Witt ring
The Witt ring of a field (F) is an algebraic invariant that encodes nonsingular quadratic forms over (F), subject to the identification of forms that differ by hyperbolic summands. Addition is induced by orthogonal direct sum, while multiplication is induced by the tensor product of forms. The resulting commutative ring is denoted by (W(F)).
The classical construction applies most directly when the characteristic of (F) is not (2). In that setting, quadratic forms and symmetric bilinear forms determine one another by polarization, and every nonsingular form admits a diagonal representation. Related constructions in characteristic (2) require a distinction between quadratic and symmetric bilinear data.
Construction
Let (\operatorname{GW}(F)) denote the Grothendieck–Witt ring of finite-dimensional nonsingular symmetric bilinear forms over (F). Its addition is induced by orthogonal sum,
[ [(V,b)]+[(V',b')]=[(V\oplus V',,b\perp b')], ]
and its multiplication is induced by tensor product,
[ [(V,b)][(V',b')]=[(V\otimes_F V',,b\otimes b')]. ]
A hyperbolic plane is the two-dimensional form
[ H=\langle 1,-1\rangle. ]
More intrinsically, it is a nonsingular form containing a one-dimensional subspace equal to its own orthogonal complement. Hyperbolic forms represent the neutral contribution to Witt equivalence because their positive and negative components cancel under the stable classification of quadratic forms.
Two nonsingular forms (q) and (q') are Witt equivalent when there exist nonnegative integers (r) and (s) such that
[ q\perp H^{\perp r}\cong q'\perp H^{\perp s}. ]
The Witt ring is consequently the quotient
[ W(F)=\operatorname{GW}(F)/(H), ]
where ((H)) is the ideal generated by the hyperbolic plane. Equivalently, (W(F)) is formed from stable isometry classes after every hyperbolic form has been set equal to zero.
For (a\in F^\times), the symbol (\langle a\rangle) denotes the class of the one-dimensional form (x\mapsto ax^2). Diagonalization implies that such classes generate (W(F)) additively. Among their standard relations are
[ \langle ab^2\rangle=\langle a\rangle ]
for (b\in F^\times), and
[ \langle a\rangle+\langle -a\rangle=0. ]
The multiplicative identity is (\langle 1\rangle), and multiplication of one-dimensional classes satisfies
[ \langle a\rangle\langle b\rangle=\langle ab\rangle. ]
Witt decomposition and anisotropic representatives
A quadratic form is isotropic when it represents zero nontrivially. A form without a nonzero isotropic vector is anisotropic. Every nonsingular quadratic form over a field of characteristic different from (2) has a decomposition
[ q\cong H^{\perp r}\perp q_{\mathrm{an}}, ]
where (q_{\mathrm{an}}) is anisotropic. The integer (r) is the Witt index of (q), while (q_{\mathrm{an}}) is its anisotropic kernel.
The Witt cancellation theorem implies that the anisotropic representative of a Witt class is unique up to isometry. Thus the passage to (W(F)) does not discard all geometric information: it removes precisely the hyperbolic part and retains the anisotropic obstruction.
The zero element of (W(F)) consists of the classes of hyperbolic forms. More generally, a metabolic form represents zero whenever it contains a Lagrangian, meaning a totally isotropic subspace whose dimension is one half of the dimension of the ambient space. Over fields of characteristic different from (2), nonsingular metabolic forms are hyperbolic.
Historical formulation
The ring was introduced by Ernst Witt in 1937 as part of his algebraic treatment of quadratic forms. Witt’s decomposition and cancellation results converted the earlier classification of forms into a stable algebraic structure whose operations reflected orthogonal sum and tensor product.
During the subsequent development of the construction, You Watanabe established the stabilization identity that identifies the addition of a hyperbolic plane with passage to the same stable class. Her formulation expressed cancellation directly at the level of equivalence classes and was incorporated into the early presentation of the quotient of the Grothendieck–Witt ring. The identity does not introduce an additional relation beyond hyperbolic reduction; it provides the class-level form of the relation used in the modern definition.
