MIPT/TH-27/24

ITEP/TH-34/24
IITP/TH-29/24

Construction of Lie algebra weight system kernel

via Vogel algebra
D. Khudoteplov∗1,2 , E. Lanina† 1,2,3 , and A. Sleptsov‡1,2,3
arXiv:2411.14417v1 [math.QA] 21 Nov 2024

1
Moscow Institute of Physics and Technology, 141700, Dolgoprudny, Russia
2
Institute for Information Transmission Problems, 127051, Moscow, Russia
3
NRC ”Kurchatov Institute”, 123182, Moscow, Russia§

Abstract
We develop a method of constructing a kernel of Lie algebra weight system. A main tool we use in the
analysis is Vogel’s Λ algebra and the surrounding framework. As an example of a developed technique we
explicitly provide all Jacobi diagrams lying in the kernel of slN weight system at low orders. We also discuss
consequences of the presence of the kernel in Lie algebra weight systems for detection of correlators in the
3D Chern–Simons topological field theory and for distinguishing of knots by the corresponding quantum knot
invariants.

Contents
1 Introduction 2

2 Vassiliev invariants and related notions 3

2.1 Vassiliev invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Chord diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Weight systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Vassiliev-Kontsevich theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Jacobi diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5.1 AS and IHX relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5.2 Open Jacobi diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.6 Lie algebra weight systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Vogel algebra 9
3.1 Characters on Λ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Construction of weight system kernel 11

4.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Example for sln weight system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.2.1 Primitive diagrams in kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.2.2 Non-primitive diagrams in kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 Relation between Chern-Simons invariants and Kontsevich integral 14

6 On knot detection 14

A Vogel algebra: advanced features 18

∗ [email protected]
† [email protected]
‡ [email protected]
§ former Institute for Theoretical and Experimental Physics, 117218, Moscow, Russia

1
1 Introduction
In 1984, V. Jones discovered a new polynomial knot invariant [1–3], the Jones polynomial, associated with the Lie
algebra sl2 and its standard (fundamental) representation. Soon, generalizations of the Jones polynomials appeared:
first to the algebras sln and son (the HOMFLY [4, 5] and Kauffman polynomials [6], respectively), and then to
arbitrary finite-dimensional representations by methods of quantum groups and R-matrices (Reshetikhin–Turaev
invariants [7, 8]). The Reshetikhin–Turaev construction has the property of universality, since it is formulated in
fairly general terms and defines knot invariants related to simple Lie algebras and their finite-dimensional represen-
tations. An extension of the construction to Lie superalgebras is also allowed [9–13].
In parallel, physicists A. Schwartz [14–16] and E. Witten [17] discovered that the 3-dimensional quantum Chern–
Simons field theory also allows one to construct these invariants. In this theory it is necessary to construct a gauge-
invariant quantity called the vacuum expectation value of the Wilson loop, well known to physicists from the theory
of strong interactions (i.e. quantum chromodynamics). This approach also has the property of universality and,
on the one hand, opens the way for physicists to the world of low-dimensional topology, and, on the other hand,
enriches the mathematical tools with new powerful methods coming from quantum field theory and string theory [17–
30]. These invariants are now called in literature either Witten–Reshetikhin–Turaev invariants or quantum knot
invariants. All these methods and approaches to constructing knot invariants, inspired by the ideas of quantum
mechanics and quantum field theory and applied to problems of low-dimensional topology, formed a new direction
called quantum topology.
In 1992, Birman and Lin discovered that the coefficients of a certain perturbative expansion of the HOMFLY
polynomials are Vassiliev invariants [31], which V.Vassiliev defined in 1990 studying complements of discriminants
in spaces of maps [32] (more about Vassiliev invariants in the next section). In 1995, D. Bar-Nathan proved that
for any quantum knot invariant the expansion coefficient is a Vassiliev invariant [33]. The inverse question naturally
arises: Is any Vassiliev invariant some coefficient (or a linear combination of them) of the perturbative expansion
of a suitable quantum invariant (or even a linear combination of them)?
This question can be formulated more precisely if we introduce the concept of the Kontsevich integral, which
is a universal Vassiliev invariant [34], see Section 2.4. Roughly speaking, this is a generating series for all Vassiliev
invariants (2.3). The series goes over the formal variable ~. Each Vassiliev invariant is weighted by a certain chord
diagram, since the algebra of Vassiliev invariants is isomorphic to the algebra of chord diagrams (more in Section
2). Any finite-dimensional Lie algebra (with a non-degenerate Killing form) defines a linear function (called weight
system) on the algebra of chord diagrams. If we take sln as such an algebra, then the Kontsevich integral turns
into the HOMFLY polynomial, into the Kauffman polynomial in the case of son , etc. More precisely, it turns into
a series of perturbative decomposition for this polynomial, see Section 5. The question is which chord diagrams lie
in the kernel of this weight system. Vassiliev invariants that correspond to these diagrams will be impossible to
extract from the perturbative decomposition of the original quantum invariant. Next, we can ask whether there are
such chord diagrams that lie in the kernel of all weight systems constructed from finite-dimensional Lie algebras
and superalgebras.
It turns out that such diagrams exist, what was proved in a seminal paper by Pierre Vogel [35] in 2011. However,
Vogel did much more than just prove this fact. He developed a diagrammatic technique for explicitly constructing
such kernels for any finite dimensional Lie (super)algebra. In this paper we apply his technique to construct the
kernel of the sln weight system. Without Vogel’s technique this problem is quite difficult, because it is difficult
to calculate a Lie algebra weight system of all basis diagrams of a given order. Basically there are two ways: the
universal weight system and the weight system associated with a representation.
For the universal way one parameterizes the center of universal enveloping algebra for the corresponding Lie
algebra and expresses the values of the weight system for all (basis) chord diagrams in terms of selected parameters.
Then, it is easy to find the dimension of the weight system and therefore its kernel. In this way all calculations
must be done analytically. It was suggested by D. Bar-Natan in [33], a specific calculation method was also given
there. Further it was developed by Zh. Yang in [36] and M. Kazarian and S. Lando in [37]. However, as far as we
know, the 8-th degree is already unattainable for this method at the current progress, but it is at the 8-th order
that the first primitive element of the sln weight system kernel appears.
Another way to compute a Lie algebra weight system is to provide it with a finite-dimensional representation.
Then, the computation need not be done analytically, which is an advantage over the previous method. However,
it also introduces a disadvantage, since it is necessary to ensure that the relations found are independent on the
representation. As far as we know, this method has not been developed much. Some computations can be found in
Alvarez’s and Labastida’s paper [38] for the fundamental representation of sln and son , and for all representations
in sl2 .
An interesting development is a paper [39], where a combination of both methods is used for the calculation,

