Reductions and degenerate limits of Yang-Baxter

maps with 3 × 3 Lax matrices

P. Adamopoulou⋆ , T.E. Kouloukas† and G. Papamikos∗

⋆
Maxwell Institute for Mathematical Sciences and Department of Mathematics,
Heriot-Watt University
†
School of Computing and Digital Media, London Metropolitan University
∗
School of Mathematics, Statistics and Actuarial Science, University of Essex
arXiv:2501.01210v1 [nlin.SI] 2 Jan 2025

E-mail: p.adamopoulou@hw.ac.uk, t.kouloukas@londonmet.ac.uk

g.papamikos@essex.ac.uk

Abstract
We generalise a family of quadrirational parametric Yang-Baxter maps with 3×3
Lax matrices by introducing additional essential parameters. These maps pre-
serve a prescribed Poisson structure which originates from the Sklyanin bracket.
We investigate various low-dimensional reductions of this family, as well as de-
generate limits with respect to the parameters that were introduced. As a result,
we derive several birational Yang-Baxter maps, and we discuss some of their in-
tegrability properties. This work is part of a more general classification of Yang-
Baxter maps admitting a strong 3 × 3 Lax matrix with a linear dependence on
the spectral parameter.

Mathematics Subject Classification: 16T25, 37J10, 14E05

Keywords: Yang-Baxter equation, birational maps, Lax matrices,
discrete dynamical systems, symplectic maps, Liouville integrability.
1 Introduction
1.1 Yang-Baxter maps and Lax matrices
In [8], the study of set theoretical solutions to the Yang-Baxter (YB) equation [4, 25]
was proposed, with such solutions now known as YB maps. A map R : X ×X → X ×X ,
with X any set, is called a YB map [6,23] if it satisfies the set theoretical YB equation

R12 ◦ R13 ◦ R23 = R23 ◦ R13 ◦ R12 . (1)

Rij , for i 6= j ∈ {1, 2, 3}, denotes the action of the map R on the i and j factors of
X × X × X and identically on the remaining factor, i.e. R12 = R × idX where idX is
the identity map over X . While in general the set X can be any set, here we assume
that it is an algebraic variety over a field F of characteristic zero. A parametric YB
map [23, 24] is a YB map which depends on parameters a, b ∈ P ⊂ Fd , acting as
R : (X × P) × (X × P) → (X × P) × (X × P), with


R((x, a), (y, b)) = (u(x, a, y, b), a), (v(x, a, y, b), b) := (u, v) . (2)

We will refer to the parameters a, b ∈ P as YB parameters and denote such a YB map

simply by Ra,b : (x, y) 7→ (u, v), i.e. as a map from X × X to itself. The YB maps
that we consider are all birational maps. We call a birational map (x, y) → (u, v)
quadrirational [2, 17] or non-degenerate if both maps u(·, y) and v(x, ·) are birational
isomorphisms of X to itself.
Of particular relevance are YB maps which arise from refactorisation problems of
Lax matrices. A matrix L depending on x ∈ X , a ∈ P and another parameter λ ∈ F,
is called a Lax matrix [22, 23] for a parametric YB map Ra,b : (x, y) 7→ (u, v), if

L(u, a, λ)L(v, b, λ) = L(y, b, λ)L(x, a, λ) . (3)

If the refactorisation problem (3) is equivalent to (u, v) = Ra,b (x, y), then L is called
strong Lax matrix. The maps Ra,b obtained in this way are all birational but not
necessarily quadrirational. In addition, if the equation

L(x̂, a, λ)L(ŷ, b, λ)L(ẑ, c, λ) = L(x, a, λ)L(y, b, λ)L(z, c, λ) (4)

implies the unique solution x̂ = x, ŷ = y, ẑ = z, then it follows that the map Ra,b is a
YB map [14].
Various YB maps and their generalisations in higher dimensions or over asso-
ciative algebras (but not necessarily commutative) have been constructed recently,
see [1, 5, 7, 10–12, 16] and references therein. In this work, we focus on refactorisation
problems of 3 × 3 Lax matrices. Starting from an 18-dimensional YB map, which we
call principal parametric YB map, we derive lower-dimensional multi-parametric YB
maps via several types of reductions. The obtained maps admit invariant quantities
(first-integrals) that Poisson commute with respect to an r-matrix Poisson structure
(Sklyanin bracket). After a degenerate limit is considered, the resulting maps lose
their quadrirationality and become birational. These reduced birational maps can
be thought of as vectorial and multi-parametric generalisations of the Adler-Yamilov
(AY) map which is related to the nonlinear Schrödinger equation [3].

1
1.2 Poisson Yang-Baxter maps with binomial Lax matrices
To construct multidimenional Yang-Baxter maps it is very natural to study the solu-
tions of the refactorisation problem,
L(U, a, λ)L(V, b, λ) = L(Y, b, λ)L(X, a, λ) (5)
with respect to U and V and with Lax matrices L(X, a, λ) which are first degree
polynomials in the spectral parameter λ,
L(X, a, λ) = X − λKa . (6)
Here, X, Y, U, V are generic elements in gln (F), while K : Fd → GLn (F) is a d-
parametric family of commuting matrices and Ka , Kb denote the values K(a), K(b)
respectively. In [14, 15], solutions of this refactorisation problem were presented which
satisfy the Yang-Baxter equation. We can express these solutions recursively as
n
! n !−1
X X
U = −f0 (X; a)I − (−1)i fi (X; a)Mi−1 (−1)i fi (X; a)Mi−1 Ka ,
i=1 i=1
(7)
V = Ka−1 (Y Ka + Kb X − UKb ) ,
where M0 = I, N0 = 0, M1 = (Y Ka + Kb X)Kb−1 , N1 = −Y Kb−1 Ka and
Mi = M1 Mi−1 + N1 Ni−1 , Ni = N1 Mi−1 , for i = 2, . . . , n.
Here, the functions fi , for i = 0, . . . n, are defined by the coefficients of the polynomial
paλ (X) = det(X − λKa ), by the expression:
paλ (X) = (−1)n fn (X, a)λn + (−1)n−1 fn−1 (X, a)λn−1 + · · · + (−1)f1 (X, a)λ + f0 (X, a) ,
with fn (X, a) = det Ka and f0 (X, a) = det X.
The solution (7) satisfies the additional conditions
fi (U, a) = fi (X, a), fi (V, b) = fi (Y, b), i = 0, . . . , n , (8)
or equivalently the condition det(UKb − Y Ka ) 6= 0 (or det(Ka V − Kb Y ) 6= 0). The
corresponding map Ra,b : (X, Y ) 7→ (U, V ) defined by (7), is a quadrirational Yang-
Baxter map. Furthermore, Ra,b is a Poisson map with respect to the Sklyanin bracket
[20]  
P
{L(X, a, λ1 ) , L(X, a, λ2 )} =
⊗ , L(X, a, λ1 ) ⊗ L(X, a, λ2 ) , (9)
λ1 − λ2
and {L(X, a, λ1 )⊗, L(Y, b, λ2 )} = 0, where P (x ⊗ y) = y ⊗ x. The functions fi , along
with all elements of Ka and Kb , are Casimirs for this Poisson bracket. Hence, the
invariant conditions (8) allow us to further reduce the 2n2 -dimensional map Ra,b to a
2n(n − 1)-dimensional symplectic Yang-Baxter map on the level sets
C = {(X, Y ) : fi (X) = αi , fi (Y ) = βi , i = 0, . . . n − 1} ⊂ gln (F) × gln (F) , (10)
where αi and βi represent additional YB parameters.
In this paper, we study the case where Ka and Kb are 3 × 3 diagonal matrices. Our
analysis covers all cases of binomial Lax matrices with diagonalisable higher-degree
term, as equation (5) remains invariant under conjugation with a constant matrix.
We will investigate lower dimensional reductions and specific limits leading to non-
quadrirational Yang-Baxter maps.

2
2 Yang-Baxter maps with 3 × 3 Lax matrices
2.1 The principal parametric Yang-Baxter map
We consider the refactorisation problem (5) for generic 3 × 3 matrices X, Y, U, V and
nonzero diagonal 3-parametric matrices Ka , Kb . Using the scaling symmetry L → rL,
with r ∈ F\{0}, of equation (5) one can rescale any of the parameters of Ka and Kb
to 1, without loss of generality. Nevertheless, in this section we keep all parameters in
Ka , Kb arbitrary, as elements of a projective space, and we will use the rescaling when
we consider certain reductions in later sections.
We start with the general 3 × 3 Lax matrix of the form L(X, a, λ) = X − λKa with
λ ∈ F and    
x11 x12 x13 a1 0 0
X = x21 x22 x23  , Ka =  0 a2 0  , (11)
x31 x32 x33 0 0 a3
where X ∈ gl3 (F) and a = (a1 , a2 , a3 ) is an element of the F-projective plane P2 (F).
The Sklyanin bracket (9) implies the following linear Poisson bracket between the
variables xij
{xij , xkl } = ai xkj δli − aj xil δkj , (12)
while {xij , ak } = 0. This Poisson bracket admits six linearly independent Casimir
functions a1 , a2 , a3 , f0 , f1 , f2 , where fi are defined by the coefficients of the polynomial
paλ (X) = det(X − λKa ), i.e.,

f0 (X, a) = det X,
f1 (X, a) = a3 (x11 x22 − x12 x21 ) + a2 (x11 x33 − x13 x31 ) + a1 (x22 x33 − x23 x32 ),
(13)
f2 (X, a) = a2 a3 x11 + a1 a3 x22 + a1 a2 x33 ,
f3 (X, a) = det Ka ,

In this case, the refactorisation problem (5), implies uniquely an 18-dimensional

Poisson Yang-Baxter map
Ra,b : (X, Y ) 7→ (U, V )
defined by (7), which can be reduced to a 12-dimensional symplectic YB map Rā,b̄
on C, the intersection ∩2i=0 fi−1 (αi ) × fi−1 (βi ) ⊂ gl3 (F) × gl3 (F), where αi and βi
are additional Yang-Baxter parameters taking values in F. Here we denote by
ā = ((a1 , a2 , a3 ), (α0 , α1 , α2 )) ∈ P2 (F) × F3 and similarly for b̄. This parametric family
of YB maps Rā,b̄ , containing ten effective parameters, constitutes a generalisation of
the family derived in Proposition 4.4 of [15] where the case Ka = Kb = I was consid-
ered. In the following sections, we show that further lower dimensional reductions of
this map are possible.

