**************************** * * * Update notes * * * * PYTHIA version 6.4 * * * **************************** (Last updated 30 January 2007) PYTHIA version 6.4 is a direct continuation of version 6.3; actually 6.400 is identical with 6.327. Therefore it should not be a big operation for the normal user to run the program. PYTHIA 6.4 comes with an updated big manual, reflecting the status as of version 6.400. These update notes document what is new in each subsequent subversion. ----------------------------------------------------------------------- 6.400 : 25 March 2006 - This version is identical with Pythia 6.327. Readers should therefore turn to the Pythia 6.3 update notes to trace the evolution of the program. - The only changes are updates of character strings, 6.3 -> 6.4. 6.401 : 16 April 2006 - Bug fix in PYRAND: when the Les Houches Accord interface is used with more than 50 processes, some of them would be mapped into a range of PYTHIA process numbers assigned to SUSY processes, where KFPR(ISUB,1) and KFPR(ISUB,2) would be overwritten. - Modification of string fragmentation, introducing new tolerance parameter. PARU(14): (D=2.) when passing string corners, the (mis)match of transverse momentum directions may need to be compensated by using momentum fractions x outside the allowed range 0 < x < 1, by having a slightly negative x. Occasionally the x can become quite negative, and then rarely give strange results. The new parameter sets limits how far outside the allowed range one may go before rejecting the current try and restarting the fragmentation of the current string. - The size of the /PYPART/ commonblock has been expanded, by replacing the current MAXNUP size parameter by a new MAXNUR one with size 1000. This should avoid some problems with overflows. - Bug correction: for the MSTP(95) > 3 options the colour reconnection in PYFSCR was not called from PYEVNW. 6.402 : 2 May 2006 - PYFSCR: bugfix and updates (performing colour reconnections in the new underlying-event framework). A bug for MSTP(95)=2,3 caused an unintended suppression of connections between free colour octets. This has now been fixed, so the scenario works as described in the manual. In addition, a new option has been added, by MSTP(95)=6,7, where only partons with one and only one free colour tag are allowed to initiate a string piece. This amounts to saying that the string collapse is driven by free triplet charges. Finally, the parameter PARP(78) now also works for MSTP(95)>1, and controls the strength of the colour reconnections. In addition, the reconnection probability grows as a function of the number of multiple interactions, to reflect the possibility of a higher amount of reconnections occurring in collisions with a higher amount of strings. The full description of the PYFSCR parameters is thus now: MSTP(95) : CR Type = 1 : old cut-and-paste reconnections, handled in PYMIHK = 2 : Type I(no gg loops); hadron-hadron only = 3 : Type I(no gg loops); all beams = 4 : Type II(gg loops) ; hadron-hadron only = 5 : Type II(gg loops) ; all beams = 6 : Type S ; hadron-hadron only = 7 : Type S ; all beams Types I and II are described in Sandhoff+Skands, in hep-ph/0604120. Type S is driven by starting only from free triplets, not octets. PARP(78) : CR Strength The probability to keep a given colour-dipole pair depends on PARP(78) and the number of interactions, MINT(31): PKEEP = (1D0-PARP(78))**(MINT(31)) ; PARP(78) -> 1D0 : Full strength. PARP(78) -> 0D0 : Zero strength. No reconnections. NOTE: New min-bias/UE tunes with these new models are also available from the web page. - Speeded-up machinery for minimum-bias events, ISUB = 95. Specifically, the simultaneous choice of impact parameter b and pT for the hardest interaction has been optimized. The key trick is to split the b range in two, roughly where the probability for an interaction of two hadrons passing through each other has dropped to 1/2. With such a split it is feasible to pick b once and for all, and then to use two different formalisms for the pT selection. At small b, one includes the "Sudakov" exponent to generate a trial sequence of falling pT's until accepted. If the sequence reaches pT = 0 the generation is restarted at maximum. At large b the trial pT's are selected unordered over the whole pT range, and then the "Sudakov" exponent is used as (part of) the acceptance weight. This technical change should not affect physics. Internally, two new variables are used: MINT(39) : region of b value in current event. = 0 : not defined (e.g. for MSTP(82) = 1 or 2). = 1 : low-b region. = 