6.2 Binomial Coefficients and Counting Methods 303
Next, a combinatorial identity.
Example.Letmandnbe two integers,m≤n− 21. Prove the identity
n− 21
∑
k=m
(
n
2 k+ 1
)(
k
m
)
= 2 n−^2 m−^1
(
n−m− 1
m
)
.
Solution.The solution is a Fubini-type argument. Consider the setPof pairs(A, B),
whereAis a subset of{ 1 , 2 ,...,n}with an odd number of elementsa 1 <a 2 <···<
a 2 k+ 1 andBis a subset of{a 2 ,a 4 ,...,a 2 k− 2 ,a 2 k}withmelementsb 1 <b 2 <···<bm.
For a givenkthere are
( n
2 k+ 1
)
such subsetsA, and for eachAthere are
(k
m
)
subsetsB,
so the left-hand side of the identity is the number of elements ofPcounted by choosing
Afirst.
Let us count the same number choosingBfirst. Note that if(A, B)∈P, thenB
contains no pairs of consecutive numbers. More precisely,B ={b 1 ,b 2 ,...,bm}⊂
{ 2 , 3 ,...,n− 1 }withbi+ 1 −bi≥2.
FixB 0 , a set with this property. We want to count the number of pairs(A, B 0 )inX.
Choosec 0 ,c 1 ,...,cmsuch that
1 ≤c 0 <b 1 ,b 1 <c 1 <b 2 ,...,bi<ci<bi+ 1 ,...,bm<cm≤n.
Then for any subsetEof{ 1 , 2 ,...,n}{c 0 ,b 1 ,c 1 ,b 2 ,...,bm,cm}there is a uniqueA
such that(A, B 0 )∈PandE⊂A.
Indeed, if(A, B 0 )∈PandE⊂Awe have to decide whichci’s are inA. Since
the setDi ={x ∈A | bi <x<bi+ 1 }must contain an odd number of elements
for each 0 ≤ i ≤ m+1 (withb 0 =0,bm+ 1 = n+ 1 ), and the setDi is either
{x∈E|bi<x<bi+ 1 }or{x∈E|bi<x<bi+ 1 }∪{ci}, the parity condition on the
cardinality ofDidecides whethercibelongs toA. It is now clear that the number of pairs
(A, B 0 )inPis the same as the number of subsets of{ 1 , 2 ,...,n}{c 0 ,b 1 ,...,bm,cm},
and the latter is 2n−^2 m−^1.
How many subsetsBwithmelements of{ 2 , 3 ,...,n− 1 }do not contain consec-
utive numbers? IfB={b 1 <b 2 <···<bm}is such a set, letB′={b 1 − 1 ,b 2 −
2 ,...,bm−m}. It is easy to see thatB′is an (arbitrary) subset of{ 1 , 2 ,...,n−m− 1 }
withmelements, and for each such subsetB′ ={b′ 1 <b′ 2 <···<bm′}, by letting
bi=b′i+i, we obtain a setBas above. Hence the number of suchB’s is
(n−m− 1
m
)
, and by
choosingBfirst we count the number of elements inPas 2n−^2 m−^1
(n−m− 1
m
)
. The identity
is proved.
Using similar ideas solve the following problems.
879.Find in closed form
1 · 2
(
n
2
)
+ 2 · 3
(
n
3
)
+···+(n− 1 )·n
(
n
n