Scribd
Upload
8 views3 pages

Lecture Notes 9

The document discusses quadratic forms, providing definitions and equivalences between homogeneous quadratic polynomials, symmetric matrices, symmetric bilinear forms, and homogeneous quadratic functions. It explores properties such as rank, discriminants, and the conditions under which quadratic forms represent elements of a field. The Hasse-Minkowski theorem is highlighted, establishing a local-global principle for quadratic forms over the rational numbers.

Uploaded by

Leonardo Méndez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
8 views3 pages

Lecture Notes 9

The document discusses quadratic forms, providing definitions and equivalences between homogeneous quadratic polynomials, symmetric matrices, symmetric bilinear forms, and homogeneous quadratic functions. It explores properties such as rank, discriminants, and the conditions under which quadratic forms represent elements of a field. The Hasse-Minkowski theorem is highlighted, establishing a local-global principle for quadratic forms over the rational numbers.

Uploaded by

Leonardo Méndez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

18.

782 Introduction to Arithmetic Geometry Fall 2013


Lecture #9 10/3/2013

9.1 Quadratic forms


We assume throughout k is a field of characteristic different from 2.
Definition 9.1. The four equivalent definitions below all define a quadratic form on k.

1. A homogeneous quadratic polynomial f ∈ k[x1 , . . . , xn ].


2. Associated to f is a symmetric matrix A ∈ P k n×n whose entries (aij ) correspond to the
coefficients of xi xj in f via f (x1 , . . . , xn ) = i,j aij xi xj .1 Conversely, every symmetric
matrix defines a homogeneous quadratic polynomial.
3. Each symmetric matrix A defines a symmetric bilinear form B : k n × k n → k via
Bf (x, y) = xt Ay, where x and ydenote column vectors. It is symmetric, since

B(x, y) = xt Ay = (xt Ay)t = y t At x = y t Ax = B(y, x),

and it is bilinear, since for any a ∈ k and x, y, z ∈ k n we have

B(ax + y, z) = (ax + y)t Az = (axt + y t )Az = axt Az + y t Az = aBf (x, z) + B(y, z).

Conversely, if B is a symmetric bilinear form, and e1 , . . . , en are basis vectors, the


matrix A = (aij ) defined by aij = B(ei , ej ) is symmetric.
4. The function f : k n → k obtained by evaluating a homogeneous quadratic polynomial
is a homogeneous quadratic function. In terms of the corresponding bilinear form
B(x, y), we have f (x) = B(x, x). Conversely, we can recover B(x, y) from f (x) via
f (x + y) − f (x) − f (y)
B(x, y) = .
2
We thus have canonical isomorphisms between four sets of objects: homogeneous quadratic
polynomials, symmetric matrices, symmetric bilinear forms, and homogeneous quadratic
functions. We use the symbol f to refer to both a homogeneous quadratic polynomial and
its evaluation function, and we use the symbols A and B to refer to the associated matrix
and bilinear form.
The definition of a symmetric bilinear form B : V × V → k makes sense over any finite
dimensional k-vector space V , and we can define the corresponding homogeneous function
f : V → k abstractly as f (v) = B(v, v). If we then choose a basis for V we can compute
the symmetric matrix A whose coefficients define a homogeneous quadratic polynomial.
Symmetric bilinear forms can be viewed as a generalization of inner products to arbitrary
fields. Inner products are also required to satisfy B(v, v) > 0 for any nonzero vector v, but
this only makes sense if k is an ordered field.2 In general, symmetric bilinear forms are
allowed to vanish on nonzero vectors (indeed, the zero map is a symmetric bilinear form).
1
Note that for i 6= j this means that if fij is the coefficient of xi xj then aij = aji = fij /2, so that
fij xi xj = aij xi xj + aji xj xi . This is slightly unpleasant but makes everything else work nicely.
2
An ordered field is a field with a total ordering ≤ that satisfies a ≤ b ⇒ a+c ≤ b+c and a, b > 0 ⇒ ab > 0.
In such a field 0 cannot be written as a sum of nonzero squares. This is a severe restriction; it rules out all
fields of positive characterstic, all p-adic fields, the complex numbers, and most number fields.

Andrew V. Sutherland
The group GLn (k) of invertible n × n matrices over k acts on the space of quadratic
forms as a linear change of variables. If T is any invertible linear transformation on V , and
A is the matrix of a quadratic form f on V , then we have

f (T v) = (T v)t A(T v) = v t (T t AT )v

where T t AT is a symmetric matrix that defines another quadratic form.


Definition 9.2. Two quadratic forms f and g are equivalent if g(v) = f (T v) for some
T ∈ GLn (k). This defines an equivalence relation on the set of all quadratic forms of the
same dimension over the field k.
Note that, in general, the matrices T t AT and T −1 AT are not the same, this GLn (k)
action is not the same as its action by conjugation. In particular, equivalent symmetric
matrices need not be similar, as can be seen by the fact that equivalent matrices may have
different determinants:

det(T t AT ) = det(T t ) det(A) det(T ) = det(T )2 det(A).

