AMEGIC integrator broken for ee->nj (n>2)

[marek]

Amegic's default integrator seems to be broken for ee->nj (n>2) setups. eg.

(run){

EVENTS = 10000 ME_SIGNAL_GENERATOR Amegic FRAGMENTATION Off

}(run)

(beam){

BEAM_1 = 11; BEAM_ENERGY_1 = 45.6; BEAM_2 = -11; BEAM_ENERGY_2 = 45.6;

}(beam)

(processes){

Process 11 -11 -> 93 93 93; Order_EW 2; Scales VAR{sqr(91.2)}; End process;

}(processes)

(selector){

NJetFinder 3 5. 0. 1. 1

}(selector)

The integrator's efficiency drops below 1e-5% before the first integration step. Running this setup with INTEGRATOR=4 at least comes as far being optimized, then is performing nicely. INTEGRATOR=0 performs nicely and converges fastest.

I have checked these findings with all releases as far back as 1.2.0 with the same result. Any ideas?

[Stefan Hoeche]

Channel generator 6 produces threshold integrators for propagator masses. In the case of ee->3j with the NJetFinder the thresholds are zero because NJetFinder does not set minimum invariant masses. The proper solution will be to use AMEGIC_INTEGRATOR=4, which may be inefficient at first but avoids the fake threshold problem.

[Stefan Hoeche]

Fixed on trunk r25669.