Folding defect ane Toda eld theories

Craig Robertson

April 2013
Abstract
A folding process is applied to fused a
(1)
r
defects to construct defects for the non-simply laced
ane Toda eld theories of c
(1)
n
, d
(2)
n
and a
(2)
2n
at the classical level. Support for the hypothesis
that these defects are integrable in the folded theories is provided by the observation that
transmitted solitons retain their form. Further support is given by the demonstration that
energy and momentum are conserved.

craig.robertson@durham.ac.uk
a
r
X
i
v
:
1
3
0
4
.
3
1
2
9
v
1


[
h
e
p
-
t
h
]


1
0

A
p
r

2
0
1
3
Contents
1 Introduction 2
1.1 Ane Toda eld theory and classical defects . . . . . . . . . . . . . . . . . . . . 3
1.2 Layout of this paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Bivalent foldings of a
(1)
r
5
2.1 The bivalent foldings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Case 1: h even, k even . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Case 2: h even, k odd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.3 Case 3: h odd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Solitons and folding 10
3.1 a
(1)
r
solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Folding a
(1)
r
solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Tau functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.1 One soliton solution of a
(1)
r
. . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.2 Two soliton solution of a
(1)
r
. . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.3 Single solitons in the folded theories . . . . . . . . . . . . . . . . . . . . . 13
3.3.4 Other solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Energy-momentum of solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.1 Coupling dependence of energy, momentum and mass . . . . . . . . . . . 19
3.4.2 Calculation of energy and momentum of solitons . . . . . . . . . . . . . . 21
3.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 Fusing defects 25
4.1 Two defect system and time delays . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 A fused a
(1)
r
defect and energy-momentum conservation . . . . . . . . . . . . . . . 27
4.2.1 Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2.2 Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5 The folded defect 32
5.1 Folding the Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.2 Content of the defect equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.3 Time delay of solitons through the defect . . . . . . . . . . . . . . . . . . . . . . 34
5.3.1 Algebraic constraints with such d and

d . . . . . . . . . . . . . . . . . . . 35
5.4 Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.5 Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.5.1 Proof of the relations (5.22) and (5.23) . . . . . . . . . . . . . . . . . . . . 38
5.5.2 Proof of (5.27) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6 Discussion 41
A Tau function identications 43
A.1 c
(1)
n
solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
A.2 d
(2)
n
solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
A.3 a
(2)
2n
solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1
B a
1
embeddings in a
(1)
2n1
and folded theories 45
B.1 p = n soliton as an a
1
soliton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
B.2 A less obvious embedding of a
1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
C Equivalence of a priori and a posteriori folding and ways to write D

47
C.1 Gradients of the defect potential . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
C.1.1 A more useful way to write D

. . . . . . . . . . . . . . . . . . . . . . . . 51
C.2 Comparison of the a priori and a posteriori defect conditions . . . . . . . . . . . 52
D The folded defect in a
(2)
2
53
D.1 Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
D.2 Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
D.3 Comparison to the type II defect . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
D.4 Algebraic constraints and a posteriori folding . . . . . . . . . . . . . . . . . . . . 57
1 Introduction
The ane Toda eld theories (ATFTs) have seen an upsurge of interest due to the discovery
of integrable defects. Thus far only for the a
(1)
r
[1, 2] and a
(2)
2
[3] models do integrable defects
exist in the literature- even at the classical level- despite the search for defects in ATFT having
been initiated a decade ago [1]. Compare this to the discovery of solitons in ATFT, where the
construction of a
(1)
r
and d
(1)
4
solitons [4] was rapidly followed by solitons in the other models [5].
The dierent ATFTs have similar properties (e.g., they all possess soliton solutions [5]) so it
is expected that defects should exist for all of the ATFTs- as such, an overarching goal in this
eld is to nd and investigate the properties of all of the possible defects.
Folding, or reduction, is a powerful tool which allows properties of non-simply laced theo-
ries to be found using properties of simply laced theories
1
which are often easier to work with.
Indeed, folding has previously been put to use to nd, in the non-simply laced theories, solitons
[5, 6] and also scattering matrices at the classical [7] and quantum [8] levels. Note that folding
has not previously been applied to the study of defect ATFTs.
In this paper, a folding process, originally described in [9], is used to obtain candidate inte-
grable defects for the c
(1)
n
, d
(2)
n
and a
(2)
2n
series of non-simply laced ATFTs by making use of
a
(1)
r
solitons and defects. Two methods, which turn out to be linked, are used to suggest that
these defects should be classically integrable. Firstly, what happens when a soliton is sent
through the defect; secondly, in what circumstances does the defect conserve energy and mo-
mentum. An interesting by-product of the folding process is that it allows for the construction
of multisolitons, breathers and fusing rules for these folded theories (c
(1)
n
, d
(2)
n
and a
(2)
2n
)- some-
thing not explicitly considered previously in the literature.
What this paper does not do is provide defects for any of the ATFTs which do not fall un-
der the umbrella of a
(1)
r
, though it does suggest that if the appropriate defects can be found
1
The simply laced ATFTs are the a
(1)
r
,d
(1)
s
,e
(1)
6
,e
(1)
7
and e
(1)
8
theories. They are distinguished in having all roots
of the same length, conventionally

2; although in the a
1
case the choice of = 1 is more conventional. The
non-simply laced theories all have more than one length of root present.
2
for the other simply laced theories (d
(1)
s
,e
(1)
6
,e
(1)
7
and e
(1)
8
) then folding may be applied to give
the defects of the other non-simply laced theories. Note that only one type of defect is given
for each of c
(1)
n
, d
(2)
n
and a
(2)
2n
- there is no claim that other defects should not exist in these theories.
In order to set notation, and inform the reader unfamiliar with ATFT, a short summary of
ATFT with defects is given below.
1.1 Ane Toda eld theory and classical defects
To each ane Dynkin diagram there is a corresponding ane Toda model [10]. In 1+1 dimen-
sions a working denition of ATFT is given by the Lagrangian
L(u) =
1
2
u u
1
2
u

U(u) (1.1)
where u is an r component vector living in the root space described by the ane Dynkin diagram
and is a Lorentz scalar. The potential is given by [4, 5]
U(u) =
m
2

2
r

j=0
n
j
_
e

j
u
1
_
. (1.2)
In (1.2), {
i
} (i = 1, . . . , r), are the positive simple roots of the root space while
0
=

r
j=1
n
j

j
is the lowest root of the root space, corresponding to the extra node of the ane
Dynkin diagram. It is the case that

r
j=0
n
j

j
= 0, so conventionally n
0
= 1, while the other
marks {n
j
} are characteristic of the underlying Lie algebra.
The constant m sets a mass scale, which has no importance for the classical discussions herein
so will be set to unity, m = 1. is the coupling constant- it will be convenient to also set = 1
but this will need to be justied rst (see section 3.4). The potential here is chosen such that
the zero solution has zero energy.
The Euler-Lagrange equations applied to (1.1) give
u u

= U
u
(1.3)
where U
u
denotes the gradient of the potential with respect to the vector u.
Guided by the Lagrangian description of a sine-Gordon defect [1]; Bowcock, Corrigan and
Zambon suggested an Ansatz for defects in the other ATFTs in [2] (subsequently referred to as
a type I defect [3])
L = (x)L
u
+(x)L
v
+(x)
_
1
2
uA u +uB v +
1
2
vC v D(u, v)
_
. (1.4)
Equation (1.4) describes a defect situated at x = 0 with A,B and C constant matrices. L
u
and
L
v
are bulk Lagrangians of the form (1.1) for the elds u and v respectively. In practice u and
v always correspond to the same root data.
The Euler-Lagrange equations applied to (1.4) give at x = 0 the conditions
u

= A u +B v D
u
v

= C v +B
T
u +D
v
(1.5)
3
as well as the bulk equations of motion.
In [2] a defect Lax pair, taking into account (1.5), was constructed and it was shown that
only the a
(1)
r
models may incorporate such a defect and remain integrable. It was subsequently
shown in [11] that (1.4) conserves modied energy and momentum only for a
(1)
r
and that the
conditions imposed are the same as for the Lax pair- i.e., in this case energy-momentum con-
servation implies integrability. From either approach it is found that
A = C = 1 B
D(u, v) = d
r

j=0
e
1
2

(
B
T
u+Bv
)
+d
1
r

j=0
e
1
2

j
B(uv)
(1.6)
and so the defect can be classied by B along with the rapidity parameter d. There are two
possibilities for B, which are
B =
r

j=1
(
j

j+1
)
T
j
(1.7)
and its transpose
B =
r

j=1
(
j

j1
)
T
j
(1.8)
where {
i
} are the fundamental weights of a
(1)
r
, dened by
i

j
=
ij
for i, j = 1, . . . , r and

0
= 0. In this paper B will be taken to mean (1.7), with B
T
used if the defect is classied by
(1.8).
In [2] the eect of the defect on a one soliton solution was examined. Using a one soliton
Ansatz (of species p, say) for u and for v revealed that a B defect gives v a time delay of

p
=
ie

+d
p
2
ie

+d

p
2
(1.9)
while a B
T
defect gives a time delay of

p
=
ie

p
2
ie

d
p
2
(1.10)
where is the rapidity of the soliton and = e
2i
r+1
. It should be noted that the time delay is
dependent on the species of soliton. In particular, the species p soliton and the species h p
soliton receive dierent delays, meaning that neither B nor B
T
type I defects are compatible
with the folding considered in this paper.
The framework for defects was extended in [3] to include integrable defects for a
(2)
2
by means of
what is referred to as a type II defect with an Ansatz of the form
L = (x)L
u
+(x)L
v
+(x) (u v 2(u v) D(u, v, )) (1.11)
4
where u and v are now scalar elds in either a
1
or a
(2)
2
(the form of D depends upon which
theory is being looked at) while is a scalar eld which only exists at the defect, known as an
auxiliary eld.
It was also noticed that a system containing a type I B defect and a type I B
T
defect with the
same parameter d in a
(1)
2
would give the same overall time delay to the species 1 and species 2
solitons
2
. As such, solitons possessing the a
(2)
2
symmetry in the eld u would also possess it in
the eld v. This opens up the possibility that (1.11) for a
(2)
2
somehow arises from fusing two
type I a
(1)
2
defects. This possibility is formulated and generalised to all ATFTs obtainable by
folding a
(1)
r
in this paper.
1.2 Layout of this paper
Section 2 describes the foldings of a
(1)
r
which will be used in this paper, which may be referred
to as bivalent foldings since the roots of the folded theories are obtained by identifying roots
of a
(1)
r
pairwise (contrast with e.g., d
(1)
4
a
(2)
2
by which one of the folded roots is obtained
by identifying four of the roots of d
(1)
4
). Such foldings were also considered in [9]. Section
3 describes how solitons of the folded theories may be obtained from a
(1)
r
solitons via these
folding processes. The relevant Hirota tau functions [12] are obtained. Where the folding being
done is non-canonical (in the sense of [9], i.e., foldings not found in [13]) the tau functions are
compared to those found in [6], which were obtained by canonical folding of d
(1)
s
models. The
energy and momentum of solitons are then considered. An argument is given which suggests
that at this classical level there is no need to consider just imaginary coupling. Section 4
describes a fused a
(1)
r
defect and its eects on solitons and energy and momentum conservation.
Section 5 considers folded defects and how they aect the folded solitons. Conservation of
modied energy and momentum is then proven for the folded defects. This section is perhaps
the most important, and opens up a number of possible future research directions. Section 6
gives the conclusions and outlook for this discourse. The appendices contain some additional
calculations, including some explicit calculations in a
(2)
2
.
2 Bivalent foldings of a
(1)
r
In this section the folding process used in later sections is formulated. It was noted in [9] that
numerous foldings of this type are possible for a given a
(1)
r
and whilst other ways to fold a
(1)
r
are
considered in [9] and [14], no extra folded theories arise in this way. Here the bivalent folding
is done explicitly by means of a parameter k Z.
This paper places its foundations in the rm ground of a
(1)
r
, so it is useful to rst state some of
the properties of a
(1)
r
. The roots of a
(1)
r
obey

i

j
= 2
ij

i(j+1)

i(j1)
(2.1)
for i, j = 0, . . . , r. It is useful to extend this range for i and j by identifying the indices modulo
the Coxeter number h. For a
(1)
r
, h = r +1, so e.g.,
2

r1
. For a
(1)
r
the roots conventionally
2
P. Bowcock, private communication. Also mentioned in [3]
5
all have length

2 so the Cartan matrix is given by

C
ij
=
2

j

j

i

j
=
i

j
i.e., (2.1). Note that there is no sum over j. The eigenvalues of the Cartan matrix play an
important role in the construction of solitons.
Out of the r + 1 roots {
0
, . . . ,
r
} any choice of r of them form a basis for the root space.
Conventionally, the fundamental weights {
i
} are given by

i

j
=
ij
, i, j = 1, . . . , r
with
0
= 0 so they form a dual basis to {
1
, . . . ,
r
}. However, it is equally valid to dene a
set of fundamental weights by e.g.,

i

j
=
ij
, i, j = 0, 1, . . . , q 1, q + 1, . . . , r
with
q
= 0. Note then that in (1.7) and its transpose the sum may be extended to include
j = 0 and then whichever set of fundamental weights is used B and B
T
remain the same as
before. Note also that the relation

i
= 2
i

i1

i+1
is also valid for any choice of {
i
}.
The ane Toda eld, u, lives on the a
(1)
r
root space so a natural basis is given by the simple
roots {
1
, . . . ,
r
}: u = u
1

1
+. . . u
r

r
. However, one can consider an extended representation
u =
0
u
0
+
1
u
1
+. . . +
r
u
r
where only r of the r + 1 components are independent. Noting then that u
i
=
i
u, it is
reasonable to consider u
q
= 0 for some q = 0 whilst retaining u
0
. This innocuous observation
plays a role when considering the folding here in its full generality.
2.1 The bivalent foldings
The foldings considered here are bivalent in the sense that the roots of the folded theory will be
obtained by identifying all roots of a
(1)
r
pairwise. Specically, the roots
i
and
h+ki
in a
(1)
r
will
be paired up for all i. Which folded theory is obtained depends upon whether h is even (r odd)
or odd (r even) and also on whether k is even or odd. That there are a number of distinct ways
to identify the roots whilst obtaining a folded theory was already noted in [9]. Roots in the
folded theories will be labelled by

.
2.1.1 Case 1: h even, k even
This case includes the only canonical folding in this paper (i.e., the only one also found in [13])
when k = 0, that is a
(1)
2n1
c
(1)
n
(h = 2n). The following Dynkin diagrams illustrate this
folding.
6
e
e e e e e
e
e e e e e

Z
Z

Z
Z
a
(1)
2n1
-
6
?
e u u u u u e
c
(1)
n
Figure 1: a
(1)
2n1
c
(1)
n
White nodes correspond to roots of length

2 and black nodes length 1. The nodes in the above

diagram have purposely not been labelled. For the a
(1)
2n1
picture
0
could really be placed any-
where, though conventionally it would be placed on the left. To go from a to c diagrammatically
the nodes that are directly above/below each other are identied.
Here two nodes are self-identied, so there exist two solutions to
i
=
h+ki
; the self-identied
nodes are k
2
and h+k
2
= k
2
+n
. Upon folding the labelling of nodes is no longer democratic
and to get the right Cartan matrix for the folded theory, the extra root

0
must be identied
with one of these self-identied nodes.
The choice made here is