2 : high-b region. VINT(139) : current b value. Note that no rescaling to physical units is performed, but only to an average b of 1 in minimum-bias events, and thereby below 1 for high-pT events. (Related to the jet pedestal effect.) Thus it only has a relative meaning in the context of comparing different events inside the same sample (i.e. generated with the same parameter values). - Initialization of multiple interactions did not take into account that the old and new multiple interactions scenarios use somewhat different factorization scale choices. This gave partly misleading initialization information for the new scenarios, but hardly affects the events actually generated for Tevatron or LHC applications. It could be more visible if tunes are extended to lower energies - PYSLHA: update. When reading in SLHA decay tables for Higgs bosons, a message is printed to alert the user to the fact that the h/H/A->gg decay width is also used to compute the cross section for the inverse process: gg -> h/H/A - PYFEYN: update. The previous interface to FeynHiggs became obsolete with FeynHiggs version 2.3, where a number of parameters in the relevant FeynHiggs call routines were changed from DOUBLE PRECISION to DOUBLE COMPLEX. The corresponding change has now been introduced in PYFEYN as well, so that earlier Pythia versions are compatible with FeynHiggs 2.2, while Pythia versions from 6.402 onward are compatible with FeynHiggs 2.3. - PYSUGI: small bug fix, removed a CLOSE statement that could lead to crashes if no file had previously been opened. - Production channels g g and gamma gamma -> Higgs rely on knowledge of the Higgs -> g g and gamma gamma partial widths. If SLHA is used to replace the internal decay tables the program would be looking in the wrong place. This has now been corrected. Note, however, that it is not possible to study e.g. g g -> Higgs if the new decay tables do not provide the Higgs -> g g decay channel. 6.403 : 7 June 2006 - The Pythia 6.4 Physics and Manual is now published in T. Sjostrand, S. Mrenna and P. Skands, JHEP05 (2006) 026 [hep-ph/0603175]. From now on, this is the only correct reference to the Pythia 6 program. - PYFEYN: further bug correction in this interface. - Typo in the speeded-up multiple-interactions treatment in 6.402 may affect results for the impact-parameter profile MSTP(82) = 5 when PARP(83) < 2, and is now corrected. - Insert Planck scale in PYSUGI in appropriate units: AMPL = 2.4 E18 GeV = {8 pi G_newton}^{1/2}. - Remove superfluous comma in PYSUGI printout statement. - Move default initialization of the PYLH3P and PYLH3C commonblocks to BLOCK DATA PYDATA for consistency with Fortran standard. - PYSIGH: An additional possible classification of basic scattering processes now exists, for the special case of Weak Boson Fusion and g g or q qbar -> Q Qbar Higgs processes: ISJETS = 2: Processes for which the matrix element does contains jets at leading order, but which are not included in ISJETS=1 since they are not of a QCD origin, as for instance the forward jets in WW/ZZ fusion to Higgs, qq -> qqH. The processes classified as ISJETS=2 are currently ISUB=121-124, 173, 174, 178, 179, 181, 182, 186, 187. The maximum scale for initial state radiation in these processes is set equal to the factorisation scale, see MSTP(32) and MSTP(39). 6.404 : 10 August 2006 - Changes to handle the new proposed Les Houches Event Files (LHEF) format [hep-ph/0609017]. 1) The existing two sample routines UPINIT and UPEVNT, which read the PYTHIA old separate initialization and event files, have been commented out (so the code is still there, just in case...). 2) The new (uncommented) UPINIT and UPEVNT routines handle an LHEF, for initialization and event read-in. As before, the file to be used must be opened by the user in the main program, and the file number set in MSTP(161) and MSTP(162). Previously, with two separate input files, MSTP(161) and MSTP(162) were different. Now they are the same, but both must still be set. (As before, the routines that come with PYTHIA are only suggestions. You are free to supply your own routines and link them instead. The difference is that now there will be default code that eventually will work right away with several different parton-level generators.) 