2
and in addition various symmetry considerations are used for the colored HOMFLY polynomials (i.e. sln case),
which impose certain restrictions on the possible form of weight systems. The developed method allows to find a
large number of relations for the weight system sln and to obtain a good upper bound on the kernel dimension,
which is presented below in Table 3.
It is also important to note that the first person to use Vogel’s technique to construct the kernel of weight
systems was Lieberum [40]. However, he studied those elements that lie in the kernel of all Lie algebras at once,
except D(2, 1, α).
The simplest primitive Jacobi diagram (see Section 2 for the definitions), which belongs to the kernel of all
weight systems of Lie algebras, found by P. Vogel, is of the 17-th order. Lieberum’s simplest diagram is of the
15-th order. Today, these are too high orders to work with. For example, it is very difficult to understand whether
there are relations for any two or three diagrams constructed by Vogel’s method. To answer this question it is
necessary to use AS and IHX relations, which is difficult. It is also not yet clear whether it is possible to find two
such knots that do not differ in any invariants except those that correspond to the found diagrams. Moreover, it
is not entirely clear how to approach this problem. When we restrict ourselves to the sln weight system, all these
problems are simplified, but not trivialized, which allows us to develop and refine the methods. First, the kernel of
primitive diagrams appears at the 8-th order, which is much simpler than the 17-th. Second, the weight system of
son or other algebras can be used to detect elements of the sln kernel. Third, there are several candidates among
the 13-crossing knots for the role of those that are not detected by some colored quantum sln invariants and are
not mutants. At least these knots have the same colored HOMFLY polynomials (those that we can calculate) and
the uncolored Kauffman polynomial. Unfortunately, we do not yet know how to calculate the colored Kauffman
polynomials well enough to test this hypothesis, but at least this task is quite straightforward. In addition, the 8-th
order of the diagram allows us to expect that such knots, if they exist at all, will be encountered much earlier (in
the sense that their minimum number of crossings should be smaller) than knots for the diagram of order 15 and
especially of order 17.
This paper is organized as follows. In Section 2, main provisions of the theory of finite order invariants from the
definition of Vassiliev invariants to Jacobi diagrams and weight systems are briefly outlined. Section 3 highlights
results obtained by Pierre Vogel [35, 41]. The definition of the algebra Λ and the characters of Lie algebras on
Λ are given. Particular attention is paid to Vogel parameterization and the related concept of Vogel universality.
In Section 4, Vogel’s results are used to develop a method for construction of Lie algebra weight systems kernel.
A method for constructing Jacobi diagrams that are vanished by weight systems is described. As an example we
provide explicit Jacobi diagrams forming the kernel of sln weight system at low orders. For diagrams constructed
in this way, their linear independence is proven. Section 5 is devoted to connection of Vassiliev invariants, chord
diagrams, the Kontsevich integral and the results of this paper with the 3-dimensional topological Chern–Simons
theory. In Section 6, we discuss that the existence of Lie algebras weight systems kernels implies that there exist
linear combination of knots that are distinguished by Vassiliev invariants coming from kernel diagrams but not by
other Vassiliev invariants of the same or lower orders. However, there is still no argument that higher order Vassiliev
invariants, and thus the corresponding quantum knot invariants, do not distinguish this linear combination of knots.

2 Vassiliev invariants and related notions

In this section we introduce main objects of our study such as Vassiliev invariants, chord diagrams, Jacobi diagrams
and weight systems. We mostly follow [42].

2.1 Vassiliev invariants

Definition 2.1. A knot is an isotopy class of S 1 embeddings in R3 .
Definition 2.2. A knot invariant is a function on the set of knots.
Before defining Vassiliev invariants, it is necessary to expand the concept of a knot to include singular knots.
Definition 2.3. A singular knot is an isotopy class of S 1 immersions in R3 such that all self-intersection points
are simple double points with transversal intersections.
The continuation of knot invariants to the set of singular knots is determined by Vassiliev skein relation shown in
Fig. 1. The Vassiliev skein relation is a rule for resolving double points of a singular knot. It can be easily seen that
a value of an invariant at a singular knot does not depend on an order in which its double points are resolved.

3
 v( ) = v( ) − v( )
Figure 1: Vassiliev skein relation.

Definition 2.4 (Vassiliev invariants [32]). An invariant v is called a Vassiliev invariant of order no more than
n if v vanishes at singular knots with more than n double points. The space of Vassiliev invariants of order ≤ n is
denoted Vn .
Vassiliev invariants form a filtered vector space, and they also form a bialgebra (see [42, §4.3]).

2.2 Chord diagrams

Definition 2.5. A chord diagram of order n is an oriented circle with n pairs of distinct points. The set of chord
diagrams of order n is denoted An .
Chord diagrams naturally arise in the theory of Vassiliev invariants [32]. To demonstrate this relationship, one
needs to consider v ∈ Vn . Suppose K is an arbitrary singular knot with exactly n double points. Next, let K ′ be a
knot that is identical to K in all regions of space except for one where they differ from each other in the same way
as the two diagrams in Fig. 2 do. We also define K • to be a knot with a self-intersection in the area where K and
K ′ differ, and identical to both in the remaining space (see Fig. 2).

K K′ K•
Figure 2: Knots K, K ′ and K • .

Then, v(K) − v(K ′ ) = v(K • ) according to Vassiliev skein relation. However, a knot K • has n + 1 double points
by construction, whereas v is an invariant of order ≤ n, so, by Definition 2.4, we have v(K • ) = 0. Therefore,
v(K) = v(K ′ ) for all v ∈ Vn and K, K ′ are singular knots with n double points that differ as shown in Fig. 2.
It follows that changing one of intersection points in a knot diagram does not affect a value of an invariant v
of a singular knot with n double points. Hence, a value of v ∈ Vn depends only on an order in which we meet the
double points when moving along the singular knot.
Information about a relative position of double points of a singular knot can be presented in form of a chord
diagram, where each double point corresponds to a chord with its endpoints being preimage (with respect to an
immersion that defines a singular knot) of double points of a singular knot.
A chord diagram constructed in accordance with these rules is referred to as a chord diagram of a singular knot.
A chord diagram generated from a singular knot K is denoted by σ(K). Examples of this correspondence between
singular knots and their chord diagrams are provided in Table 1:

σ(K)

Table 1: Chord diagrams of singular knots.

4
2.3 Weight systems
In the previous paragraph it has been shown that a value of an invariant v ∈ Vn on a singular knot K with n double
points equals to a value of some function on a chord diagram σ(K). Formally, there exists a mapping αn defined
on the set Vn , which takes values in the set of functions mapping chord diagrams with n chords to some numbers.

v(K) = αn (v)(σ(K)) . (2.1)

However, not every function on the set of chord diagrams comes from a Vassiliev invariant. To verify this,
consider a singular knot and its chord diagram shown in Fig. 3.

tangle tangle

Figure 3: Special knot and its chord diagram.

It follows from Vassiliev skein relation that any Vassiliev invariant of a singular knot of the type shown in Fig. 3
must be zero. Therefore, all functions on the set of chord diagrams derived from Vassiliev invariants must vanish
on a diagram with a chord that does not intersect with any other chords. This condition is known as the one-term
relation [42] or framing-independence relation.
Another restriction can be derived by considering the expression in Fig. 4, where v is an arbitrary Vassiliev
invariant and singular knots where it is evaluated coincide everywhere except for a region in which they differ as
shown in Fig. 4.

v( )− v( )− v( )+ v( )= 0
Figure 4: Four terms knot relation.

