---
title: 2-Loop Trace Map in Topology, Dynamics & Bicategories
url: https://www.emergentmind.com/topics/2-loop-trace-map
type: topic
---

# 2-Loop Trace Map in Topology, Dynamics & Bicategories

Searching arXiv for papers specifically on the 2-loop trace map and closely related formulations.
The expression **2-loop trace map** appears in several mathematically distinct settings. In the topology of mapping class groups, it denotes the \(r=2\) summand of the refined hairy-graph trace on the Johnson cokernel, where it captures stable \(\mathrm{Sp}\)-components in degree \(6\) that are invisible to the classical Enomoto–Satoh trace [2508.19041]. In smooth dynamics and character-variety theory, the closely related phrase **2-Loop (Fibonacci) trace map** refers to the polynomial automorphism \(T(x,y,z)=(2xy-z,x,y)\), arising from \(\mathrm{SL}(2,\mathbb C)\) trace identities and preserving the Fricke–Vogt cubic [1507.07912]. In bicategorical trace theory, a **two-loop trace** is formed by composing an inner cotrace with an outer shadow-trace, producing a torus-shaped closure operation on a 2-endomorphism [2307.13070].

## 1. Johnson cokernels and the loop decomposition

In the mapping-class-group setting, one begins with a once-bordered surface \(\Sigma_{g,1}\) of genus \(g\), its mapping-class group
\[
MCG=\pi_0\,\mathrm{Diff}^+(\Sigma_{g,1}),
\]
and the \(n\)-th Johnson homomorphism
\[
\tau_n:I_{g,1}[n]\to \mathrm{Hom}(H,\wedge^{n+1}H),
\]
where \(H\cong \mathbb Q^{2g}\) is a symplectic vector space with form \((\cdot,\cdot)\). Hain’s theorem states that, in the stable range \(g\gg n\), the images of all \(\tau_n\) generate the Lie algebra
\[
h=\bigoplus_{n\ge 1} h(n)\subset \mathrm{Der}^+(\mathcal L H)
\]
of symplectic derivations of the free Lie algebra on \(H\). If one sets
\[
m=\mathrm{Lie}\langle h(1)\rangle\subset h,
\qquad
cok(n)=h(n)/m(n),
\]
then \(cok(n)\) is the degree-\(n\) Johnson cokernel [2508.19041].

The loop grading enters through the Conant–Kassabov–Vogtmann hairy-graph trace
\[
Tr:h(n)\to C_1H(n),
\]
which splits according to the number \(r\) of dotted edges:
\[
Tr=\bigoplus_{r=0}^{\lfloor n/2\rfloor+1} Tr_r:h(n)\to C_{1,r}H(n)\cong T_{r,n+2-2r}.
\]
After projection to the top-level \(\mathrm{Sp}\)-summands in each tensor factor \(H^{\otimes(n+2-2r)}\), one obtains
\[
\widetilde{Tr}_r:h(n)/m(n)\to \widetilde{\Omega}_{r,\langle n+2-2r\rangle}.
\]
Conant proved, for \(g\gg n\), that
\[
cok(n)\cong \bigoplus_{r=1}^{\lfloor n/2\rfloor+1}\widetilde{\Omega}_{r,\langle n+2-2r\rangle}.
\]
Accordingly, \(\widetilde{\Omega}_{r,\langle n+2-2r\rangle}\) is called the \(r\)-loop part of the Johnson cokernel.

The \(r=1\) term is the classical Enomoto–Satoh trace. In this case,
\[
\widetilde{\Omega}_{1,n}\cong H^{\langle n\rangle}/D_{2n},
\]
where \(D_{2n}\) is the dihedral action on \(H^{\otimes n}\), and the composite
\[
cok(n)\to H^{\otimes n}/D_{2n}\to H^{\langle n\rangle}/D_{2n}
\]
is precisely the Enomoto–Satoh trace. It is injective for \(n\le 5\), but already for \(n=6\) it has nonzero kernel.

## 2. The \(\theta\)-graph presentation of the 2-loop trace map

The 2-loop trace isolates the summand
\[
\widetilde{\Omega}_2=\bigoplus_{m\ge 0}\widetilde{\Omega}_{2,m},
\]
where \(m=n+2-4=n-2\) is the number of hairs. Let
\[
\mathcal T=\bigoplus_{m\ge 0} H^{\otimes m}
\]
be the tensor algebra on \(H\), equipped with coproduct \(\Delta\) and antipode \(\overline{(\ )}\) reversing order. For \(t,u,v,w\in \mathcal T\), one defines a 2-loop hairy graph
\[
\Theta(t,u,v,w)\in T_{2,|t|+|u|+|v|+|w|}
\]
by attaching \(t,u\) to the two horizontal edges of a \(\theta\)-graph and \(v,w\) to the two vertical loops [2508.19041].

