Non-Markovian Monte Carlo algorithm for the constrained Markovian evolution in QCD

We revisit the challenging problem of finding an efficient Monte Carlo (MC) algorithm solving the constrained evolution equations for the initial-state QCD radiation. The type of the parton (quark, gluon) and the energy fraction x of the parton exiting emission chain (entering hard process) are predefined, i.e. constrained throughout the evolution. Such a constraint is mandatory for any realistic MC for the initial state QCD parton shower. We add one important condition: the MC algorithm must not require the a priori knowledge of the full numerical exact solutions of the evolution equations, as is the case in the popular "Markovian MC for backward evolution". Our aim is to find at least one solution of this problem that would function in practice. Finding such a solution seems to be definitely within the reach of the currently available computer CPUs and the sophistication of the modern MC techniques. We describe in this work the first example of an efficient solution of this kind. Its numerical implementation is still restricted to the pure gluon-strahlung. As expected, it is not in the class of the so-called Markovian MCs. For this reason we refer to it as belonging to a class of non-Markovian MCs. We show that numerical results of our new MC algorithm agree very well (to 0.2%) with the results of the other MC program of our own (unconstrained Markovian) and another non-MC program QCDnum16. This provides a proof of the existence of the new class of MC techniques, to be exploited in the precision perturbative QCD calculations for the Large Hadron Collider.