DIS: CC FONLL-A scale variations #27
Comments
|
We forgot to specify that the issue is related only to the quarks' channels, but it is not present for the gluon channel. |
|
@felixhekhorn @alecandido could you please be more specific as to where the missing factor 2 should be? Also, from the banner I see that when the evolution is initialised you have: but then when the DIS module is initialised you have: which are clearly incompatible. produces the following banner: where you can see that the ratio |
Factor of two...The missing factor of two is just the one detected from the numbers (see the diff spoiler in @felixhekhorn OP), there is no place in APFEL's code in which we could put it to fix the code itself, because instead we believe the scale variation to be missing for the MassiveZero charged current coefficient functions, as @felixhekhorn said. Where actually they are missing for us it's hard to tell, also because there is not an actual place where the quark channel, cc, massive-zero coefficient functions are implemented, being the same of the neutral current zero-mass. Our use of APFEL (and banner inconsistencies)The code used to produce the numbers is freely available in However if the first part of the banner it's really related to the evolution I would not care so much, and also this may be the reason why the value is actually inconsistent: we are taking care of deactivating the evolution, indeed being Statement ProofTo conclude we are pretty sure on the statement that the mentioned scale variations are somehow missing in APFEL implementation, indeed we used the very same code for APFEL's call and modified our code deactivating the scale variations we are speaking about, and the results coincide perfectly, and we this setup we would have the very same discontinuity that we displayed in APFEL's numbers in the OP. |
|
@alecandido The file |
|
Sorry @vbertone while I have nothing against 3.17, I really believe we are doing the same thing, (even if still at NLO), instead I'm not sure about the implementation, because to me it's quite hard to understand completely what's going on inside APFEL in details. For this reason I tried to understand the content of the file you kindly pointed, and I made this sketch:
Since what we are interested in is not On the other hand I think you can admit that the bump we pointed out it's quite a strange behavior, and it's really strange that it happens actually at the charm threshold (note: we are reporting values for So: are you able to compute that quantity with APFEL (the points we reported) for |
|
Hi @alecandido. Your understanding of the ZM sector of It is surely possible that there is something that is not correctly implementend but the bump that you observe (that to my understanding is at the bottom threshold) has a possible explanation related to the fact that in APFEL, when you change the factorisation scale w.r.t. Q, you are also implicitly moving the heavy quark thresholds and this introduces a discontinuity at the threshold also at NLO, which does not automatically means a bug. Of course, if this is the issue there are ways to avoid it. If you agree, I would do a benchmark using the toyLH set that is implemented in APFEL. I'll send you the numbers asap. |
|
@alecandido while working on this I realised that, since you don't want to consider the evolution, using the |
|
@vbertone first of all: thank you for your help in the investigation. Then, keeping the foces on the
I agree with you that a bump is not a bug itself, but it is a quite clear clue that something is happening, and I would call it a bug if no one is able to explain that behaviour (and maybe you are, so let's keep going).
For this investigation a Q2 independent PDF would be really better, so please manually install the LHAPDF file that I'm attaching (it is an sbar-only version of toyLH, in the LHAPDF format). You can load it simply specifying |
|
(in case you ever need it, even if I don't think so, the |
|
@alecandido here is a plot for F2charm as a function of Q at x = 10^{-3} using the |
|
Hi @vbertone, thank you for the plot you produced.
It's not that we simply noticed something weird we didn't understand and we just claimed a bug. We worked for a full day on it, checking the implementation in The problem we faced it's that a huge code written by someone else, with a not so familiar design, may be difficult to occasionally explore, so we tried to find the bug (or inconsistency) ourselves, we failed, and we simply reported that the numbers produced by APFEL are consistent with a wrong implementation of If possible we would be pleased to find the bug in APFEL, because of two reasons:
So if not calling it "a bug" let's call it "a problem". Of course, I will never be sure of what it's right, but there is also a confidence level for the SM itself... |
|
There is also another thing I didn't consider in my previous comment: perspective. From our perspective it's rather easy to change our code and silence a channel, but not being so familiar with APFEL this is not true for the other side. This is why we are using an uncommon PDF, in order to select some information without having to touch the code. Looking at the mirror: probably for you it would be easier to be sure not to use the gluon channel running |
|
Dear @alecandido, I fully understand all you wrote and I would be glad to help you understand the reason of the discrepancy taking your perspective as much as possible. As a matter of fact, I did use the PDF set you provided me with for producing the plot I shared with you, which effectively turns off the gluon channel. Unfortunately, though, I cannot see any bug in APFEL related to factorisation-scale variations in CC F2c, as you claim. I'm not sure there is much more I can do on my side. |
|
I see your point @vbertone, and it's perfectly true that you did your best with the information we provided. It's evident from your plot that you're not experiencing any discontinuity while we are, so it's undeniable that we are doing two different things.
