13.11.2014 Views

proofs

proofs

proofs

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

208 Identities versus bijections<br />

thought should convince you that this is just the number of ways to write n<br />

as a sum<br />

n = n 1 · 1 + n 2 · 2 + n 3 · 3 + . . .<br />

= 1 + . . . + 1 + 2 + . . . + 2 + 3 + . . . + 3 + . . . .<br />

} {{ } } {{ } } {{ }<br />

n 1 n 2 n 3<br />

So the coefficient is nothing else but the number p(n) of partitions of n.<br />

Since the geometric series 1+x k +x 2k 1<br />

+. . . equals<br />

1−x<br />

, we have proved<br />

k<br />

our first identity:<br />

∏ 1<br />

1 − x k = ∑ p(n)x n . (4)<br />

k≥1<br />

n≥0<br />

6 = 5 + 1<br />

6 = 3 + 3<br />

6 = 3 + 1 + 1 + 1<br />

6 = 1 + 1 + 1 + 1 + 1 + 1<br />

Partitions of 6 into odd parts: p o(6) = 4<br />

7 = 7<br />

7 = 5 + 1 + 1<br />

7 = 3 + 3 + 1<br />

7 = 3 + 1 + 1 + 1 + 1<br />

7 = 1 + 1 + 1 + 1 + 1 + 1 + 1<br />

7 = 7<br />

7 = 6 + 1<br />

7 = 5 + 2<br />

7 = 4 + 3<br />

7 = 4 + 2 + 1.<br />

The partitions of 7 into odd resp. distinct<br />

parts: p o(7) = p d (7) = 5.<br />

1<br />

What’s more, we see from our analysis that the factor<br />

1−x<br />

accounts for<br />

k<br />

1<br />

the contribution of k to a partition of n. Thus, if we leave out<br />

1−x<br />

from k<br />

the product on the left side of (4), then k does not appear in any partition<br />

on the right side. As an example we immediately obtain<br />

∏<br />

i≥1<br />

1<br />

1 − x 2i−1 = ∑ n≥0<br />

p o (n)x n , (5)<br />

where p o (n) is the number of partitions of n all of whose summands are<br />

odd, and the analogous statement holds when all summands are even.<br />

By<br />

∏<br />

now it should be clear what the n-th coefficient in the infinite product<br />

k≥1 (1 + xk ) will be. Since we take from any factor in (3) either 1 or x k ,<br />

this means that we consider only those partitions where any summand k<br />

appears at most once. In other words, our original product (1) is expanded<br />

into<br />

∏<br />

+ x<br />

k≥1(1 k ) = ∑ p d (n)x n , (6)<br />

n≥0<br />

where p d (n) is the number of partitions of n into distinct summands.<br />

Now the method of formal series displays its full power. Since 1 − x 2 =<br />

(1 − x)(1 + x) we may write<br />

∏<br />

k≥1(1 + x k ) = ∏ k≥1<br />

1 − x 2k<br />

1 − x k = ∏ 1<br />

1 − x 2k−1<br />

k≥1<br />

since all factors 1 − x 2i with even exponent cancel out. So, the infinite<br />

products in (5) and (6) are the same, and hence also the series, and we<br />

obtain the beautiful result<br />

p o (n) = p d (n) for all n ≥ 0. (7)<br />

Such a striking equality demands a simple proof by bijection — at least that<br />

is the point of view of any combinatorialist.

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

Saved successfully!

Ooh no, something went wrong!