=a_0^k n(n-1)...(n-k+1)+a_1^k n(n-1)...(n-k)+...+a_k^k n(n-1)...(n-2k+1).$$ Then the above system reduces to $$a_i^k={a_i^{k-1}\over i-k}$$ for $i < k$ and $a_k^k$ is determined by the condition that $f_{2k}^{(k)}=0$. Clearly the determination of $a_k^k$ is the difficult part. But if the sequence 3,19 etc. shows up elsewhere, that should be suggestive.

Addendum: If all the $a_i^k$ are expressed in terms of $a_i^i$, then the recurrence relation for the $a_k^k$ becomes the same as for the Taylor coefficients of the reciprocal Bessel function mentioned in Barry Cipra's link.

share|cite|improve this answer
Thanks. It is good idea to change basis – Melania Apr 28 '12 at 5:21

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.