We are going to define a binary operation on polynomials of one variable
modifying the definition of a resultant.
Let \(f(t)=1+a_1t+\cdots \) and \(g(t)=1+b_1t+\cdots \) belong to the multiplicative semi-group \(1+t\mathbb {F}[t]\) of polynomials with coefficients in a field \(\mathbb {F}\) and the constant term equal to 1. Define a convolution \(f\star g\) as a polynomial with the constant term 1 and having as roots the products of one root of \(f\) and one of \(g\). In other words, suppose that \(f(t)=\prod _i(1-\lambda _it)\) and \(g(t)=\prod _j(1-\mu _jt)\)
with \(\lambda _i,\mu _j\) belonging to the algebraic closure of the field \(\mathbb F\). Then
\[f\star g\,(t) = \prod _{ij}(1-t\lambda _i\mu _j)=\prod _ig(\lambda _it)=\prod _jf(\mu _jt).\]
The convolution can also be expressed in term of the resultant, namely
\[f\star g\,(t)=\mathtt {res}_z(f(z),z^{\deg g}g(t/z)).\]
To give an equivalent definition, consider the ring \(\mathbb {F}[x,y]/(f(x))+(g(y))\) and denote by \(\hat {x}^{-1}\) and \(\hat {y}^{-1}\) the multiplication in this ring by \(x^{-1}\) and \(y^{-1}\), respectively. In the standard basis they are given by matrices
with entries in \(\mathbb {F}\). Then
\[f\star g\,(t)=\det (1-t\hat {x}^{-1}\hat {y}^{-1}).\]
In this definition it is explicit that the coefficients of \(f\star g\) are polynomial functions of those of \(f\) and \(g\).
The fourth definition works for \(\mathbb {F}=\mathbb {C}\) and shows the relation to the tame symbol. Consider a curve \(\gamma \) around zero on the complex plane sufficiently small in order not to surround any root of \(f(z)\). The convolution can be defined by
the formula (see P.Deligne [2] , formula 2.7.2)
\[f\star g\,(t)=\{f(z),g(t/z)\}_\gamma =\exp \left (\frac {1}{2\pi i}\int _{\gamma }\ln f(z) d\ln g(t/z)\right )\]
valid for \(t\) so small that all roots of \(g(t/z)\) are inside the curve \(\gamma \).
The convolution enjoys the following properties obvious from the definitions:
- \(\deg (f\star g)=\deg f \deg g\),
- \(f\star 1=1,\)
- \(f\star (1-t)=f,\)
- \((1-at)\star (1-bt)=(1-abt),\)
- \(f\star g = g\star f,\)
- \(f\star (g_1g_2)=(f\star g_1)(f\star g_2)\).
These properties imply that the semi-group \(1+t\mathbb {F}[t]\) is a commutative semi-ring with respect to the multiplication as a semi-ring addition and convolution as a semi-ring multiplication. The multiplicativity property
4 is just the expression of the distributive law of the semi-ring.
The following property is also an easy consequence of the definition:
- A. The set \(1+t^{n}\mathbb {F}[t]\) is an ideal.
This property implies that the convolution can be extended to the group of formal power series \(1+t\mathbb {F}[[t]]\) providing it with a ring structure. This ring is called the ring of the
universal or
big Witt
vectors and is denoted by \(W(\mathbb {F})\), see
[3] , section 9.