3) A new routine PYLHEF has been added to produce an LHE file from a parton-level PYTHIA run output. Recall that, previously, PYUPEV had to be used instead of PYEVNT to produce parton-level-only events and write them to file MSTP(161), and a call to PYUPIN had to be added at the end to write initialization information to file MSTP(162), both files to be opened by the user. (The PYUPIN call has to come at the end, since it is only then that cross sections are known.) By adding a call to PYLHEF after the PYUPIN call, the two files will be combined into one single LHE file. MSTP(163) : (D=0) file unit number on which the LHE file will be written. The corresponding file must be opened by the user in the main program. MSTP(164) : (D=0) fate of input files when an LHE file is created by PYLHEF. = 0: the input files, associated with units MSTP(161) and MSTP(162), are removed. = 1: the input files are kept unchanged, although the information is now duplicated in the new file. 4) Note: it would have been possible to avoid the use of two intermediate files, by a judicious use of dummy cross sections when the beginning of the file is first written, with the correct ones substituted after a REWIND. This seemed like overkill, given the small use of PYTHIA as a parton-level-only generastor. The current facility to write LHE files is mainly intended to test the LHEF structure. 5) The precision with which lifetime and spin are written in PYUPEV has been reduced, in particular since spin is not assigned by PYTHIA and thus always defaults to 9. = unassigned,. and lifetime is usually assumed 0. = vanishing. - The processes 381-388 for QCD processes modified by technicolor interactions are reclassified as being QCD processes in PYSIGH (except for top production), i.e. multiple interactions and initial-state radiation are restricted to be softer than the hard process itself. - The colour octet wavefunction-at-the-origin default values for charmonia and bottomonia in PARP(141)-PARP(150) are now set according to the values in P. Nason et al., in "Standard model physics (and more) at the LHC", eds. G. Altarelli and M.L. Mangano, CERN 2000-004, p. 231 [hep-ph/0003142] as described in talk by M. Bargiotti in the HERA4LHC workshop, CERN, 8 June 2006. - The values of the parton densities, xf(x,Q2) of the two incoming partons to the hard interaction (if any) are now stored in PARI(29) and PARI(30), respectively. (And internally also in VINT(39) and VINT(40). The XSFX array, which contains this kind of information, is overwritten for each new multiple interaction, and is therefore not so useful.) Since before the related flavours are stored in MSTI(15), MSTI(16), the x values in PARI(33) and PARI(34), and the Q scale in PARI(23). - The size NMXHEP of the HEPEVT commonblock is now stored in MSTU(8) (for informational purposes). - The two dummy subroutines FHSETPARA and FHIGGSCORR have been updated to take complex variables, cf. update note 6.402 for PYFEYN. Irrelevant in practice, but introduced for consistency. 6.405 : 31 August 2006 - Mistake in 6.404 when processes 381-388 were to be classified as QCD processes (an .OR. instead of an .AND.) lead to all processes being classified as QCD-type, and thereby, in some cases, to the maximum scale for ISR and MI being set incorrectly. - Several processes involving photons in the initial, intermediate or final state have been assigned ISJETS=3 in PYSIGH. This mean that the maximum shower scale is set equal to the hard-process scale, typically pT, rather than the full energy. Thereby (in most cases) the old behaviour from Pythia 6.2 should be recovered for these processes. The ones relevant for hadron colliders are 10, 18, 19, 20, 83, 106 and 114. The ones for incoming photon beams are 33 - 36, 54, 58, 69, 70, 80, 84, 85, 107, 108, 110 and 131 - 140. - A few processes have been added to the ISJETS = 2 class introduced in 6.403, i.e. WW/ZZ fusion: 5, 8, 71, 72, 73, 76, 77, 351 and 352. - Showering and PDF scales for elastic and diffractive processes explicitly set zero, for cosmetics reasons only. - New option MSTP(123) = 3 gives same bahaviour as = 2 but does not print out warnings. 6.406 : 14 September 2006 - In 6.402 a bug was fixed, where previously an incorrect PDF scale choice was used at initialization of the new multiple interactions scenario: intended to have pT^2 but got pT^2 + pT_0^2, as in the old scenario. The mistake did not have so dramatic consequences, since the intended scale was used during event generation, so it mainly affected the initialization of the impact-parameter picture. When the bug was fixed in 6.402, unfortunately the fix came to overwrite the initialization of the old scenario, which therefore came to suffer from the opposite problem, of being initialized with a PDF scale of pT^2 instead of the pT^2 + pT_0^2 intended for this scenario. This is now fixed. - Second-order matching formulae for alpha_strong at flavour thresholds introduced, following W.J. Marciano, Phys. Rev. D29 (1984) 580. 