0
= k
2
implying that

i
=
k
2
+i
+k
2
+2ni
2
.
For the a
(1)
2n1
eld u fold by identifying uk
2
= 0; uk
2
+i
= uk
2
+2ni
=

i
2
for i = 1, . . . , n 1 and
uk
2
+n
=
n
. So
u =
2n1

j=0
u
j

j
=
n

j=0

j
k
2
+j
+k
2
+2nj
2
.
In above, note that when j =
h
2
= n, both a
2n1
roots there are the same, so that term is
merely
n

n+
k
2
. The notation is that u is an unfolded eld, and is a folded c
(1)
n
eld, so
=
n

j=0

j
.
The claim is that, rstly, {

i
} are the roots of c
(1)
n
and, secondly, that the Lagrangian (1.1) for
a
(1)
2n1
folds to that of c
(1)
n
. For the rst part of the claim, by (2.1)

j
=
k
2
+i
+k
2
+2ni
2

k
2
+j
+k
2
+2nj
2
=
1
2
_
2
ij

i(j+1)

i(j1)
+ 2
i(2nj)

i(2n+1j)

i(2n1j)
_
7
so the inner products are
(i, j)

j
(0, 0) 2
(s, s) s = 0, n 1
(n, n) 2
(s, s + 1) s = 0, n 1
1
2
(0, 1) 1
(n 1, n) 1
with all other cases giving zero. These are the inner products of c
(1)
n
, conrming the veracity of
the folding. With this observation it is clear that the kinetic terms of the Lagrangian (1.1) are
as required when folding u . Less obvious is the potential: for a
(1)
2n1
, (1.2) gives
U(u) =
2n1

j=0
(e

j
u
1)
since all of the marks of a
(1)
r
obey n
j
= 1. This can be rewritten
U = e
u
k
2
+
n1

j=1
_
e
u
k
2
+j
+e
u
k
2
+2nj
_
+e
u
k
2
+n
2n .
Note that k
2
+i

j
= k
2
+2ni

j
=

j
, so folding gives the potential
= e

k
2
+
n1

j=1
_
e

k
2
+j
+e

k
2
j
_
+e

k
2
+n
2n
= e

0
+ 2
n1

j=1
e

j
+e

n
2n
=
n

j=0
n
j
_
e

1
_
(2.2)
with {n
i
} the marks of the c
(1)
n
algebra. Thus, whichever even k is used this folding prescription
sends the a
(1)
2n1
ATFT to the c
(1)
n
ATFT.
2.1.2 Case 2: h even, k odd
This case is quite dierent, in as much as all nodes are identied with one other. This case
gives a
(1)
2n1
d
(2)
n
(despite the notation d
(2)
n
has just n nodes instead of n + 1). There are
equivalent choices of

0
since the Dynkin diagram looks the same from both ends. The choice

0
=

k+1
2
+
k1
2
+2n
2
is made, which includes
0
when k = 1. The diagram that illustrate this
are
8
e e e e e e
e e e e e e
a
(1)
2n1
-
6
?
er u u u u re
d
(2)
n
Figure 2: a
(1)
2n1
d
(2)
n
The black-in-white node has length
1

2
.
The identication of roots in the folded theory is chosen to be

i
=
k+1
2
+i
+k1
2
+2ni
2
and so
u =
2n1

j=0
u
j

j
=
n1

j=0

j
k+1
2
+j
+k1
2
j
2
=
n1

j=0

j
The normalisation of the d
(2)
n
roots here is non-canonical so this could be viewed with suspicion,
however the inner products
(i, j)

j
(0, 0)
1
2
(s, s) s = 0, n 1
(n, n)
1
2
(s, s + 1)
1
2
are precisely half of the standard inner products of d
(2)
n
. The remedy, if conventional normali-
sation is required, is simply to rescale the roots

i
.
The potential obtained by this folding is
=
n1

j=0
_
e

k+1
2
+j
+e

k1
2
+2nj
_
2n
= 2
n1

j=0
n
j
_
e

1
_
(2.3)
since all the marks are equal to one in d
(2)
n
. This potential obtained is twice the standard d
(2)
n
potential. The factor of two may be removed in the action by an isotropic space-time rescaling:
x
x

2
, t
t

2
. Such a rescaling does not aect the kinetic terms, so the standard d
(2)
n
ATFT
Lagrangian is obtained.
2.1.3 Case 3: h odd
This case covers a
(1)
2n
a
(2)
2n
, so h = 2n + 1. Since the indices on the roots of a
(1)
2n
are identied
modulo this odd Coxeter number, there is no longer a notion of odd and even indices (e.g.,
9

1

2n+2
). Due to this, k odd and k even give the same folding, and so the choice will be made
that k is even (to ensure that all foldings are considered, a suitable range is k = 0, 2, 4, . . . , 4n).
This case contains one self-identied node, k
2
which will be chosen to correspond to

0
.
e
e e e e e e
e e e e e e

Z
Z
a
(1)
2n
-
6
?
e u u u u u re
a
(2)
2n
Figure 3: a
(1)
2n
a
(2)
2n
There are n + 1 roots in the folded theory and the identication made is

i
=
k
2
+i
+k
2
+2n+1i
2
.
The inner products are
(i, j)

j
(0, 0) 2
(p, p) p = 0, n 1
(n, n)
1
2
(p, p + 1) p = 0
1
2
(0, 1) 1
which are the correct inner products for a
(2)
2n
.
The potential becomes
=
n

j=0
n
j
_
e

1
_
(2.4)
with n
0
= 1 and n
i
= 2 for i = 0. Exactly as required for the folding to give a
(2)
2n
ATFT.
Thus it is shown that these foldings are valid, as expected from [9]. Given this, it is rea-
sonable to consider the construction of solitons for c
(1)
n
, d
(2)
n
and a
(2)
2n
using these foldings (as this
is the only folding process considered in this paper, these ATFTs will be collectively referred to
as the folded theories).
3 Solitons and folding
This section details the construction of solitons in the folded theories, all of which are a
(1)
r
solitons
possessing a certain symmetry (namely
i
=
k+hi
). In this context, a
(1)
r
solitons must be
introduced rst.
10
3.1 a
(1)
r
solitons
Hirota methods [12] are used to nd soliton solutions in the ATFTs. For this paper the starting
point will be a
(1)
r
, as it is the theories obtained from folding this that are of interest. For a
general ATFT the soliton solutions may be written in the form [5]
u =
1

j=0

j
ln
j
(3.1)
where
i
=
2

i
with no sum implied and the ATFT corresponds to an ane Dynkin diagram
with r + 1 nodes.
The tau functions {
j
} depend only upon the root data (i.e., which ATFT is being consid-
ered) and do not depend upon the coupling. In order to nd the tau functions (3.1) is used in
the equation of motion (1.3) along with a decoupling [4]. In general the equation to be solved
is

i
_

i

i

2
i

i

i
+
2
i
_
n
i
_
_
r

j=0

i
j
1
_
_

2
i
= 0 . (3.2)
Note that the kinetic terms are always in Hirota bilinear form
i

i

2
i

i

i
+
2
i
=
1
2
_
D
2
t
D
2
x
_

i
, but the potential is only in true bilinear form for a
(1)
r
. For a
(1)
r
,
i
= n
i
= 1 for all i, so (3.1)
becomes
u =
1

j=0

j
ln
j
(3.3)
while the equation the tau functions must obey simplies greatly to

i

i

2
i

i

i
+
2
i
=
i1

i+1

2
i
. (3.4)
3.2 Folding a
(1)
r
solitons
In all cases folding is achieved by identifying
i
with
k+hi
, so to fold the a
(1)
r
ATFT set
u
i
= u
k+hi
. In the soliton Ansatz (3.3) this is tantamount to having
i
=
k+hi
. Any
a
(1)
r
soliton with this property is also a soliton of the folded theory. A summary of the tau
function identications required for the folded solitons to t the Ansatz (3.1) is given here:
c
(1)
n
d
(2)
n
a
(2)
2n

0
k
2
(k+1
2
)
1
2
k
2

i
i = 0, n 1, n k
2
+i
k+1
2
+i
k
2
+i

n1
k
2
+n1
(k+1
2
+n1
)
1
2
k
2
+n1

n
k
2
+n
(k
2
+n
)
1
2
The details may be found in the appendix A.
The possibility of obtaining solitons in the folded theories is dependent on the relation
i
=

k+hi
having solutions for all i, so the functions will need to be examined. It is already
known that for c
(1)
n
in the case k = 0 that such a relation is possible [5].
11
3.3 Tau functions
3.3.1 One soliton solution of a
(1)
r
The tau functions for the one soliton solution (of species p) of a
(1)
r
are of the form

j
= 1 +
pj
E
p
with = e
2i
r+1
, and p = 1, . . . , r (note that p = 0 just gives the trivial solution), so that

p
encompasses the (r + 1)-th roots of unity (where r + 1 = h, the Coxeter number). The
expression E
p
is given by E
p
= e
a
p
xb
p
t+c
p
, c
p
is constant: its imaginary component determines
the topological charge of the soliton under consideration- there are up to r+1 sectors of diering
topological charge [15, 6]. In order to avoid singularities in the soliton solution (avoiding
j
= 0)
this imaginary part of c
p
should be chosen carefully. The real part of c
p
gives the position of the
centre of mass of the soliton at time t = 0 and may be chosen arbitrarily. a
p
=
p
cosh and
b
p
=
p
sinh with R() > 0 for a right-moving soliton ( is the rapidity- or at least R() is).

p
is taken to be the positive square root of the pth eigenvalue of the matrix (NC)
ij
= n
i

i

j
.
For a
(1)
r
these are merely the ordered eigenvalues of the Cartan matrix of the (ane) Lie algebra,
given by

2
p
= 4 sin
2
_
p
r + 1
_
(3.5)
with most eigenvalues appearing twice. p = 0,
2
0
= 0 (in all cases), and p = n,
2
n
= 4, (in the
case r = 2n 1) are the only eigenvalues which arent degenerate. It should be noted that the
square masses of the a
(1)
r
solitons are proportional to these eigenvalues, so it is not erroneous to
refer to {
p
} as the masses of the solitons. One thing to note is that the {
p
} in the folded
theories obtained from a
(1)
r
in this paper are exactly the same, only that the degeneracy has
been removed (the masses of the basic soliton solutions in the folded theories are typically
{2
p
}). This ought to be contrasted to the cases of d
(2)
n
and a
(2)
2n
when folded from d
(1)
s
theories-
there the eigenvalues of NC for d
(1)
s
need to be rescaled to obtain those of the folded theories [6].
It will be convenient to rewrite the tau function in a slightly dierent way as

j
= 1 +
pj+p

k
E
p
(3.6)
which is equivalent to redening the imaginary part of c
p
. For the one soliton solution this has
no purpose, but for the two soliton solution this reparametrisation will prove useful.
Note at this point that there is but one non-trivial soliton of a
(1)
r
which is compatible with
folding (i.e., has
j
=
h+kj
), which is the n soliton of a
(1)
2n1
with k even. In that case

j
= 1 + (1)
j
E =
2n+kj
, since (1)
2n+k
= 1. This soliton is the species n single soliton of
c
(1)
n
, as well as being a bona de a
1
soliton- more on a
1
embeddings is included in appendix B.
3.3.2 Two soliton solution of a
(1)
r
The tau functions of a two soliton solution (one of type p; the other of type q) in a
(1)
r
have the
form

j
= 1 +
pj
E
p
+
qj
E
q
+A
(pq)

(p+q)j
E
p
E
q
(3.7)
12
where there is now an interaction parameter
A
(pq)
=
(a
p
a
q
)
2
(b
p
b
q
)
2

2
pq
(a
p
+a
q
)
2
(b
p
+b
q
)
2

2
p+q
(3.8)
and a
p
=
p
cosh
1
, a
q
=
q
cosh
2
, etc. (i.e., there are two rapidities to be considered). If
p and q are the same, and both rapidities are the same, then (3.8) vanishes, and in fact (3.7)
describes a one soliton solution (from this point of view the n soliton of c
(1)
n
may also be thought
of as a two soliton solution of a
(1)
2n1
).
For a
(1)
r
solitons to also be solitons of the folded theories it must be the case that
j
=
k+hj
.
After folding (3.7) is to be interpreted as a one soliton solution, so the constituent solitons must
be given the same centre of mass, R(c
p
) = R(c
q
), as well as the same rapidity,
1
=
2
.
Also, there is little chance of satisfying
j
=
k+hj
if the quadratic term in E depends on j:
the conclusion is that
p+q
= 1 and so q = h p. Consequently, the solitons being paired up
possess the same mass, meaning that a
p
= a
q
, b
p
= b
q
. Thus, E
p
and E
q
may only dier in
I(c).
From (3.8), with
1
=
2
,
A
(p(hp))
=

2
2p
4a
2
p
4b
2
p
=
4 sin
2
_
2p
r+1
_
16 sin
2
_
p
r+1
_ = cos
2
_
p
r + 1
_
A . (3.9)
Thus, the tau functions compatible with folding (to a one soliton folded solution) all possess
the form

j
= 1 +
_

pj
+
pkpj
_
E
p
+A
pk
E
2
p
(3.10)
which in a
(1)
r
navely appears to have the interpretation of a species p a
(1)
r
soliton joined to a
species hp soliton which has its topological charge shifted into a dierent sector. It turns out
however, that the topological charges of the folded solitons only depend on whether k is odd or
even (and even then only when h is even) and so all possibilities are actually encompassed by
k = 0 and k = 1. Nonetheless, this paper will continue to work with arbitrary k.
3.3.3 Single solitons in the folded theories
Here the folded solitons (3.10) with tau function identications given in section 3.2 are compared
to the tau functions listed by McGhee in [6], thereby conrming that (3.10) are indeed the bona
de single soliton solutions of the folded theories
c
(1)
n
tau functions
In this case k is even, and as identied in section 3.2,

j
= k
2
+j
, so

j
= 1 +
pk
2
_

pj
+
pj
_
E
p
+A
pk
E
2
p
.
This does not match [6] unless k = 0; however the freedom to shift I(c
p
) allows E
p

pk
2
E
p
to give

j
= 1 + 2 cos
_
pj
n
_
E
p
+ cos
2
_
p
2n
_
E
2
p
13
which does match [6]. That changing k only shifts E
p
claries the comment at the end of section
3.3.2, i.e., any even k gives the same folded soliton possessing the same topological charges.
d
(2)
n
tau functions
This case is signicantly more interesting to verify as the folding process previously used to nd
d
(2)
n
solitons made use of d
(1)
n+1
solitons [5, 6], rather than those of a
(1)
2n1
(note that d
(1)
n+1
and
a
(1)
2n1
do possess the same Coxeter number h = 2n)- the identication of roots required for the
canonical folding d
(1)
n+1
d
(2)
n
is given diagrammatically in the discussion, section 6.
In this case k is odd with

0
=
_
k+1
2
,

n1
=
_
k+1
2
+n1
and

j
= k+1
2
+j
for j = 1, . . . , n2.
In general
k+1
2
+j
= 1 +
pk
2
_

p(j+
1
2
)
+
p(j+
1
2
)
_
E
p
+A
pk
E
2
p
.
The k dependence can then be scaled away by E
p
E
p

pk
2
. Further scaling E
p
by E
p

E
p

A
=
E
p
cos(
p
2n
)
gives
k+1
2
+j
= 1 + 2
cos
_
p(2j+1)
2n
_
cos
_
p
2n
_ E
p
+E
2
p
which already matches

j
in [6] for j = 0, n 1
3
. For the tau functions corresponding to the
end nodes,
k+1
2
= 1 + 2E
p
+E
2
p
= (1 +E
p
)
2
=