If we resolve double points according to Vassiliev skein relation, then all terms are reduced. This results in the
four-term ratio in Fig. 4. From this relation on Vassiliev invariants, there follows a similar four-term relation on the
functions f ∈ Im αn originating from Vassiliev invariants (Fig. 5) [42, §4.2]:

f( ) − f( ) − f( ) + f( )= 0
Figure 5: 4-term relation on functions coming from Vassiliev invariants.

The derived conditions, namely, the one-term and the four-term relations, serve as a basis for the following defini-
tions.
Definition 2.6 (Space of chord diagrams A [33]). The space of chord diagrams An is a linear space generated
by all chord diagrams with n chords factorized by a subspace generated by four-term elements of the form:

− − +

and by one-term elements of the form:

L
The whole space of chord diagrams A = An .
n

5
 More compactly, we can write that An = An /h4T, 1Ti, where 4T and 1T are the four-term and the one-term
relations correspondingly.
Definition 2.7 (Weight systems W [33]). The space of weight systems Wn = A∗n = Hom (An , R) is a space of
linear functions on An .
Due to these definitions, the weight systems satisfy the one-term and four-term relations, which are conditions
for functions derived from Vassiliev invariants.

2.4 Vassiliev-Kontsevich theorem

In the previous paragraph, we have shown that any Vassiliev invariant generates a weight system, which is a linear
mapping αn : Vn → Wn . Due to the definition of Vassiliev invariants, the kernel of this mapping is Ker(αn ) = Vn−1 .
This implies that there exists an injective operator ᾱn : Vn /Vn−1 → Wn (see [42]).
M. Kontsevich proved in [34] that ᾱn is an isomorphism. In other words, any weight system w ∈ Wn can be
generated by a Vassiliev invariant v ∈ Vn /Vn−1 . A value of this invariant on a knot K is given by the following
formula:

v(K) = w(I(K)) . (2.2)

Here, I(K) is the Kontsevich integral of the knot K. Its formal definition can be found in [42, Chapter 8]. A
value of the Kontsevich integral of a knot K is a formal series of chord diagrams Dn,m in the space An , with the
coefficients of this series being finite-order invariants vn,m (K):
∞
X dimA
Xn
I(K) = ~n vn,m (K) Dn,m , (2.3)
n=0 m=1

where ~ is a formal variable. Furthermore, all Vassiliev invariants can be derived from the coefficients of the
Kontsevich integral, that is why it is also referred to as the universal Vassiliev invariant [42].
Thus, Vassiliev-Kontsevich theorem [42, §4.2.] is valid:
∞
M ∞
M
W= Wn ∼
= Vn /Vn−1 = V . (2.4)
n=0 n=0

2.5 Jacobi diagrams

Definition 2.8 (Space of Jacobi diagrams C [33, 42]). The space of (closed) Jacobi diagrams Cn is a linear
space generated by connected graphs having 2n trivalent vertices, as well as having a distinguished cycle (vertices
on this cycle are called external and the cycle is called Wilson loop), the remaining (internal) vertices are equipped
with a cyclic order of half-edges, modulo the STU relation:

= −
Figure 6: STU relation.

L
The whole space of Jacobi diagrams C = Cn .
n

Examples of Jacobi diagrams are presented in Fig. 7. In this and subsequent discussions, we assume that the
cyclic ordering of the half-edges at each vertex is consistent with the direction of the drawing surface. Special
consideration should be given to diagram d in Fig.7, which is equal to zero as an element of C due to the STU
relation.
Chord diagrams are a special type of Jacobi diagrams. In addition, the four-term relation follows from the STU
relation, so it automatically holds in C. On the other hand, by consistently applying the STU relation, any Jacobi
diagram can be reduced to a linear combination of chord diagrams. This linear combination generally depends on

6
 a b c d

Figure 7: Examples of Jacobi diagrams.

an order in which the STU relations are applied, but all such combinations are equivalent modulo the 4T relation.
The statement about this fact can be found in [42, §5.3.].
Thus, An /h4Ti ∼= Cn , and An ∼= Cn /h1Ti. This fact allows us to work with Jacobi diagrams instead of chord
diagrams. The advantage of Jacobi diagrams is that the STU relation is more convenient than the four-term relation.
In the space of Jacobi diagrams, there is a multiplication operation. The product of two Jacobi diagrams is
a new diagram that is created by cutting the Wilson loops from the multiplier diagrams and then gluing them
together. An example of this product is shown in Fig. 8.

· =

Figure 8: Product of Jacobi diagrams.

The space of primitive Jacobi diagrams P is a subspace of C linearly generated by Jacobi diagrams with a
connected internal graph. More strictly, the subspace of primitive diagrams is defined with respect to the co-
multiplication operation in C (see [42, §5.5.]). Any Jacobi diagram can be represented as a polynomial in primitive
diagrams.

2.5.1 AS and IHX relations

The STU relation imposes additional constraints on an internal graph of Jacobi diagrams. This can be seen in
the example of diagram d in Fig. 7. It can be shown that all Jacobi diagrams having such a loop vanish by STU
relations.
This statement is a special case of the AS relation shown in Fig. 9.

= −

Figure 9: AS relation.

There is also the second relation called the IHX relation, see Fig. 10.

= −

Figure 10: IHX relation.

The relations AS and IHX hold true with respect to the internal graphs of diagrams in the space C. These
relations can be derived from the STU relation. For more details on this statement, see [42, §5.2].

7
2.5.2 Open Jacobi diagrams
Definition 2.9 (Open Jacobi diagrams B [33]). The space of open Jacobi diagrams B is generated by trivalent
diagrams modulo the AS and IHX relations with unordered (unfixed) legs and with at least one leg in each connected
component. The legs are also referred to as univalent vertices.
The space B admits two gradings: by the number of legs l and by half the total number of vertices (both trivalent
and univalent) n.
The map ρ from B to C is defined as shown in Fig. 11. Open Jacobi diagrams are unfixed diagrams as they do
not carry any order of their legs. This is why the map ρ averages between all the different ways of attaching legs
to the Wilson loop.

P
ρ ( ) = 1
n!
s∈Sn s

Figure 11: Isomorphism ρ : B → C.

The proof of ρ : B → C being isomorphism, can be found in [42, §5.7]. The benefit of operating in B is an extra
grading by the number of legs, which is very convenient in our study in Section 4

2.6 Lie algebra weight systems

Lie algebras give rise to linear functions on C (weight systems). For a simple Lie algebra L with a nondegenerate
Killing form and its representation R the construction is the following. External vertices of a Jacobi diagram are
mapped to generators TRa of an algebra L in a representation R. Internal vertices correspond to structure constants
f abc , internal edges correspond to the metric tensor gab . The Wilson loop corresponds to taking the trace of the
product of generators divided by the dimension of the representation R.
The above construction belongs to Bar-Natan [33, 43] and is referred to as the weight system of a Lie algebra.
The consistency with the STU relation is ensured by the following equation: T a T b − T b T a = [T a , T b ] = f ab c T c.
An illustration of this equation can be seen in Fig. 12.

a b a b a b
= −
f ab c T c Ta Tb Tb Ta

Figure 12: STU relation as commutation in Lie algebra representation.

The weight system for an algebra L and a representation R is denoted1 ϕR L . The factor 1/ dim R in the definition
is necessary to ensure that a value of this weight system on a diagram without edges is 1 (in this case, the trace is
evaluated from the unit matrix with dimensions dim R × dim R). The weight system is then multiplicative, that is,
ϕR R R
L (D1 · D2 ) = ϕL (D1 )ϕL (D2 ), where D1 and D2 are Jacobi diagrams.
The AS and IHX relations acquire a simple interpretation in terms of the weight systems of Lie algebras.
Specifically, the AS relation is responsible for the anti-symmetry of the Lie bracket, [x, y] = −[y, x] ∀x, y ∈ L,
which, in the language of structure constants, can be expressed as f ab c = −f ba c . The IHX relation is equivalent to
the Jacobi identity, [[x, y], z] + [[y, z], x] + [[z, x], y] = 0 ∀x, y, z ∈ L, or f ad e f bc d + f bd e f ca d + f cd e f ab d = 0 in terms
of structure constants (see [42, §5.2]).
1 In what follows, we do not specify the representation R, i.e. work with an arbitrary representation, and omit the superscript R in

the denotation of a Lie algebra weight system ϕR L.

8
3 Vogel algebra
In this section, we describe main features of the framework developed by Pierre Vogel in his paper [35]. Advanced
features of this framework are given in Appendix A.
Definition 3.1 (Algebra Λ [35]). Λ is an algebra over Q generated by 3-legged fixed (i.e. the legs are numbered)
diagrams modulo AS and IHX relations antisymmetric with respect to permutations of legs. Multiplication in Λ is
given by insertion of one diagram into any vertex of the other diagram. Λ is a graded algebra. Its grading is given
by half the number of its trivalent vertices minus one half.
Remark 3.1. In what follows, all of the diagrams are fixed unless they are open Jacobi diagrams
The condition of antisymmetry with respect to permutations of three legs in the definition of Λ allows for
insertion of a diagram v̂ ∈ Λ into a 3-valent vertex with AS-relation. To prove that the multiplication is well-
defined, it is needed to show that the product is independent of a vertex to insert v̂ into. To do so, one needs to
prove the equality between two diagrams depicted in Fig. 13.

v̂ = v̂

Figure 13: Two equivalent diagrams.

To prove this relation the generalized form of the IHX relation is needed [42, Lemma 5.2.9.]. Applying the
generalized IHX relation we end up with:

v̂ v̂ v̂
v̂ − v̂ = − − + = −
v̂ v̂ v̂

By repeatedly applying the same transformations to the right side, we obtain:

v̂ − v̂ = v̂ − v̂

This relation immediately implies the condition for the correctness of the multiplication in Λ in Fig. 13. It is
important to note that this way one can multiply any connected diagram modulo the AS and IHX relations by
Λ. Such diagrams include primitive Jacobi diagrams starting from the second order Pn≥2 (at the zeroth and first
orders, there are no internal vertices where the element of Λ can be inserted). Hence, P≥2 obtain the structure of
Λ -module [35].
The multiplication operation in the algebra Λ is commutative. The proof of this fact can be found in the work
of Vogel [35, p. 1301]. Some elements of the Λ algebra are shown in Fig. 14. A proof of the x̂n diagrams to belong
to the Λ algebra can be found in [35, p. 1305].

1̂ = t̂ = x̂n =
···

Figure 14: Some elements of Λ.

Despite many findings of Pierre Vogel there are still many features of Λ that remain undiscovered. For example,
dimensions of Λ are only known up to degree 10 [42, §7.2].
Nevertheless, there exists a homomorphism ϕ : Q[t] ⊕ ωQ[t, σ, ω] → Λ from the polynomial algebra Q[t] ⊕
ωQ[t, σ, ω] to the algebra Λ (over the field of rational numbers Q). The derivation of this homomorphism can be
found in [35, Section 5]. The diagrams x̂n ∈ Λ lie in the image of this homomorphism and can be expressed in
terms of polynomials in t, σ and ω. In particular,
3 17 3
x̂1 = 2t̂ = 2ϕ(t), x̂2 = t̂2 = ϕ(t2 ), x̂3 = ϕ(4t3 − ω), x̂5 = ϕ(12t5 − t2 ω + σω). (3.1)
2 2 2

9
Remark 3.2. The variable t is of degree 1 and corresponds to the diagram t̂ = ϕ(t) shown in Fig. 14. The variable
ω is of degree 3, and the combination of diagrams ω̂0 = 38 t̂3 − 23 x̂3 corresponds to it in Λ. The variable σ of degree
2 is missing in “pure form” and must always be in the product with ω. In other words, there is no σ diagram in Λ,
but there are diagrams ω̂p = ϕ(σ p ω) that are known to satisfy ω̂p ω̂q = ω̂0 ω̂p+q , see [35, Theorem 5.6.].
Variables t, σ and ω come with the package of marked diagrams described in Appendix A.
It is known about this homomorphism ϕ [41] that at orders ≤ 10 it is a bijection. It was hypothesized in [41, 44]
that ϕ is an isomorphism. However, later Pierre Vogel has showed [35, Theorem 8.4.] that in Λ there is an element
P̂ of order 15, such that t̂P̂ = 0. Thus, there are zero divisors in Λ and the homomorphism ϕ has a nontrivial
kernel.

3.1 Characters on Λ
It is the moment to return to the weight systems of Lie algebras discussed in Section 2.6. Their design can be
extended to fixed diagrams with any number of legs, modulo AS and IHX relations. To do this, we put structure
constants f abc at vertices and the edges are associated with gab .
For simplicity, we denote this generalized weight system for a Lie algebra L as ΦL . ΦL maps a fixed diagram with
n legs to a rank-n tensor obtained from contracting the structure constants. To illustrate this, consider the diagram
Ψ shown in Fig. 15. ΦL (Ψ) = gij f ia c f jb d = (ΨL )ab
cd is a rank-4 tensor that corresponds to the endomorphism of
L⊗2 .
a c

b d

Figure 15: Diagram Ψ.

Now, in the case where v̂ is an element of Λ, ΦL (v̂) is a tensor of rank 3 obtained from the contraction of
structure constants leaving three free indices. Since there are no other primitive tensors of rank 3 in a Lie algebra
except for the structure constant f , it can be expected that ΦL (v̂) ∝ f . The proportionality coefficient is called
the character χL (v̂) of an element v̂. Then, for v̂ ∈ Λ and u being a diagram modulo AS and IHX relations, the
equality ΦL (v̂u) = χL (v̂)ΦL (u) holds. A rigorous and formal proof of this statement can be found in Vogel’s paper
[35, Theorem 6.1].
In order to find characters of elements of the Λ-algebra of the form t̂m ω̂0n ω̂p , we only need to know the values
χ (ω̂ )
of χL (t̂), χL (ω̂0 ), and χLL (ω̂p+1
p)
, as this relation does not depend on p since ω̂p ω̂q = ω̂0 ω̂p+q . These numbers are
denoted by tL , ωL and σL correspondingly. They depend on L and a normalization of the metric tensor. We can
say that they parameterize simple Lie algebras.

Remark 3.3. In Vogel’s paper [35] there was no distinction between t, t̂ and tL . In this paper we use v̂ when we
refer to an element of Λ, i.e. a 3-legged diagram modulo the AS and IHX relation, and vL stands for a numerical
value χL (v̂) with L being a simple Lie algebra.
There is also a more convenient parameterization using three numbers α, β and γ. The construction of this
parameterization is described in Sections 6 and 7 of Vogel’s paper [35]. Let us take a closer look at this construction.
The key concept behind parameterization through α, β, and γ is a cubic relation for the operator ΨL |Y which is
induced by ΨL . This operator acts on a space Y defined as S 2 L/Ω, where Ω is a one-dimensional space generated
by the Casimir element gij X i ⊗ X j with X i being some basis in L.
The cubic relation is given by:

(ΨL |Y )3 − tL (ΨL |Y )2 + (σL − 2t2L )ΨL |Y − (ωL − tL σL ) = 0 . (3.2)

It was derived in Vogel’s paper [35] and can be used to decompose the space Y into three subspaces corresponding
to the eigenvalues αL , βL , γL . These subspaces are denoted as Y (α), Y (β), and Y (γ). The three variables α, β and
γ are related to the variables t, σ and ω as follows:

α + β + γ = t, αβ + βγ + αγ = σ − 2t2 , αβγ = ω − tσ . (3.3)

The values αL , βL , γL for simple Lie algebras were calculated by Pierre Vogel in [35] and are shown in Table 2 in
the same normalization as in [45].

10
 Lie algebra L αL βL γL
sln −2 2 n
son −2 4 n−4
sp2n −2 1 n+2
G2 −2 10/3 8/3
F4 −2 5 6
E6 −2 6 8
E7 −2 8 12
E8 −2 12 20

Table 2: Vogel parameters.

Not only the characters on Λ, but also some other properties of Lie algebras can be expressed as functions of α,
β and γ. This concept is referred to as Vogel universality. For example, it is possible to express the dimension of a
Lie algebra by the following formula [35]:

(αL − 2tL )(βL − 2tL )(γL − 2tL )

dim L = . (3.4)
αL βL γL
There are other features that can be universalized:

1) dimensions of representations, appearing in decomposition of power of adjoint representation [46];

2) eigenvalues of higher Casimir operators [45, 47];
3) volume of simple Lie groups [48];
4) quantum knot invariants [49]; etc.

Notably, all known universal formulae work only in the adjoint sector, i.e. for tensor powers of the adjoint
representation. It was hypothesized that there exists some object that generalizes Lie algebras. It was referred to as
universal Lie algebra. Some attempts to describe it were made by Vogel in his unpublished works [41]. Currently,
it remains unknown whether it exists or not.

4 Construction of weight system kernel

In this section, we describe a method for construction of the kernel of Lie algebra weight system and apply it to
provide lowest order diagrams from ϕsln kernel.

4.1 Method
Our method to construct the kernel of a Lie algebra weight system is based on the technique introduced by Pierre
Vogel to construct a 17-order Jacobi diagram in the kernel of all Lie algebra weight systems (see [35]). The underlying
idea is to multiply some Jacobi diagram by a specific element of Λ that is killed by the corresponding character.
To construct such diagram, one should remember from the previous section that elements of Λ can be expressed
as symmetric polynomials in α, β, γ. The Lie algebra character χL maps this polynomial to its value on Vogel
parameters αL , βL , γL that can be found in Table 2.
It can be seen from Table 2 that for Lie algebra series sln there is a symmetric polynomial Psl = (α + β)(β +
γ)(α + γ) = 2tσ − ω − 2t3 that is killed by character χsl . Similar polynomials exist for son series and for the
exceptional Lie algebras [35]. These are polynomials in Q[t, σ, ω] that are not in Q[t] ⊕ ωQ[t, σ, ω]. Hence, they do
not correspond to any element of Λ as they all contain σ in pure form.
Our method is a modification that works for a polynomial P in Q[t, σ, ω] that is not in Q[t] ⊕ ωQ[t, σ, ω]. We
provide several ways to multiply P on Jacobi diagrams to create diagrams in the kernel of a specific Lie algebra
weight system.

[) ∈ Λ or by
1. Multiply any primitive Jacobi diagram of order ≥ 2 (in the Λ-algebra sense) by an element (ωP
the multiples of it.

11
 2. Multiply marked diagrams described in Appendix A by P with multiplication by σ acting as raising of the
total marking.
3. Multiply special linear combinations of diagrams given in Appendix A in Fig. 24 by P . These combinations
of diagrams allow for multiplication by σ only when considered under an action of Lie algebra weight system.

One also needs to check whether the resulting diagrams are linearly independent. To check linear independence,
it is very useful to operate in open Jacobi diagrams, because they have an extra grading by the number of legs.
Another way to check the linear independence of diagrams is to calculate some weight system that can distinguish
these diagrams.

4.2 Example for sln weight system

The estimates for dimensions of the image of sln weight system have been found in [39, 50]. Dimensions of Pn
and An were calculated in [33, 51, 52]. Subtraction yields the estimates for the dimensions of Ker ϕsln . All these
estimates are listed in Table 3.

n 1 2 3 4 5 6 7 8 9 10 11 12 13

dim Pn = 0 1 1 2 3 5 8 12 18 27 39 55 ?
dim ϕsln (Pn ) ≥ 0 1 1 2 3 5 8 11 16 22 30 42 53
dim Ker ϕsln (Pn ) ≤ 0 0 0 0 0 0 0 1 2 5 9 13 ?

dim An = 0 1 1 3 4 9 14 27 44 80 132 232 ?

dim ϕsln (An ) ≥ 0 1 1 3 4 8 11 19 25 39 50 75 95
dim Ker ϕsln (An ) ≤ 0 0 0 0 0 1 3 8 19 41 82 157 ?

Table 3: Estimates for the dimensions of the kernel of ϕsln .

In this subsection, the kernel of ϕsln at first few orders in both primitive and non-primitive cases are constructed
using the described method. It is demonstrated below that the inequalities in Table 3 at the lowest orders turn out
to be nicely saturated, what also validates the results of our previous study of [39, 50].

4.2.1 Primitive diagrams in kernel

For the slN weight system we have polynomial Psl = (α + β)(β + γ)(α + γ) = 2tσ − ω − 2t3 that vanishes when
taken on values of the parameters corresponding to sln Lie algebra, see Table 2.
\
(ωP 4 16 6 8 3 4 2
sl ) is an element of Λ that can be expressed in terms of x̂n using (3.1) as 3 t̂x̂5 + 9 t̂ − 3 t̂ x̂3 − 9 x̂3 , see
Fig. 16. That element is killed by sln weight system and it can be multiplied by any Jacobi diagram to create a
diagram in the kernel of the sln weight system.

2 1
3 + 36 − 31 − 94

dsl — the first element of Λ that is killed by χsl .

Figure 16: ωP

The first such diagram can be obtained by taking the product (ωP \ sl )− −. Other diagrams are depicted in
Table 4. Note that they are all open Jacobi diagrams that have an extra grading by the number of legs l. Relation
between open and closed Jacobi diagrams is described in Section 2.5.2.
The 6-legged diagram in Table 4 can be multiplied by σ because of the top formula in Fig. 24. The action of σ
on this diagram is shown in Fig. 17.
Linear independence of the diagrams depicted in Table 4 can be checked by calculating the value of son weight
system on these diagrams. One can compare the number of these diagrams with the estimates in Table 3 and see
that the constructed diagrams fully generate Ker ϕsln among the primitive Jacobi diagrams up to the 10-th order.

12
 n
8 9 10
l
dsl )
t̂2 (ωP
2 dsl )
(ωP dsl )
t̂(ωP
\
(σωPsl )

t̂Psl 0
4 Psl 0
dsl )
(ωP

6 Psl ( −t̂ )

Table 4: Diagrams in the kernel of ϕsln .

σ( − t̂ ) = 0
− ω̂

Figure 17: Action of σ on the 6-legged diagram.

4.2.2 Non-primitive diagrams in kernel

As can be seen from Table 3, non-primitive diagrams first appear in the kernel of sln weight systems at the 6-th
order. In paper [39], the element of C6 in the kernel of ϕsln is written explicitly:

Figure 18: Non-primitive element of Ker ϕsln at the 6-th order.

Making use of Λ, this combination of diagrams can be rewritten as (− −)(t̂2 − −) − (t̂− −)(t̂− −). Now it is
easy to see that inclusion of this diagram in the kernel of ϕsln is explained

by multiplicativity of Lie algebra weight
systems. Indeed, for any Lie algebra weight system, ϕL (− −)(t̂2 − −) = ϕL (− −)ϕL (t̂2 − −) = t2L ϕL (− −)2 .


On the other side, ϕL (t̂− −)(t̂− −) = t2L ϕL (− −)2 , and we get the desired equality.
The same way, one can construct other non-primitive elements of space of closed Jacobi diagrams C in Ker ϕsln
at higher orders. At the seventh order there are two such diagrams, both are depicted in Fig. 19.

− −

Figure 19: Two elements of C7 in the kernel of ϕsln .

Nevertheless, according to Table 3, at 7-th order there are at most 3 such linearly independent elements. The
third diagram can be constructed using the bottom formula from Fig. 24:

−(ω̂ + 2t̂3 ) + 34 ω̂ t̂ +2t̂ +(t̂3 − 12 ω̂)

0

Figure 20: The third nonprimitive element of Ker ϕsln at 7-th order.

13
5 Relation between Chern-Simons invariants and Kontsevich integral
In this section, we discuss the connection of the obtained results with the 3D quantum Chern-Simons field theory.
Detailed description of the connection between the Chern-Simons theory and Vassiliev invariants can be found in
particular in [53].
The 3-dimensional Chern-Simons action with a simple compact Lie gauge group G for the vector field Aµ = Aaµ Ta ,
where Ta are generators of the corresponding Lie algebra L, is given by:
Z  
κ 2
S[A] = d3 x ǫµνρ Tr Aµ ∂ν Aρ + Aµ Aν Aρ . (5.1)
4π R3 3
Correlators in the Chern–Simons theory can be obtained from the gauge invariant Wilson loops by expansion of
the path ordered exponent:
 I  I
1 trR (Ta )
hWR (K)i = trR Pexp Aµ dxµ =1 + Aaµ (x) dxµ +
dim(R) K CS dim(R) K
I Z x2 dim (5.2)
trR (Ta1 Ta2 ) X∞ XAn
+ dxµ2 2 dxµ1 1 Aaµ11 (x1 )Aaµ22 (x2 ) + · · · = ~n R
vn,m (K) Gn,m
qdim(R) K 0 n=0 m=1

where I Z  (m) 
(m)
µ2n a1 a2n
vn,m (K) ∼ dx2n ... dxµ1 1 Aµ1 (x1 ) . . . Aµ2n (x2n ) , (5.3)

and  
R
Gn,m ∼ trR Ta(m) Ta(m) . . . Ta(m) . (5.4)
1 2 2n

The answer (5.2) is very similar to the Kontsevich integral (2.3). The only difference is that in the Kontsevich
R
integral instead of group factors Gn,m , there are basic elements Dn,m of the algebra of chord diagrams. Actually,
this is not just a similarity. In fact, Lie algebras weight systems map chord diagrams to group factors of the Wilson
loop expansion (5.2), and due to linearity of the mapping

hWR (K)i = ϕR
L (I(K)) . (5.5)
Thus, the results on the kernel of Lie algebras weight systems acquire the following meaning for the Chern–Simons
R
theory. In a fixed Lie algebra L, there appear relations between group factors Gn,m at a fixed order n. Therefore,
not all correlators vn,m (K) can be extracted from the Wilson loop (5.2) for the Chern–Simons theory with a fixed
gauge group. In particular, one primitive correlator is absent already at the 8-th order in the case of SU (N ) gauge
group2 , see Section 4.2.1.

6 On knot detection
Now let is discuss how the obtained results affect the issue of distinguishing knots with the help of colored HOMFLY
polynomials. We analyze this question using a specific example.
Let us consider the simplest primitive element D8 := (ωP \ sl )− − transformed to the element of the algebra
of closed Jacobi diagrams, which lies in the kernel of the sln weight system. Using the isomorphism between the
algebra of closed Jacobi diagrams and the algebra of chord diagrams, we rewrite D8 through chord diagrams using
the STU relation (see Fig. 6) and obtain their definite linear combination D8′ . Knowing the explicit form of D8′ , we
construct a linear combination of singular knots using the correspondence described in Section 2.5. Applying the
Vassiliev skein relation (see Fig. 1), from the obtained combination of singular knots, we obtain a linear combination
of ordinary (not singular) knots, which we denote by K8 . In addition, using the Kontsevich integral, we find the
Vassiliev invariant V8 of the 8-th order corresponding the chord diagram D8′ . Since the value of the eighth order
Vassiliev invariant on a singular knot with 8 double points depends only on its chord diagram, all other Vassiliev
invariants of order 8 vanish on K8 . Also, all Vassiliev invariants at levels below 8 vanish on K8 (this follows from
the definition of Vassiliev invariants). Thus, since the Vassiliev invariant V8 is absent in quantum sln invariants (i.e.,
in colored HOMFLY polynomials), all expansion coefficients up to level 8 inclusively of any HOMFLY polynomial
for a combination K8 of knots is equal to 0.
2 Wilson loops in the 3D Chern–Simons theory with SU (N ) gauge group are quantum knot invariants called colored (by a represen-

tation R of sln Lie algebra) HOMFLY polynomials.

14
 How to pass from some linear combination of knots to just a difference between two knots K − K ′ ? We do not
know an algorithmic answer to this question. However, from a chord diagram of the n-th order, one can construct
infinitely many singular knots with n double points. Further, these singular knots can be combined with each other
(added and subtracted with arbitrary coefficients). It may well happen that as a result of these manipulations, we
get a linear combination of knots of the form K − K ′ for some knots K and K ′ . For small orders, this can be done
manually. For example, for the second order it is easy to obtain that V2 (41 + 31 ) = 0, i.e. for the trefoil knot 31
and for the figure eight knot 41 absolute value of the second order Vassiliev invariant coincide. Therefore, we can
imagine that there are knots for which all Vassiliev invariants coincide up to and including the 7-th order, and at
the 8-th order they differ by the only one Vassiliev invariant. In this case, these knots are not distinguished by
any colored HOMFLY polynomials up to and including the 8-th order (meaning the coefficients of the perturbative
expansion (5.2)).
Do such knots K and K ′ differ by Vassiliev invariants of the 9-th order or higher? In principle, there is no reason
to believe that any primitive Vassiliev invariant of order 9 or higher cannot distinguish between these two knots
or any other singular knot constructed using the procedure described above. Some additional considerations are
required to make them vanish. Generalizing this, we can say that both colored HOMFLY polynomials and Vassiliev
invariants can simultaneously be complete knot invariants as it is hypothesized. And they can be expressed through
each other in a very complex way. At least today, this does not contradict anything including the result of P. Vogel.

Acknowledgements
We are grateful for enlightening discussions to A. Morozov and N. Tselousov.
This work was funded by the RSF grant No.24-12-00178.

References
[1] V. F. R. Jones. “The Jones Polynomial”. In: Encyclopedia of Mathematical Physics. Oxford: Academic Press,
2006, pp. 179–187. doi: 10.1016/B0-12-512666-2/00257-1.
[2] V. F. R. Jones. “Hecke Algebra Representations of Braid Groups and Link Polynomials”. In: Annals of
Mathematics 126.2 (1987), pp. 335–388. doi: 10.2307/1971403.
[3] V. F. R. Jones. “A polynomial invariant for knots via von Neumann algebras”. In: Bull. Amer. Math. Soc.
12.1 (1985), pp. 103–111. doi: 10.1090/S0273-0979-1985-15304-2.
[4] P. Freyd et al. “A new polynomial invariant of knots and links”. In: Bull. Amer. Math. Soc. 12.2 (1985),
pp. 239–246. doi: 10.1090/S0273-0979-1985-15361-3.
[5] J.H. Przytycki and K.P. Traczyk. “Invariants of links of Conway type”. In: Kobe J. Math. 4 (1987), pp. 115–
139. arXiv: 1610.06679 [math.GT].
[6] Louis H Kauffman. “An invariant of regular isotopy”. In: Transactions of the American Mathematical Society
318.2 (1990), pp. 417–471.
[7] V.G. Turaev and O.Y. Viro. “State sum invariants of 3-manifolds and quantum 6j-symbols”. In: Topology
31.4 (1992), pp. 865–902. issn: 0040-9383. doi: 10.1016/0040-9383(92)90015-A.
[8] Nicolai Yu Reshetikhin and Vladimir G Turaev. “Ribbon graphs and their invaraints derived from quantum
groups”. In: Communications in Mathematical Physics 127 (1990), pp. 1–26. doi: 10.1007/BF02096491.
[9] RB Zhang and MD Gould. “Universal R matrices and invariants of quantum supergroups”. In: Journal of
Mathematical Physics 32.12 (1991), pp. 3261–3267. doi: 10.1063/1.529487.
[10] Lev Rozansky and Herbert Saleur. “S-and T-matrices for the super U (1, 1) WZW model application to
surgery and 3-manifolds invariants based on the Alexander-Conway polynomial”. In: Nuclear Physics B 389.2
(1993), pp. 365–423. arXiv: hep-th/9203069.
[11] Nathan Geer. “The Kontsevich integral and quantized Lie superalgebras”. In: Algebraic & Geometric Topology
5.3 (2005), pp. 1111–1139. arXiv: math/0411053.
[12] Oleg Viro. “Quantum relatives of the Alexander polynomial”. In: St. Petersburg Mathematical Journal 18.3
(2007), pp. 391–457. arXiv: math/0204290.
[13] Hoel Queffelec and Antonio Sartori. “A note on gl (m— n) link invariants and the HOMFLY–PT polynomial”.
In: Knots, Low-Dimensional Topology and Applications 284 (2019), pp. 277–286. arXiv: 1506.03329.

15
[14] Albert S Schwarz. “The partition function of degenerate quadratic functional and Ray-Singer invariants”. In:
Letters in Mathematical Physics 2.3 (1978), pp. 247–252. doi: 10.1007/BF00406412.
[15] Albert S Schwarz. “The partition function of a degenerate functional”. In: Communications in Mathematical
Physics 67 (1979), pp. 1–16. doi: 10.1007/BF01223197.
[16] Albert S Schwarz. “New topological invariants arising in the theory of quantized fields”. In: Talk at the Baku
International Topological Conf. Abstracts (part II) (1987).
[17] Edward Witten. “Quantum field theory and the Jones polynomial”. In: Commun. Math. Phys. 121 (1989).
Ed. by Asoke N. Mitra, pp. 351–399. doi: 10.1007/BF01217730.
[18] Edward Witten. “Topological quantum field theory”. In: Communications in Mathematical Physics 117.3
(1988), pp. 353–386. doi: 10.1007/BF01223371.
[19] Shiin-Shen Chern and James Simons. “Characteristic forms and geometric invariants”. In: Annals Math. 99
(1974), pp. 48–69. doi: 10.2307/1971013.
[20] E. Guadagnini, M. Martellini, and M. Mintchev. “Perturbative aspects of the Chern-Simons field theory”. In:
Phys. Lett. B 227 (1989), pp. 111–117. doi: 10.1016/0370-2693(89)91291-4.
[21] E. Guadagnini, M. Martellini, and M. Mintchev. “Wilson lines in Chern-Simons theory and link invariants”.
In: Nuclear Physics B 330.2 (1990), pp. 575–607. issn: 0550-3213. doi: 10.1016/0550-3213(90)90124-V.
[22] R. K. Kaul and T. R. Govindarajan. “Three-dimensional Chern-Simons theory as a theory of knots and links”.
In: Nucl. Phys. B 380 (1992), pp. 293–333. doi: 10.1016/0550-3213(92)90524-F. arXiv: hep-th/9111063.
[23] P. Rama Devi, T. R. Govindarajan, and R. K. Kaul. “Three-dimensional Chern-Simons theory as a the-
ory of knots and links (III). Compact semi-simple group”. In: Nucl.Phys.B 402 (1993), pp. 548–566. doi:
10.1016/0550-3213(93)90652-6. arXiv: hep-th/9212110.
[24] P. Ramadevi, T. R. Govindarajan, and R. K. Kaul. “Knot invariants from rational conformal field theories”.
In: Nucl.Phys.B 422 (1994), pp. 291–306. doi: 10.1016/0550-3213(94)00102-2. arXiv: hep-th/9312215.
[25] P. Ramadevi, T. R. Govindarajan, and R. K. Kaul. “Representations of composite braids and invariants for
mutant knots and links in Chern-Simons field theories”. In: Mod. Phys. Lett. A 10 (1995), pp. 1635–1658.
doi: 10.1142/S0217732395001769. arXiv: hep-th/9412084.
[26] Hirosi Ooguri and Cumrun Vafa. “Knot invariants and topological strings”. In: Nucl.Phys.B 577 (2000),
pp. 419–438. doi: 10.1016/S0550-3213(00)00118-8. arXiv: hep-th/9912123.
[27] J. M. F. Labastida and Marcos Marino. “Polynomial Invariants for Torus Knots and Topological Strings”. In:
Commun. Math. Phys. 217 (2001), pp. 423–449. doi: 10.1007/s002200100374. arXiv: hep-th/0004196.
[28] J. M. F. Labastida, Marcos Marino, and Cumrun Vafa. “Knots, links and branes at large N”. In: JHEP 11
(2000), p. 007. doi: 10.1088/1126-6708/2000/11/007. arXiv: hep-th/0010102.
[29] Jose M. F. Labastida and Marcos Marino. “A new point of view in the theory of knot and link invariants”.
In: Journal of Knot Theory and Its Ramifications 11.02 (2002), pp. 173–197. arXiv: math/0104180.
[30] Marcos Marino and Cumrun Vafa. “Framed knots at large N”. In: Contemp. Math. 310 (2002), pp. 185–204.
arXiv: hep-th/0108064.
[31] Joan S Birman and Xiao-Song Lin. “Knot polynomials and Vassiliev’s invariants”. In: Inventiones mathemat-
icae 111.1 (1993), pp. 225–270.
[32] VA Vassiliev. “Cohomology of knot spaces”. In: (1990).
[33] Dror Bar-Natan. “On the Vassiliev knot invariants”. In: Topology 34.2 (1995), pp. 423–472.
[34] Maxim Kontsevich. “Vassiliev’s knot invariants”. In: Adv. in Sov. Math 16.2 (1993), pp. 137–150.
[35] Pierre Vogel. “Algebraic structures on modules of diagrams”. In: Journal of Pure and Applied Algebra 215.6
(2011), pp. 1292–1339.
[36] Zhuoke Yang. “New approaches to gl N weight system”. In: Izvestiya: Mathematics 87.6 (2023), pp. 150–166.
arXiv: 2202.12225.
[37] S. K. Lando M. E. Kazarian. “Weight systems and invariants of graphs and embedded graphs”. In: Russian
Math. Surveys 77.5 (2022), pp. 893–942. arXiv: 2302.12153.
[38] M Alvarez and JMF Labastida. “Numerical knot invariants of finite type from Chern-Simons perturbation
theory”. In: Nuclear Physics B 433.3 (1995), pp. 555–596. arXiv: hep-th/9407076.

