09.12.2012 Views

Concrete mathematics : a foundation for computer science

Concrete mathematics : a foundation for computer science

Concrete mathematics : a foundation for computer science

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

(Attention, <strong>computer</strong><br />

scientists:<br />

The stack interpretation<br />

now applies<br />

with respect to an<br />

m-ary operation,<br />

instead of the binary<br />

multiplication<br />

considered earlier.)<br />

7.5 CONVOLUTIONS 347<br />

whose total sum is +l . Such sequences can be called m-Raney sequences. If<br />

there are k occurrences of (1 - m) and mn + 1 - k occurrences of +l , we have<br />

k(l-m)+(mn+l-k) = 1,<br />

hence k = n. There are (“t+‘) sequences with n occurrences of (1 - m) and<br />

mn + 1 - n occurrences of +l, and Raney’s lemma tells us that the number<br />

of such sequences with all partial sums positive is exactly<br />

mn+l -<br />

I<br />

-<br />

mn<br />

=<br />

1<br />

( n > mn+l ( n > (m-l)n+l’<br />

(7.66)<br />

So this is the number of m-Raney sequences. Let’s call this a Fuss-Catalan<br />

number Cim,“‘, because the sequence (&“‘) was first investigated by N.I.<br />

Fuss [log] in 1791 (many years be<strong>for</strong>e Catalan himself got into the act). The<br />

ordinary Catalan numbers are C, = Cr’.<br />

Now that we k:now the answer, (7.66), let’s play “Jeopardy” and figure<br />

out a question that leads to it. In the case m = 2 the question was: “What<br />

numbers C, satisfy the recurrence C, = xk CkCnPiPk + (n = O]?” We will<br />

try to find a similar question (a similar recurrence) in the general case.<br />

The trivial sequence (+1) of length 1 is clearly an m-Raney sequence. If<br />

we put the number (1 -m) at the right of any m sequences that are m-Raney,<br />

we get an m-Raney sequence; the partial sums stay positive as they increase<br />

to +2, then +3, . . . , fm, and fl . Conversely, we can show that all m-Raney<br />

sequences (ae, . . . , ~a,,) arise in this way, if n > 0: The last term a,,,,, must<br />

be (1 - m). The partial sums sj = a0 +. . + aj- 1 are positive <strong>for</strong> 1 < j 6 mn,<br />

and s,, = m because s,, + a,,,,, = 1. Let kl be the largest index 6 mn such<br />

that Sk, = 1; let k2 be largest such that skz = 2; and so on. Thus ski = j<br />

and sk > j, <strong>for</strong> ki cc k 6 mn and 1 < j 6 m. It follows that k, = mn, and<br />

we can verify without difficulty that each of the subsequences (ae, . . . , ok, -I),<br />

(ok,, . . . , okJPi), . . . , (ok,,-, , . . . , ok,,, -1) is an m-Raney sequence. We must<br />

have kl = mnl + 1, k2 - kl = mn2 + 1, . . . , k, - k,_l = mn, + 1, <strong>for</strong><br />

some nonnegative integers nl, n2, . . . , n,.<br />

There<strong>for</strong>e (“‘t’-‘) & is the answer to the following two interesting ques-<br />

tions: “What are the numbers Cim’ defined by the recurrence<br />

<strong>for</strong> all integers n?” “If G(z) is a power series that satisfies<br />

G(z) = zG(z)" + 1, (7.68)<br />

what is [z”] G(z)?”

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!