0
= 1 +E
p
and
k+1
2
+n1
= 1 + 2
cos
_
p(2n1)
2n
_
cos
_
p
2n
_ E
p
+E
2
p
= 1 + 2(1)
p
E
p
+E
2
p
= (1 + (1)
p
E
p
)
2
=

n1
= 1 + (1)
p
E
p
in agreement with [6]. Note that the eigenvalues
p
of d
(2)
n
are exactly those inherited from
a
(1)
2n1
so do not have to be rescaled, unlike with the canonical folding [6].
a
(2)
2n
tau functions
This case is equally interesting, with the canonical folding being d
(1)
2n+2
a
(2)
2n
, instead of
a
(1)
2n
a
(2)
2n
- the canonical folding is illustrated in the discussion.
In this case k may be taken to be even, with

n
=
_
k
2
+n
, whilst

j
= k
2
+j
for j = 0, . . . , n1.
From (3.10)
k
2
+j
= 1 +
pk
2
_

pj
+
pj
_
E
p
+A
pk
E
2
p
3
Note that [6] uses the wrong Coxeter number.
14
so, shifting E
p

E
p

pk
2
cos(
p
2n+1
)
results in
k
2
+j
= 1 + 2
cos
_
2pj
2n+1
_
cos
_
p
2n+1
_E
p
+E
2
p
which matches

j
in [6] when j = n (though ordering is reversed- note that a
(2)
2n
is self-dual)
4
.
When j = n the result is
k
2
+n
= 1 + 2
cos
_
2np
2n+1
_
cos
_
p
2n+1
_E
p
+E
2
p
= 1 + 2(1)
p
E
p
+E
2
p
= (1 + (1)
p
E
p
)
2
=

n
= 1 + (1)
p
E
p
.
This matches [6] but with the ordering reversed. Again the eigenvalues obtained need not be
rescaled.
Thus the previously known single solitons of each of the folded theories are recovered. The
relative simplicity of a
(1)
r
allows easier construction of the multisoliton solutions than via d
(1)
s
,
which would otherwise be necessary.
3.3.4 Other solutions
Once the basic soliton solutions of the folded models have been found, multisolitons in these
models can be constructed; however, this requires knowledge of the generally complicated inter-
action parameters of the folded model. This problem may be obviated by instead constructing
these multisolitons in the a
(1)
r
model. This was noted in [16], but was only applied to c
(1)
n
.
A formula for multisoliton solutions in a
(1)
r
is [4]

j
=
1

1
=0
. . .
1

N
=0
exp
_
_
N

i=1

i
ln
_

p
i
j
E
p
i
_
+

1i<jN

j
ln
_
A
(p
i
p
j
)
_
_
_
. (3.11)
Using (3.11), the two soliton solution of the folded theory can be constructed by means of a
four soliton solution of a
(1)
r
. To obtain a general two soliton solution in the folded theory the
four a
(1)
r
soliton solution will arise from the single soliton tau functions:

1
j
= 1 +
pj
E
p
type p with rapidity
1
,

2
j
= 1 +
p(kj)
E
p
type h p with rapidity
1
,

3
j
= 1 +
qj
E
q
type q with rapidity
2
,

3
j
= 1 +
q(kj)
E
q
type h q with rapidity
2
.
4
There is an error in [6] in the cosine in the numerator.
15
Then, by (3.11), the four soliton a
(1)
r
solution has tau functions

j
= 1 +
_

pj
+
p(kj)
_
E
p
+
_

qj
+
q(kj)
_
E
q
+A
(12)

pk
E
2
p
+A
(34)

qk
E
2
q
+
_
A
(13)

pj+qj
+A
(14)

pj+q(kj)
+A
(23)

p(kj)+qj
+A
(24)

p(kj)+q(kj)
_
E
p
E
q
+A
(12)

pk
_
A
(13)
A
(23)

qj
+A
(14)
A
(24)

q(kj)
_
E
2
p
E
q
+A
(34)

qk
_
A
(13)
A
(14)

pj
+A
(23)
A
(24)

p(kj)
_
E
p
E
2
q
+A
(12)
A
(13)
A
(14)
A
(23)
A
(24)
A
(34)

pk+qk
E
2
p
E
2
q
.
This is a solution of the folded theory provided that
j
=
k+hj
. For this to be possible it must
be the case that
A
(13)
= A
(24)
A
(14)
= A
(23)
and so the two soliton folded solution has

j
= 1 +
_

pj
+
p(kj)
_
E
p
+
_

qj
+
q(kj)
_
E
q
+A
(12)

pk
E
2
p
+A
(34)

qk
E
2
q
+A
(13)
_

pj+qj
+
p(kj)+q(kj)
_
E
p
E
q
+A
(14)
_

pj+q(kj)
+
p(kj)+qj
_
E
p
E
q
+A
(12)
A
(13)
A
(14)

pk
_

qj
+
q(kj)
_
E
2
p
E
q
+A
(34)
A
(13)
A
(14)

qk
_

pj
+
p(kj)
_
E
p
E
2
q
+A
(12)
A
(34)
_
A
(13)
_
2
_
A
(14)
_
2

pk+qk
E
2
p
E
2
q
. (3.12)
Note that there is no need to convert these tau functions to those that are conventional from
(3.1) as these tau functions with the Ansatz (3.3) and appropriate identication of roots are
the solitons of the folded theories. Note that (3.12) contains four interaction parameters- a fact
that is not obvious, should one wish to construct folded solitons using the folded theory as a
starting point.
Using that a
p
=
p
cosh
1
, a
q
=
q
cosh
2
, b
p
=
p
sinh
1
, b
q
=
q
sinh
2
and denoting the
two rapidities by
1
= + and
2
= gives the interaction parameters as
A
(13)
=

2
pq
+ (
p
+
q
)
2
sinh
2
(
p

q
)
2
cosh
2

(
p
+
q
)
2
cosh
2
(
p

q
)
2
sinh
2

2
p+q
(3.13)
A
(14)
=

2
p+q
+ (
p
+
q
)
2
sinh
2
(
p

q
)
2
cosh
2

(
p
+
q
)
2
cosh
2
(
p

q
)
2
sinh
2

2
pq
(3.14)
A
(12)
= cos
2
_
p
r + 1
_
(3.15)
A
(34)
= cos
2
_
q
r + 1
_
. (3.16)
Among these two soliton solutions, an interesting case is when the relative rapidity
1

2
= 2
between the solitons is imaginary. This can result in a breather (if p = q), and can also result
in an interaction parameter being singular (namely (3.13)) resulting in fusion.
16
Fusing rules in the folded theories
It is known, certainly in the simply laced theories, that the rapidities of a two soliton solution
may be chosen such that fusion occurs, resulting in a one soliton solution [4, 17, 18]. The two
soliton solution (3.12) is used here to show that fusion occurs in the same vein as in the simply
laced theories.
In the folded theory a one soliton solution of species s has the form

j
= 1 +
_

sj
+
sksj
_

E + cos
2
_
s
r + 1
_

sk

E
2
(3.17)
and so the aim of this section is to show that (3.12) may be reduced to this form. For this
to be the case it will be necessary that the highest order term in (3.12), (

A)
pk+qk
E
2
p
E
2
q
, is
equal to the highest order term in (3.17), cos
2
_
s
r+1
_

sk

E
2
. This necessitates the identications
s = p +q (identied modulo h) and

E = E
p
E
q
. This still leaves unwanted terms in (3.12), but
shifting E
p

_
A
(13)
_

1
2
E
p
and E
q

_
A
(13)
_

1
2
E
q
gives

j
= 1 +
_
A
(13)
_

1
2
_

pj
+
p(kj)
_
E
p
+
_
A
(13)
_

1
2
_

qj
+
q(kj)
_
E
q
+
_
A
(13)
_

1
2
A
(12)

pk
E
2
p
+
_
A
(13)
_

1
2
A
(34)

qk
E
2
q
+
_

pj+qj
+
p(kj)+q(kj)
_
E
p
E
q
+A
(14)
_
A
(13)
_

1
2
_

pj+q(kj)
+
p(kj)+qj
_
E
p
E
q
+A
(12)
A
(14)
_
A
(13)
_

1
2

pk
_

qj
+
q(kj)
_
E
2
p
E
q
+A
(34)
A
(14)
_
A
(13)
_

1
2

qk
_

pj
+
p(kj)
_
E
p
E
2
q
+A
(12)
A
(34)
_
A
(14)
_
2

pk+qk
E
2
p
E
2
q
so all the unwanted terms may be removed provided that the relative rapidity between the p
and q solitons is chosen such that there is a pole in A
(13)
, which leaves the tau functions

j
= 1 +
_

sj
+
s(kj)
_

E +A
(12)
A
(34)
_
A
(14)
_
2

sk

E
2
and so it needs to be shown that A
(12)
A
(34)
_
A
(14)
_
2
= cos
2
_
s
r+1
_
and that

E is valid for a
species s soliton, i.e.,

E = e

s
cosh

x
s
sinh

t+ c
.
For there to be a pole in A
(13)
the denominator of (3.13) must vanish, resulting in
= i
(p +q)
2(r + 1)
i

. (3.18)
The choice of the plus sign is made in what follows (the sign makes no real dierence). The
rapidity dierence between the solitons 2 is i times the fusing angle. Having xed the eect
17
on A
(14)
is
A
(14)
=
cos
_
(p+q)
r+1
_
+ cos(2

)
cos
_
(pq)
r+1
_
+ cos(2

)
=
cos
_
s
r+1
_
cos
_
p
r+1
_
cos
_
q
r+1
_
and so indeed
A
(12)
A
(34)
_
A
(14)
_
2
= cos
2
_
s
r + 1
_
is satised as required.
The nal step in checking that fusion has occurred is to show that

E has the desired form,

E = e

s
cosh

s
sinh

+ c
,

E = E
p
E
q
= exp [(
p
cosh( +) +
q
cosh( )) x
(
p
sinh( +) +
q
sinh( )) t +c
p
+c
q
]
so it must be the case that c = c
p
+c
q
while the prefactor of x gives

p
cosh( +) +
q
cosh( ) = 4 sin
_
(p +q)
2(r + 1)
_
cos
_
(p q)
2(r + 1)
_
cos
_
(p +q)
2(r + 1)
_
cosh
+ 4i cos
_
(p +q)
2(r + 1)
_
sin
_
(p q)
2(r + 1)
_
sin
_
(p +q)
2(r + 1)
_
sinh
=
s
cosh
_
+i
(p q)
2(r + 1)
_
and similarly the prefactor of t gives
s
sinh
_
+i
(pq)
2(r+1)
_
. Thus,

= +i
(p q)
2(r + 1)
is the expression for the rapidity of the one soliton solution which arises from such a fusion.
Note that an alternative choice of rapidities

p
= +i
q
r + 1

q
= i
p
r + 1
will result in the species p +q (modulo h) soliton obtained by fusion to have rapidity .
Breathers in the folded theories
The existence of breather solutions in ATFT has been known for some time [17] and have been
considered in Hirota form for a
(1)
r
[19] and d
(1)
4
[20]. In this section the breathers of the folded
theories c
(1)
n
, d
(2)
n
and a
(2)
2n
are constructed- they are a
(1)
r
breathers which respect
j
=
k+hj
.
Equation (3.12) can describe folded breather solutions. This will happen when the constituent
solitons are of the same species p = q with the same centre of mass R(c
1
) = R(c
2
) (where
18
E
p
E
1
and E
q
E
2
) and with imaginary rapidity dierence, = i

(where
1
= + ,

2
= ). Note that 2

must be less than the fusing angle.

Taking p = q and such gives simple forms for the interaction parameters
A
(12)
= A
(34)
= cos
2
_
p
r + 1
_
A
(13)
=
sin
2

cos
2
_
p
r+1
_
cos
2

A
(14)
=
cos
2
_
p
r+1
_
sin
2

cos
2

so there are three interaction parameters involved and the tau functions take the form

j
= 1 +
_

pj
+
p(kj)
_
(E
1
+E
2
) +A
(12)

pk
(E
2
1
+E
2
2
)
+A
(13)
_

2pj
+
2p(kj)
_
E
1
E
2
+ 2A
(14)

2pk
E
1
E
1
+A
(12)
A
(13)
A
(14)

pk
_

pj
+
p(kj)
_
E
1
E
2
(E
1
+E
2
)
+
_
A
(12)
A
(13)
A
(14)
_
2

2pk
(E
1
E
2
)
2
.
These tau functions have spacetime dependence through E
1
+E
2
, E
2
1
+E
2
2
and E
1
E
2
. Denoting
C =
c
1
+c
2
2
and i =
c
1
c
2
2
and taking the rest frame = 0 gives
E
1
+E
2
e

p
cos x+C
cos (
p
sin t + )
E
2
1
+E
2
2
e
2
p
cos x+2C
cos (2
p
sin t + 2)
E
1
E
2
e
2
p
cos x+2C
so it is readily seen that all time dependence in the rest frame is oscillatory, as expected.
Note also, that in the cases of d
(2)
n
and a
(2)
2n
these breather tau functions are also the tau
functions of particular breathers in d
(1)
s
.
The validity of the results here is further evidenced by consideration of the soliton energies
and momenta.
3.4 Energy-momentum of solitons
3.4.1 Coupling dependence of energy, momentum and mass
Consider the Lagrangian for bulk ATFT (1.1) at general coupling with m = 1
L =
1
2
u u
1
2
u

+
1

2
r

j=0
n
j
_
e

j
u
1
_
.
Then the energy and momentum densities are obtained via the stress tensor
T

=
L
(

u)

L (3.19)
19
where is the at metric of 1+1 dimensional space. In standard coordinates used here x
0
= t,
x
1
= x so
00
= 1,
11
= 1,
10
=
01
= 0; though lightcone coordinates can be more useful in
examining conserved charges. In particular,
T
00
=
1
2
u
2
+
1
2
u
2
+U(u) = E
T
11
=
1
2
u
2
+
1
2
u
2
U(u)
T
01
= T
10
= uu

= P
where E is the energy density and P the momentum density.
Now, consider soliton solutions at general coupling (3.1) for an r component ane Toda eld
u =
1

i=0

j
ln
j
=
1

u
with u denoting the soliton solution obtained if is chosen to be unity (note that the tau func-
tions do not depend upon the coupling). Then, for such a solution the energy and momentum
densities are
E(u) =
1

2
_
_
1
2

u

u +
1
2
u

j
n
j
_
e

j
u
1
_
_
_
=
1

2
E( u) (3.20)
P(u) =
1

2
_

u u

_
=
1

2
P( u) . (3.21)
Since the coupling dependence just boils down to a prefactor
1

2
, it is clear that the total energy
and momentum in the bulk, which are just the integrals of (3.20) and (3.21) over space, similarly
only depends on the coupling via the same prefactor
E(u) =
1

2
E( u) , P(u) =
1

2
P( u) .
In [4] the coupling was taken to be imaginary, = i (so E(u) =
1

2
E( u), P(u) =
1

2
P( u))
and the energy and momentum of a species p soliton in a
(1)
r
were found to be
E(u) =
2h

2
a
p
=
4(r + 1)

2
sin
_
p
r + 1
_
cosh()
P(u) =
2h

2
b
p
=
4(r + 1)

2
sin
_
p
r + 1
_
sinh()
which are both real and positive, provided that the rapidity is real, and give the soliton mass
M =
_
E
2
P
2
=
2h

p
=
4(r + 1)

2
sin
_
p
r + 1
_
.
Since E(u) and P(u) are both positive in the imaginary coupling case the conclusion is that
E( u) and P( u), the energy and momentum for = 1, are both real but negative (the same is
true for all real couplings).
20
From a purely classical standpoint, there is no a priori reason not to allow unit coupling,
= 1. If this causes the wrong sign of energy and momentum to appear one might suppose
that the denitions of energy and momentum used are wrong. Classically, there is no prob-
lem with rescaling the action- rescaling does not aect the Euler-Lagrange equations obtained
(provided there is no boundary contribution to the action). The particular choice

