HERWIG Parameter Settings

Tuning Monte Carlo event generators to data, particularly the large volume of very precise results available from e+e- machines, is a quite non-trivial exercise. Below we collect together some recent HERWIG parameter tunings which have kindly been made available to us by the various Collaborations.

If you make use of these tunings please give credit to the appropriate experiment.

Version 6.1

The default parameter settings in the release version 6.1 were not tuned to data, they corresponded to a tuned set appropriate only to version 5.8. It was, however, anticipated that such a tuning would be necessary to obtain good agreement with data. So far, we only have details of one tuning: In this table, a blank entry means that the parameter was not tuned, while D means the parameter was tuned but did not change from its default value.
Parameter  Default  ALEPH  DELPHI  L3  OPAL 
QCDLAM  0.180 
RMASS(13)  0.750 
CLMAX  3.35 
CLPOW  2.00 
PSPLT(1)  1.00 
PSPLT(2)  1.00  0.33 
CLSMR(1)  0.00  0.40 
CLSMR(2)  0.00 
PWT(3)  1.00 
PWT(7)  1.00 
SNGWT  1.00 
DECWT  1.00  0.70 

Version 5.9

The default parameter settings in the release version 5.9 were not tuned to data, they corresponded to a tuned set appropriate only to version 5.8. It was, however, anticipated that such a tuning would be necessary to obtain good agreement with data. A number of such tunings have kindly been communicated to us, we thank the authors for making them available.  
Parameter  Default  ALEPH I  ALEPH II  DELPHI  L3  OPAL  SLD 
QCDLAM  0.180  0.179  0.173  0.177  0.15 
RMASS(13)  0.750  0.68  0.645  0.70 
CLMAX  3.35  3.04  3.025  3.006  3.75 
CLPOW  2.00  2.033  1.30 
PSPLT  1.00  (0.87)  0.984  0.85 
CLSMR  0.00  0.40  0.35 
PWT(3)  1.00  0.88 
PWT(7)  1.00  0.80 
SNGWT  1.00  0.70 
DECWT  1.00  0.745  0.50 

Version 5.8

The following parameter tunings were compiled by the QCD event generators working group of the LEP 2 workshop: CERN yellow report CERN 96-01 vol.2 p.127.
Parameter  Default  ALEPH  DELPHI  L3  OPAL  SLD 
QCDLAM  0.180  0.149  0.163  0.170  0.160 
RMASS(13)  0.750  0.726  0.650  0.750  0.750 
VPCUT  0.40  1.00  0.40  0.50  0.40 
CLMAX  3.35  3.90  3.48  3.20  3.40 
CLPOW  2.00  2.00  1.49  1.45  1.30 
CLSMR  0.00  0.56  0.36  0.00  0.35 
PWT(3)  1.00  1.00  0.83  1.00  1.00 
PWT(7)  1.00  1.00  0.74  1.00  1.00 
DECWT  1.00  1.00  0.77  1.00  1.00 


This is Lambda_QCD; it sets the scale for the running strong coupling. At high momentum fractions (x or z) it can be identified with the fundamental QCD scale Lambda-MSbar (5 flavours). However, this relation does not necessarily hold in other regions of phase space, since higher order corrections are not treated precisely enough to remove renormalization scheme ambiguities.

See S. Catani, G. Marchesini and B.R.Webber, Nucl. Phys. B349 (1991) 635.


This is the effective gluon mass used during hadronization. To allow for non-perturbative gluon splitting RMASS(13) must be greater than 2*RMASS(1) or 2*RMASS(2).

The light quark and gluon masses acts as lower cut-offs on the parton shower evolution. They can be simulataneously set to zero provided that the shower cut-off parameters VQCUT and VGCUT are large enough. The condition to avoid divergences in the shower is:

1/Q(i) + 1/Q(j) < 1/QCDL3

where Q(i)=RMASS(i)+VQCUT or RMASS(i)+VGCUT for i=q or g and QCDL3 is Lambda_QCD(3-flavours) as derived from QCDLAM.


Is the shower cut-off for photons, its default value is SQRT(S), which implies no photon radiation.

After experimental cuts results show little sensitivity to VPCUT in the range 0.1 to 1.0 GeV.


These two parameters determine the maximum allowed cluster mass before forced cluster splitting occurs, according to:


Smaller values of CLPOW will increase the yield of heavier clusters (and hence of baryons) for heavy quarks, without affecting light quarks much.


This determines the mass distribution in the cluster splitting CL1 --> CL2 + CL3 when CL1 is above the maximum allowed mass. The masses of CL2 and CL3 are generated uniformly in Mass**PSPLT.

PSPLT is more influential for heavy flavour clusters. Accordingly, in version 6.1 this parameter was converted into a two-dimensional array, with PSPLT(1) controlling light (d,u,s,c) quark clusters, PSPLT(2) those that contain a b-quark.


During cluster splitting in the CLDIR=1 (default) option, a hadron which contains a quark which came from the perturbative shower (as opposed to the non-perturbative gluon splitting) retains knowledge of this quark's direction. CLSMR controls a Gaussian smearing of the hadron direction about the original constituent quark direction. Strictly the distribution is exponential in 1-COS(THETA), with mean CLSMR; increasing CLSMR increases the decorrelation.

These parameters are particularly important for b-hadrons. Accordingly, in version 6.1 they were converted into two-dimensional arrays, with CLDIR(1) and CLSMR(1) controlling light (d,u,s,c) quark clusters, CLDIR(2) and CLSMR(2) those that contain a b-quark.

PWT(3) and PWT(7)

These are a priori weights for choosing a strange quark or a diquark when splitting a cluster to form hadrons.


These are a priori weights for the relative production rates of singlet (Lambda-like) and decuplet baryons comparred to octet baryons.

Analogous weights, REPWT, exist for same flavour meson of different internal S & L quantum numbers.

Ian Knowles (mailto:I.Knowles@ed.ac.uk) and Bryan Webber (mailto:webber@hep.phy.cam.ac.uk)

30 May 2000