flag
Vincia:GluonSplittingCorrection
(default = on
)option
on : Use the factor PAri. Default
since it greatly improves the shower's approximation to matrix
emelents.
option
off : Don't use the factor PAri.
Included mostly for testing purposes.
mode
Vincia:NLL
(default = 1
; minimum = 0
; maximum = 2
)option
0 : Off. Scale variations in
αs are not compensated
at all in the shower evolution.
option
1 : Scale Cancellation Only. Scale variations
in αs
are compensated up to 1st order by matching to the explicit
b0-dependent terms in the second-order (NLL)
antenna functions.
option
2 : (Reserved for future use.) Full matching to
second-order antenna functions.
Note: VINCIA uses a special color-factor normalization in which CF=8/3, and hence the difference CA-CF=1/NC2.
In the soft limit, the full radiation pattern from a color-connected chain of partons, qg...gqbar, can be written as a sum of leading-color dipole-antennae spanned between each of the color-connected partons, plus a negative qqbar antenna spanned directly between the endpoint quarks, proportional to CA/NC2. VINCIA contains a few different options for taking this antenna into account at various levels, as follows:
mode
Vincia:NLC
(default = 1
; minimum = -1
; maximum = 2
)option
-1 : Strict LC. All gluon emission color factors forced
equal to CA. No NLC correction.
option
0 : Ordinary LC. The color factors are taken
from the chosen antenna set. I.e., they will typically be
CF for qqbar antennae, CA for gg antennae, and something inbetween
for qg antennae. No NLC correction.
option
1 : Eikonal NLC. A negative Eikonal suppressed by 1/NC2
is spanned between the two endpoint quarks of each qg...gqbar chain
and is absorbed systematically into the radiation off the LC
ones. Has been checked to correctly generate the subleading-color
1/ε2 term
of the full 1-loop qqbar→qgqbar splitting function, and also
improves the agreement with full-color tree-level Z→4,5,6-parton
matrix elements relative to option 0.
option
2 : Full NLC. A negative qqbar antenna (Full = Eikonal +
single-log terms) suppressed by 1/N2
is spanned between the two endpoint quarks of each qg...gqbar chain
and is absorbed systematically into the radiation off the LC ones.
Has been checked to correctly
generate both the subleading-color 1/ε2
and 1/ε terms of
the full 1-loop qqbar→qgqbar splitting function, and also further improves
the agreement with full-color tree-level Z→4,5,6-parton matrix
elements relative to option 1.
Note: switching these corrections on may reduce the speed of the
shower algorithm by up to a factor of 2.
Since the subleading-color antenna spanned between the q and qbar is not positive definite, it cannot be treated at the same footing as the leading-color ones. It can, however, be systematically absorbed into the leading-color ones, effectively reducing their color factor from CA towards something closer to CF, with the exact magnitude of the correction depending on the phase space point.
Expressed in dimensionless variables, each LC antenna is modified as follows
where N=3 for QCD and the factor fj partitions the full correction among the LC contributions. By default in VINCIA, it is given by
where again the eikonals must be expressed as dimensionless functions,
Practically, the implementation relies on a guess-and-reweight strategy, in which branchings are first generated according to leading-color antennae, but then the resulting configurations are reweighted by vetos, now taking into account the subleading correction which is always negative.
This matching is applied at every branching step (exponentiated) and so is in principle active to all orders. However, since the final state of subleading-color branchings are represented as planar, the approach cannot be used to reach formal 1/NC^4 accuracy.