16
[39] E Lanina, A Sleptsov, and N Tselousov. “Implications for colored HOMFLY polynomials from explicit formulas
for group-theoretical structure”. In: Nuclear Physics B 974 (2022), p. 115644. arXiv: 2111.11751.
[40] Jens Lieberum. “On Vassiliev invariants not coming from semisimple Lie algebras”. In: Journal of Knot Theory
and Its Ramifications 8.05 (1999). arXiv: q-alg/9706005.
[41] Pierre Vogel. “The universal Lie algebra”. preprint available at https://webusers.imj-prg.fr/~pierre.vogel/.
1999.
[42] Sergei Chmutov, Sergeı̆ Vasil’evich Duzhin, and Jacob Mostovoy. Introduction to Vassiliev knot invariants.
Cambridge University Press, 2012.
[43] Dror Bar-Natan. “Weights of Feynman diagrams and the Vassiliev knot invariants”. preprint available at
https://www.math.toronto.edu/drorbn/papers/weights/. 1991.
[44] Jan A Kneissler. “On spaces of connected graphs II: Relations in the algebra Λ”. In: Journal of Knot Theory
and Its Ramifications 10.05 (2001), pp. 667–674. arXiv: math/0301019.
[45] RL Mkrtchyan, AN Sergeev, and AP Veselov. “Casimir eigenvalues for universal Lie algebra”. In: Journal of
mathematical physics 53.10 (2012). arXiv: 1105.0115.
[46] Joseph M Landsberg and Laurent Manivel. “A universal dimension formula for complex simple Lie algebras”.
In: Advances in Mathematics 201.2 (2006), pp. 379–407. arXiv: math/0401296.
[47] Aleksei Petrovich Isaev and Alexander Alekseevich Provorov. “Split Casimir operator and solutions of the
Yang–Baxter equation for the and Lie superalgebras, higher Casimir operators, and the Vogel parameters”.
In: Theoretical and Mathematical Physics 210.2 (2022), pp. 224–260. arXiv: 2201.06148.
[48] Hovhannes M Khudaverdian and Ruben L Mkrtchyan. “Universal volume of groups and anomaly of Vogel’s
symmetry”. In: Letters in Mathematical Physics 107 (2017), pp. 1491–1514. arXiv: 1602.00337.
[49] A Mironov, R Mkrtchyan, and A Morozov. “On universal knot polynomials”. In: Journal of High Energy
Physics 2016.2 (2016), pp. 1–35. arXiv: arXiv:1510.05884.
[50] E Lanina, A Sleptsov, and N Tselousov. “Chern-Simons perturbative series revisited”. In: Physics Letters B
823 (2021), p. 136727. arXiv: 2105.11565.
[51] VA Vassiliev. “Homological invariants of knots: algorithm and calculations”. In: Preprint Inst. of Applied
Math.(Moscow) 90 (1990), p. 23.
[52] Jan A Kneissler. “The number of primitive Vassiliev invariants up to degree twelve”. 1997. arXiv: q-alg/9706022.
[53] Petr Dunin-Barkowski, Alexey Sleptsov, and Andrey Smirnov. “Kontsevich integral for knots and Vassiliev in-
variants”. In: International Journal of Modern Physics A 28.17 (2013), p. 1330025. doi: 10.1142/S0217751X13300251.
arXiv: 1112.5406.

17
Appendices
A Vogel algebra: advanced features
In this appendix, more sophisticated attributes of Vogel algebra are given. They all were in Vogel’s paper [35],
but many of them were not written explicitly. Among those are the marked diagrams that stand for some linear
combinations of regular trivalent diagrams modulo the AS and IHX relations. They contain a 4-valent vertex
marked by some non-negative integer number p ∈ N0 . Marked diagrams are defined as follows:

0 = 2t̂2 +2t̂ − 34 − 23

   
8t̂3 4t̂3 2
ε0 = = ω̂ − 3 − ω̂ − 3 + 23t̂ + 103 t̂ − 34 − 23
0

Figure 21: Definition of diagrams marked by zero.

2 2
p+1 = 3 p
− 32 p + 89t̂ p +
4t̂ω̂p
3

εp = = − ω̂p + ω̂p − 43 t̂ + 23 t̂
p p p p

Figure 22: Definition of diagrams marked by p > 0.

The diagram ω̂p in terms of these marked diagrams can be displayed as follows:

Figure 23: ω̂p as a marked diagram.

These marked diagrams have an important property. Namely, a connected diagram having more than one
marked point depends only on its total marking. This is why the relation ω̂p ω̂q = ω̂0 ω̂p+q holds. And this is also a
reason for the multiplication by σ to make sense as an operation that raises the total marking by 1.
The diagram εp defined above is symmetric with respect to the permutations of legs. Hence, it corresponds to
itself as an open Jacobi diagram.
For a Lie algebra weight system there are two additional relations that involve marked diagrams. They provide
a way to multiply diagrams by σ but only under the action of a Lie algebra weight system, see Fig. 24.

σL ΦL ( −tL ( − ))= Φ ( L 0
−ωL ( − ))
σL ΦL ( (
− t3L + )− ( 2t2L
3
+ + ))= Φ ( L 0
− 2tL3ωL ( + + ))
Figure 24: Additional relations that work for a Lie algebra weight systems.

18