As a vector space, \(\widetilde{\Omega}_2\) is generated by the \(\Theta(t,u,v,w)\), multilinear in each slot, subject only to four relations:

| Relation | Formula |
|---|---|
| Q-multilinearity | \(\Theta\) is \(\mathbb Q\)-multilinear in \(t,u,v,w\in\mathcal T\) |
| Core-symmetry | \(\Theta(t,u,v,w)=\Theta(\bar v,\bar w,\bar t,\bar u)=\Theta(\bar u,\bar t,\bar w,\bar v)\) |
| Handle-balance | \(\Theta(t,u,v,wa)-\Theta(t,u,av,w)=\Theta(t,au,v,w)-\Theta(ta,u,v,w)\) |
| Generalized IHX | \(\Theta(t,u,v,w)+\Theta(\bar t',t''u,\bar v',wv'')+\Theta(tu'',\bar u',w''v,\bar w')=0\) |

Here \(\Delta(t)=t'\otimes t''\), and similarly for the other tensor inputs. Under the trace construction, these \(\Theta\)-graphs furnish a basis of the 2-loop part in each homological degree.

The refined trace
\[
\widetilde{Tr}_2:cok(n)\to \widetilde{\Omega}_{2,n-2}
\]
is manifestly \(\mathrm{Sp}(2g)\)-equivariant, because the \(H\)-colorings transform in the defining representation of \(H\), and it shifts grading by \(4\), with hair number \(m=n-2\).

## 3. Degree \(6\), detection beyond Enomoto–Satoh, and refined versus unrefined traces

The first nontrivial role of the 2-loop trace map occurs in degree \(6\). In the stable range \(g\gg 6\), one has
\[
T_{2,4}=C_{1,2}H(6)\cong
6[4]\oplus 9[31]\oplus 12[2^2]\oplus 9[21^2]\oplus 6[1^4]
\quad\text{(GL-types)},
\]
and after quotienting by the \(\beta\)-image,
\[
\widetilde{\Omega}_{2,4}\cong
2[4]\oplus 3[31]\oplus 3[2^2]\oplus 3[21^2]\oplus 2[1^4].
\]
On the other hand, Theorem 1.3 of Morita–Sakasai–Suzuki identifies the kernel of the 1-loop trace in degree \(6\) as
\[
\mathrm{Ker}(Tr_1)= [1^4]_{Sp}\oplus [1^2]_{Sp}\oplus [0]_{Sp}.
\]
The 2-loop trace vanishes on no nonzero vector from each of these three summands, so
\[
\widetilde{Tr}_2|_{\mathrm{Ker}\,Tr_1}:h(6)\cap \mathrm{Ker}\,Tr_1\to \widetilde{\Omega}_{2,4}
\]
detects all of them. Explicit maximal-vector hairy graphs \(X_S\), \(X_{S'}\), and \(X_{S''}\) are written down so that their 2-loop traces lie in the canonical maximal-vectors of \([0]\), \([1^2]\), and \([1^4]\), respectively. Together with injectivity of the full trace \(Tr=Tr_1\oplus Tr_2\oplus Tr_3\oplus Tr_4\) in degree \(6\), this yields the theorem that, for \(g\gg 6\),
\[
(Tr_1\oplus Tr_2):cok(6)\to \widetilde{\Omega}_{1,6}\oplus \widetilde{\Omega}_{2,4}
\]
is injective [2508.19041].

The comparison with Conant’s earlier construction is structurally important. The unrefined target \(\Omega_r\) is strictly smaller than \(\widetilde{\Omega}_r\), and the corresponding map
\[
Tr^C:h\to \Omega=\bigoplus_r \Omega_r
\]
vanishes on \(m\). For \(r=2\), \(Tr^C_2\) factors through the 1-loop map \(Tr^C_1\): there is a natural dotted-edge insertion map
\[
\Phi:\Omega_1\to \Omega_2
\]
such that
\[
\Phi\circ Tr^C_1 = 3\,Tr^C_2.
\]
Thus \(Tr^C_2\) carries no new information beyond the 1-loop trace. By contrast, the refined hairy-graph construction \(\widetilde{Tr}_2\) does not factor through \(\widetilde{Tr}_1\), and it genuinely detects new 2-loop obstructions.

