< daughter1%kappasq = min(x*qbarsq/(1.0d0 - x),calQ1**2,mother%kappasq) < daughter2%kappasq = min((1.0d0-x)*qbarsq/x,calQ2**2,mother%kappasq)where calQ1 or calQ2 is the absolute value of the three momentum of the corresponding daughther produced by a primary splitting and qbarsq is the virtuality of the splitting and mother%kappasq had been previously assigned based on the topology of the three initial parton momenta. This has a problem for x --> 1 for the first line and x --> 0 for the second line. The new version is
> daughter1%kappasq = min(x*qbarsq/(1.0d0 - x),calQ1**2,ckappa*qbarsq) > daughter2%kappasq = min((1.0d0 - x)*qbarsq/x,calQ2**2,ckappa*qbarsq)where ckappa is a constant, 4 by default. When qbarsq --> 0, the daughter kappasq values go to zero. There is also a change in the starting value for kappa for splitting the soft gluon that the program creates to model radiation from the the three jets of the born graph.
< theshower%ptn(4)%kappasq = lsq * theshower%onemthrustbecomes
> theshower%ptn(4)%kappasq = ckappasoft * lsqwhere lsq is the square of the three momentum of the soft gluon and ckappasoft = 1/10. There was no reason to involve the thrust of the born event here.
The numerical effect appears to be negligible.