L =
_

2
||
2
_
(1.1)
=
_

2
||
2
_
_
_
1
2
u u
1
2
u

2
r

j=0
n
j
_
e

j
u
1
_
_
_
(3.22)
will result in real, positive energy and momentum for solitons with any complex coupling. For
the species p soliton of a
(1)
r
the expectation is
E =
2h
||
2
a
p
=
2h
||
2

p
cosh() (3.23)
P =
2h
||
2
b
p
=
2h
||
2

p
sinh() (3.24)
M =
2h
||
2

p
=
4(r + 1)
||
2
sin
_
p
r + 1
_
. (3.25)
3.4.2 Calculation of energy and momentum of solitons
A construction rst found in [21] and applied in a similar way to here in [19] involves the
observation that for solitons (3.19) may be written in the form
T

=
_

_
C (3.26)
such that
T
00
= C

(3.27)
T
11
=

C (3.28)
T
01
= T
10
= +

C

. (3.29)
In particular this means that

2
C = T
00
T
11
= 2U =
2
||
2
r

i=0
n
i
_
e

i
u
1
_
where the rescaled Lagrangian (3.22) has been used. Up to a constant, the solution to this is
given by
C =
2
||
2
r

j=0

j
ln
j
(3.30)
since

2
C =
2
||
2
r

j=0

j

j

j

2
j

j

j
+
2
j

2
j
=
2
||
2
r

i=0
n
i
_
e

i
u
1
_
21
where (3.2) has been used.
Having C is useful as it allows the energy and momentum to be calculated much more eas-
ily. That is,
E =
_

T
00
dx =
_
C

x=
(3.31)
P =
_

T
01
dx =
_

C
_

x=
(3.32)
so it is clear that the actual values of E and P depend upon the particular form of the tau
functions of the soliton/multisoliton being considered.
3.4.3 Results
In what follows the coupling is scaled to = 1 and the context will be a
(1)
r
so
j
= 1 j.
Noteworthy is the observation that C does not change when the theory being considered is one
of those bivalently folded from a
(1)
r
: i.e., if
j
=
k+hj
in a
(1)
r
then
C = 2
r

j=0
ln
j
= 2

j
ln

j
where the primed identications are those of the folded theory and the range of j in the folded
theory depends upon which folding is being considered- the number of terms is the number of
nodes on the folded Dynkin diagram (see appendix A for tau function identications).
With this observation it is clear that folding in this manner does not change the energy-
momentum of the solitons- the a
(1)
r
soliton with the symmetry
j
=
k+hj
has the correct
energy-momentum for the folded theory.
Single solitons of a
(1)
r
The tau functions are given by (3.6), so for a species p soliton

j
= 1 +
pj+p

k
E
p
=

j
=
a
p

pj+p

k
E
p
1 +
pj+p

k
E
p
and

j

j
=
b
p

pj+p

k
E
p
1 +
pj+p

k
E
p
.
For a soliton with real rapidity a
p
is a positive quantity and for a xed time E
p
= Ke
a
p
x
for
some constant K. Thus,
lim
x

j
= 0
lim
x

j

j
= 0
lim
x

j
= a
p
lim
x

j

j
= b
p
22
implying that
C

= 2
r

j=0
a
p
= 2(r + 1)a
p
= 2ha
p

C|

= 2
r

j=0
b
p
= 2hb
p
and so
E = 2ha
p
= 2h
p
cosh()
P = 2hb
p
= 2h
p
sinh()
M = 2h
p
in agreement with (3.23), (3.24) and (3.25).
Double soliton solution in a
(1)
r
The tau functions are given by (3.7) for the case of a two soliton solution arising from a species
p and species q single solitons. As long as the interaction parameter is non-zero (it is only zero
for p = q = n in a
(1)
2n1
which is a single soliton solution) and a
p
and a
q
are both positive, the
limit of
j
and its derivatives will be dominated by the E
p
E
q
term as x as it will grow the
fastest. Thus it is found that
lim
x

j
= 0
lim
x

j

j
= 0
lim
x

j
= a
p
+a
q
lim
x

j

j
= b
p
b
q
giving
C

= 2
r

j=0
(a
p
+a
q
) = 2h(a
p
+a
q
)

C|

= 2
r

j=0
(b
p
+b
q
) = 2h(b
p
+b
q
)
and so
E = 2h(a
p
+a
q
)
P = 2h(b
p
+b
q
)
M
2
= (2h)
2
_

2
p
+
2
q
+ 2
p

q
cosh(
p

q
)
_
.
Thus, it is seen that the square mass of the solution is relativistically invariant (depending as it
does only on the dierence of rapidities) and, for real rapidities, is at a minimum when
p
=
q
-
in that case the result is M = M
p
+M
q
.
23
Folded solitons
With the exception of p = q = n in c
(1)
n
the folded single solitons are all given by two soliton
solutions of a
(1)
r
with q = h p and
p
=
q
= . The result is then simply
E = 4ha
p
P = 4hb
p
M = 4h
p
.
I.e., the folded single solitons have twice the masses of the a
(1)
r
single solitons. This is to be ex-
pected because, from the a
(1)
r
point of view, two a
(1)
r
solitons are placed together with the same
mass and rapidity.
For two soliton solutions in the folded theories the relevant tau functions are given by (3.12)
which for large x will again be dominated by the highest order term in E, namely E
2
p
E
2
q
, so
lim
x

j
= 0
lim
x

j

j
= 0
lim
x

j
= 2a
p
+ 2a
q
lim
x

j

j
= 2b
p
2b
q
resulting in
E = 2h(2a
p
+ 2a
q
) = 4h(
p
cosh
p
+
q
cosh
q
) (3.33)
P = 4h(
p
sinh
p
+
q
sinh
q
) (3.34)
M
2
= 4 (2h)
2
(
2
p
+
2
q
+ 2
p

q
cosh(
p

q
)) . (3.35)
Of particular interest are the breather solutions and the fusing rules, both of which occur
at imaginary rapidity dierence. For breathers the requirement is p = q so
p
=
q
, while

p
= + i

and
q
= i

. So, if is real the dominant term in the tau function as x

is once again the E
2
p
E
2
q
term, giving
E = 8h
p
cosh cos
P = 8h
p
sinh cos
while the mass of the breather is
M = 4 cos (2h
p
) = 4 cos M
p
and thus it is seen that this is indeed a bound state of four a
(1)
r
solitons, all of the same mass
M
p
, since the mass M < 4M
p
when || > 0.
Fusion of two folded solitons will occur when
p

q
=
i(p+q)
h
so the choice will be made that

p
= +i
q
h
,
q
= i
p
h
24
so (3.33) and (3.34) give
E = 4h(
p
cosh
p
+
q
cosh
q
) = 2h
p+q
cosh
P = 4h(
p
sinh
p
+
q
cosh
q
) = 2h
p+q
sinh .
For calculating the mass it must be noted that
2
p
+
2
q
+ 2
p

q
cosh(i
(p+q)
h
) =
2
p+q
and so it
follows that
M = 4h
p+q
which is precisely the mass of the species p +q folded soliton- this is entirely expected and adds
weight to the assertion that fusion has occured.
In summary, the masses are given by
Type of soliton highest order of E m =
mass
2h
a
(1)
r
type p 1
p
most folded single solitons, type p 2 2
p
breather of two type p folded solitons 4 4
p
cos
These solitons are directly proportional to the masses of the fundamental Toda particles (dif-
fering by a factor of 2h) and possess the same fusing rules.
In subsequent sections the rescaled Lagrangian (3.22) will not be used but rather there is a
reversion to use of (1.1), though with = 1. This shall not cause any issues as only energy and
momentum dierences will be looked at.
4 Fusing defects
This section deals with fusing defects, i.e., combining two defects (both of type I in the language
of [3]) to form a single (type II) defect. There are two kinds of (type I) integrable defect known
in a
(1)
r
[2] which are referred to in the introduction as B and B
T
defects. One kind of defect may
be transformed into the other kind by considering the roots of the algebra {
i
} in a reversed
order- this is reminiscent of the folding procedure which identies u
i
with u
k+hi
.
4.1 Two defect system and time delays
The starting point here will be consideration of a
(1)
r
with two unfused defects: a B defect at
x = a < 0 and a B
T
defect at x = 0.
u v

x = a x = 0
Figure 4: Two unfused defects in a
(1)
r
25
The Lagrange density describing this system may be written as
L = (a x)L
u
+(a x)
_
1
2
uA u +uB +
1
2
A D
(1)
(u, )
_
+(x a)(x)L

+(x)
_

1
2
A +B
T
v
1
2
vA v D
(2)
(, v)
_
+(x)L
v
(4.1)
where A = 1 B and B + B
T
= 2 with L
u
being a Lagrange density in the form of (1.1); L
v
and L

similarly. In section 2 it was argued that B may be written

B = 2
r

j=0
(
j

j+1
)
T
j
(4.2)
where
q
= 0 for some q (depending on how the fundamental weights are dened) and B
T
is
simply the transpose of B.
The defect potentials are given by
D
(1)
= d
r

j=0
e
1
2

j(
B
T
u+B
)
+
1
d
r

j=0
e
1
2

j
B(u)
(4.3)
D
(2)
=

d
r

j=0
e
1
2

j(
B
T
v+B
)
+
1

d
r

j=0
e
1
2

j
B(v)
(4.4)
where d and

d are free parameters, referred to here as the defect parameters- they may be
interpreted as relating to the rapidities of solitons when
p
and

p
have poles/zeroes.
In addition to the bulk ane Toda equations of motion, the Euler-Lagrange equations of (4.1)
give the defect conditions
u

= A u +B D
u
|
x=a
(4.5)

= A +B
T
u +D
(1)

|
x=a
(4.6)

= A +B
T
v D
(2)

|
x=0
(4.7)
v

= A v +B +D
v
|
x=0
. (4.8)
In (4.5), (4.8) and subsequently, where D is not given a superscript it will be taken to mean
D = D
(1)
+D
(2)
.
Consider evolving a right-moving species p soliton through this system. An Ansatz is made
as in [2] that the soliton should retain its form, only picking up a time delay and phase shift
from the defect, so consideration of the defect on the left means that
u =
r

j=0

j
ln
u
j
=
r

j=0

j
ln

j
26
where the tau functions have the form of (3.6)

u
j
(a) = 1 +
pj+p

k
E
p
(a)

j
(a) = 1 +
pj+p

p
E
p
(a) .
By considering (4.5) and (4.6) (take the inner product of each with
i
) it is found, as in [2]
that
p
is given by (1.9). The time delay to the soliton from this defect is then given by
t =
1
b
p
ln |
p
| while the phase of
p
may be absorbed into I(c
p
) (possibly aecting the topo-
logical charge). Note the possibility of a pole in (1.9) hints at the possibility of solitons being
absorbed and emitted- this change in soliton number is not surprising as the x = a defect
conditions give a Backlund transformation for a
(1)
r
(the x = 0 defect conditions give another
Backlund transformation).
The other defect at x = 0 produces a similar delay

j
(0) = 1 +
pj+p

p
E
p
(0)

v
j
(0) = 1 +
pj+p

p
E
p
(0)
where

p
is given by (1.10). The eld on the right, v then diers from what it would in the no
defect case by the factor
p

p
.
The general two soliton solution of a
(1)
r
has tau functions given by (3.7). The consituent solitons
(species p and species q) are delayed by the defects independently so the eect of the defects is

u
j
= 1 +
pj
E
p
+
qj
E
q
+A
(pq)

(p+q)j
E
p
E
q

j
= 1 +
pj

p
E
p
+
qj

q
E
q
+A
(pq)

(p+q)j

q
E
p
E
q

v
j
= 1 +
pj

p
E
p
+
qj

q
E
q
+A
(pq)

(p+q)j

q
E
p
E
q
. (4.9)
This has been veried the lowest non-trivial order (i.e., consideration of (4.5)-(4.8) at order
E
p
E
q
). Its veracity may also be argued in terms of the commutability of Backlund transforma-
tions: evolving the species p and species q solitons through a defect and then combining into a
two soliton solution ought to give the same as combining into a two soliton solution and then
evolving through the defect.
It turns out that (4.9) retains its relevance when the defects are fused and folded, this is
argued in the following sections.
4.2 A fused a
(1)
r
defect and energy-momentum conservation
Fusing is done at the Lagrangian level by taking a 0 in (4.1), leaving
L = (x)L
u
+(x)L
v
+(x)
_
1
2
uA u +uB vB
1
2
vA v D
(1)
D
(2)
_
. (4.10)
27
x = 0

u v
Figure 5: The fused a
(1)
r
defect
There no longer exists any bulk for the eld ; it is eectively trapped in the defect and hence
may be referred to as an auxiliary eld. Note also that (4.10) bears some similarity to (1.11)
though there is no direct coupling of the bulk elds u and v- this is to be expected here as the
fused defect arises from two separated defects and locality dictates that u could only couple to
v via the eld . In fact, [22] and [23] use a modied type II Ansatz which eliminates the direct
coupling of the bulk elds found in [3].
The Euler Lagrange equations of (4.10) give now the bulk equations for u and v as well as
the defect conditions at x = 0
u

= A u +B D
u
(4.11)
v

= A v +B +D
v
(4.12)
B
T
u +D
(1)

= B
T
v D
(2)

. (4.13)
There are now three vector equations instead of the four of (4.5)-(4.8) and it is seen that (4.13)
also arises from identication (4.6) and (4.7) when a = 0. Thus it follows that any solution to
(4.5)-(4.8) with a = 0 must also solve (4.11), (4.12) and (4.13) and so the time delays previously
found and (4.9) remain valid for the fused defect.
Conservation of a modied energy and momentum for the fused defect is now shown- this
strongly indicates classical integrability of the fused defect, but is mainly done in order to set
the scene for energy-momentum conservation of the folded defects.
4.2.1 Energy conservation
Consider the bulk contribution to the energy, i.e., the integral of T
00
of the stress tensor which
derives from (4.10).
E =
_
0

1
2
u u +
1
2
u

+U dx +
_

0
1
2
v v +
1
2
v

+V dx
where U and V are the bulk potentials for the elds u and v respectively (the eld , before
fusing, would have a potential denoted X). Taking the time derivative of this and using the
bulk equations of motion; u u

= U
u
, v v

= V
v
; gives

E = u u

v v

|
x=0
28
provided that the elds are constant at spatial innity. This is now amenable to the use of the
defect conditions (4.11), (4.12) and (4.13)

E = u u

v v

|
x=0
= u(A u +B D
u
) v (A v +B +D
v
)
=
_
B
T
u B
T
v
_
u D
u
v D
v
= u D
u
v D
v
D

=

D
So, the quantity E + D, the bulk plus the defect energy, is conserved. In some sense this is
trivial, as applying a Legendre transformation to the Lagrange density (4.10) and performing
spatial integration gives the quantity E +D as the Hamiltonian of the system.
It would appear that energy conservation makes no constraint on D, but there is one sub-
tle implication in interpreting D as an energy contribution. The energy can be viewed as the
sum of a positive and a negative helicity term, i.e., the sum of the lightcone momenta P
+
and
P

(with a possible factor depending on notation)- hence, the defect potential D (indeed D
(1)
and D
(2)
separately- from energy conservation of type I defects) may also be written as the
sum of a positive and a negative helicity term. The form of D
(1)
and D
(2)
are already known
and given by (4.3) and (4.4) and each is already written in two parts- the interpretation here is
that the terms with prefactor d and

d have positive helicity while the d
1
and

d
1
terms have
negative helicity- the opposite labelling is equally possible. This interpretation of the helicities
is a useful guide for the momentum conservation argument. Momentum is given by the dier-
ence of the lightcone momenta P
+
P