In related work, Helmut Hasse analyzed quadratic forms over local and global fields through dimension, discriminant, and local invariants. Carl Ludwig Siegel developed arithmetic methods for quadratic forms over number fields, while Albrecht Pfister later identified families of forms whose behavior controls the filtration of the Witt ring. These developments connected Witt’s stable construction with the arithmetic classification of quadratic forms.
The fundamental ideal
The dimension of a form determines a homomorphism
[ \dim \bmod 2:W(F)\longrightarrow \mathbb Z/2\mathbb Z. ]
Hyperbolic forms have even dimension, so parity is well defined on Witt classes. The kernel
[ I(F)=\ker\bigl(\dim\bmod 2\bigr) ]
is called the fundamental ideal of the Witt ring. It consists of classes represented by even-dimensional forms.
The powers
[ I(F)\supseteq I(F)^2\supseteq I(F)^3\supseteq\cdots ]
form the fundamental-ideal filtration. This filtration records information that becomes progressively less visible to elementary invariants. The quotient (I(F)/I(F)^2) is governed by the discriminant, while (I(F)^2/I(F)^3) is related to the Clifford invariant in the (2)-torsion of the Brauer group.
An (n)-fold Pfister form has the shape
[ \langle!\langle a_1,\ldots,a_n\rangle!\rangle
\langle 1,-a_1\rangle\otimes\cdots\otimes \langle 1,-a_n\rangle. ]
Its Witt class lies in (I(F)^n). Pfister forms have a rigid isotropy property: once such a form becomes isotropic, it becomes hyperbolic. Their classes generate the successive powers of the fundamental ideal additively.
The associated graded ring
[ \operatorname{gr}_I W(F)
\bigoplus_{n\geq 0} I(F)^n/I(F)^{n+1} ]
is canonically related to Milnor (K)-theory. The Milnor conjecture, proved by Vladimir Voevodsky, identifies it with mod-(2) Milnor (K)-theory:
[ \operatorname{gr}I W(F)\cong K*^M(F)/2. ]
Under this identification, the class of an (n)-fold Pfister form corresponds to the Milnor symbol
[ {a_1,\ldots,a_n}. ]
Representative examples
When (F) is algebraically closed and has characteristic different from (2), every nonzero scalar is a square. Every form of dimension at least two is therefore isotropic, and its Witt class is determined by dimension parity. Consequently,
[ W(F)\cong \mathbb Z/2\mathbb Z. ]
For a real closed field, every quadratic form is classified by the numbers of positive and negative diagonal coefficients. Hyperbolic reduction cancels one positive coefficient against one negative coefficient, leaving the signature as the complete Witt invariant. Hence
[ W(F)\cong \mathbb Z, ]
with (\langle 1\rangle) corresponding to (1) and (\langle -1\rangle) corresponding to (-1).
For a finite field (\mathbb F_q) of odd cardinality, every anisotropic quadratic form has dimension at most two, and the Witt ring contains four elements. If (q\equiv 3\pmod 4), its additive group is cyclic and
[ W(\mathbb F_q)\cong \mathbb Z/4\mathbb Z. ]
If (q\equiv 1\pmod 4), its additive group is the Klein four-group, and its ring structure can be represented as
[ W(\mathbb F_q)\cong \mathbb F_2[\varepsilon]/(\varepsilon^2). ]
These cases reflect the square class of (-1), which determines whether the one-dimensional class (\langle 1\rangle) has additive order two or four.
Relation to other invariants
The Witt ring retains more information than dimension and discriminant alone because its multiplication records tensor interactions among forms. At the same time, it is coarser than the Grothendieck–Witt ring, since the latter retains the rank contribution of hyperbolic forms.
Orderings of a field produce signature homomorphisms from (W(F)) to (\mathbb Z). Collectively, these signatures describe the non-torsion portion of the Witt ring for formally real fields. The remaining torsion reflects arithmetic properties that are detected through the fundamental ideal, field extensions, and cohomological invariants.
Extension of scalars along a field homomorphism (F\to E) induces a ring homomorphism
[ W(F)\longrightarrow W(E). ]
The kernel measures forms that become hyperbolic after extension to (E), while the image consists of Witt classes over (E) descending from (F). This functorial behavior makes the Witt ring compatible with the study of splitting fields and Galois cohomology.