2.2 Reduction to 8-dimensional Yang-Baxter map

The 9-dimensional Poisson manifold L := {L(X, a, λ) : a constant}, equipped with
the Sklyanin bracket (12), has rank six. However, the rank of this Poisson structure
can be reduced to four by imposing constraints on xij that result in the vanishing of

3
matrix minors. We denote by Mijk,lmn the minor formed by deleting rows i, j, k and
columns l, m, n from the matrix in (9). We consider the minors

2
M789,125 = −a1 a2 a23 a2 (x213 x21 − x13 x11 x23 ) + a1 (x13 x22 x23 − x12 x223 ) ,

2
M589,127 = −a21 a2 a3 a3 (x212 x23 − x12 x13 x22 ) + a2 (x13 x12 x33 − x213 x32 , (14)
M478,478 = −a21 a22 a23 (x12 x23 x31 − x13 x21 x32 )2 .
We notice that the system of equations M789,125 = 0, M589,127 = 0, M478,478 = 0 is
linear in x11 , x31 , x32 , and for x13 , x23 6= 0 implies the solution
x13 x21 a1 (x13 x22 − x12 x23 )
x11 = + ,
x23 a2 x13
x21 x33 a3 x21 x22 a3 x21 x12
x31 = − + , (15)
x23 a2 x23 a2 x13
x12 x33 a3 x12 x22 a3 x212 x23
x32 = − + .
x13 a2 x13 a2 x213
Substituting relations (15) in the Casimir functions (13) we obtain the following set
of reduced rational Casimirs ι∗ fi = fi ◦ ı : F6 → F
(x13 x22 − x12 x23 )2 (a2 x13 (a1 x23 x33 + a3 x13 x21 ) + a1 a3 x12 x223 )
ι∗ f0 = ,
a22 x313 x23
(x13 x22 − x12 x23 )(2a2 x13 (a1 x23 x33 + a3 x13 x21 ) + a1 a3 x23 (x12 x23 + x13 x22 ))
ι∗ f1 = ,
a2 x213 x23
a1 a3 x12 x23 a2 a3 x13 x21
ι∗ f2 = a1 a2 x33 + 2a1 a3 x22 − + .
x13 x23
(16)
Here we define ı to be the inclusion map F6 ֒→ gl3 (F)
ι : (x12 , x13 , x21 , x22 , x23 , x33 ) 7→ X = (xij ),
with x11 , x31 , x32 given by (15).
In what follows, for simplicity we will denote all ι∗ fi by fei . We also denote M =
Img(ι), i.e. the set of reduced matrices in L, with x11 , x31 , x32 defined by (15). The
following proposition holds:
Proposition 2.1. M is a Poisson submanifold of L of rank four. Furthermore, the
discriminant of the cubic polynomial in λ of ι∗ paλ vanishes, i.e.

4fe0 fe23 − fe12 fe22 + 4fe3 fe13 − 18fe0 fe1 fe2 fe3 + 27fe32 fe02 = 0. (17)
Proof. By direct computation we can show that for x13 , x23 6= 0, the inclusion map ι
defined by (15) is Poisson with respect to the Poisson bracket (12) (and the induced
bracket on M). Hence, M is a Poisson submanifold of L and by substituting (15)
in (9) reduces the rank of the Poisson structure matrix to four. The coefficients of
the pullback ι∗ paλ are the reduced Casimirs ι∗ fi := fei given in (16). The rank of
the Jacobian matrix of fe0 , fe1 , fe2 is equal to two, and therefore there is one functional
relation between them. Relation (17) can be obtained using elimination algorithms.

4
 As we mentioned in the proof of Proposition (2.1), on the submanifold M there
are two functionally independent Casimirs. Solving the system fe1 = α1 , fe2 = α2 for
x22 and x33 we obtain the following expressions
c1 c2 x12 x23
x22 = + + ,
a3 a3 x13
c1 2c2 a3 x13 x21 a3 x12 x23 (18)
x33 = − − − ,
a2 a2 a1 x23 a2 x13
where c1 and c2 depend on the level sets of the Casimirs and the matrix Ka as follows:
p
α2 α22 − 3α1 a1 a2 a3
c1 = , c2 = ± . (19)
3a1 3a1
Using relations (18) in (15) we obtain the reduced expressions for x11 , x31 , x32
a1 (c1 + c2 ) x13 x21
x11 = + ,
a2 a3 x23
3c2 x21 a3 x13 x221 a3 x12 x21
x31 =− − − , (20)
a2 x23 a1 x223 a2 x13
3c2 x12 a3 x12 x21 a3 x212 x23
x32 =− − − .
a2 x13 a1 x23 a2 x213
After the reduction, the Poisson brackets (12) between the remaining variables
x12 , x13 , x21 , x23 take the form
x12 x23 x13 x21
{x13 , x21 } = a1 x23 , {x12 , x21 } = a1 − a2 , {x12 , x23 } = −a2 x13 , (21)
x13 x23
with all other brackets vanishing. Similarly, after the reduction (18)-(20) the Lax
matrix (6) takes the form X̃ − λKa where X̃ is given by
 a1 (c1 +c2 )

a2 a3
+ x13x23x21 x12 x13
 x21 c1 +c2
+ x12x13x23 x23 
 a3 .
2
a3 x13 x21 2
a3 x12 x23
3c2 x21 a3 x12 x21 3c2 x12 a3 x12 x21 c1 −2c2 a3 x13 x21 a3 x12 x23
− a2 x23 − a1 x2 − a2 x13 − a2 x13 − a1 x23 − a2 x2 a2
− a1 x23 − a2 x13
23 13

The change of variables

x12 = −a2 x2 X1 , x13 = X1 , x21 = −a1 x1 X2 , x23 = X2 , (22)

brings the brackets (21) to the canonical form, i.e.

{x1 , X1 } = 1 , {x2 , X2 } = 1 , {x1 , x2 } = 0 , {X1 , X2 } = 0 , (23)

while the reduced Lax matrix X̃ − λKa takes the form

L(x, X, p, λ) = (24)
 
a1 (c1 +c2 )
a2 a3 − a1 x1 X1 − λa1 −a2 x2 X1 X1
 c1 +c2 
 −a1 x1 X2 a3 − a2 x2 X2 − λa2 X2 ,
3a1 c2 x1 c1 −2c2
a2 − a1 a3 x1 (x1 X1 + x2 X2 ) 3c2 x2 − a2 a3 x2 (x1 X1 + x2 X2 ) a2 + a3 (x1 X1 + x2 X2 ) − λa3

5
where x = (x1 , x2 ), X = (X1 , X2 ) and p = (c1 , c2 , a1 , a2 , a3 ).
The matrix refactorisation problem

L(u, U , p, λ)L(v, V , q, λ) = L(y, Y , q, λ)L(x, X, p, λ), (25)

with q = (d1 , d2 , b1 , b2 , b3 ) has a unique solution for u, U , v, V in terms of x, X, y, Y

given by
 
b3 b3 C1 a3 a3
(u1 , u2) = (y1 , y2) + x1 − y1 , x2 − y2 ,
a3 a3 D1 b1 b2
    (26)
a1 a2 C2 a3 a3
(v1 , v2 ) = x1 , x2 − b1 a1 ( x1 − y1 ), a2 ( x2 − y2 ) ,
b1 b2 D2 b1 b2

where
C1 = a2 a3 b1 (d1 − 2d2 ) − b1 b2 b3 (c1 − 2c2 ),
b2 b3
C2 = (c1 + c2 ) − a2 (d1 + d2 ),
a3 (27)
D1 = a2 a3 b2 b3 (a3 x1 − b1 y1 )Y1 + a2 a3 b1 b3 (a3 x2 − b2 y2 )Y2 + C1 + 3a2 a3 b1 d2 ,
D2 = a2 b1 b2 b3 (a3 x1 − b1 y1 )X1 + a2 b1 b2 b3 (a3 x2 − b2 y2 )X2 − C1 − 3a2 a3 b1 d2 ,

and
(a1 x1 − b1 v1 )X1 + (a1 y1 − a3 v1 )Y1 (a2 x2 − b2 v2 )X2 + (a2 y2 − a3 v2 )Y2
U1 = , U2 = ,
a1 u1 − b3 v1 a2 u2 − b3 v2
(b1 u1 − b3 x1 )X1 + (a3 u1 − b3 y1 )Y1 (b2 u2 − b3 x2 )X2 + (a3 u2 − b3 y2 )Y2
V1 = , V2 = .
a1 u1 − b3 v1 a2 u2 − b3 v2
(28)

Finally, by direct computation we can prove the following proposition:

Proposition 2.2. The map

Rp,q : ((x1 , x2 , X1 , X2 ), (y1, y2 , Y1 , Y2 )) 7→ ((u1 , u2 , U1 , U2 ), (v1 , v2 , V1 , V2 )) , (29)

with ui , Ui , vi , Vi defined by (26)-(28) is a parametric quadrirational Yang-Baxter map

with strong Lax matrix (24). Furthermore, Rp,q is symplectic with respect to

ω = dx1 ∧ dX1 + dx2 ∧ dX2 + dy1 ∧ dY1 + dy2 ∧ dY2 .

