Two-loop corrections to partition function of Pohlmeyer-reduced theory for $AdS_5 x S^5$ superstring

Pohlmeyer reduction of AdS_5 x S^5 superstring leads to a set of equations of motion following from an action containing a bosonic Sp(2,2) x Sp(4)/[SU(2)]^4 gauged WZW term, an integrable potential and a fermionic part coupling bosons from the two factors. The original superstring and the reduced model are in direct correspondence at the classical level but their relation at the quantum level remains an open question. As was found earlier, the one-loop partition functions of the two theories computed on the respective classical backgrounds match; here we explore the fate of this relation at the two-loop level. We consider the example of the reduced theory solution corresponding to the long folded spinning string in AdS. The logarithm of the AdS_5 x S^5 superstring partition function computed on the spinning string background is known to be proportional to the universal scaling function which depends on the string tension ~ \l^{1/2} where \l=\lambda is `t Hooft coupling. Its "quantum" part is f(\l) = a1 + \l^{-1/2} a2 + ... where the one-loop term is a1 = - 3 \ln 2 and the two-loop term is minus the Catalan's constant, a2=-K. We find that the counterpart of f(\l) in the reduced theory is f'(k) =a1' + 2 k^{-1} a2' + ..., where k is the coupling of the reduced theory. Here the one-loop coefficient is the same as in the string theory, a1'= a1, while the two-loop one is a2' =a2 - 1/4 (a1)^2. Remarkably, the first Catalan's constant term here matches the string theory result if we identify the two couplings as k= 2\l^{1/2}. Nevertheless, the presence of the additional (a1)^2 ~ (\ln 2)^2 term implies that a relation between the two quantum partition functions (if any) is not a simple equality. Similar results are found in the case of AdS_3 x S^3 superstring theory where a1= - 2 \ln 2 and a2=0, while in the corresponding reduced theory a1'=a1, a2'= a2- 1/4 (a1)^2.