6.407 : 25 September 2006 - New option for Les Houches Event Files. MSTP(165) : (D=0) writing or not of an extra line with parton density information for each event, using the format suggested in the LHEF writeup: #pdf id1 id2 x1 x2 scalePDF xpdf1 xpdf2 which here translates into #pdf MSTI(15) MSTI(16) PARI(33) PARI(34) PARI(23) PARI(29) PARI(30) = 0 : no. = 1 : yes, provided that MSTP(162) > 0. Note: currently such information is not read back in when the UPEVNT routine is used. It is read by Pythia8, however. 6.408 : 24 October 2006 - Colour Reconnections: The new colour annealing models described in the 6.402 update notes above were previously only available with the new interleaved underlying-event framework (MSTP(81)=21). To help extricate genuine colour-reconnection effects from changes in the shower and other underlying-event aspects, the colour annealing models can now also be used as an 'afterburner' to the old underlying-event framework (i.e. with MSTP(81)=1). The models are switched on, as before, using MSTP(95), with values between 2 and 6 currently valid. The strength is regulated with 0D0 < PARP(78) < 1D0. Since the models do increase the time it takes to generate an event, it may be of interest to note that MSTP(95)=6 is somewhat faster than the other models. In order not to 'double count' colour reconnections, we also recommend setting the colour connection parameter PARP(85)=0D0 when using the old framework with these models. - Pythia Tunes for Underlying Event (UE) and Min-Bias (MB): Several sets of widely used PYTHIA parameter settings ('tunes') have been collected and made more easily available through the auxiliary routine PYTUNE(ITUNE), which should be called before the call to PYINIT. At this time, the following values of ITUNE are recognized: ITUNE NAME 0 Default : No settings changed => linked Pythia version's defaults. ====== Old UE, Q2-ordered showers ======================================== 100 A : Rick Field's Tune A 101 AW : Rick Field's Tune AW 102 BW : Rick Field's Tune BW 103 DW : Rick Field's Tune DW 104 DWT : Rick Field's Tune DW with slower UE energy scaling 105 QW : Rick Field's Tune QW (NB: needs CTEQ6.1 pdfs externally) 106 ATLAS : Arthur Moraes' ATLAS tune 107 ACR : Tune A modified with annealing CR ====== New UE, Q2-ordered showers ======================================== 200 IM 1 : Intermediate 1: new UE, Q2-ordered showers, annealing CR ====== New UE, interleaved pT-ordered showers, annealing CR ============== 300 S0 : Sandhoff-Skands Tune 0 301 S1 : Sandhoff-Skands Tune 1 302 S2 : Sandhoff-Skands Tune 2 303 S0A : S0 with "Tune A" UE energy scaling 304 NOCR : New UE "best try" without colour reconnections. 305 Old : New UE, original (primitive) colour reconnections Note 1: care should be taken when interpreting results obtained with these tunes. They have not all been tuned to the same data, and not to the same level of sophistication. For more details about each tune / model, see the list of parameters and useful references printed by PYTUNE for each specific model. (This printout can be switched off with MSTU(13)=0.) Note 2: care has been taken so that these parameter settings should also work when used 'in sequence', i.e. when several different models are initialized one after the other during a single Pythia run, but we have not explicitly verified this for all possible combinations. If running multiple models in sequence you are therefore advised first to make a few quick sanity checks first, comparing e.g. to what you get when initializing only one model in each Pythia run. Note 3: the PYTUNE routine also contains tune parameters for the GAL and SCI scenarios of the Uppsala group. These options can only be run with a modified version of PYTHIA 6.215, however, as described in the routine. If you want to use those, you therefore have to extract PYTUNE and link it together with the modified PYTHIA version on the Uppsala web site. The GAL and SCI models cannot be used in sequence with other PYTUNE calls. - SLHA Decay Tables. Bug fix in PYSLHA. When reading in SLHA decay tables for 3-body decays, erroneous behaviour could occur since Pythia expects a specific ordering in the list of decay products while the SLHA standard does not impose such ordering. There were a few cases (e.g. sparticle in SLHA file appearing as decay product 2, with absolute KF code of number 3 larger than that of number 1, or coloured particle appearing as decay product 2) where PYSLHA failed to correctly reorder the decay products. This has now been amended. - External particles, QNUMBERS, PARTICLE. Improvements in the read-in of external particles. PYSLHA is now more intelligent when forming the name of antiparticles, replacing '+' by '-' etc for charged states, rather than just appending 'bar'. PYSLHA now also deals better with reading mass spectra for completely generic external states. Finally, to test out alternative naming schemes (QNUMBERS may be included in SLHA2 but the standard has not yet been agreed upon), Pythia will understand BLOCK PARTICLE to mean the same thing as BLOCK QNUMBERS. 