Classification of anisotropic spaces over finite fields
Statement
Suppose is a finite field of characteristic not equal to two and is a symmetric bilinear form on a finite-dimensional vector space over . Suppose, further, that is anisotropic with respect to , i.e., for any nonzero vector , . Then, the dimension of is either 1 or 2. The complete classification is given below.
For describing the classification, we denote by the size of the field. is a prime power (specifically, a power of the field characteristic) and completely determines the field up to isomorphism. Since the characteristic is not two, is odd.
One-dimensional spaces
If is one-dimensional, there are two possibilities for the form up to equivalence:
| Representative matrix for symmetric bilinear form | All possible matrices for symmetric bilinear form |
|---|---|
| , (total of possibilities) | |
| whre and is not a square | , (total of possibilities) |
Two-dimensional spaces when field size is 1 mod 4
If is 1 mod 4, then, up to equivalence, there is a unique anisotropic two-dimensional space. A representative matrix is:
where is a quadratic nonresidue. More generally, we can choose any diagonal matrix where one of the diagonal entries is a quadratic residue and the other is a quadratic nonresidue.
Two-dimensional spaces when field size is 3 mod 4
If is 1 mod 4, then, up to equivalence, there is a unique anisotropic two-dimensional spaces. Representatives for these are:
Facts used
- Anisotropic form can be represented by diagonal matrix
- Product of subsets whose total size exceeds size of group equals whole group (the way we use this generalizes the idea behind every element of a finite field is expressible as a sum of two squares)
- Quotient of determinants of equivalent symmetric bilinear forms is a square
Proof
Given: A finite field of size , a -dimensional vector space over , a symmetric bilinear form on that is anisotropic.
To prove: and the classification above holds.
Proof:
| Step no. | Assertion/construction | Facts used | Given data used | Previous steps used | Explanation |
|---|---|---|---|---|---|
| 1 | There is a basis for , say , in which the matrix for is diagonal with nonzero diagonal entries. In other words, we have for and . | Fact (1) | is anisotropic | ||
| 2 | For any tuple that is not the zero tuple, the sum is nonzero. In fact, this is precisely the condition for the form to be anisotropic. | is anisotropic | Step (1) | The indicated sum is where . This sum is nonzero because is anisotropic. | |
| 3 | Let for . Each is the union of with either the set of squares or the set of non-squares in , depending on whether is a square. | Step (1) | By Step (1), , and the rest follows immediately. | ||
| 4 | Each has size | Step (3) | Step-direct | ||
| 5 | If , then , i.e., every element of can be written as . Further, if the element of is nonzero, we must have that are not both zero. | Fact (2) | Step (4) | Fact-step combination direct (note the word product in the fact is here used for the sum, i.e., with respect to the additive structure) | |
| 6 | If , then there is a tuple of the form such that the sum is zero, a contradiction with Step (2). This forces . | Steps (1), (2), (5) | [SHOW MORE] | ||
| 7 | If , then there are two inequivalent possibilities: those where is a square and those where it is not. | Fact (3) | [SHOW MORE] | ||
| 8 | If and is 1 mod 4, there is a unique possibility, with representative matrix where is a quadratic nonresidue. | Steps (1), (2) | [SHOW MORE] | ||
| 9 | If and is 3 mod 4, there is a unique possibility, with representative matrix | Steps (1), (2) | [SHOW MORE] |