asymptotics of Q’s similar to (2.20), they form a closed system of equations alternative to
the Pµ-system.
One can also relate µ’s and ω’s with the help of Q-functions with four indices Q
ab|ij
defined as a determinant,
Q
ab|ij
=
Q
a|i
Q
a|j
Q
b|i
Q
b|j
. (2.22)
Then we have
µ
ab
=
1
2
Q
−
ab|ij
ω
ij
. (2.23)
In the next section we will discuss how to change the large u asymptotics to describe
the generalized cusp.
2.2 Adapting the QSC for the cusp anomalous dimension
In this section we will discuss the modifications in the QSC which are needed to describe
the generalized quark-antiquark potential. Below we will only discuss the vacuum state,
i.e. the Wilson line with L scalar insertions at the cusp (the extension for more general
insertions should be straightforward).
The Quantum Spectral Curve equations of [2, 3] in N = 4 SYM can be deduced
from the TBA equations or the corresponding T- and Y-systems with special analyticity
assumptions. In our case the TBA equations for the generalised cusp are almost the same
as the original TBA system. The Y-system and T-system equations are exactly the same
as for the original problem. Thus it is natural to expect that the QSC equations should
also be the same to a large extent. In the TBA there are only two important differences:
the extra boundary dressing phase supplementing the BES phase, and the twists which
appear as chemical potentials and introduce the angles φ, θ into the problem.
3
We do not
derive the QSC from the Thermodynamic Bethe ansatz, rather we will put forward and
motivate a proposal which is consistent with several highly nontrivial checks, leaving little
doubt as to its correctness.
First, we expect to have the same set of Q-functions and auxiliary functions such as
µ
ab
as in the original problem. All of them will satisfy the same functional relations, for
instance the Pµ-system equations (2.1)–(2.4) or the QQ relations are unchanged. However
some analyticity properties will change, as we will discuss below, and in particular the
P-functions acquire an extra cut going from u = 0 to infinity. In addition, the large u
asymptotics clearly need to be modified. Indeed, the twists in the boundary conditions
typically correspond to imposing exponential rather than powerlike asymptotics for the
Q-functions (see e.g. [3] and references therein). In our case the angle θ is naturally related
to the S
5
part of the geometry, which qualitatively corresponds to the P-functions, so
roughly speaking we expect P
a
∼ e
±θu
at large u. Similarly, the angle φ is associated to
AdS
5
leading to Q
i
∼ e
±φu
. This argument is also supported by the expectation that P’s
and Q’s should be related in the classical limit to the quasimomenta for S
5
and AdS
5
,
3
There is also an extra symmetry requirement on the Y-functions of the TBA, namely they should be
invariant under the exchange of the two wings of the Y-system with a simultaneous reflection u → −u, i.e.
Y
a,s
(u) = Y
a,−s
(−u), see [33, 34] for details.
– 8 –