, so the defect contribution to momentum is expected

to consist of the dierence of a positive and a negative helicity term.
4.2.2 Momentum conservation
Here it is shown that the fused defect conserves momentum. Navely one might not expect
any sort of momentum conservation from a defect ATFT given that placing the defect explicitly
breaks spatial translation invariance. Nonetheless, [11] shows that not only do a
(1)
r
defects admit
a modied conserved momentum, the conservation of energy and momentum alone is enough
to specify the Lagrangian with the result being that the defect is integrable, given the Lax pair
analysis of [2]. It is not guaranteed that momentum conservation in other types of defect implies
integrability, but given the similarities of the dierent ATFTs, there is reason to believe that
this is true.
The bulk contribution to the momentum is given by the integral of T
01
, so
P =
_
0

u u

dx +
_

0
v v

dx .
Taking the time derivative and using the bulk equations of motion gives

P =
1
2
u u +
1
2
u

U
1
2
v v
1
2
v

+V |
x=0
(4.14)
so long as the elds and potentials are constant (in vacuum) at spatial innity. The aim again
is to show that there is a conserved quantity by showing that

P is a total time derivative of a
29
quantity that exists at x = 0, so the rst step is to remove the spatial derivatives via (4.11) and
(4.12) to obtain
u

=
_
uA
T
+ B
T
D
u
_
(A u +B D
u
)
= uA
T
A u + B
T
B +D
2
u
+ 2 B
T
A u 2 uA
T
D
u
2 B
T
D
u
(4.15)
and similarly
v

=
_
vA
T
+ B
T
+D
v
_
(A v +B +D
v
)
= vA
T
A v + B
T
B +D
2
v
+ 2 B
T
A v + 2 vA
T
D
v
+ 2 B
T
D
v
. (4.16)
Noting that A
T
A = B
T
B 1 = BB
T
1, B
T
A = AB
T
, (4.15) and (4.16) give
u

= u
_
BB
T
1
_
u v
_
BB
T
1
_
v + 2 A
_
B
T
u B
T
v
_
+D
2
u
D
2
v
2 uA
T
D
u
2 vA
T
D
v
2
_
B
T
D
u
+B
T
D
v
_
= u
_
BB
T
1
_
u v
_
BB
T
1
_
v +D
2
u
D
2
v
2 uA
T
D
u
2 vA
T
D
v
2
_
B
T
D
u
+AD
(1)

+B
T
D
v
+AD
(2)

_
(4.17)
where (4.13) has been used to substitute for B
T
u B
T
v. It is clear that (4.17) contains the
terms u u + v v, cancelling terms in the momentum expression, but there are still terms
which are quadratic in time derivatives to deal with, i.e., uBB
T
u vBB
T
v; these terms can be
dealt with by squaring and equating both sides of (4.13). As a general principle, it is sensible
to pair up D
(1)
with the eld u and D
(2)
with the eld v which is why (4.13) is presented in its
given form. Squaring both sides of (4.13) gives
uBB
T
u vBB
T
v = 2 uBD
(1)

2 vBD
(2)

D
(1)

D
(1)

+D
(2)

D
(2)

(4.18)
and consequently, (4.14) reduces to

P =
1
2
_
u u v v +u

_
U +V
= U +V +
1
2
_
D
2
u
D
(1) 2

+D
(2) 2

D
2
v
_
u
_
A
T
D
u
+BD
(1)

_
v
_
A
T
D
v
+BD
(2)

_

_
B
T
D
u
+AD
(1)

+B
T
D
v
+AD
(2)

_
. (4.19)
For (4.19) to be a total time derivative strong constraints must be placed on the form of D.
Fortunately, the form of D here is already known from [2] and the fusing process so it becomes
a matter of testing that D has the right properties.
With the identication of helicities of terms in D made in section 4.2.1, as well as D being
30
the sum of D
(1)
and D
(2)
, the defect potential, (4.3) plus (4.4), splits into four parts given by
D
(1)+
= d
r

j=0
e
1
2

j(
B
T
u+B
)

j=0
a
j
D
(2)+
=

d
r

j=0
e
1
2

j(
B+B
T
v
)

j=0
a
j
D
(1)
=
1
d
r

j=0
e
1
2

j
B(u)

j=0
b
j
D
(2)
=
1

d
r

j=0
e
1
2

j
B
T
(v)
=
1

d
r

j=0
e
1
2

j
B(v)

j=0

b
j
. (4.20)
Note now that D
(2)
has precisely the same form as D
(1)
only with v instead of u; consequently
any property of D
(2)
may be drawn on by analogy to the corresponding property of D
(1)
. Note
now that
D
(1)+
u
=
r

j=0
1
2
B
j
a
j
, D
(1)+

=
r

j=0
1
2
B
T

j
a
j
D
(1)
u
=
r

j=0
1
2
B
T

j
b
j
, D
(1)

=
r

j=0
1
2
B
T

j
b
j
with similar expressions involving D
(2)
. This gives the relations
B
T
D
+
u
= BD
(1)+

u
= D
(1)

B
T
D
+
v
= BD
(2)+

v
= D
(2)

(4.21)
and so, using (4.21), (4.19) reduces to

P = U +V +
1
2
_
D
2
u
D
(1) 2

+D
(2) 2

D
2
v
_
u
_
D
+
u
D

u
_
v
_
D
+
v
D

v
_

_
D
+

_
.
so, in order for momentum conservation to work and give a result that is rather expected, D
must be quadratically related to U and V such that
1
2
_
D
2
u
D
(1) 2

+D
(2) 2

D
2
v
_
= U V .
Noting that

i
B
T

j
=
_

_
2 if i = j,
2 if i = j 1
0 otherwise

i
B
j
=
_

_
2 if i = j,
2 if i = j + 1
0 otherwise
and using the relations (4.21) and the form of D in (4.20) it is seen that
1
2
_
D
2
u
D
(1) 2

_
= D
+
u
D

u
D
(1)+

D
(1)

=
1
2
r

i,j=0

i
B
T

j
a
i
b
j
=

i
a
i
b
i

i
a
i
b
i+1
= U X
31
where X is the would-be bulk potential for . A similar relation holds for D
(2)
, implying that
D
(2) 2

D
2
v
= XV and so it is true that
1
2
_
D
2
u
D
(1) 2

+D
(2) 2

D
2
v
_
= U V which nally
gives

P =
_

D
+

_
meaning that there is a conserved momentum given by P + D
+
D

. It is conjectured that
for the folded case that the same quantity, with the bulk elds u and v replaced in D by their
folded counterparts, is conserved, i.e., the aim is to take

P and show that it in fact is equal to

_

D
+

_
. This turns out to be true, but the calculation is a lot more involved than for
the unfolded case.
One thing to note is that an analogous analysis holds if in the rst instance in (4.1) the B
defect is taken to be to the right of the B
T
defect. This perhaps should come as no sur-
prise as one may appeal to commutability given that the defect conditions describe Backlund
transformations.
5 The folded defect
This section discusses of the folding of the fused defect system and by consideration of soliton
time delays then energy and momentum conservation it strongly suggests the existence of clas-
sically integrable c
(1)
n
, d
(2)
n
and a
(2)
2n
defects. The process and the outcome are perhaps the two
most salient aspects of this paper.
5.1 Folding the Lagrangian
There is from the outset an ambiguity over what is meant by the folding of the fused defect
system of section 4.2; this ambiguity is investigated in appendix C. Here the folding being con-
sidered is a priori folding: folding the Lagrangian (4.10).
The aim is to have the folded system describe a defect in the folded theory. In the bulk
that means that the elds u and v must be folded, but at the defect it is not clear what should
be done with the auxiliary eld . Consideration of the time delays of soliton solutions, section
5.3, suggests that the eld should not be folded. One might then expect the folded Lagrangian
to be the same as (4.10) but with u and v . This is true but there is at least one thing
that diers from nave expectations which occurs with the self-couplings at the defect: uA u and
vA v.
Note that all of the folded roots,

i
, resulting from bivalent foldings have the form of

l
+
k+hl
2
,
with i =
k
2
+l for c
(1)
n
and a
(2)
2n
and i =
k+1
2
+l for d
(2)
n
. Now observe that
(
i
+
h+ki
)B(
j
+
h+kj
) =
i
B
j
+
h+ki
B
h+kj
+
i
B
h+kj
+
h+ki
B
j
=
1
2
(
i
B
j
+
h+ki
B
h+kj
+
i
B
h+kj
+
h+ki
B
j
)
+
1
2
_

h+ki
B
T

h+kj
+
i
B
T

j
+
h+ki
B
T

j
+
i
B
T

h+kj
_
= (
i
+
h+ki
) (
j
+
h+kj
)
where use has been made of the fact that
i
B
j
=
h+ki
B
T

h+kj
(= 2
i,j
2
i,j+1
) and also
that B +B
T
= 2.
32
Since A = 1 B it must then be the case that

i
A

j
=

i
B

j
= 0
and thus
A

= A

= 0 . (5.1)
Thus, the folded defects contain no self-coupling kinetic terms at x = 0. This means that the
Lagrangian (4.10) folds to
L = (x)L

+(x)L

+(x)
_
B B D
(1)
(, ) D
(2)
(, )
_
(5.2)
which ts the modied type II framework of [22].
In vector form the Euler-Lagrange equations are thus

=
p
B D

(5.3)

=
p
B +D

(5.4)
B
T

+D
(1)

= B
T

D
(2)

(5.5)
The subscript p found in (5.3) and (5.4) in front of B denotes projected. What is meant by
this is explained in appendix C. Also tackled in appendix C is the relation between the folded
and unfolded gradients, i.e., comparison of D
u
to D

, the results of this analysis are summarised

in the next section.
5.2 Content of the defect equations
In appendix C the equations (5.3), (5.4) and (5.5) are examined in greater depth. Here a sum-
mary is given of the most salient features, as the full detail here would obscure the clarity of
the following sections.
Universal expressions are found for D

and D

which are
D
+

=
h1

j=0
1
4
_
B
j
+B
T

k+hj
_
a
j
(5.6)
D

=
h1

j=0
1
4
_
B
T

j
+B
k+hj
_
b
j
(5.7)
D
+

=
h1

j=0
1
4
_
B
j
+B
T

k+hj
_
a
j
(5.8)
D

=
h1

j=0
1
4
_
B
T

j
+B
k+hj
_

b
j
(5.9)
33
while the gradient of D with respect to is given by
D
+

j
1
2
B
T

j
(a
j
+ a
j
) (5.10)
D

j
1
2
B
T

j
_
b
j
+

b
j
_
. (5.11)
Examination of the components of (5.5) gives the algebraic constraints
D

i
+D

k+h1i
= 0
which may be put in the form
a
i
a
k+hi
+ a
i
a
k+hi
b
i
+b
k+hi

b
i
+

b
k+hi
= 0 . (5.12)
5.3 Time delay of solitons through the defect
The picture here is similar to that of section 4.2 as regards how fusing defects aects the time
delays of the solitons. There it is argued that since any solution to (4.5)-(4.8) with a = 0 also
solves the fused equations (4.11), (4.12) and (4.13), the tau functions pick up factors given
by (4.9). From the argument of a posteriori folding given in appendix C it is also true that
any solution to (4.11), (4.12) and (4.13) which has the symmetry of a folded eld, namely
u
i
= u
k+hi
, v
i
= v
k+hi
, must then satisfy the a posteriori folded equations (C.1), (C.2) and
(C.3). Consequently, (4.9) should be re-examined in the case of a folded soliton on the left, i.e.,
q = h p, E
q
=
pk
E
p
. This gives

j
=
u
j
= 1 +
_

pj
+
pkpj
_
E
p
+A
kj
E
2
p
(5.13)

j
= 1 +
_

pj

p
+
pkpj

hp
_
E
p
+A
kj

hp
E
2
p
(5.14)

j
?
=
v
j
= 1 +
_

pj

p
+
qj

hp

hp
_
E
p
+A
kj

hp

hp
E
2
p
(5.15)
where A = cos
2
_
p
r+1
_
and the delay factors are given by

p
=
ie

+d
p
2
ie

+d

p
2
,
hp
=
ie

p
2
ie

d
p
2

p
=
ie

p
2
ie

d
p
2
,

hp
=
ie

+

d
p
2
ie

+

d

p
2
. (5.16)
It is clear that (5.13) represents a folded soliton, this is the choice that is made. Equally clear
is that

j
=

k+hj
, since
p
=
hp
except for p = n in c
(1)
n
- consequently it is not possible
to fold the auxiliary eld , despite the algebraic constraints (5.12) implying that may be
thought of as having the same number of degrees of freedom as a folded eld has.
What is desired is that the eld on the right, v represents a folded soliton, i.e.,
v
j
=
v
k+hj
, so
that v may be replaced by and the a posteriori defect equations satised. For this to be true
it must be the case that

p
=
hp

hp
34
in which case every may be absorbed into the denition of E
p
. Note that this condition is the
same condition as having the a
(1)
r
single soliton species p and species h p solutions receiving
the same delay through the fused defect. Thus, the condition for the soliton on the right of the
defect to be in the folded theory is
0 =
p

hp

hp
=
1
denom.
_
e
2
_

p
_
_
d
2

d
2
__
where denom. is the common denominator obtained by multiplying all of the denominators of
(5.16) together. So the defect represented by the Lagrangian (5.2) is only likely to be integrable
if

d = d (5.17)
as this is what is required for the soliton solution (5.15) to be compatible with folding. Therefore,
there are two possibilities then that give folded solitons for
When

d = d it is the case that
p

p
=
hp

hp
= 1. i.e., all of the solitons receive
a trivial time delay- the defect does nothing. Indeed in this case if = is imposed
then the defect part of the Lagrangian (5.2) vanishes- so there is no defect there. The
interpretation of this is that the second defect is the anti-defect of the rst- fusing them
causes annihilation.
When

d = d there is a non-trivial time delay (dierent for each p) so this should represent
a bona de defect which does not destroy the form of the solitons (a strong constraint,
suggesting that the defect is integrable). Note that if and d are real then the time-delay
is real too- there is no change of topological charge.
5.3.1 Algebraic constraints with such d and

d
The algebraic constraints from (5.12) have the undesirable property of mixing helicities for
general d and

d. As such, only if the positive and negative helicity terms in (5.12) separately
vanish will there be no mixing of helicities.
Before folding the terms in D are
a
i
= de
u
i
u
i+1
+
i

i1
a
i
=

de
v
i
v
i+1
+
i

i1
b
i
= d
1
e
u
i
u
i1

i
+
i1

b
i
=

d
1
e
v
i
v
i1

i
+
i1
.
If

d = d then
a
i
a
k+hi
+ a
i
a
k+hi
= d
_
e
u
i
u
i+1
+
i

i1
e
u
k+hi
u
k+h+1i
+
k+hi

k+h1i
e
v
i
v
i+1
+
i

i1
e
v
k+hi
v
k+h+1i
+
k+hi

k+h1i
_
b
i
b
k+hi
+

b
i

b
k+hi
= d
1
_
e
u
i
u
i1

i
+
i1
e
u
k+hi
u
k+h1i

k+hi
+
k+h1i
e
v
i
v
i1

i
+
i1
e
v
k+hi
v
k+h1i

k+hi
+
k+h1i
_
where the and are strictly correlated- the top set of relations refer to

d = d, while the
bottom to

d = d.
35
Folding the defect eld theory, in all cases considered involves identifying u
i
= u
k+hi
and
v
i
= v
k+hi
. Rather than identify the components of and , due to the subtlety in folding to
c
(1)
n
(normally u
i
=

i
2
, but in a
(1)
2n1
c
(1)
n
the identication is uk
2
+n
=

k
2
+n
), the folded elds
will be left in terms of {u
i
} and {v
i
}, thus
a
i
a
k+hi
+ a
i
a
k+hi
= d
_
e
u
i
u
i+1
+
i

i1
e
u
i
u
i1
+
k+hi

k+h1i
e
v
i
v
i+1
+
i

i1
e
v
i
v
i1
+
k+hi

k+h1i
_
whilst
b
i
b
k+hi
+

b
i

b
k+hi
= d
1
_
e
u
i
u
i1

i
+
i1
e
u
i
u
i+1

k+hi
+
k+h1i
e
v
i
v
i1

i
+
i1
e
v
i
v
i+1

k+hi
+
k+h1i
_
= d
1
e

i
+
i1

k+hi
+
k+h1i
_
e
u
i
u
i1
+
k+hi

k+h1i
e
u
i
u
i+1
+
i

i1
e
v
i
v
i1
+
k+hi

k+h1i
e
v
i
v
i+1
+
i

i1
_
= d
2
e

i
+
i1

k+hi
+
k+h1i
(a
i
a
k+hi
+ a
i
a
k+hi
) .
Thus, for either identication,

d = d
b
i
b
k+hi
+

b
i

b
k+hi
= d
2
e

i
+
i1

k+hi
+
k+h1i
(a
i
a
k+hi
+ a
i
a
k+hi
)
and so the positive and negative helicity algebraic constraints are equivalent
a
i
a
k+hi
+ a
i
a
k+hi
= 0 b
i
b
k+hi
+

b
i

b
k+hi
= 0 .
In light of (5.12) then, the algebraic constraints become
D
+

i
+D
+

k+h1i
= 0
D

i
+D

k+h1i
= 0
resulting in
a
i
a
k+hi
+ a
i
a
k+hi
= 0 (5.18)
b
i
b
k+hi
+

b
i

b
k+hi
= 0 . (5.19)
That the condition for solitons to be able to pass through the defect in the folded model without
losing their form, i.e., (5.17), is precisely what is needed to prevent the algebraic constraints from
mixing helicities is one of the most interesting outcomes of this analysis. This is particularly
salient as the momentum conservation argument in section 5.5 relies upon (5.18) and (5.19) being
separately true- this is the link between the two methods that is alluded to in the introduction.
From the consideration of the solitons, it is thus assumed for the purposes on showing energy
and momentum conservation that (5.17) is true,

d = d.
36
5.4 Energy conservation
For both energy and momentum conservation, the a priori folded defect equations; (5.3), (5.4)
and (5.5); will be used. The process here is analogous to that of the energy conservation of the
fused defect in section 4.2.1. The bulk energy is then modied by the defect such that