The map (29) admits four functionally independent invariants which can
be obtained from the characteristic polynomial of the monodromy matrix
L(y, Y , q, λ)L(x, X, p, λ). The two simplest invariants obtained in this way are

I1 = x1 X1 + y1 Y1 , I2 = x2 X2 + y2 Y2 . (30)

Remark 2.3. For ai = bi = 1 and particular choices of the parameters ci , di , the YB

map (29) reduces to non-degenerate Boussinesq and Goncharenko-Veselov maps [9,15].

6
2.3 Reduction to a 4-dimensional Yang-Baxter map
In this section, we consider a further folding reduction of the Yang-Baxter map (29)
on an invariant manifold. We impose the relations
x1 = x2 := x , X1 = X2 := X , a1 = a2 := a ,
(31)
y1 = y2 := y , Y1 = Y2 := Y , b1 = b2 := b ,
which are consistent with the map since relations (26)-(28) imply
u1 = u2 , U1 = U2 , v1 = v2 , V1 = V2 .
Thus, for p = (c1 , c2 , a, a3 ) and q = (d1 , d2, b, b3 ), the map (29) is reduced to a 4-
dimensional quadrirational YB map
Rp,q : ((x, X), (y, Y )) 7→ ((u, U), (v, V ))
on the invariant manifold
N = {((x, x, X, X), (y, y, Y, Y )) : x, X, y, Y ∈ F} ,
where
b3 e1  a3
b3 C  a Ce2  a3 
u= y+ x − y , v = x − ab x−y ,
a3 a3 De1 b b De2 b
(32)
(ax − bv)X + (ay − a3 v)Y (bu − b3 x)X + (a3 u − b3 y)Y
U= , V = ,
au − b3 v au − b3 v
ei , D
and C e i are obtained from (27) under the reduction (31).
The reduced symplectic structure on N is
ω = dx ∧ dX + dy ∧ dY ,
and Rp,q is symplectic with respect to ω. Furthermore, map Rp,q admits the reduced
Lax matrix
 c1 +c2 
a3
− axX − λa −axX X
L(x, X, p, λ) =  −axX c1 +c2
a3
− axX − λa X ,
2 2 c1 −2c2
3c2 x − 2aa3 x X 3c2 x − 2aa3 x X a
+ 2a3 xX − λa3
(33)
which can be obtained from (24) by imposing the reduction (31).
Using the trace of the monodromy matrix associated to Lax matrix (33) we obtain
two functionally independent invariants of the map (32). One of the invariants is
I1 = xX + yY , which is I1 (or I2 ) under the reduction (31), while the other has the
form
I2 = a00 11 01 10 01 2 21 2 11
11 xX + a00 yY + a10 xY + a01 Xy + XY (a21 x + a01 y + a11 xy), (34)
where the coefficients akl i j k l
ij of the monomials x X y Y depend on the parameters of the
map (32). For simplicity, we have omitted the exact dependence of the coefficients akl
ij
on the parameters of the map. Moreover, the invariants I1 and I2 Poisson commute with
respect to the canonical Poisson structure {x, X} = 1 and {y, Y } = 1 and therefore
the map (32) is a 4-dimensional symplectic quadrirational map that is also integrable
in the Liouville sense.

7
3 Degenerate Limits
3.1 A single degenerate limit
We are interested in studying zero limits for certain parameters involved in the maps
derived in Section 2, which effectively result in degenerations for the maps introduced
in [15]. We call such limits degenerate. In particular, in this section we focus on the
degenerate limit a3 → 0.
We consider the limit a3 → 0 of expressions (18). For both values of c2 given in
(19), the level set α2 of the Casimir f2 can be chosen so that the branch of the square
root is such that the limit a3 → 0 of x22 results in a unique well-defined expression.
The same limit of x33 is regular. The obtained expressions are the following
a2 α1 x12 x23 α2
lim x22 = + , lim x33 = . (35)
a3 →0 2α2 x13 a3 →0 a1 a2
Substituting formulas (35) in the expressions (15) for x11 , x31 , x32 and taking the now
regular limit a3 → 0 we obtain
a1 α1 x13 x21 α2 x21 α2 x12
x11 = + , x31 = , x32 = . (36)
2α2 x23 a1 a2 x23 a1 a2 x13
Hence, in the limit a3 → 0 the Lax matrix X̃ − λKa takes the form
 a1 α1 x13 x21 
  2α2
+ x23 − λa1 x12 x13
lim X̃ − λKa =  x21 a2 α1
2α2
+ x12x13x23 − λa2 x23  . (37)
a3 →0 α2 x21 α2 x12 α2
a1 a2 x23 a1 a2 x13 a1 a2

