A Vassiliev invariant, also called a finite type invariant, is a knot invariant whose extension to singular knots vanishes whenever the number of ordinary double points exceeds some fixed nonnegative integer. The notion was independently introduced by V. Vassiliev and M. Goussarov around 1989 (Duzhin 1999). Vassiliev's approach is based on the study of discriminants in the (infinite-dimensional) spaces of smooth maps from one manifold into another. By definition, the discriminant consists of all maps with singularities.
For example, consider the space of all smooth maps from the circle into three-dimensional Euclidean
space .
If
is an embedding (i.e., has no singular points), then
it represents a knot. The complement of the set
of all knots is the discriminant
. It consists of all smooth
maps from
into
that have singularities, either local, where
,
or nonlocal, where
is not injective. Two knots are equivalent
iff they can be joined by a path in the space
that does not intersect the
discriminant. Therefore, knot types are in one-to-one
correspondence with the connected components
of the complement
,
and knot invariants with values in an Abelian
group
are nothing but cohomology classes from
. The filtration
of
by subspaces corresponding to singular knots with
a given number of ordinary double points
gives rise to a spectral sequence, which contains,
in particular, the spaces of finite type invariants.
Birman and Lin (1993) have contributed significantly to the simplification of Vassiliev's original techniques. In particular, they explained the relation between Jones
polynomials and finite type invariants (Peterson 1992, Birman and Lin 1993, Bar-Natan
1995) and emphasized the role of the algebra
of chord diagrams. In fact, substituting the power
series for
as the variable in the Jones polynomial yields
a power series whose coefficients
are Vassiliev invariants (Birman and Lin 1993). Kontsevich (1993) proved the first
difficult theorem about Vassiliev invariants with the help of the Kontsevich
integral. Bar-Natan undertook a thorough study of Vassiliev invariants; in particular,
he showed the importance of the algebra of Feynman diagrams and diagrams with uni-
and tri-valent vertices (Bar-Natan 1995). Bar-Natan (1995) gives a detailed account
of the theory.
Expressed in simple terms, Vassiliev's fundamental idea is to study the prolongation of knot invariants to singular
knots--immersions having a finite number of ordinary
double points. Let
denote the set of equivalence
classes of singular knots with
double points and no other singularities. The following definition
is based on a recursion which allows to extend a knot
invariant from
to
,
then to
,
etc., and thus finally to the whole of
. For oriented unframed knots,
let
be a rational-valued knot invariant. Its Vassiliev
prolongation
is defined by the rules
1. ,
and
2. Vassiliev's skein relation, illustrated below.
The right-hand side of Vassiliev's skein relation refers to the two resolutions of the double point--positive and negative. A crucial observation is that each of them
is well-defined (does not depend on the plane projection used to express this relation).
A knot invariant is called a Vassiliev invariant of order
if its prolongation
vanishes on all knots with more
than
ordinary double points. For example, the simplest
nontrivial Vassiliev invariant
has the following explicit description. Let
be an arbitrary knot diagram
of the given knot
and
an arbitrary distinguished point on
, different from all crossings. Then
where the summation spreads over all pairs of crossing points such that (1) during one complete turn of the diagram in
the positive direction starting from point
the points
and
are encountered in the order
, and (2) the four corresponding passages through these
crossing points are underpass, overpass, overpass, and underpass, respectively. The
numbers
,
stand for the local writhe at points
and
, defined according to the above illustration.
The th
coefficient of the Conway
polynomial is a Vassiliev invariant of order at most
and, in particular, the second coefficient
coincides with
.
Vassiliev invariants are at least as strong as the following polynomial knot invariants: Alexander, Jones,
Kauffman, and HOMFLY
polynomials. This means that if two knots and
can be distinguished by such a polynomial, then there is
a Vassiliev invariant that takes different values for
and
.
The set of all -valued Vassiliev invariants
, where
consists of invariants of order at most
, forms a vector space over
the rational numbers with the increasing filtration
.
With
,
its associated graded space
is a Hopf
algebra. Each quotient
is naturally the dual
vector space of the degree-
component of the unframed algebra
of chord diagrams, which imposes the one-term
relation (framing-independence or 1T relation) and four-term
relations (Kneissler 1997, p. 3).
The dimensions of the vector spaces
of Vassiliev invariants of order at most
for
to 12 are summarized in the following table (Kneissler 1997,
p. 5; OEIS A007473).
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | |
| 1 | 1 | 2 | 3 | 6 | 10 | 19 | 33 | 60 | 104 | 184 | 316 | 548 |
The Kontsevich integral defines a universal Vassiliev invariant over the real numbers. Each finite type invariant is obtained by applying a linear functional to a finite-degree truncation of this invariant.
Duzhin (1999) records the following foundational questions about Vassiliev invariants, raised in 1990.
1. Is it true that Vassiliev invariants distinguish knots? In other words, given two nonequivalent knots and
, is it always possible to indicate a finite type invariant
such that
?
2. Is it true that Vassiliev invariants can detect knot orientation? More specifically, is there a knot
and a finite type invariant
such that
, where
differs from
by a change of parameterization that reverses the orientation?