Beruflich Dokumente
Kultur Dokumente
n
k
Algebraic solution:
X X
X
X n X n X
n
n
n
n
2
k
=
k
+ (nk)
=
k
+ (nk)
=
n
= n2n .
k
k
nk
k
k
k
k
So
X n
k
= n2n1 .
k
k
kk k
Algebraic solution: Since k 2 = 2 k2 + k1 , by the previous result the sum
has value
n(n + 1)2n2 .
Combinatorial solution: The sum counts the number of ways to choose a
committee of arbitrary size k, a president and a vice-president who may
or may not be the same person, but who must be on the committee from
n people. If the president and VP are the same then there are n choices
of pres and VP and 2n1 choices of committee; if they are different, there
are n2 n choices of leaders and 2n2 choices of committee. The answer
follows.
The solution for x3 is similar.
n
k 2k
Algebraic solution:
X
k n
0 = (1 1) =
(1)
k
k
2k =
X n
k
Adding these two equations, the odd k terms on the right cancel out, while
the even terms are doubled. Dividing out the factor of 2, we get
X n
k1
2
=
.
2k
k
X n
4k
X n
+
n
n
n
+
+
4k + 1
4k + 2
4k + 3
n
n
n
0=
4k
4k + 1
4k + 2
4k + 3
k
X n
n
n
n
(1 + i)n =
+i
i
4k
4k + 1
4k + 2
4k + 3
k
X n
n
n
n
n
i
+i
.
(1 i) =
4k
4k + 1
4k + 2
4k + 3
k
n 2
k
n
k
m
rk
n+m
r
2n = 1.
nk =
f (k, r),
n=1
which would solve the problem. But the identity itself is easily proven by
induction.
(This is Putnam 2005 B4.)
Please peruse the three solutions at http://209.222.159.179/amc/a-activities/a7problems/putnam/-pdf/2005s.pdf, as well as the solution below.
We will produce a bijection between the set of n-tuples of integers whose
absolute values sum to at most m and the set of m-tuples whose absolute
values sum to at most n.
Suppose x1 , x2 , . . . , xn are integers whose absolute values sum to at most
m. We associate to this n-tuple of integers a collection of triples (pi , qi , i ),
defined as follows. Let xn1 , xn2 , . . . , xnr denote the nonzero integers xi ,
taken in order (so n1 < n2 < < nr ). For 1 i r, let pi = ni be the
Pi
index of the i-th nonzero integer, and let qi = k=1 |xn1 | be the sum of
the first i nonzero integers. Finally, choose i = 1 to have the same sign
as xni .
4
The r triples (pi , qi , i ) have the following properties, which we call properties Pm,n . The values pi and qi are positive integers, and i = 1.
Both pi and qi are increasing sequences; that is, p1 < p2 < < pr and
q1 < q2 < < qr . Also, we have the bounds pr n, qr m.
Given r triples with these properties, we can see that there is a unique
x1 , x2 , . . . , xn to which it is associated. Indeed, we require that xk = 0 for
all k not equal to any of the pi . For each pi , we take xpi = i (qi qi1 ),
with the convention that q0 = 0. This is possible because pi n and the
pi are all distinct. Since qi > qi1 all these integers are nonzero. And the
sum of their absolute values is exactly qr m. Thus we have produced a
bijection between the sequences of integers xi satisfying the bound in the
problem and the sequences of triples satisfying Pm,n .
It remains to show that the number of sequences of triples satisfying Pm,n
is the same as the number of sequences of triples satisfying Pn,m . But this
is clear, as interchanging the ps and the qs produces a bijection between
these two sets.