The change of variables (22) is not affected by the limit a3 → 0, and in the variables
x1 , x2 , X1 , X2 the Lax matrix (37) takes the form
 a1 α1 
2α2
− a1 x1 X1 − λa1 −a2 x2 X1 X1
e
L(x, X, p, λ) =  −a1 x1 X2 a2 α1
− a2 x2 X2 − λa2 X2  , (38)
2α2
α2
− a2 x1 − αa12 x2 α2
a1 a2

with p = (α1 , α2 , a1 , a2 ). The refactorisation problem associated to the Lax ma-

trix (38) results in the following birational (but non quadrirational) YB map R ep,q :
(x1 , x2 , X1 , X2 , y1 , y2, Y1 , Y2 ) → (u1 , u2, U1 , U2 , v1 , v2 , V1 , V2 ) where
a1 a2 β2
ui = yi ,
b1 b2 α2
b1 b2 α2 b1 b2 bi k
Ui = Yi − Xi ,
a1 a2 β2 a1 a2 β2 (b1 y1 X1 + b2 y2 X2 ) − α2 β2
(39)
ai a1 a2 ai k
vi = xi + yi
bi a1 a2 α2 (b1 y1 X1 + b2 y2 X2 ) − α22
bi
Vi = Xi ,
ai
for i = 1, 2 and with k = α2 β12β−α
2
1 β2
. This 8-dimensional YB map admits four function-
ally independent polynomial invariants. Two of these are given in (30), which remain

8
unaffected by the limit a3 → 0. Hence, we have Ie1 = x1 X1 + y1 Y1 , Ie2 = x2 X2 + y2 Y2 .
Additionally, from the spectrum of the monodromy matrix, we obtain the following
two invariants:

Ie3 = −a21 a2 b21 b2 (α2 β1 x1 X1 + α1 β2 y1 Y1 ) − a1 a22 b1 b22 (α2 β1 x2 X2 + α1 β2 y2 Y2 )

   (40)
+ 2α2 β2 b1 b2 (a1 x1 Y1 + a2 x2 Y2 ) − β2 a1 a2 (b1 y1 X1 + b2 y2 X2 ) − α2 ,

and
 
Ie4 = a1 a2 b1 b2 α2 β1 (x1 X1 + x2 X2 ) + α1 β2 (y1 Y1 + y2 Y2 ) + 2α22 β2 (b2 x1 Y1 + b1 x2 Y2 )
+ 2α2 β22 (a2 y1 X1 + a1 y2 X2 ) − 2a1 a2 b1 b2 α2 β2 (x1 X1 + x2 X2 )(y1 Y1 + y2 Y2 ).
(41)

The folding reduction (31) can also be applied to the map (39). Indeed, in this case
the map will simplify to the four dimensional YB map R ea,b : (x, X, y, Y ) → (u, U, v, V )
given by

a2 β2 b2 α2 b3 k
u= y, U = Y − X,
b2 α2 a2 β2 2a2 bβ2 yX − α2 β2
(42)
b a a3 k
V = X, v = x+ 2 y.
a b 2a bα2 yX − α22

The invariants of map (39) will reduce to invariants of map R ea,b given in (42). Under
the folding reduction, both invariants Ie1 , Ie2 will become equal to eI = xX + yY , and
similarly both invariants Ie3 and Ie4 will take the same form (up to a scaling factor),
resulting in the following invariant
e = a2 b2 (α2 β1 xX + α1 β2 yY ) + 2α2 β2 (bα2 xY + aβ2 yX) − 4a2 b2 α2 β2 xXyY.
J (43)

Both invariants eI, J

e can be obtained from the trace of the monodromy associated to
the Lax matrix (38) after applying the reduction (31). The invariants eI and J
e Poisson
commute therefore the map (42) is symplectic and Liouville integrable.

Remark 3.1. The 4-dimensional map (42) can also be obtained from the refactorisa-
tion of the Lax (33) by taking the limit a3 → 0.

Remark 3.2. The choice of parameters

a = b = 1, α2 = β2 = −1, α1 = 2A, β1 = 2B,

results in the map

A−B A−B
u = y, U =Y − X, v =x+ y, V = X, (44)
1 + 2yX 1 + 2yX

which after the rescaling of the variables x, y, u, v by a factor of 12 and the flip x ↔ X,
y ↔ Y , u ↔ U and v ↔ V becomes the well-known AY map [3].

9
3.2 A double limit to a vectorial Adler-Yamilov map
In this section we aim to demonstrate that a vectorial generalisation of the Adler-
Yamilov map (vAY) together with its standard Lax matrix, can be obtained from the
Lax matrices that we study after taking the double limit a2 → 0 and a3 → 0 in Ka .
Since in this case the only nonzero parameter of Ka is a1 , without loss of generality
we can rescale it to one. Therefore, in this section we have that Ka = e11 .
The double limit (a2 , a3 ) → (0, 0) of the Casimir f2 in (13) exists and implies that
f2 = α2 ≡ 0. This means that the iterated single limits commute. Taking first the
limit of f2 when a2 → 0 we obtain that a3 x22 = α2 , while if we take first the limit
a3 → 0 it follows that a2 x33 = α2 . Hence, when a2 , a3 → 0, both x22 and x33 have
to be constants. The case x22 = x33 = 0 leads to a trivial refactorisation problem,
therefore we consider the case x22 = x33 = 1 which occurs only when a2 = a3 = ǫ → 0.
If we restrict on the level set α1 = 1 of the Casimir f1 , then we obtain the constraint
x23 x32 = 0. We assume that x23 = x32 = 0. Finally, under the above assumptions, the
Casimir f0 leads to the following constraint

f0 := x11 − x12 x21 − x13 x31 = a, (45)

which we use to solve for x11 .

