I re-built Genie without --enable-vhe-extension, I edited 
config/UserPhysicsOptions.xml to have GVLD-Emax = 4000 (GeV), and re-generated 
splines with -e 4000.

The disease is similar:
Now the total cross-section goes crazy at E ~ 3100 GeV (still nu_mu on C12), and 
similar errors and warnings appear in the output. Those same splines now have a 
zero at E=4000. They each have at least 4 valid points above E=2000, so 
interpolating the spline up to E=2000 will be fine; for now I'll provisionally 
assume it's OK up to that energy, and explore things further.

Tom Roberts


On 6/10/11 6/10/11 - 10:17 AM, Tom Roberts wrote:
> Generating the free-nucleon splines and then the nuclear splines went OK, but
> there are puzzles.
>
> The nu_mu total cross-section goes crazy just above 1500 GeV (see plot). All I
> did was modify config/UserPhysicsOptions.xml to have GVLD-Emax = 2000 (GeV), and
> use gmkspl with -e 2000 to generate all the splines (freenuc first, then
> nuclei). I did not re-build the code, so I still have --enable-vhe-extension.
>
> Looking at the output from this command:
> gmkspl -p 12,14,16,-12,-14,-16 -t 1000060120 -e 2000 -o 1000060120.xml
>
> I see many hundreds of errors like this:
> 1307662008 ERROR GridDim : [n] <UnifGridDimension.cxx::ClosestPoint (109)> :
> Non-positive step (=0)
>
> I see many hundreds of warnings like this:
> 1307662008 WARN FMap : [n] <FunctionMap.cxx::Value (74)> : No point with UId = 8
>
> And I see six errors like this, all for anti-neutrinos:
> 1307662036 ERROR IntLst : [n]
> <QELInteractionListGenerator.cxx::CreateInteractionListCharmCC (206)> :
> Returning NULL InteractionList for init-state: nu-pdg:-14;tgt-pdg:1000060120;
>
>
> Note that I saw similar errors before changing GVLD-Emax to 2000 (i.e. when it
> was 300). Moreover, the spline file was written, and seems to work (below 1500
> GeV).
>
>
>
>
> Many splines have their last 4 points at E={1290,1493,1728,2000}; some are
> E={932,1203,1551,2000}.
>
> Many of the splines seem to be approximately constant in the range 1000<E<2000,
> but some are increasing ~ proportional to E. I strongly suspect the problem is
> the following splines -- they have their largest value in the next-to-last point
> E=1728 and a zero at E=2000, which is surely wrong:
> genie::AivazisCharmPXSecLO/CC-Default
> /nu:-12;tgt:1000060120;N:2112;q:-1(s);proc:Weak[CC],DIS;charm:incl;
> genie::AivazisCharmPXSecLO/CC-Default
> /nu:-12;tgt:1000060120;N:2112;q:-3(s);proc:Weak[CC],DIS;charm:incl;
> genie::AivazisCharmPXSecLO/CC-Default
> /nu:-12;tgt:1000060120;N:2212;q:-1(s);proc:Weak[CC],DIS;charm:incl;
> genie::AivazisCharmPXSecLO/CC-Default
> /nu:-12;tgt:1000060120;N:2212;q:-3(s);proc:Weak[CC],DIS;charm:incl;
>
> There may be others with this disease, I stopped looking. These are among the
> largest cross-sections at E=1728. So it looks like there is a bug that affects
> the highest-energy point. Given the nature of cubic splines, it seems reasonable
> that this erroneous zero starts affecting the values around E=1500, as in the plot.
>
> I wonder if I can work around this by setting GVLD-Emax to 4000 but only using
> it up to 2000 (to get at least 4 points above the highest energy used, so the
> value of the highest point is irrelevant). Or is there something more serious
> happening?
>
> Does anybody know what is going on?
>
> [On my 4-core Mac Pro, it took 2.5 hours to generate the freenuc
> splines (using 3 cores), and then 2 hours to generate splines for
> all 6 neutrinos on 143 nuclei (using 4 cores). I put each nucleus
> in its own file and have code that only loads spline files for
> the nuclei actually used, greatly speeding up initialization.]
>
>
> Tom Roberts
>
>
>
> On 6/9/11 6/9/11 - 3:51 PM, Constantinos Andreopoulos wrote:
>> Dear Tom.
>>
>> On 9 Jun 2011, at 20:55, Tom Roberts<[log in to unmask]> wrote:
>>
>>> Costas:
>>>
>>> Thank you. That sounds simple enough.
>>>
>>> I believe that after changing config/UserPhysicsOptions.xml I must