6.409 : 13 November 2006 - Division in DATA statements in PYALPS removed, owing to problems for some compilers. - PYTIME modified so no time information at all by default, again owing to problems for some compilers. 6.410 : 30 January 2007 - Bug corrected in intialization of Les Houches colour tags in /PYCTAG/. The number of colour tags, NCT, was not consistently reset to zero for each new event, leading to a possible slowing-down and eventually crashing of the program for certain specific settings, e.g. for Tune ACR (using the PYTUNE preset 107). NCT is now explicitly reset to zero at the beginning of each event. - Added warning in PYINOM to clarify the treatment of chargino and neutralino masses in extreme cases. When the internal PYTHIA code is used to calculate the Supersymmetric neutralino and chargino masses, it is prohibited for the lightest chargino to be lighter than the lightest neutralino. When this would otherwise happen, the lightest chargino mass is forced to be larger than the lightest neutralino mass plus two times the pion mass. This choice is based on the fact that gauge boson loop corrections dominate over Supersymmetric corrections for many cases (see Nucl. Phys. B543: 47-72,1999 for a discussion). To avoid confusion about this, an explicit warning is now printed whenever this happens. We stress that the calculation of sparticle masses and mixing angles from input soft Supersymmetry breaking parameters is (for the most part) treated at only the tree level in PYTHIA. Currently, a large number of dedicated (external) codes exist to calculate the Supersymmetric spectrum at higher orders. These can be interface with PYTHIA using the SLHA interface. With these tools, the interested user can study, for example, the case of chargino decays to a single pion or a quasi-stable chargino. - An additional warning statement is added to PYAPPS to remind the user whenever highly approximate formulae are used to calculate the Supersymmetric particle spectrum. This code was and is only intended for debugging purposes. - Change in PYSLHA, for read-in of particle masses via the SLHA. Earlier, PYSLHA would only read the mass given in BLOCK MASS for particles either unknown to PYTHIA or relevant to the (N)MSSM. Now, the mass of any particle can be read from BLOCK MASS. For instance the mass of a Z' (KF=32) could not previously be read in via SLHA. This is now possible. - Change in PYSLHA, for read-in of BSM states via the SLHA. The introduction of new states via the proposed SLHA BLOCK QNUMBERS (or, alternatively, BLOCK PARTICLE) has been made more general and especially particle names are formed in a more stable way. - Changes in PYSLHA, for read-in of decay tables via the SLHA. 1) Behaviour made more stable against inconsistent inputs. DECAY statements without accompanying explicit decay channels are ignored, unless the WIDTH is zero, in which case the particle in question is set stable. Previously, if a DECAY line contained a non-zero width for a particle but no explicit decay channels were given, errors would be printed and program execution aversely affected. 2) For a known resonance, DECAY entries are now simply placed at the end of the internal decay table, and MDCY(KC,2) and MDCY(KC,3) updated accordingly. Previously, the existing decays were first removed and all other particle decays moved down. This had the advantage of optimizing memory usage, but also meant that the decay channels of all particles following the modified one were renumbered, which could lead to problems for some applications. \ - An event that fills up and tries to overflow the whole PYJETS commonblock can enter a loop (posibly at times infinite) where many unsuccessful tries are made, especially for the new PYEVNW machinery. Therefore both PYEVNW and PYEVNT have been modified so that such overflow problems leads to the current event being thrown away and a new one generated. -----------------------------------------------------------------------