E =

|
x=0
=

p
B

D

p
B

D

=
_
B
T

B
T

=

D
so once again the quantity E+D is conserved. Note that it is safe to drop the p subscript above
as

and

both lie in the folded root space.
5.5 Momentum conservation
The momentum conservation by itself is the most involved of the calculations presented in this
paper, but much simplied by employment of the relations (5.6)-(5.11) of section 5.2 along with
the helicity conserving algebraic constraints (5.18) and (5.19). The bulk momentum is modied
such that

P =
1
2
_

_
+ (5.20)
where and are the folded bulk potentials.
The rst step in showing this momentum is conserved is to use the squares of (5.3) and (5.4),
giving

=
_
p
B
T
D

_
(
p
B D

)
=
p
B
T
p
B 2 B
T
D

+D
2

and

=
_
p
B
T
+D

_
(
p
B +D

)
=
p
B
T
p
B + 2 B
T
D

+D
2

.
In both cases it is not known what the rst term,
p
B
T
p
B , is in general; nor indeed is it even
clear what it means. However, this issue drops out of the momentum conservation argument
since

= 2
_
B
T
D

+B
T
D

_
+D
2

D
2

. (5.21)
At this stage progress can be made by anticipating the nal answer to be

P =
_

D
+

_
which requires a term
_
D
+

_
. It is evident from the defect conditions; (5.3), (5.4)
and (5.5); that the only place where terms dependent on can appear in (5.20) stems from

. The conclusion is then that

B
T
D
+

+B
T
D
+

= D
+

(5.22)
B
T
D

+B
T
D

= D

. (5.23)
37
If (5.22) and (5.23) are true then (5.5) may be rewritten, noting that B
T
is invertible, as

+D
+

=

D
+

+D

(5.24)
Note that the left-hand side of (5.24) is not equal
_
B
T
_
1
times the left-hand side of (5.5)-
some rearrangement is necessary. Squaring both sides of (5.24) gives

= 2

_
D
+

_
2

_
D
+

_
D
+

_
2
+
_
D
+

_
2
(5.25)
and hence (5.21) and (5.25) may be combined to give

= 2
_

D
+

_
+ 4D
+

4D
+

to wit (5.20) reduces to

P =

D
+
+

D

+ 2D
+

2D
+

+ . (5.26)
Therefore, a modied momentum of P +D
+
D

is conserved, provided
2D
+

2D
+

= (5.27)
This argument would prove momentum conservation but there are three equations which must
be proven which rely upon the particular form of D, i.e., (5.22), (5.23) and (5.27).
5.5.1 Proof of the relations (5.22) and (5.23)
In order to prove (5.22) and (5.23), all of the relations (5.6)-(5.11) will be required, along with
the algebraic constraints (5.18) and (5.19). The rst step is to note (5.10), given by
D
+

i
1
2
B
T

j
(a
j
+ a
j
)
and to put the left-hand side of (5.22) into this form. Thus,
B
T
D
+

+B
T
D
+

j
1
4
B
T
_
Ba
j
+B a
j
+B
T
a
k+hj
+B
T
a
k+hj
_

j
=

i
1
4
B
T
_
(B +B
T
)a
j
+ (B +B
T
) a
j
+B
T
(a
k+hj
a
j
) +B
T
( a
k+hj
a
j
)
_

j
=

j
1
2
B
T

i
(a
j
+ a
j
) +

j
1
4
B
T
B
T
(a
k+hj
a
j
+ a
k+hj
a
j
)
i
=

j
1
2
B
T

j
(a
j
+ a
j
) = D
+

where B +B
T
= 2 has been used along with (5.18).
The situation for (5.23) is analogous:
D

j
1
2
B
T

j
_
b
j
+

b
j
_
38
while
B
T
D

+B
T
D

j
1
4
B
T
_
B
T
b
j
+B
T

b
j
+Bb
k+hj
+B

b
k+hj
_

j
=

j
1
4
B
T
_
(B +B
T
)b
j
+ (B +B
T
)

b
j
+B(b
k+hj
b
j
) +B(

b
k+hj

b
j
)
_

j
=

j
1
2
B
T

j
_
b
j
+

b
j
_
+

j
1
4
B
T
B
_
b
k+hj
b
j
+

b
k+hj

b
j
_

j
=

j
1
2
B
T

j
_
b
j
+

b
j
_
= D

.
5.5.2 Proof of (5.27)
Here the nal relation for momentum conservation is proven
2D
+

2D
+

= .
Before folding, the bulk potentials are given by
U(u) =
r

j=0
a
j
b
j
X() =
r

j=0
a
j
b
j+1
=
r

j=0
a
j

b
j+1
V =
r

j=0
a
j

b
j
and so
U X =
1
2
r

i,j=0

i
B
T

j
a
i
b
j
V X =
1
2
r

i,j=0

i
B
T

j
a
i

b
j
resulting in
U V =
1
2
r

i,j=0

i
B
T

j
_
a
i
b
j
a
i

b
j
_
. (5.28)
The equation (5.28) can now be folded to give the right-hand side of (5.27) as
=
1
2
r

i,j=0

i
B
T

j
_
a
i
b
j
a
i

b
j
_
(5.29)
where a
i
, a
i
, b
i
and

b
i
are now written in terms of the folded elds.
39
The left-hand side of (5.27) can be rewritten using (5.6)-(5.9) to give
2D
+

2D
+

i,j
1
8
_
a
i
b
j
a
i

b
j
_
_

i
B
T
B
T

j
+
k+hi
BB
k+hj
+
i
B
T
B
k+hj
+
k+hi
BB
T

j
_
=

i,j
1
8
_
a
i
b
j
a
i

b
j
_
_

i
B
T
(B
T
+B)
j
+
k+hi
B(B +B
T
)
k+hj

i
B
T
B
j

k+hi
BB
T

k+hj
+
i
B
T
B
k+hj
+
k+hi
BB
T

j
_
=

i,j
1
4
_
a
i
b
j
a
i

b
j
_
_

i
B
T

j
+
k+hi
B
k+hj
_
+

i,j
1
8
_
a
i
b
j
a
i

b
j
_
(
i

k+hi
) BB
T
(
k+hj

j
)
=
1
2

i,j
_
a
i
b
j
a
i

b
j
_
_

i
B
T

j
_
+

i,j
M
ij
_
a
i
b
j
a
i

b
j
_
= +

i,j
M
ij
_
a
i
b
j
a
i

b
j
_
where use has been made of the facts that
k+hi
B
k+hj
=
i
B
T

j
and that B+B
T
= 2. In
the last step (5.29) has been used. The quantity M
ij
is dened as
M
ij
=
1
8
(
i

k+hi
) BB
T
(
k+hj

j
)
so what now remains to be shown is that

i,j
M
ij
_
a
i
b
j
a
i

b
j
_
= 0 . (5.30)
First note that
M
(k+hi)j
= M
i(k+hj)
= M
ij
.
Using (5.18) then implies that

i
M
ij
a
i
=

i
M
ij
a
k+hi

i
M
ij
a
i
+

i
M
ij
a
k+hi
=

i
M
(k+hi)j
a
k+hi

i
M
ij
a
i

i
M
(k+hi)j
a
k+hi
=

i
M
ij
a
i
2

i
M
ij
a
i
=

i
M
ij
a
i
=

i
M
ij
a
i
while (5.19) analogously gives

j
M
ij
b
j
=

j
M
ij

b
j
.
40
Thus, (5.30) is true since

i,j
M
ij
a
i
b
j
=

i,j
M
ij
a
i
b
j
= +

i,j
M
ij
a
i

b
j
and hence
2D
+

2D
+

= .
In conclusion, (5.26) reduces to

P =
_

D
+

_
and so P +D
+
D

is a conserved quantity,
certainly when

d = d. Hence, energy and momentum are conserved by the folded defects. Mo-
mentum conservation in particular is a strong constraint, perhaps even implying integrability, so
there is more than a hope that (5.2) with

d = d represents integrable defects for c
(1)
n
, d
(2)
n
and a
(2)
2n
.
It is perhaps the soliton time delay argument that gives the stronger evidence of integrabil-
ity: preservation of the forms of the solitons by the defect hints at the conservation of an
innite number of charges.
6 Discussion
Conclusions
The main conclusion of this paper is that certain a
(1)
r
defects may be folded to give defects in
the folded models of c
(1)
n
, d
(2)
n
and a
(2)
2n
with Lagrangian description in the form (5.2). There are
two strong reasons to believe that such defects are integrable.
Firstly, energy and momentum conservation, when applied to a type I a
(1)
r
defect, are in
themselves enough to force the defect to be integrable [11]. It is certainly plausible that
the same holds for these folded defects.
Secondly, there is the rather simpler argument of classical scattering of solitons o the
defect. This argument alone is suggestive of integrability as the solitons retain their form
and so ought to conserve an innite number of charges.
Though not the main focus, this paper also furthers the work of [9] by using the relatively simple
arena of a
(1)
r
to construct solitons and breathers in c
(1)
n
, d
(2)
n
and a
(2)
2n
. The breather solutions in
particular might not have been published before.
There do remain a couple of loose ends at the classical level. One regards the interpreta-
tion of the auxiliary eld, , in the folded defect. Since the algebraic constraints may be used
to reduce the number of degrees of freedom of to that of a folded eld, it seems highly likely
that one may construct similar defects by starting in the folded model, i.e., without any refer-
ence to a
(1)
r
. Another is that it is as yet unclear whether or not the folded defect conditions
((5.3),(5.4),(5.5)) represent a Backlund transformation in the folded theory- such as was found
for a
(1)
r
type I defects in [2].
41
Future directions
One clear extension of this work is to consider the transmission matrices of the folded defects
in the quantum picture. The transmission matrices for type II defects in the a
(1)
r
theories are
found in [23], where the Lagrangian considered bears more than a supercial similarity to (5.2).
One would hope that in a similar way in which [8] presents the folded bulk S-matrices in terms
of the unfolded ones, the folded transmission matrices may be found.
Another direction is to try to nd defects for the other simply laced ATFTs (d
(1)
s
, e
(1)
6
, e
(1)
7
and e
(1)
8
) which then might be folded such that all ATFTs are covered. In fact, there are possi-
bly some implications already for d
(1)
s
defects in this paper due to the use of the non-canonical
foldings a
(1)
2n1
d
(2)
n
and a
(1)
2n
a
(2)
2n
. For both of these folded theories the canonical way to
fold is from a d series ATFT rather than an a series one, so there is the question of whether a
d
(1)
s
defect ATFT, should such a thing exist, might be folded to give something of the form of
(5.2).
One thing of note is that d
(2)
n
and a
(2)
2n
solitons are particular two soliton solutions of a
(1)
r
while
d
(2)
n
and a
(2)
2n
defects are constructed from particular two defect solutions of a
(1)
r
. It happens
to be the case that d
(2)
n
and a
(2)
2n
solitons are particular single soliton solutions of d
(1)
s
and so
perhaps (5.2) stems from a single defect of d
(1)
s
, it then being of type II in having one auxiliary
vector eld. It remains to be seen whether or not this speculation is true.
e
e
e e e e e
e
e

Z
Z

Z
Z
d
(1)
n+1
-
6
?
6
?
u e e e e e u
d
(2)
n
Figure 6: d
(1)
n+1
d
(2)
n
e
e
e e
e
e
e
e
e

Z
Z

Z
Z
Z
Z
Z
Z

6
6
?
?
6
?
-
d
(1)
2n+2
re u u e
a
(2)
2n
Figure 7: d
(1)
2n+2
a
(2)
2n
One can also look at the possibility of using defects to nd more general integrable boundary
conditions. The possibility of a boundary with an auxiliary eld was considered in [24]; while
the paper [25] considers a defect fused to a GhoshalZamolodchikov type boundary [26] in
a
1
. Classically integrable boundary conditions, which happen to be highly restrictive (no free
42
parameters), have been known for the folded theories c
(1)
n
, d
(2)
n
and a
(2)
2n
, for some time [27]. There
is the possibility of fusing the defects here to such boundaries to give more general boundary
conditions, though it is not immediately clear whether or not integrability would be preserved.
Acknowledgements
The author wishes to thank Peter Bowcock and Ed Corrigan for their guidance and thought-
provoking discussions. This work was supported by an STFC studentship.
A Tau function identications
This appendix gives the calculations for the tau function identications found in section 3.2.
A.1 c
(1)
n
solitons
a
(1)
2n1
(h = 2n) is folded to c
(1)
n
by setting k
2
+i
= k
2
+2ni
where k is even. Making this
identication on the tau functions of u gives
= k
2
ln k
2

n1

j=1
2
k
2
+j
+k
2
+2nj
2
ln k
2
+j
k
2
+n
ln k
2
+n
=

0
ln k
2

n1

j=1
2

j
ln k
2
+j

n
ln k
2
+n
=
n

j=0

j
ln

j
which is precisely the soliton Ansatz of c
(1)
n
[5], where

i
=
2

i
. The identication made is

i
= k
2
+i
= k
2
+2ni
.
The usual convention of
0
= 0 may be obtained by adding to the quantity 0 =

n
i=0
n

0
ln

0
,
resulting in

0
= 0

i
= ln
_

2
i

2
0
_
i = 1, . . . , n 1

n
= ln
_

0
_
.
43
A.2 d
(2)
n
solitons
In this instance k is odd and folding is achieved by setting k+1
2
+i
= k1
2
+2ni
. This identica-
tion on u gives
=
n1

j=0
2
k+1
2
+j
+k1
2
+2nj
2
ln k+1
2
+i
=
n1

j=0
2

i
ln k+1
2
+j
=
n1

j=0

j
ln

j
resulting in a d
(2)
n
soliton. The identication required is

j
= k+1
2
+j
= k1
2
+2nj
, but only
for j = 1, . . . , n 2. There is an issue that

0
= 4 so the identication must be that
2 ln k+1
2
= 4 ln

0
, with a similar situation for

n1
. The resulting tau function identica-
tions are thus

0
= (k+1
2
)
1
2
= (k1
2
+2n
)
1
2
and

n1
= (k+1
2
+n1
)
1
2
= (k+1
2
+n
)
1
2
.
Setting
0
= 0 as standard gives

0
= 0

i
= 2 ln
_

2
0
_
i = 1, . . . , n 2

n1
= 4 ln
_

n1

0
_
.
A.3 a
(2)
2n
solitons
This situation is similar but with a
(1)
2n
(h = 2n + 1) and as explained in section 2.1.3, k can be
restricted to being even. Folding is achieved by setting k
2
+i
= k
2
+2n+1i
. This identication
on u gives
= k
2
ln k
2

j=1
2
k
2
+j
+k
2
+2n+1j
2
ln k
2
+j
=

0
ln k
2

j=1
2

j
ln k
2
+j
=
n

j=0

j
ln

j
the soliton Ansatz for a
(2)
2n
. This time identify

i
= k
2
+i
= k
2
+2n+1i
, but only for i =
0, . . . , n 1. It is the case that 2 ln k
2
+n
= 4 ln

n
, thus

n
= (k
2
+n
)
1
2
= (k
2
+n+1
)
1
2
.
44
Setting
0
= 0 gives

0
= 0

i
= 2 ln
_

0
_
i = 1, . . . , n 1

n
= 2 ln
_

2
n

0
_
.
B a
1
embeddings in a
(1)
2n1
and folded theories
It is the a
1
ATFT which is the most applied in other contexts, due to the presence of well known
sine-Gordon solitons. Any progress in nding integrable ane Toda defects in the physical set-
ting is likely, in the rst instance, to apply to a
1
. For this reason theories with a
1
embeddings
are of particular interest.
As noted in section 3.3.1, the p = n soltion of a
(1)
2n1
satises the a
1
equations of motion
(indeed it is possible to fold a
(1)
2n1
directly to a
1
by identifying all even indexed roots

0
=
1
n
(
0
+
2
+. . . +
2n2
) with each other; and all odd indexed roots

1
=
1
n
(
1
+
3
+. . . +
2n1
)
with each other) [9]. There is also a less obvious embedding when n = 2m is even, i.e., in a
(1)
4m1
.
B.1 p = n soliton as an a
1
soliton
The roots in a
1
will be taken to have length

2 here. For the a

1
soliton (3.3) becomes
u =
1
ln
_

0
_
=
1
ln
_
1 +E
1 E
_
and so

u
U =
0
e

0
u
+
1
e

1
u
=
1
(e

1
u
e

0
u
) =
1
8E(1 +E
2
)
(1 E
2
)
2
whilst
u =
1
2bE
1 E
2
= u =
1
2b
2
E(1 +E
2
)
(1 E
2
)
2
= u u

=
1
8E(1 +E
2
)
(1 E
2
)
2
.
u u

=
u
U
using that
2
= 4, the only non-trivial eigenvalue for a
1
.
45
For a
(1)
2n1
(species n soliton,
2
n
= 4)
u =

j odd

j
ln
_

0
_
=

j odd

j
ln
_
1 +E
1 E
_

u
U =
2n1

j=0

j
e

j
u
=

j odd

j
_
1 +E
1 E
_
2
+

j even

j
_
1 E
1 +E
_
2
=

j odd

j
8E(1 +E)
2
(1 E
2
)
2
u u

j odd

j
8E(1 +E
2
)
(1 E
2
)
2
since

j even

j
=

j odd

j
. Each of the n odd indexed components of u satisfy the equation
of an a
1
soliton. Identifying roots merely gives a factor of n multiplying the a
1
soliton equation.
One may question whether the folded theories retain this a
1
embedding. Indeed it is ques-
tionable from the outset for the case of d
(2)
n
as there is no node to correspond to the p = n
soliton. Consider the two soliton solution in a
(1)
2n1
compatible with folding, i.e. equation (3.10)

j
= 1 +
_

pj
+
p(kj)
_
E
p
+A
pk
E
2
p
.
Now taking p = n it is the case that, by (3.9), A = cos
2
_

2
_
= 0, while
p
= 1. The result is

j
= 1 + (1)
j
_
1 + (1)
k
_
E
p
.
If k is even, such that the folding is a
(1)
2n1
c
(1)
n
, this gives

j
= 1 + 2(1)
j
E
which, upon rescaling E
E
2
, gives exactly tha a
1
tau functions.
If k is odd, however, such that the folding is a
(1)
2n1
d
(2)
n
the result is
j
= 1 for all j,
which is merely the trivial solution. The conclusion is that this a
1
embedding is absent from
d
(2)
n
but present still in c
(1)
n
.
B.2 A less obvious embedding of a
1
It would appear that there are no sine-Gordon solitons embedded in d
(2)
n
, however it is in fact
possible to nd an a
1
embedding when n is even.
In a
(1)
4m1
, setting p = m in equation (3.10) results in

j
= 1 +
_
i
j
+i
kj
_
E +
1
2
i
k
E
2
46
where this is a d
(2)
2m
soliton when k is odd. Since i
4
= 1 the only distinct possibilities are
represented by k = 1 and k = 3. If k = 1 it is found that

0
=
_
1 +
1
2
(1 +i)E
_
2
=
1
=
4
=
5
= . . .

2
=
_
1
1
2
(1 +i)E
_
2
=
3
=
6
=
7
= . . . .
Allowing
1
2
(1 + i)E E results in the squares of the tau functions of the a
1
soliton. Since
soliton solutions only depend logarithmically on the tau functions this just gives a factor of two
and identifying nodes as in [9] gives an a
1
soliton. If k = 3 the resultant tau functions are

0
=
_
1 +
1
2
(1 i)E
_
2
=
3
=
4
=
7
= . . .

1
=
_
1
1
2
(1 i)E
_
2
=
2
=
5
=
6
= . . . .
So
1
2
(1 i)E E results again in an a
1
soliton.
In conclusion: if n is odd then a
(1)
2n1
possesses one a
1
embedding which is inherited by c
(1)
n
but not by d
(2)
n
. If n is even a
(1)
2n1
possesses two a
1
embeddings- one of which is inherited by
c
(1)
n
; the other by d
(2)
n
.
C Equivalence of a priori and a posteriori folding and ways to
write D

As is mentioned in section 5.1, there is an ambiguity in what is meant by the folding of the
fused defect. There are two dierent points of view to consider.
1. The folding may be applied to the Lagrangian (4.10) as is done in section 5.1. This folding
is a priori in the sense that it is done before varying the action.
2. The folding may be applied a posteriori to the defect equations (4.11), (4.12) and (4.13),
after varying the unfolded action.
The a posteriori equations are given by

= A

+B D
u
(C.1)

= A

+B +D
v
(C.2)
B
T

+D
(1)

= B
T

D
(2)

(C.3)
giving
a priori folding a posteriori folding

=
p
B D

= A

+B D
u

=
p
B +D

= A

+B +D
v
B
T

+D
(1)

= B
T

D
(2)

B
T

+D
(1)

= B
T

D
(2)

Although the third equation on each side matches up, there are a number of critical dierences
between the two sets of equations:
47
The apparent self coupling found in the a posteriori equations and not in the a priori
ones.
The quantities D
u
and D
v
are still used in the a posteriori folded case even after folding
has occurred. What is meant by this is that the expressions are the same as D
u
and D
v
respectively, but with the identications made on the components of u and v such that
D
u
and D
v
are written with and , respectively, in the exponents. These quantities do
not match D

and D

, as the meaning of the gradient is changed by folding. In summary:

u

= D
u
D

.
Perhaps the most fundamental dierence is in the number of equations obtained when
considered in component form. Whilst the variation of the action required to obtain the
a posteriori equations is always done on the full a
(1)
r
root space, the rst two equations of
the a priori case are obtained by varying in the folded root space- or with respect to only
symmetric congurations of the a
(1)
r
root space. Thus, there are always fewer equations of
motion in the a priori case- so it should be the true that if the folding processes are not
completely equivalent, then at least the a posteriori case will contain the a priori case
and possibly some extra information.
The presence of the subscript p before the B in the rst two a priori equations is there
to emphasise that the B found in those equations only makes sense in components when
projected onto the folded root/weight space. It is valid to consider (
i
+
k+hi
) B since

i
+
k+hi
lies in the folded root space, but it is not valid to consider (
i

k+hi
) B
since
i

k+hi
does not lie in the folded root space. For the a posteriori case both
projections may be considered as they both lie on the unfolded a
(1)
r
root space.
By rst tackling the form of D

and D

it will become possible to compare each pair of equations

above in turn.
C.1 Gradients of the defect potential
In section 2 it is seen that despite all of the possible bivalent foldings of a
(1)
r
having major dier-
ences, all of the folded roots did take the universal form of

i
+
k+hi
2
. It would be helpful to be
able to write the gradients of the potential in a universal way, rather than having to consider
c
(1)
n
, d
(2)
n
and a
(2)
2n
separately.
First consider the folded eld written in a dierent manner to the standard one of section 2.
For c
(1)
n
and a
(2)
2n
this is
=
k
2
+n

j=
k
2
+1

j
+
k+hj
2

j
with

k
2
+i
=
i
.
Similarly, if the folded theory is d
(2)
n
the eld is
=
k+1
2
+n1

j=
k+1
2
+1

j
+
k+hj
2

j
48
with

k+1
2
+i
=
i
.
The reason for this relabelling is to make contact to the components u
i
and u
k+hi
in the
defect potential D. Folding is, in all but one instance, achieved by setting u
i
= u
k+hi
=

i
2
(the exception being that uk
2
+n
=

k
2
+n
in c
(1)
n
) and so the derivative of D with respect to the
component

i
is given by
D

i
=
1
2
_
D
u
i
+D
u
k+hi
_
(C.4)
which is true for all of the identications, including the exception mentioned above. Given this,
D

i
may be found explicitly using the relations (4.20). The u dependence only appears in D
(1)
,
and it will be necessary to split the positive and negative helicity parts, so the quantities of
interest are
D
(1)+
= d
r

j=0
e
1
2

j(
B
T
u+B
)
= d
r

j=0
e
u
j
u
j+1
+
j

j1
=
r

j=0
a
j
D
(1)+
=
1
d
r

j=0
e
1
2

j
B(u)
=
1
d
r

j=0
e
u
j
u
j1

j
+
j1
=
r

j=0
b
j
with the identications
a
j
= e
u
j
u
j+1
+
j

j1
and b
j
= e
u
j
u
j1

j
+
j1
.
Thus, written in this compact notation it is seen that
D
+
u
i
= a
i
a
i1
, D
+
u
k+hi
= a
k+hi
a
k+h1i
D

u
i
= b
i
b
i+1
, D

u
k+hi
= b
k+hi
b
k+h+1i
so, explicitly, (C.4) may be rewritten as
D
+

i
=
1
2
(a
i
a
i1
+a
k+hi
a
k+h1i
)
D

i
=
1
2
(b
i
b
i+1
+b
k+hi
b
k+h+1i
) (C.5)
which is true of all the bivalent foldings.
One must now consider how D

is related to D

i
and hope that it may be written in a similarly
universal form. Consider rst the unfolded case of D
u
. As mentioned in section 2, u may be
written as
u =
r

j=0
u
j

j
with one of the components, u
q
, set to zero. The reciprocal basis is then just the set of funda-
mental weights {
i
} that has
q
= 0. Thus, the gradient with respect to u may be written

u
=
r

j=0

u
j
49
and so
D
u
=
r

j=0

j
D
u
j
.
For the folded eld it is a similar picture. If is in c
(1)
n
or a
(2)
2n
the eld is
=
n

j=1

j
and so
D

=
n

j=1

j
D

j
(C.6)
where

j
=
ij
for i, j = 1, . . . , n (

0
= 0), meaning that {

i
} are the fundamental weights
of the folded theory. If is instead in d
(2)
n
, the same analysis holds but with an upper limit to
the sum of n 1 rather than n.
One can use the expression for

i
found in section 2 to write

i
in terms of the weights of
a
(1)
r
, with all of the folded theories; c
(1)
n
, d
(2)
n
and a
(2)
2n
; having dierences resulting in three dif-
ferent possible expressions for D

. These expressions are not written here, as it can be shown

that they are all encompassed by the universal form
D

=
1
2
h1

j=0
(
j
+
k+hj
) D

j
. (C.7)
The expression (C.7) counts all the non-zero terms twice apart from the j =
k
2
+n case of c
(1)
n
.
The results (C.5) can now be used, resulting in
D
+

=
1
4
h1

j=0
(
j
+
k+hj
) (a
i
a
i1
+a
k+hi
a
k+h1i
) (C.8)
D

=
1
4
h1

j=0
(
j
+
k+hj
) (b
i
b
i+1
+b
k+hi
b
k+h+1i
) . (C.9)
While (C.5) is useful in comparing the a priori and a posteriori defect equations, the form of
(C.7) is not easy to work with when considering the momentum conservation of the folded defect,
which requires, among other things, the quadratic quantity D
2

to be calculated. Working out

the inner products between the weights of a
(1)
r
, which would need to be done for D
2

, can be
done on a case by case basis, but it is not clear how to do this in general so an easier to use
form of D

is a necessity.
50
C.1.1 A more useful way to write D

For the unfolded elds, the gradients of D

(1)
are already given in section 4.2.2 in an alternative
form. The gradients of D can similarly be written
D
+
u
=

j
1
2
B
j
a
j
, D

u
=

j
1
2
B
T

j
b
j
D
+
v
=

j
1
2
B
j
a
j
, D

v
=

j
1
2
B
T

b
j
D
+

j
1
2
B
T

j
(a
j
+ a
j
) , D

j
1
2
B
T

j
_
b
j
+

b
j
_
. (C.10)
However, it is unclear from (4.20) what the analogous expressions for the gradients of D with
respect to the folded elds and are. Here the form of D

is proposed to be
D
+

=
h1

j=0
1
4
_
B
j
+B
T

k+hj
_
a
j
(C.11)
D

=
h1

j=0
1
4
_
B
T

j
+B
k+hj
_
b
j
. (C.12)
The equations (C.11) and (C.12) are now shown to be true. The form of B must be used from
(4.2) along with its transpose, i.e.,
B = 2
r

j=0
(
j

j+1
)
T
j
B
T
= 2
r

j=0
(
j

j1
)
T
j
which results in
B
i
+B
T

k+hi
= 2 (
i

i+1
+
k+hi

k+hi1
) (C.13)
B
T

i
+B
k+hi
= 2 (
i

i1
+
k+hi

k+hi+1
) . (C.14)
The rst of these identities, (C.13) appears in D
+

, so reordering sums in two ways gives

D
+

=
1
2
h1

j=0

j
(a
j
a
j1
+a
k+hj
a
k+hj+1
)
D
+

=
1
2
h1

j=0

h+kj
(a
j
a
j1
+a
k+hj
a
k+hj+1
)
which may be combined to give
D
+

=
1
4
h1

j=0
(
j
+
k+hj
) (a
j
a
j1
+a
k+hj
a
k+hj+1
)
51
which matches (C.8), thus proving that (C.11) is correct. In exactly the same way (C.14) can
be used to show that (C.12) matches (C.9).
Having these universal forms for the gradients will allow momentum conservation to be shown
for all of the folded theories using the same argument. Everything that has been shown here
for D

has exact equivalents for D

, the dierence being that while a

i
a
i
, b
i


b
i
;
hence, there is no need to prove anything for D

separately.
All of the information needed to compare the a priori defect conditions ((5.3), (5.4) and (5.5))
to the a posteriori defect conditions ((C.1), (C.2) and (C.3)) is given by (C.7) with (C.4).
C.2 Comparison of the a priori and a posteriori defect conditions
The rst equations to compare are those obtained from the Euler-Lagrange equation for the
bulk fold on the left ( for the a priori case; u for the a posteriori case), i.e., (5.3) and (C.1).

=
p
B D

= A

+B D
u
.
There are clear dierences between the equations, particularly as regards the number of equa-
tions represented in component form. The a priori equation on the left only has meaning in the
folded root space while the a posteriori equation on the right has meaning in all of the a
(1)
r
root
space. For both types of folding to match it should at least be the case that (C.1) has the same
projection onto the folded root space as (5.3). Projection can be made onto the folded root
space by taking the inner product with
i
+
k+hi
; the left-hand sides match up so what needs
to be compared is
(
i
+
k+hi
)RHS(5.3) (
i
+
k+hi
)
p
B (
i
+
k+hi
) D

= (
i
+
k+hi
)B
1
2

j
(
i
+
k+hi
) (
j
+
k+hj
)
_
D
u
j
+D
u
k+hj
_
= (
i
+
k+hi
)B D
u
i
D
u
k+hi
with
(
i
+
k+hi
)RHS(C.1) (
i
+
k+hi
)A

+ (
i
+
k+hi
)B (
i
+
k+hi
) D
u
= (
i
+
k+hi
)B

j
(
i
+
k+hi
)
j
D
u
j
= (
i
+
k+hi
)B D
u
i
D
u
k+hi
.
These match, but there could be extra information contained in (C.1) in the part of the a
(1)
r
root
space which is not on the folded space. This extra information may actually be contained else-
where, perhaps in the time derivatives of the algebraic constraints (5.12) found in the equation
of motion. Some evidence for this supposition is provided for the a
(2)
2
case in appendix D.
The equations (5.4) and (C.2) give a completely similar picture to the above analysis so there
is no need to consider them explicitly.
Also clear is that the equations (5.5) and (C.3) are exactly the same, so the equation of
52
motion is unchanged by the dierent folding processes. Examining the components of (5.5) a
set of algebraic constraints is found:

_
B
T

B
T

_
=
i

_
D
(1)

D
(2)

_
=
i
D

i+1

i
+

i+1
= D

i
and

k+h1i

_
B
T

B
T

_
=
k+h1i
D

i
+

i+1
+

i+1
= D

k+h1i
and consequently
D

i
+D

k+h1i
= 0 (C.15)
where D = D
(1)
+D
(2)
. In this form the constraints are valid for all of the bivalent foldings. It
is worth checking how these constraints appear in terms of a
i
, a
i
, b
i
and

b
i
which involves
D

i
= a
i
a
i+1
+ a
i
a
i+1
b
i
+b
i+1

b
i
+

b
i+1
.
For each folding the number of constraints is
c
(1)
n
the n + 1 relations become n as the relation for i =
k
2
+ n is a trivial identity 0 = 0.
The remaining n relations are not all linearly independent but n1 of them are- so there
are n1 algebraic constraints- reducing the number of degrees of freedom of from 2n1
down to n.
d
(2)
n
: n of the n + 1 relations are independent so has (2n 1) n = n 1 degrees of
freedom remaining.
a
(2)
2n
: n independent constraints so has 2n n = n degrees of freedom remaining
Using the constraints, degrees of freedom in the auxiliary eld may be removed such that the
root space dimensionality of is the same in each case as the dimensionality of the folded elds
and .
In each case the relations (C.15) may be written in a more useful form
a
i
a
k+hi
+ a
i
a
k+hi
b
i
+b
k+hi

b
i
+

b
k+hi
= 0 .
Since each a
i
is a positive helicity term and each b
i
has negative helicity, the algebraic constraints
generally mix helicity which is an undesirable property. The remedy to this problem is provided
in section 5.3 which examines the passing of solitons through the defect.
D The folded defect in a
(2)
2
This appendix deals with the particular case of a
(2)
2
, also known as the T it eica or the Bullough
Dodd model, when considering the defect of section 5. It is taken that k = 0 in terms of the
53
bivalent foldings, and also that

d = d to give the non-trivial defect, so in components the elds
are
=

1
+
2
2

=

1
+
2
2

=
1

1
+
2

2
.
The ambiguity over the meaning of and will disappear since everything will be expressed
in terms of components henceforth. In components the Lagrangian (5.2) is then
L = (x)L

+(x)L

+(x) (( )
2
D) (D.1)
with
L

=
1
4

1
4

= e

+ 2e

2
and an entirely analogous picture for L

.
The defect potential is given by
D = d
_
e

2
+e

1
+
2
+e

2
+e

1
+
2
+ 2e

1
_
+
1
d
_
e

2
+
2
+e

1
+e

2
+
2
+e

1
+ 2e

2
_
. (D.2)
The defect equations of (D.1) at x = 0 are

= 2
2
2D

(D.3)
D

1
= 0 (D.4)

= D

2
(D.5)

= 2
2
+ 2D

. (D.6)
The rst thing to note now is the algebraic constraint (D.4). Since

d = d it is expected that the
algebraic constraint does not mix helicities (i.e., D
+

1
= D

1
= 0) and indeed that is the case.
Explicitly, using (D.2), the situation is
D

1
=
_
d +d
1
e

2
_
_
2e

1
e

1
+
2
e

1
+
2
_
= 0
so it is clear that D
+

1
and D

1
vanish separately. The algebraic constraint may be written as
e
2
1
=
1
2
e

2
_
e

2
+e

2
_
(D.7)
meaning that the
1
degree of freedom may be removed such that the auxiliary eld has just
one degree of freedom,
2
. One aspect of using (D.7) to remove the
1
terms in the exponentials
in the potential D is that it results in terms coupling to directly- this situation is also seen
54
in the type II defects of [3].
Also notable about the potential is that
D
+

2
= 2D
+

+ 2D
+

(D.8)
D

2
= 2D

2D

(D.9)
where the rst relation (D.8) is seen naturally in the form of (D.2); but the second, (D.9),
requires use of the algebraic constraint D

1
= 0 (or, alternatively, (D.7)).
Due to (D.8) and (D.9), the defect condition (D.5) may be rewritten as

+ 2D
+

2D

=

2D
+

+ 2D

. (D.10)
Energy and momentum conservation are now provided in components
D.1 Energy conservation
The situation here is precisely that of section 5.4 in a
(2)
2
. In components the energy conservation
equation is

E =
1
2

1
2

.
Using (D.3), (D.6) and then (D.5) gives

E =
1
2

1
2

1
D

1

2
D

2
=

D
with zero being added in the form of
1
D

1
. Thus E +D is conserved as required.
D.2 Momentum conservation
The a
(2)
2
component form version of the argument of section 5.5 is

P =
1
4
_

_
+ .
Squaring (D.3) and (D.6) give
1
4
_

_
= 2
2
(D

+D

) +D
2

D
2

=
2
_
D
+

2
D

2
_

1
_
D
+

1
D

1
_
+D
2

D
2

by (D.8), (D.9) and the fact that D

+

1
D

1
= 0. Similarly, squaring both sides of (D.10) and
equating results in
1
4
_

_
=

_
D
+

_
D
+

_
+
_
D
+

_
2

_
D
+

_
2
55
and thus the momentum conservation equation reduces to

P =
_

D
+

_
+ 4D
+

4D
+

+ .
It is seen now that this is of the expected form since
4D
+

4D
+

= e

+
_
e

_
e
2
1
+
2
= e

+
_
e

_
2
e

2
+e

2
= e

+ 2e

2
e

2e

2
= .
Hence, P + D
+
D

is conserved. Of course it is known already from section 5.5 that it is

conserved so this is nothing other than an explicit verication; however, the a
(2)
2
Lagrangian
(D.1) does have a simplicity that allows easier examination of a couple of other properties.
D.3 Comparison to the type II defect
As mentioned in the introduction, the paper [3] introduces the type II defect Lagrangian (1.11)
for a
(2)
2
(or alternatively for a
1
). It is further asserted in [3] and [22] that, at least in the a
1
case, there is an interpretation of a type II defect as two fused type I defects. Here (D.1) is
compared to that of (1.11) with the purpose of arming that this is also true, at least at the
classical level in a
(2)
2
.
The aim is to obtain (1.11)
L = (x)L
u
+(x)L
v
+(x) (u v 2(u v) D(u, v, ))
with the particular form of D valid for the a
(2)
2
defect in [3], by means of making elementary
identications on the elds of (D.1), viz.
L = (x)L

+(x)L

+(x)
_
( )
2

D(, ,
2
)
_
with the notation

D used for the eventual comparison to the D of (1.11).
To match the bulk Lagrange densities of [3] (L
u
, L
v
), the rst identications needed are
u = 2
v = 2
x

2x , t

2t
so at an intermediate stage of comparison (D.1) becomes
L = (x)L
u
+(x)L
v
+(x)
_
2 (u v)
2

D(2u, 2v,
2
)
_
and this Lagrangian can thus be put into the form of (1.11) by dening
=
2

1
4
(u +v) .
56
There is one further thing to note and that is that this Lagrangian has a gauge symmetry
+f(u v)
where f is any function. Thus (D.1) becomes
L = (x)L
u
+(x)L
v
+(x)
_
u v 2(u v)

D
_
2u, 2v,
1
4
(u +v) +f(u v)
__
.
It can be shown that

D with this argument agrees the (1.11) D specied in [3] for a specic
choice of gauge.
D.4 Algebraic constraints and a posteriori folding
This a
(2)
2
case has been examined from the point of view of a priori folding thus far, but there is
the concern that the a
(2)
2
versions of (C.1) and (C.2) may contain extra information not found
in the a priori case. Indeed extra defect conditions are found which are
(
1

2
) (C.1) 0 =

+ 4
1
2
2
D
u
1
+D
u
2
(D.11)
(
1

2
) (C.2) 0 =

+ 4
1
2
2
+D
v
1
D
v
2
. (D.12)
An apparently extra condition may also be obtained in the a priori (and a posteriori ) folded
case from the algebraic constraint (D.7). For (D.7) to be true for all times it must be that case
that
d
dt
_
2e
2
1

2
_
=
d
dt
_
e

2
+e

2
_
and so
1
2
(4
1
2
2
)
_
2e
2
1

2
_
=
1
2
_

2
+

e

2
_
.
Applying (D.7) to this results in
_

+ 4
1
2
2
_
e

2
+
_

+ 4
1
2
2
_
e

2
= 0 . (D.13)
So the situation is that (D.13) is true in the a priori folded case and this might just contain the
same information as (D.11) and (D.12). What is certainly true is that (D.13) can be veried
by means of (D.11) and (D.12); applying these to (D.13) gives
_

+ 4
1
2
2
_
e

2
+
_

+ 4
1
2
2
_
e

2
= (D
u
1
D
u
2
) e

2
+ (D
v
1
+D
v
2
) e

2
= 2de

1
+
2
_
2e
2
1

2
_
e

2
e

2
_
e

+e

_
+d
1
e

1
_
2e
2
1

2
_
e

2
e

2
_
+e

_
= 0
where the algebraic constraint (D.7) is used in the nal step.
57
References
[1] P. Bowcock, E. Corrigan, C. Zambon, Classically integrable eld theories with defects, Int.
J. Mod. Phys. A19 (2004) 82; arXiv:hep-th/0305022
[2] P. Bowcock, E. Corrigan, C. Zambon, Ane Toda eld theories with defects, JHEP
01(2004)056; arXiv:hep-th/0401020
[3] E. Corrigan, C. Zambon, A new class of integrable defects, J. Phys. A42 (2009) 475203;
arXiv:0908.3126
[4] T.J. Hollowood, Solitons in ane Toda eld theories, Nucl. Phys. B384 (1992) 523
[5] N.J. MacKay, W.A. McGhee, Ane Toda solitons and automorphisms of Dynkin diagrams,
Int. J. Mod. Phys. A8 (1993) 2791; arXiv:hep-th/9208057
[6] W.M. McGhee, On the topological charges of the ane Toda solitons, PhD thesis, Durham
(1994); etheses.dur.ac.uk/5482/
[7] H.W. Braden, E. Corrigan, P.E. Dorey, R. Sasaki, Ane Toda eld theory and exact S-
matrices, Nucl. Phys. B338 (1990) 689
[8] S.P. Khastgir, S-matrices of non-simply laced ane Toda theories by folding; arXiv:hep-
th/9702170
[9] S.P. Khastgir, R. Sasaki, Non-canonical folding of Dynkin diagrams and reduction of ane
Toda theories, Prog. Theor. Phys. 95 (1996) 503; arXiv:hep-th/9512158
[10] P. Dorey, Exact S-matrices; arXiv:hep-th/9810026
[11] E. Corrigan, C. Zambon, Comments on defects in the a
r
Toda eld theories, J. Phys. A42
(2009) 304008; arXiv:0902.1307
[12] R. Hirota, Direct methods in soliton theory, in Soliton: ed. R.K. Bullough and P.S. Caudrey
(1980)
[13] D.I. Olive and N. Turok, The symmetries of Dynkin diagrams and the reduction of Toda
eld equations, Nucl. Phys. B215 (1983) 470
[14] R. Sasaki, Reduction of the ane Toda eld theory, Nucl. Phys. B383 (1992) 291
[15] W.A. McGhee, The Topological Charges of the a
(1)
n
Ane Toda Solitons, Int. J. Mod. Phys.
A9 (1994) 2645; arXiv:hep-th/9307035
[16] Z. Zhu, D.G. Caldi, Multi-soliton solutions of ane Toda models, Nucl. Phys. B436 (1995)
659; arXiv:hep-th/9307175
[17] D.I. Olive, N. Turok, J.W.R. Underwood, Ane Toda solitons and vertex operators, Nucl.
phys B409 (1993) 509; arXiv:hep-th/9305160
[18] R.A. Hall, Ane Toda solitons and fusing rules, PhD thesis, Durham (1994); ethe-
ses.dur.ac.uk/5596/
[19] U. Harder, A.A.P Iskandar, W.A. McGhee, On the breathers of a
(1)
n
ane Toda eld theory,
Int. J. Mod. Phys. A10 (1995) 1879; arXiv:hep-th/9409035
58
[20] A.A.P. Iskandar, On breathers in ane Toda theories, PhD thesis, Durham (1995); ethe-
ses.dur.ac.uk/5236/
[21] D.I. Olive, N. Turok, J.W.R. Underwood, Solitons and the energy-momentum tensor for
ane Toda theory, Nucl. Phys. B401 (1993) 663
[22] E. Corrigan, C. Zambon, A transmission matrix for a fused pair of integrable defects in the
sine-Gordon model, J. Phys. A43 (2010) 345201; arXiv:1006.0939
[23] E. Corrigan, C. Zambon, Integrable defects in ane Toda eld theory and innite dimen-
sional representations of quantum groups, Nucl. Phys. B848 (2011) 545; arXiv:1012.4186
[24] P. Baseilhac, G.W. Delius, Coupling integrable eld theories to mechanical systems at the
boundary, J. Phys. A34 (2001) 8259; arXiv:hep-th/0106275
[25] E. Corrigan, C. Zambon, Innite dimension reection matrices in the sine-Gordon model
with a boundary, JHEP 06(2012)050; arXiv:1202.6016
[26] S. Ghoshal, A. Zamolodchikov, Boundary S-matrix and boundary state in two-dimensional
integrable quantum eld theory, Int. J. Mod. Phys. A9 (1994) 3841; arXiv:hep-th/9306002
[27] P. Bowcock, E. Corrigan, P.E. Dorey, R.H. Rietdijk, Classically integrable boundary condi-
tions for ane Toda eld theories, Nucl. Phys. B445 (1995) 469; arXiv:hep-th/9501098
59