>>> re-generate all of the splines. But the posted gxspl-freenuc-v2.6.2.xml only
>>> goes up to 200 GeV. So how do I generate the free nucleon splines? Do I just
>>> do this:
>>> unset GSPLOAD
>>> gmkspl -p 12,14,16,-12,-14,-16 -t 1000000010,1000010010 -e 2000 \
>>> -o freenuc-v.2.6.2.xml
>>> and then set GSPLOAD=$PWD/freenuc-v.2.6.2.xml for the other targets?
>>
>> Yes.
>>
>> Generating the free nucleon splines takes long time (integrating the
>> differential cross section for every modelled process -about 100 or so- for
>> every energy for every initial state).
>> What I do is to split the task and submit multiple jobs at a batch farm. Then
>> I combine the XML outputs using the gspladd utility (see manual).
>> See in src/scripts/production/batch (or in
>> http://projects.hepforge.org/genie/trac/browser/trunk/src/scripts/production/batch
>> ). The script you want is named submit-xsec_freenuc.pl .
>> I am using it at the RAL/PPD Tier2 which is using PBS. But the script also
>> supports LSF so you should be able to adjust it easily and use it at FNAL.
>>
>> With the gmkspl utility you can use the "-n" option to set the number of knots
>> in each spline. By default gmkspl uses logarithmically spaced knots, starting
>> from the energy threshold of each process. It uses (by default) 15 knots per
>> decade. Use a sensible number to avoid getting an enormous XML file,
>> especially if you plan to add multiple nuclear targets.
>>
>> cheers
>> Costas
>>
>> --
>> Dr Constantinos Andreopoulos
>> Staff Scientist
>> STFC, Rutherford Appleton Lab
>> http://costas.andreopoulos.eu
>>
>> Sent from my iPhone
>>
>>
>>>
>>>
>>> Note I will be running Genie for the neutrinos from 2 TeV muon decays, so
>>> different neutrino types and energies are all mixed together. (I am
>>> interfacing Genie to another simulation program that handles everything
>>> except the neutrino interactions, tracking neutrinos with greatly enhanced
>>> interaction lengths in regions of interest.)
>>>
>>>
>>> Tom Roberts
>>>
>>>
>>> On 6/9/11 6/9/11 - 1:23 PM, Costas Andreopoulos wrote:
>>>> Just increase in the limit at the very end of the
>>>> config/UserPhysicsOptions.xml .
>>>> I do not think that the simulation quality deteriorates very fast in high
>>>> energies.
>>>> Results at ~2 TeV should be sensible. But, I've never run the simulation at
>>>> 2 TeV.
>>>> Let me know if there are problems.
>>>>
>>>> You don't have to use the --enable-vhe-extension. This is experimental and,
>>>> currently, it just adds a Glashow resonance generator (PeV scales).
>>>> This option should (but it currently doesn't) adjust the DIS generator for
>>>> use in higher-energies
>>>> (use a lower minimum Bjorken x cutoff value, use a PDF set other than the
>>>> Bodek-Yang which is tuned
>>>> to high-x low-Q2 data, switch-on the propagator term in the DIS cross
>>>> section etc... ).
>>>>
>>>> Also if you are running GENIE at 2 TeV then (for reasons of computational
>>>> efficiency)
>>>> you may want to disable all modes other than DIS.
>>>>
>>>> Best,
>>>> Costas
>>>>
>>>> On 8 Jun 2011, at 22:13, Tom Roberts wrote:
>>>>
>>>>> I want to investigate neutrino backgrounds from a muon collider, so I need
>>>>> neutrino energies up to about 2 TeV.
>>>>>
>>>>> What is the status of Genie's very high energy extension? How do I use it?
>>>>>
>>>>> In my .configure command I included --enable-vhe-extension. But in gmkspl
>>>>> when I try to go above 300 GeV it says "Refusing to exceed validity range:
>>>>> Emax = 300".
>>>>>
>>>>> As I am investigating backgrounds, I am primarily concerned with the total
>>>>> cross-section, and not the details of the secondaries (as long as they are
>>>>> not crazy). I realize there are no experimental data at such high energies,
>>>>> and there are theoretical uncertainties.
>>>>>
>>>>> Is there anything simple I can do that will give results up to 2 TeV that
>>>>> aren't crazy (even if they are not known to be accurate)?
>>>>>
>>>>>
>>>>> Tom Roberts
>>>>