Homework 5 | Due February 12 (Friday) 1

Read : Lax, Chapter 5, pages 44–54.

1. Let Sn denote the set of all permutations of {1, . . . , n}.

(a) Prove that sgn(π1 ◦ π2 ) = sgn(π1 ) sgn(π2 ).

(b) Prove that sgn(τ ) = −1 for all transpositions τ ∈ Sn .
(c) Let π ∈ Sn , and suppose that π = τk ◦ · · · ◦ τ1 = σ` ◦ · · · ◦ σ1 , where τi , σj ∈ Sn are
transpositions. Prove that k ≡ ` mod 2.

Solution:
(a) Recall that
P (xπ(1) , · · · , xπ(n) ) = sgn(π)P (x1 , · · · , xn ) (1)
Q
where P (x1 , · · · , xn ) is the descriminant i<j (xi − xj ).
We have:
P (xπ1 ◦π2 (1) , · · · , xπ1 ◦π2 (n) ) = sgn(π1 ◦ π2 )P (x1 , · · · , xn )
= sgn(π1 )P (xπ2 (1) , · · · , xπ2 (n) ) (2)
= sgn(π1 ) sgn(π2 )P (x1 , · · · , xn )
which proved the claim.
(b) Consider the transposition τ = (ij) of i and j = i + m with i < j, which sends
(1, 2, · · · , n) to:
(1, 2, · · · , i + m, · · · , i, · · · , n) (3)
Then in the permuted desriminant, the terms that change sign for moving i to the location
of i + m are:
xi − xi+1 , xi − xi+2 , · · · xi − xi+m (4)
i.e., xi was to the left of xi+1 , · · · , xi+m , but now is to the right to them. Similarly, the
terms that change sign for moving i + m to i are:

xi − xi+m , xi+1 − xi+m , · · · , xi+m−1 − xi+m (5)

They add up to a total of 2m − 1 (since the two lists share xi − xi+m ) terms, which means:

P (xτ (1) , · · · , xτ (n) ) = −P (x1 , · · · , xn ) (6)

We have then shown that sgn(τ ) = −1.

(c) Using conclusion of (a) and (b) together, we see that:

sgn(π) = sgn(τk ◦ · · · ◦ τ1 ) = sgn(τk ) · · · sgn(τ1 ) = (−1)k (7)

Similarly, we have:

sgn(π) = sgn(σl ◦ · · · ◦ σ1 ) = sgn(σl ) · · · sgn(σ1 ) = (−1)l (8)

We have (−1)k = (−1)l , or k ≡ l mod 2.

Math 8530 | Linear Algebra | Spring 2021 | M. Macauley

Homework 5 | Due February 12 (Friday) 2

2. Let f be a non-degenerate symmetric bilinear form over an n-dimensional vector space X.

That is, for all nonzero x ∈ X, there is some y ∈ X for which f (x, y) 6= 0. Consequently,
fixing any nonzero x ∈ X defines a nonzero dual vector

f (x, −) ∈ X 0 , f (x, −) : y 7−→ f (x, y).

(a) Prove that the map Lf : X → X 0 given by Lf : x 7→ f (x, −) is an isomorphism.

(b) Let x1 , . . . , xn be a basis for X. Express the dual basis `1 , . . . , `n in this form. That
is, find g for which Lg : xi 7→ `i .
(c) Show how to construct another basis y1 , . . . , yn such that f (xi , yj ) = δij .
(d) Conversely, prove that if BX = {x1 , . . . , xn } and BY = {y1 , . . . , yn } are sets of vectors
in X with f (xi , yj ) = δij , then BX and BY are bases for X.

(a) We prove that Lf sends a basis of X to a basis of X 0 . Let us pick a basis (x1 , · · · , xn )
for X, then we claim that (f (x1 , −), · · · , f (xn , −)) is a basis of X 0 . We only need to prove
they are linearly independent, since then the fact that there are n of them, and span a
n-dimensional subspace of X 0 , which also has dimension n means they span the whole
of X 0 . Now, given any linear combination of these linear functions that equals the zero
function:
n
X
λi f (xi , −) = 0 (not the zero vector, but the zero linear function!) (9)
i=1

Since f is bilinear, we have:

n
X n
X
f( λi xi , −) = 0 ⇒ f( λi xi , y) = 0, ∀y ∈ X (10)
i=1 i=1
Pn
But f
Pn is non-degenerate, i=1 λi xi = 0 (for otherwise, there must exist y ∈ X such that
f ( i=1 λi xi , y) 6= 0). Since (x1 , · · · , xn ) is a basis of X, we infer that λi = 0, i = 1, · · · , n.
This means that f (xi , −)’s are linearly independent.
(b) Since `1 , · · · , `n are dual to x1 , · · · , xn , we see that:

Lg (xi ) : xj 7→ `i (xj ) = g(xi , xj ) = δij , ∀i, j ∈ {1 · · · , n} (11)

Since g is bilinear, it is uniquely determined by the above equations (i.e., its values on a
set of basis).
(c) It is not difficult to see that f (xi , −)’s serve as the dual basis of yj ’s, and therefore
using conclusion of (b), we see that:

Lg : yi 7→ f (xi , −) (12)

Starting from xi ’s, we can reach yi ’s by:

L−1
g ◦ Lf : xi 7→ f (xi , −) 7→ yi (13)

Math 8530 | Linear Algebra | Spring 2021 | M. Macauley

Homework 5 | Due February 12 (Friday) 3
Pn
(d) We prove that BX and BY are linearly independent. Given any relationship i=1 λi xi =
0, we have that:
Xn n
X
f( λi xi , yj ) = λi δij = λj = 0, j = 1, · · · , n (14)
i=1 i=1

We see that BX are linearly independent and form a basis since dim X = n. The proof is
similar for BY .