Actually doing the second should be sufficient, because we should be able to run it on our own, with your current setup and with the other one, but a further couple of eyes not conditioned by our experience it's valuable. |
|
Hi @alecandido, here is the code I've used for the plot: Let me know if you find any problem. |
|
Sorry not to reply soon (and not to be replying still). We are still interested in the investigation, and at a first sight your code it's actually reproducing the scenario we described, but right now we need some more time for doing something else, so we will come back soon on this issue. The only difference I spotted it's that you are correctly using the masses of the PDF I gave you, the only ones you had available, but instead we are actually using different values for the masses: so this shift the thresholds, e.g.: you are using a value of I'm still not giving you the full runcard only because the run was not on my machine and I need to retrieve it, maybe I could also reproduce from the runner we are always using, but before adding further entropy I would prefer to wait a little more (the excerpt I gave you belongs to the part we never change so I'm sure, but the most relevant information belongs of course to the other part). |
|
@alecandido I think this explains the discontinuity that is the genuine FONLL discontinuity at the charm threshold that is usually tamed by introducing the damping factor (that is off here). The reason is simply that F2charm (as well as FLcharm and F3charm) in the ZM-VFN scheme is by assumption set to zero below the charm threshold and switches on right above it (according to the definition of ZM-VFNS given by Collins, Wilczek, and Zee (CWZ)). Since the scale at which PDFs are computed is |
|
I know that the choice of charm and bottom masses it's not the best one, we set it before start working on scale variations, and as soon as we started working on them intensively we realized it and thought to switch. For the moment we are keeping them not to break some tests, but I'm sorry for the confusion they generate. Instead I don't understand the part on
So I still don't see where the discontinuities should come from in FONLL-A. |
|
Actually I produced the same picture of you just changing the masses: And here is the picture: Notice the discontinuity is only in the scale varied one, but also the other it's crossing all the thresholds (charm and bottom of course, not top). |
|
I think that the misunderstanding is araising from confusing the charm mass ( |
|
Probably you are right and having experienced only the NLO I'm relying too much on the continuity, even if I'm aware of that being an accident and not a rule, as you can deduce from previous FONLL paper citation. But in this particular context (NLO CC F2/3 quark channel) the Maybe I'm still making an error somewhere, and this may explain the discrepancy, but if what I said it's correct the discontinuity in APFEL it's not explained. |
|
As I said, it is the |
|
As I said before, it happens that for this particular case, since the |
|
In particular we realized that for the quark channel at NLO DIS FONLL it's trivial:
If you try to run the two options you will obtain different numbers, but the discrepancy it's only due to the different value of |
|
@vbertone can you point us where the Looks like they should be the FONLL-A |
|
Same thing for massive SC3mm, but the |
|
Sorry, I'm try to catch up. What do you mean by "Is it regular that SC2 is evaluated at a different point than SC3?" |
|
If you go to the pointed lines you will see: SC2m0CC(igrid,jxi,3,1,beta,alpha) = and SC3m0CC(igrid,Nf_FF,3,1,beta,alpha) = You see that the second argument it's not the same, and being the second argument |
|
I see, I don't remember this by heart. I need to check. |
FONLL-Ascheme (i.e. NLO) forxiF!=0there is the scale variations for the massive-zero coefficient missingnon-scale variated
scale variated
diff
APFEL banner
WARNING: FONLL-A is a VFN scheme ... setting VFNS PDF evolution WARNING: FONLL-A is NLO scheme ... setting NLO perturbative order WARNING: if there are external grids they cannot be locked ... unlocking subgridsWelcome to
/// //// //// ///_/ _/
_/ / / / / / /
//// //// /// /// _/
_/ _/ _/ _/ _/ _/
/ / / / //// ////
_____v3.0.4 A PDF Evolution Library, arXiv:1310.1394
Authors: V. Bertone, S. Carrazza, J. Rojo
Report of the evolution parameters:
QCD evolution
Space-like evolution (PDFs)
Unpolarized evolution
Evolution scheme: VFNS at N1LO
Solution of the DGLAP equation: 'exactalpha' with maximum 6 active flavours
Solution of the coupling equations: 'exact' with maximum 6 active flavours
Coupling reference value:
Pole heavy quark masses:
The matching thresholds coincide with the physical masses
muR / muF = 1.0000
Allowed evolution range [ 1.0000 : 10000.0000 ] GeV
Fast evolution enabled
Initialization of the evolution completed in 5.254 s
Report of the electroweak parameters:
Mass of the Z = 91.188 GeV
Mass of the W = 80.398 GeV
Mass of the proton = 0.9380 GeV
sin^2(thetaW) = 0.2313
GFermi = 1.16638E-05
| 0.9743 0.2253 0.0035 |
CKM = | 0.2252 0.9735 0.0410 |
| 0.0086 0.0403 0.9992 |
Z propagator correction = 0.00000
Report of the DIS parameters:
Computation in the FONLL-A mass scheme
Charged Current (CC) process
Scattering electron - proton
muR / Q = 1.0000
muF / Q = 0.5000
(No scale variation in the evolution)
Target Mass corrections disabled
FONLL damping factor disabled for all heavy quarks
Intrinsic charm disabled
The text was updated successfully, but these errors were encountered: