Exercise 1.4.3 (Vandermonde's identity)

(a)
By comparing the coefficient of z k in the polynomial identity k = 0 m + n ( m + n k ) z k = ( 1 + z ) m + n = ( 1 + z ) m ( 1 + z ) n = ( k = 0 m ( m k ) z k ) ( k = 0 n ( n k ) z k )

show that

i = 0 k ( m i ) ( n k i ) = ( m + n k ) .

(b)
Let 𝒰 and 𝒱 be disjoint sets containing m and n elements, respectively, and put 𝒮 = 𝒰 𝒱 . Show that the number of subsets 𝒜 of 𝒮 that contain k elements and that also have the property that 𝒜 𝒰 contains i elements is ( m i ) ( n k i ) . Interpret this identity combinatorially.
(c)
Show that for n 0 , k = 0 n ( n k ) 2 = ( 2 n n ) .

Answers

Proof.

(a)
All is said in the sentence. Let z be a variable (indeterminate). We calculate ( 1 + z ) m + n in two ways.

First

( 1 + z ) m + n = k = 0 m + n ( m + n k ) z k . (1)

Next

( 1 + z ) m + n = ( 1 + z ) m ( 1 + z ) n = ( i = 0 m ( m i ) z i ) ( j = 0 n ( n j ) z j ) = k = 0 m + n ( i = 0 k ( m i ) ( n k i ) ) z k ,

using the definition of multiplication of two polynomials, which gives

( i = 0 m a i z i ) ( j = 0 n b j z j ) = k = 0 m + n ( i = 0 k a i b k i ) z k .

So

( 1 + z ) m + n = k = 0 m + n ( i = 0 k ( m i ) ( n k i ) ) z k . (2)

The comparison of the two polynomials identities (1) and (2) gives

i = 0 k ( m i ) ( n k i ) = ( m + n k ) .

(Vandermonde’s identity.)

Note: ( m i ) = 0 if i > m , and ( n k i ) = 0 if k i > n . Therefore we can keep only the indices i such that i 0 , i k n , and i m , i k , so

i = max ( 0 , k n ) min ( m , k ) ( m i ) ( n k i ) = ( m + n k ) .

(b)
Here 𝒮 = 𝒰 𝒱 and = 𝒰 𝒱 , where | 𝒰 | = m , | 𝒱 | = n , so | 𝒮 | = m + n .

Now consider the set M i of subsets 𝒜 of 𝒮 containing k elements, such that 𝒜 𝒰 contains i elements, where 0 i k .

M i = { 𝒜 𝒫 ( S ) : | 𝒜 𝒰 | = i } .

(Imagine a box filled with m white balls and n black balls, and consider the draws of k balls simultaneously which contains i white balls, and so k i black balls).

If 𝒜 M i , then

𝒜 = ( 𝒜 𝒰 ) ( 𝒜 𝒱 ) , = ( 𝒜 𝒰 ) ( 𝒜 𝒱 ) .

Therefore | 𝒜 𝒱 | = k i .

Now we compute the cardinality of M i . Recall that we write 𝒫 k ( E ) for the set of subsets of E containing k elements.

Consider the map

φ { M i 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) 𝒜 ( 𝒜 𝒰 , 𝒜 𝒱 )

Since | 𝒜 𝒰 | = i and | 𝒜 𝒱 | = k i for every 𝒜 M i , the definition of φ makes sense. Moreover φ is a bijection: let

ψ { 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) M i ( R , S ) R S .

(Let ( R , S ) 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) . Then R S = and | R | = i , | S | = k i , thus | R S | = i + ( k i ) = k , and ( R S ) 𝒰 = R , where | R | = i , so R S M i .)

Then for every 𝒜 M i , ( ψ φ ) ( 𝒜 ) = ( 𝒜 𝒰 ) ( 𝒜 𝒱 ) = 𝒜 , so ψ φ = 1 M i , and for every ( R , S ) 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) ,

( φ ψ ) ( R , S ) = φ ( R S ) = ( ( R S ) 𝒰 , ( R S ) 𝒱 ) = ( R 𝒰 ) ( S 𝒰 ) , ( R 𝒱 ) ( S 𝒱 ) ) = ( R , S ) ,

because R 𝒰 , S 𝒱 and 𝒰 𝒱 = .

Since ψ φ = 1 M i and φ ψ = 1 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) , φ is bijective.

Therefore

| M i | = | 𝒫 i ( 𝒰 ) × 𝒫 k i ( 𝒱 ) | = | 𝒫 i ( 𝒰 ) | × | 𝒫 k i ( 𝒱 ) | = ( m i ) ( n k i ) .

(In less formal terms, we choose i white balls between m white balls, and k i black balls between n black balls, which gives ( m i ) ( n k i ) choices.)

This gives a new proof of the Vandermonde’s identity.

Note that 𝒮 = i = 0 k M i , since every element 𝒜 𝒮 satisfies | 𝒜 𝒰 | = i for some i [ [ 0 , k ] ] .

This union is disjoint because 0 i < i k M i M i = ( | 𝒜 𝒰 | = i and | 𝒜 𝒰 | = i are incompatible if i i ).

Therefore

( m + n k ) = | 𝒮 | = i = 0 k | M i | = i = 0 k ( m i ) ( n k i ) .

This is the Vandermonde’s identity.

(c)
If we take m = n and k = n in the Vandermonde’s identity, we obtain ( 2 n n ) = i = 0 n ( n i ) ( n n i ) = i = 0 n ( n i ) 2 ,

using the symmetry (1.10) ( n n i ) = ( n i ) .

User profile picture
2024-07-25 09:56
Comments