3. Let f be a non-degenerate symmetric bilinear form over a real n-dimensional vector space,
where 1 + 1 6= 0.

(a) Show that there exists x1 ∈ X with f (x1 , x1 ) 6= 0.

(b) Let Z1 be the nullspace of f (x1 , −). Show that f restricted to Z1 is non-degenerate.
(c) Construct a basis {z1 , . . . , zn } for X that satisfies f (zi , zj ) = δij .

(a) Since f is non-degenerate, there exists x, y ∈ X s.t. f (x, y) 6= 0. If there does not
exist x1 ∈ X s.t. f (x1 , x1 ) 6= 0, we can say f (z, z) = 0, ∀z ∈ X. In this case, we seek the
following contradictin:

:0
:0 :0


f (x + x+y) = 
y, f (x,
x)
+ f (x, y) + f (y, x) + 
f (y,
 y)= 2f (x, y) = 0 (15)
  

which leads to f (x, y) = 0 since 1 + 1 6= 0. This leads to a contradiction.

(b) Z1 is the nullspace of the following linear map:

f (x1 , −) : X → R (16)

The map is onto since we know that f (x1 , x1 ) 6= 0. Therefore, we know that X/Z1 is
isomorphic to R and that dim Z1 = n − 1. It is also clear that x1 6∈ Z1 , and therefore
X = Z1 ⊕ Rx1 (here Rx1 denotes the 1 dimensional vector subspace spanned by x1 ).
Assume that f (−, −) is degenerate on Z1 , which is to say there exists y ∈ Z1 that:

f (y, z) = 0, ∀z ∈ Z1 (17)

However, since y ∈ Z1 , we also have f (y, x1 ) = f (x1 , y) = 0. This leads to the conclusion
that f (y, z) = 0, ∀z ∈ X. However, this is impossible since f is non-degenerate on X.
(c) With the help of (a) and (b), we can construct the required basis as follows. First
pick x1 as in (a), then we have:
x1 x1
f(p ,p )=1 (18)
f (x1 , x1 ) f (x1 , x1 )

This is possible only when f (x1 , x1 ) > 0, and therefore the problem is wrong! We should
change the problem to f (zi , zj ) = ±δij , and then:
x1 x1
f( p , p ) = ±1 (19)
| f (x1 , x1 )| | f (x1 , x1 )|

Math 8530 | Linear Algebra | Spring 2021 | M. Macauley

Homework 5 | Due February 12 (Friday) 4

x1
Take this √ to be our z1 , and denote the nullspace of f (z1 , −) by Z1 . By (b), f is
| f (x1 ,x1 )|
non-degenerate on Z1 , and therefore we can repeat the same process and find us z2 such
that f (z2 , z2 ) = ±1. At the same time, we know that since z2 ∈ Z1 , we have f (z1 , z2 ) = 0.
Continue this process until you get all zi ’s (formal proof requries induction).

4. Let f be a bilinear form over a vector space X with basis {x1 , x2 }.

(a) Assume f is alternating. Determine a formula for f (u, v) in terms of each f (xi , xj )
and the coefficients used to express u and v with this basis. [Pun intented!]
(b) Repeat Part (a) but assume that f is symmetric and f (x, x) = 0 for all x ∈ X.

(a) Denote u = u1 x1 + u2 x2 , v = v1 x1 + v2 x2 , then:

2,2
X
f (u, v) = ui vj f (xi , xj ) (20)
i,j=1,1

Since f is alternating, f (xi , xi ) = 0, i = 1, 2. We then have:

f (u, v) = u1 v2 f (x1 , x2 ) + u2 v1 f (x2 , x1 ) (21)

(b) According to definition, this is exactly alternating. I don’t know what he was thinking.

5. Let X be an n-dimensional vector space over a field K.

(a) Show how any k-linear form can be expressed as a sum of a symmetric and a skew-
symmetric one. Describe the differences in the cases of char K 6= 2 and char K = 2.
(b) Give an example of a non-alternating skew-symmetric mulitlinear form.
(c) Give an example of a non-zero alternating multilinear form such that f (x1 , . . . , xk ) =
0 for some set of linearly independent vectors x1 , . . . , xk .

(a) This is not true in general for k > 2. Just skip this Problem.

Math 8530 | Linear Algebra | Spring 2021 | M. Macauley