## 4. The 2-Loop (Fibonacci) trace map in dynamics and character varieties

A different use of the phrase refers to the polynomial automorphism
\[
T:\mathbb R^3\to \mathbb R^3,
\qquad
T(x,y,z)=\bigl(2xy-z,\;x,\;y\bigr),
\]
with inverse
\[
T^{-1}(x,y,z)=(y,\;z,\;2yz-x).
\]
This map arises from trace identities in \(\mathrm{SL}(2,\mathbb C)\). If \(\alpha,\beta\) are free generators of \(F_2\) and
\[
\rho:F_2\to \mathrm{SL}(2,\mathbb C)
\]
is a representation, then, up to conjugacy, one may write
\[
x=\tfrac12\mathrm{Tr}(\rho(\alpha)),\qquad
y=\tfrac12\mathrm{Tr}(\rho(\beta)),\qquad
z=\tfrac12\mathrm{Tr}(\rho(\alpha\beta)).
\]
The classical Fricke–Vogt identity gives the cubic
\[
J(x,y,z)=x^2+y^2+z^2-2xyz,
\]
which remains invariant under the action of \(\mathrm{Aut}(F_2)\). The induced polynomial map on \((x,y,z)\)-space is precisely \(T\), and
\[
J(T(x,y,z))=J(x,y,z).
\]
Hence every level set
\[
S_V=\{(x,y,z)\in \mathbb R^3:J(x,y,z)=V\}
\]
is \(T\)-invariant [1507.07912].

The geometry of the invariant cubic depends strongly on \(V\). At \(V=0\), one obtains the Cayley cubic
\[
S_0:x^2+y^2+z^2-2xyz=0,
\]
with four ordinary conical singularities
\[
(1,1,1),\ (-1,-1,1),\ (1,-1,-1),\ (-1,1,-1).
\]
For \(V>0\), \(S_V\) is a smooth affine cubic, diffeomorphic to a four-punctured sphere; for \(V<0\) down to \(-1\), there is one compact sphere-like component together with four unbounded sheets. From the representation-theoretic viewpoint, \(S_V\) parametrizes conjugacy classes of \(\mathrm{SL}(2)\)-representations of the once-punctured torus or \(4\)-punctured sphere with fixed commutator-trace, and the coordinate ring of \(S_V\) is the classical Markoff algebra.

Its dynamical features are equally prominent. The Fibonacci substitution
\[
\alpha\mapsto \alpha\beta,\qquad \beta\mapsto \alpha
\]
induces the recurrence
\[
x_{n+1}=2x_nx_{n-1}-x_{n-2},
\]
whose second-order companion map in \((x_n,x_{n-1},x_{n-2})\)-space is exactly \(T\). For each \(V>0\), the non-wandering set of \(T|_{S_V}\) is a compact Cantor set of hyperbolic type, topologically conjugate to a subshift, with constant topological entropy
\[
\log\!\bigl(\tfrac{\sqrt5+1}{2}\bigr).
\]
For \(V\) slightly negative, \(T|_{\mathbb S_V}\) has persistent homoclinic tangencies, stochastic seas of full Hausdorff dimension, and infinitely many elliptic islands. The map is also related to the Fibonacci Hamiltonian, whose spectrum is encoded by \(T|_{S_{\lambda^2/4}}\), and at \(V=0\) its central part is semiconjugate to the toral automorphism
\[
\mathcal A=
\begin{pmatrix}
1&1\\
1&0
\end{pmatrix}
\]
through the factor map
\[
(\theta,\phi)\mapsto
\bigl(\cos 2\pi(\theta+\phi),\,\cos 2\pi\theta,\,\cos 2\pi\phi\bigr).
\]