Taking into account all the above, the reduced Lax matrix is of the form
 
a − λ + x1 X1 + x2 X2 x1 x2
L(x, X, a, λ) =  X1 1 0 , (46)
X2 0 1

where we use the notation x1j = xj−1 and xi1 = Xi−1 for i, j = 2, 3. The refactorisation
problem associated to the Lax matrix (46) implies uniquely the non-quadritational YB
map Ra,b : (x, X, y, Y ) → (u, U , v, V )
a−b
u=y− x, U = Y ,
1 + hx, Y i
(47)
a−b
V =X+ Y , v = x,
1 + hx, Y i

where h·, ·i is the standard bilinear form in F2 . From the spectrum of the monodromy
matrix we obtain the usual invariants (30)

I1 = x1 X1 + y1 Y1 , I2 = x2 X2 + y2 Y2 , (48)

together with

I3 = b hx, Xi + a hy, Y i + hx, Y i + hX, yi + hx, Xi hy, Y i . (49)

Moreover, it can be verified by a direct calculation that the determinants x1 y2 − x2 y1 ,

X1 Y2 − X2 Y1 are anti-invariants, and therefore their product

I4 = (x1 y2 − x2 y1 )(X1 Y2 − X2 Y1 ) (50)

is another invariant of the map (47).

10
Proposition 3.3. The YB map Ra,b in (47) is Liouville integrable and symplectic
with respect to the canonical symplectic structure.

Proof. The Jacobian matrix of the invariants I1 , . . . , I4 has full rank and therefore
the invariants are functionally independent. Moreover, the invariants commute with
respect to the Poisson bracket

{xi , Xi } = δij , {yi , Yi } = δij

implied by the Sklyanin bracket (9). The map (47) preserves the Poisson bracket
and therefore is symplectic with respect to the corresponding canonical symplectic
structure
X2
ω= dxi ∧ dXi + dyi ∧ dYi .
i=1

Remark 3.4. The YB map Ra,b in (47) together with its Lax matrix (46) has an
obvious generalisation to n-dimensional F-vectors. The integrability properties of the
map can be generalised to arbitrary dimensions in a similar way as the one described
in Proposition (3.3). The n-dimensional map has been derived in [12].

The reduction (31) to the invariant manifold N presented in Section (2.3) can also
be applied in this case, and reduces the map Ra,b in (47) to the standard AY map [3].
Moreover, map (47) can be obtained from the YB map R ep,q in (39) after the following
choice for the parameters

a1 = b1 = 1, a2 = b2 = 1, α2 = β2 = −1, α1 = 2a, β1 = 2b ,

and the permutation in the variables

x ↔ X, y ↔ Y , u ↔ U , v ↔ V .

This implies that the vAY map (47) admits the Lax matrix
 
x1 X1 + a + λ x1 X2 −x1
e
L(X, x, a) =  x2 X1 x2 X2 + a + λ −x2  . (51)
−X1 −X2 1

This Lax matrix is equivalent to the Lax matrix (46) as it can be directly obtained
from (46) by inversion and a similarity transformation by a permutation matrix.

4 Conclusions
In this paper we studied reductions and degenerations of a family of YB maps with
3 × 3 first-degree polynomial Lax matrices. We also introduced compatible Poisson
structures associated with the Sklyanin bracket and invariant conditions, ensuring the
Liouville integrability of all the presented maps.

11
 Yang-Baxter maps serve as fundamental building blocks for constructing higher-
dimensional discrete integrable systems. As it was shown in [23, 24], each YB map
generates an hierarchy of commuting n-dimensional transfer maps which share the
same integrals. A different variant of transfer maps occurs by considering periodic ini-
tial value problems on lattices, in correspondence with staircase initial value problems
of integrable lattice equations [18, 19]. While the dynamics of an individual YB map
sometimes may be trivial (e.g. involution), the dynamics of the associated transfer
maps display highly non-trivial behavior.
All variations of transfer maps preserve the spectrum of their corresponding mon-
odromy matrices, constituting of products of Lax matrices. In this setting, YB maps
with Lax matrices compatible with the Sklyanin bracket, as the maps presented in
this paper, hold a significant advantage as they generate Poisson transfer maps with
commutative integrals derived from the spectrum of the corresponding monodromy
matrices.
The family of the YB maps presented in this work includes all cases involving
3 × 3 Lax matrices linear in the spectral parameter and with the degree one coefficient
being a constant and diagonalisable matrix. The connection between the maps that we
obtained can be summarised in the graph below. The arrows in the graph indicate a
connection between the maps which can be a reduction, a limit, a choice of parameters,
or a simple transformation in the dynamical variables that is a symmetry of the YB
equation.
In future research, we aim to study families of YB maps and corresponding degena-
rations, with 3 × 3 Lax matrices associated with different Jordan forms of the highest-
degree terms. This will complete the classification of 3×3 binomial Lax matrices under
conjugation. Furthermore, we intend to include the 3-dimensional consistent lattice
equations associated with all these maps and investigate their integrability features.

maps dimensions

Rab 18

Rā,b̄ 12

a3 → 0
Rp,q ep,q
R vAY 8

a3 → 0
Rp,q e p,q
R AY 4

12
References
[1] P. Adamopoulou, S. Konstantinou-Rizos, and G. Papamikos, Integrable exten-
sions of the Adler map via Grassmann algebras Theor. Math. Phys. 207 553–559,
2021.