Definition 9.3. The rank of a quadratic form is the rank of its matrix; rank is clearly
preserved under equivalence. A quadratic form is non-degenerate if it has full rank, equiv-
alently, the determinant of its matrix is nonzero.
If B is the symmetric bilinear form associated to a non-degenerate quadratic form on V ,
then each nonzero v ∈ V defines a nonzero linear map w → B(v, w) (otherwise the matrix
of the form with respect to a basis including v would have a zero row).
Definition 9.4. The discriminant of a nondegenerate quadratic form with matrix A is the
image of det A in k × /k ×2 ; it is clearly preserved by equivalence.
Inequivalent forms may have the same discriminant; over C for example, every non-
degenerate form has the same discriminant (in fact all nondegenerate forms of the same
dimension are equivalent). However, quadratic forms with different discriminants cannot
be equivalent; this implies that over Q, for example, there are infinitely many distinct
equivalence classes of quadratic forms in every dimension n > 0.
A quadratic form is said to be diagonal if its matrix is diagonal.
Theorem 9.5. Every quadratic form is equivalent to a diagonal quadratic form.
Proof. We proceed by induction on the dimension n. The base cases n ≤ 1 are trivial.
Let f be a quadratic form on a vector space V , and let B be the corresponding symmetric
bilinear form. If f is the zero function then its matrix is zero, hence diagonal, so assume
otherwise and pick v ∈ k n so that f (v) 6= 0. The map x → B(x, v) is a nonzero linear map
from k n to k, hence surjective, so its kernel v ⊥ = {x ∈ V : B(x, v) = 0} has dimension
n − 1. We know that v 6∈ v ⊥ , since B(v, v) = f (v) 6= 0, so V ' hvi ⊕ v ⊥ . Thus any y ∈ V
can be written as y = y1 + y2 with y1 ∈ hvi and y2 ∈ v ⊥ . We then have

f (y1 + y2 ) = B(y1 + y2 , y1 + y2 ) = B(y1 , y1 ) + B(y2 , y2 ) + 2B(y1 , y2 ) = f (y1 ) + f (y2 ),

since B(y1 , y2 ) = 0 for any y1 ∈ hvi and y2 ∈ v ⊥ . By the inductive hypothesis, the
restriction f |v⊥ of f to v ⊥ can be diagonalized (that is, there is a diagonal quadratic form
on v ⊥ that is equivalent to f |v⊥ ), and the same is certainly true for the restriction of f to
the 1-dimensional subspace hvi, thus f can be diagonalized.
So to understand equivalence classes of quadratic forms we can restrict our attention to
diagonal quadratic forms.

Example 9.6. The quadratic form x2 + y 2 is equivalent to 2x2 + 2y 2 , since

(x + y)2 + (x − y)2 = 2x2 + 2y 2 ,

but it is not equivalent to 3x2 + 3y 2 . Indeed, for x2 + y 2 to be equivalent to αx2 + βy 2 we


must have t    2
a + c2 ab + cd
    
a b 1 0 a b α 0
= = ,
c d 0 1 c d ab + cd b2 + d2 0 β
and in particular, α and β must both be sums of squares, which 3 is not.

Thus equivalence of quadratic forms depends on arithmetic properties of the field k.

Definition 9.7. A quadratic form f on V represents a ∈ k if a lies in the image of f : V → k.


Equivalent forms necessarily represent the same elements (but the converse need not hold).

Example 9.8. The form x2 − 2y 2 represents −7 but not 0.

The constraint that x 6= 0 is critical, otherwise every quadratic form would represent 0;
the quadratic forms that represent 0 are of particular interest to us.

Theorem 9.9. If a nondegenerate quadratic form f represents 0 then it represents every


element of k.

Proof. Assume f (v) = 0 for some v ∈ V . Since f is nondegenerate, there exists w ∈ V with
B(v, w) 6= 0, and v and w must be independent, since B(v, v) = f (v) = 0 and therefore
B(v, xv) = cB(v, v) = 0 for any c ∈ k. For any x ∈ k we have

f (xv + w) = B(xv + w, xv + w) = B(v, v)x2 + 2B(v, w)x + B(w, w) = ax + b,

with a = 2B(v, w) 6= 0 and b = f (w). For any c ∈ k we can solve ax + b = c for x, proving
that f represents c = f (xv + w).

Our main goal is to prove the following theorem of Minkowski, which was generalized
to number fields by Hasse.

Theorem 9.10 (Hasse-Minkowski). A quadratic form over Q represents 0 if and only if it


represents 0 over every completion of Q, that is, over Qp for all primes p ≤ ∞.

This is an example of a local-global principle. We have an object f (in this case a


quadratic form) defined over a “global” field (in this case Q) and a certain property of
interest (in this case representing 0). Since f is defined over the global field, we can also
consider f as an object over any of the “local fields” associated to the global field (in
this case the completions of Q). If f satisfies the property of interest over the global field
then it typically must satisfy this property over every local field (this is certainly true in
our case), but the question is whether the converse holds. In the case of quadratic forms
representing 0, the answer is “yes”, but in many other cases we will see later in this the
answer is “no,” and it is a major point of interest in arithmetic geometry to understand
exactly when and how various local-global principles can fail.

You might also like