ttbar + jets error

[Generic User (don't modify these fields)]

Dear Sherpa Authors,

I have been using Sherpa to generate a sample of ttbar + jets events. The executable never completes x-section calculation however (I have left this to run for a week twice), and no events are generated. I suspect this is due to a lack of memory?

I have attached my Run.dat and the resulting output.

would you recommend doing anything differently to generate events of this kind?

Kind Regards, Chris Boddy

[Stefan Hoeche]

Dear Chris,

there are two possible solutions to your problem:

If you are happy with a pure narrow-width approach to the top- and W-decays, you can change the 'Decay'-tags in the (processes)-section to 'DecayOS', see

​http://projects.hepforge.org/sherpa/doc/SHERPA-MC-1.2.0.html#SEC56

Alternatively, you may want to help Comix optimise the phase-space integration by adding some loose cuts on the virtualities of intermediate particles. Adding a selector-section like the one below to your Run.dat should help.

(selector){

"m" -11,12 60,100; "m" 11,-12 60,100; "m" 5,-11,12 165,185; "m" -5,11,-12 165,185;

}(selector)

cf. also

​http://projects.hepforge.org/sherpa/doc/SHERPA-MC-1.2.0.html#SEC79

You can also specify a multiplicity-dependent error target, see

​http://projects.hepforge.org/sherpa/doc/SHERPA-MC-1.2.0.html#SEC67

This might be useful for higher multiplicities, which do not contribute a significant amount to the total cross section and need therefore not be known to as high precision as the leading order process.

Please let us know in case you encounter further problems

Stefan

[Generic User (don't modify these fields)]

Dear Stefan,

Thanks for your reply, I've tried several of your suggestions.

I'm now using DecayOS and set the integration error to 2% up to an 8 particle final state (apart from the the job Options are the same as those I posted before, also without the spin correlations included), to try to generate 93 93 -> 6 -6 93{2} but the calculation still doesn't complete. I've attached the new output. It seems to crash as soon as it finishes the ttbar + 1jet calculation.

Could you recommend anything further?

Kind Regards, Chris Boddy

[Generic User (don't modify these fields)]

Replying to user:

Dear Stefan,

Thanks for your reply, I've tried several of your suggestions.

I'm now using DecayOS and set the integration error to 2% up to an 8 particle final state (apart from the the job Options are the same as those I posted before, also without the spin correlations included), to try to generate 93 93 -> 6 -6 93{2} but the calculation still doesn't complete. I've attached the new output. It seems to hang as soon as it finishes the ttbar + 1jet calculation.

Could you recommend anything further?

Kind Regards, Chris Boddy

[Stefan Hoeche]

Dear Chris,

could you please try the input below? It works for me and completes the +2-jet integration within 30 mins.

Cheers

Stefan

(run){

CDXS_ITMAX 5000;

}(run)

(beam){

BEAM_1 = 2212; BEAM_ENERGY_1 = 7000; BEAM_2 = 2212; BEAM_ENERGY_2 = 7000;

}(beam)

(processes){

Process 93 93 -> 6[a] -6[b] 93{2};

DecayOS 6[a] -> 5 24[c]; DecayOS -6[b] -> -5 -24[d];

DecayOS 24[c] -> -11 12; DecayOS -24[d] -> 11 -12;

CKKW sqr(20/E_CMS); Integration_Error 0.02 {8}; End process;

}(processes)

(me){

ME_SIGNAL_GENERATOR = Comix

}(me)

[Generic User (don't modify these fields)]

Hi Stefan,

Thanks when I included that CDXS_ITMAX option the cross section calculation did complete and the events were generated.

However, the cross section appears to diverges as the number of jets in the final state increases.

I've attached the output showing the calculated cross section. I tried adding a selector

(selector){ NJetFinder 4 25.0 0.0 1.0 -1 2.5 }

which did decrease the cross section but it still increased as the number of jets increased.

Should we include any other selectors to ensure that the cross section is infra-red safe?

Many thanks, Chris

[Stefan Hoeche]

Hi Chris,

the cross sections which are calculated in Sherpa's initialisation step are not observable, cf. also ​http://projects.hepforge.org/sherpa/doc/SHERPA-MC-1.2.1.html#SEC11 especially the boldface section. The fact that these cross sections increase with multiplicity is not a problem, as they serve merely as seeds for event generation. They are reduced by Sudakov weights, attached in the parton shower step, such that the overall cross section of the leading order sample is recovered after all. For a quick review of how this works, please have a look at ​http://www.freacafe.de/dnl/talks/stefan_slac_10-04.pdf pages 12-17.

The line 'CKKW sqr(20/E_CMS);' in your setup file ensures that the individual cross sections do not diverge and that they can be combined into a meaningful inclusive event sample using the ME+PS merging algorithm. Any cuts added on top of this will make the sample more exclusive and are potentially dangerous because they can restict the phase space in an uncontrollable way. To be more precise: The NJetFinder criterion you implemented is _not_ a jet requirement at _particle level_, but at _parton level_, i.e. before showering. Showering can modify the parton kinematics and can therefore lead to a different final state, one which is not consistent with your previous cuts anymore. In other words: To generate a consistent ttbar+jets sample, you _have to_ remove the NJetFinder criterion, even if you are only interested in jets above a certain p_T (there are certain exceptions to this rule, but they do not apply to the ttbar+jets case).

Cheers

Stefan

[Stefan Hoeche]

This issue seems to be solved. If necessary, please reopen the ticket.

[Stefan Hoeche]

Milestone old deleted