[2] V.E. Adler, A.I. Bobenko, and Yu.B. Suris, Geometry of the Yang-Baxter Maps:
pencils of conics and quadrirational mappings, Comm. Anal. Geom., 12 (5), 967–
1008, 2004.

[3] V.E. Adler and R.I. Yamilov, Explicit auto-transformations of integrable chains,
J. Phys. A: Math. Gen., 27 (2), 447–492, 1994.

[4] R.J. Baxter, Exactly solved models in Statistical Mechanics, Academic Press, 1982.

[5] A.I. Bobenko and Yu.B. Suris, Integrable noncommutative equations on quad-
graphs. The consistency approach. Lett. Math. Phys., 61, 241–254, 2002.

[6] V.M. Bukhshtaber, The Yang-Baxter transformation, Russ. Math. Surv., 53 (6),
1343, 1998.

[7] A. Doliwa, Non-commutative rational Yang–Baxter maps, Lett. Math. Phys. 104,
299–309, 2014.

[8] V.G. Drinfeld, On some unsolved problems in quantum group theory, In P.P.
Kulish, editor, Quantum Groups, volume 1510 of Lecture Notes in Mathematics,
Springer, Berlin, Heidelberg, 1992.

[9] V.M. Goncharenko and A.P. Veselov, Yang-Baxter maps and matrix solitons, New
trends in integrability and partial solvability, NATO Science Series, 132, 191–197,
2004.

[10] P. Kassotakis, and T. Kouloukas, On non-abelian quadrirational Yang–Baxter

maps, J. Phys. A: Math. Theor. 55 175203, 2022.

[11] P. Kassotakis, M. Nieszporski, V. Papageorgiou and A. Tongas, Integrable two

component systems of difference equations, Proc. R. Soc. A, 476, 20190668, 2020.

[12] S. Konstantinou-Rizos and A.V. Mikhailov, Darboux transformations, finite re-

duction groups and related Yang-Baxter maps, J. Phys. A: Math. Theor., 46,
425201, 2013.

[13] S. Konstantinou-Rizos and T.E. Kouloukas, A noncommutative discrete potential

KdV lift, J. Math. Phys., 59 (6), 063506, 2018.

[14] T.E. Kouloukas and V.G. Papageorgiou, Yang-Baxter maps with first-degree-
polynomial 2 × 2 Lax matrices, J. Phys. A: Math. Theor., 42, 404012, 2009.

[15] T.E. Kouloukas and V.G. Papageorgiou, Poisson Yang-Baxter maps with binomial
Lax matrices, J. Math. Phys., 52, 073502, 2011.

13
[16] A.V. Mikhailov, G. Papamikos, and J.P. Wang, Darboux transformation for the
vector sine-Gordon equation and integrable equations on a sphere. Lett. Math.
Phys., 106, 973–996, 2016.

[17] V.G. Papageorgiou, Yu.B. Suris, A.G. Tongas, and A.P. Veselov, On quadrira-
tional Yang–Baxter maps, SIGMA, 6, 033, 2010.

[18] V.G. Papageorgiou, F.W. Nijhoff, and H.W. Capel, Integrable mappings and
nonlinear integrable lattice equations, Phys. Lett. A, 147, 106–114, 1990.

[19] G.R.W. Quispel, H.W. Capel, V.G. Papageorgiou, and F.W. Nijhoff, Integrable
mappings derived from soliton equations, Physica A, 173, 243–266, 1991.

[20] E.K. Sklyanin, Some algebraic structures connected with the Yang-Baxter equa-
tion Funct. Anal. Appl. 16 (4), 263–270, 1982.

[21] E.K. Sklyanin, Classical limits of the SU(2)-invariant solutions of the Yang-Baxter
equation, J. Math. Sci., 40, 93–107, 1988.

[22] Yu.B. Suris and A.P. Veselov, Lax matrices for Yang-Baxter maps, J. Nonlin.
Math. Phys. 10, 223–230, 2003

[23] A.P. Veselov, Yang-Baxter maps and integrable dynamics, Phys. Lett. A, 314 (3),
214–221, 2003.

[24] A.P. Veselov, Yang-Baxter maps: dynamical point of view, Combinatorial Aspects
of Integrable Systems (Kyoto, 2004) MSJ Mem., 17, 145–67, 2007.

[25] C.N. Yang, Some exact results for the many-body problem in one dimension with
repulsive delta-function interaction, Phys. Rev. Lett., 19, 1312, 1967.

14