## 5. Bicategorical two-loop trace: shadow, coshadow, and torus closure

In a closed bicategory \(C\) equipped with a shadow
\[
\langle\!\langle -\rangle\!\rangle:C(R,R)\to T
\]
and a coshadow
\[
\llangle -\rrangle:C(R,R)\to T,
\]
one can define a two-loop trace for certain maps involving dualizable 1-cells. Let \(X:I\to I\) be a right dualizable 1-cell with dual \(X^*\), coevaluation
\[
coev_X:U_I\to X\circ X^*,
\]
and evaluation
\[
eval_X:X^*\circ X\to U_I.
\]
For an endomorphism
\[
f:X\to X,
\]
the bicategorical trace is
\[
Tr_X(f)=
\big\langle\!\!\big\langle\,
eval_X\circ (f\odot id_{X^*})\circ coev_X
\big\rangle\!\!\big\rangle
\in T.
\]
Dually, if
\[
g:U_I\to Y\circ Y^*
\]
for a right dualizable \(Y\), its coshadow-cotrace is
\[
CoTr_Y(g)=
\big\langle\!\!\big\langle\,
eval_Y\circ (id_{Y^*}\odot g)\circ coev_Y
\big\rangle\!\!\big\rangle_{\cosh}
\in T
\]
[2307.13070].

The two-loop trace is obtained by combining these constructions. First one forms the inner cotrace
\[
g:=
eval_X\circ (id_{X^*}\otimes f\otimes id_{X^*})\circ coev_X
:
U_I\to X^*\circ X.
\]
Then one takes its coshadow-cotrace \(CoTr_X(g)\in T\), and finally the shadow-trace of that same composite viewed as an endomorphism of \(U_I\):
\[
T^{(2)}(f):=
\langle\!\langle
eval_{U_I}\circ \bigl(id_{U_I}\otimes CoTr_X(g)\bigr)\circ coev_{U_I}
\rangle\!\rangle.
\]
Since \(U_I\) is canonically its own dual, this simplifies to
\[
T^{(2)}(f)=
\langle\!\langle
eval_I\circ\bigl(id\otimes \llangle\,eval_X\circ(id\otimes f\otimes id)\circ coev_X\rrangle\bigr)\circ coev_I
\rangle\!\rangle.
\]
String-diagrammatically, \(f\) sits on the equator of a torus: the inner loop is formed by bending \(X^*\) around \(f\), and the outer loop is formed by closing with the shadow.

In the Morita bicategory of \(k\)-algebras, bimodules, and bimodule maps, with Hochschild-homology shadow
\[
\langle\!\langle M\rangle\!\rangle = HH_0(A;M)
\]
and Hochschild-cohomology coshadow
\[
\llangle M\rrangle = HH^0(A;M),
\]
the one-loop trace of an \(A\)-module endomorphism recovers the Hattori–Stallings trace. The two-loop construction specializes to Lipman’s residue map, and in the special case \(M=A\) it is the residue pairing of Lipman. If one instead takes the categorical trace coshadow of a \(2\)-representation, the two-loop trace is exactly Ganter–Kapranov’s \(2\)-character.

## 6. Mathematical significance and further directions

Within Johnson-cokernel theory, the 2-loop trace is the first genuinely new higher-loop obstruction beyond the classical Enomoto–Satoh 1-loop trace. In degree \(6\), it completes the description of \(cok(6)\) in the stable range: one obtains an explicit \(\theta\)-graph presentation of the entire Johnson cokernel in that degree, and the map \(Tr_1\oplus Tr_2\) is injective for \(g\gg 6\) [2508.19041].

The same methods are intended to extend further. It is stated that one can in principle push them to analyze the 2-loop summand \(\widetilde{\Omega}_{2,n-2}\) in higher \(n\), and to attempt presentations of the \(r\)-loop spaces \(\widetilde{\Omega}_r\) for \(r\ge 3\). Topologically, these graph-valued obstructions are expected to correspond to finer Johnson-style invariants of homology cylinders, many of which remain conjectural, and a proposed direction is to relate the 2-loop trace to Ohtsuki-type invariants or perturbative expansions of quantum invariants [2508.19041].

In dynamical systems, the Fibonacci trace map provides a model in which persistent homoclinic tangencies, stochastic sea of full Hausdorff dimension, and infinitely many elliptic islands are all realized for many values of the Fricke–Vogt invariant. The map has all the essential properties that were obtained previously for the Taylor–Chirikov standard map, and it can be suggested as another candidate for the simplest conservative system with highly non-trivial dynamics [1507.07912].

In bicategorical settings, the two-loop trace packages an inner cotrace and an outer trace into a single torus-level invariant, thereby linking bicategorical shadow theory to Lipman residues, Hochschild \((co)\)homology, and \(2\)-characters [2307.13070]. Across these settings, the common “2-loop” language marks a passage from one-loop trace phenomena to higher-order closures, although the ambient objects—hairy graphs, character varieties, and bicategorical duality data—are different.

Source: https://www.emergentmind.com/topics/2-loop-trace-map