Strong coupling constants are wrongly calculated

[cschwan]

This was originally reported here: NNPDF/pineappl#226, but I'm quite confident that this is a bug in Pineko, specifically these lines:

pineko/src/pineko/evolve.py

Lines 200 to 207 in f8c3261

	alphas_values = [
	4.0
	* np.pi
	* sc.a_s(
	xir * xir * muf2 / xif / xif,
	)
	for muf2 in muf2_grid
	]

which evaluate the strong coupling constant that do not correspond the renormalization scales given here:

pineko/src/pineko/evolve.py

Line 211 in f8c3261

xir * xir * mur2_grid,

I believe the patch is to change these lines as follows, with which I get the correct evolution (tested with CMS_TTB_5TEV_TOT):

@@ -201,9 +201,9 @@ def evolve_grid(
         4.0
         * np.pi
         * sc.a_s(
-            xir * xir * muf2 / xif / xif,
+            xir * xir * mur2,
         )
-        for muf2 in muf2_grid
+        for mur2 in mur2_grid
     ]
     # We need to use ekompatibility in order to pass a dictionary to pineappl
     fktable = grid.evolve(

[alecandido]

@andreab1997 can you try applying the patch above and recomputing, please?

[andreab1997]

Yes I can try but I don't understand how this can be a bug. @cschwan did you understand what is the problem?

[andreab1997]

Moreover, I got the correct predictions just removing the grid optimize which is completely unrelated to this.

[andreab1997]

So it seems that you were actually correct, applying that patch solved the problem. This is the new comparison
comparison_424.txt
As you can see there is only one problem at the beginning but I believe my eko to be bugged. Let me try to recompute it but I am pretty sure. For the rest everything seems to be fine.

[andreab1997]

No I was wrong, it seems that the ATLASDY2D8TEV is a real bug.

[cschwan]

I found the bug by printing the OperatorInfo struct passed to Grid::evolve on stderr for 1) pineappl evolve and 2) Pineko. For CMS_TTB_5TEV_TOT I find in both cases that

ren1: [
    5442.305429193529,
    7434.731381687921,
    10243.85467001917,
    14238.990475802799,
    19971.806922234402,
    28273.883344269376,
    40410.48232844362,
    58325.25318921733,
    85033.47534094655,
    125260.40013230646,
    186488.2133214792,
    280691.4902174795,
    427245.3808062111,
    657853.7431299294,
    1024996.5523865514,
    1616581.2577807596,
    2581663.421106388
]

So far so good. pineappl evolve calculates the strong coupling constant as:

alphas: [
    0.12190944075548889,
    0.11900892802232749,
    0.11617243718593613,
    0.11339821043907347,
    0.11068598682266634,
    0.1080343743829495,
    0.10544130425131869,
    0.10290650940538232,
    0.1004289992066687,
    0.09800656568331764,
    0.09563886186878709,
    0.09332518460075083,
    0.09106334734236118,
    0.08885286965181682,
    0.08669325984311221,
    0.08458246767160318,
    0.0825198335061508
]

As expected, this vector has the same size as ren1. But this isn't the case for Pineko:

alphas: [
    0.17812293619077232,
    0.17401253126774863,
    0.16998863674128645,
    0.16604972179241928,
    0.16219427022194413,
    0.15842078120174546,
    0.15472776993464313,
    0.15111376822948233,
    0.1475773249979303,
    0.14411700667916258,
    0.14073139759833034,
    0.13741910026440088,
    0.13417873561265617,
    0.13100894319682557,
    0.12790838133552054,
    0.12487572721733475,
    0.12190967696867401,
    0.11900894568808806,
    0.11617226745059381,
    0.11339839528520783,
    0.11068610112864455,
    0.10803417575788739,
    0.10544142870410592,
    0.10290668815016772,
    0.10042880081378519,
    0.09800663181814082,
    0.09563906455165323,
    0.09332500051837454,
    0.09106335918035456,
    0.08885307779316161,
    0.08669311123561792,
    0.08458243183468485,
    0.08252002918632481,
    0.08050490997306427,
    0.07853609777889299,
    0.07661263290205221,
    0.07473357216619243,
    0.07289798873031418,
    0.0711049718978489,
    0.0693536269251842
]

This can't be right, because the length of this vector doesn't match the length of ren1. Looking at the code above this makes sense: you calculate alphas at xir * xir * muf2 / xif / xif for whatever value muf2 is contained in muf2_grid, but as the corresponding scales you pass xir * xir * mur2_grid. So there's clearly a mismatch.

[cschwan]

For ATLASDY2D8TEV-aMCfast_obs_0 I get with pineappl evolve:

$ pineappl write --optimize ATLASDY2D8TEV-aMCfast_obs_0.pineappl.lz4 ATLASDY2D8TEV-aMCfast_obs_0-opt.pineappl.lz4
$ pineappl evolve ATLASDY2D8TEV-aMCfast_obs_0.pineappl.lz4 ATLASDY2D8TEV-aMCfast_obs_0.tar /tmp/deleteme.pineappl.lz4 NNPDF40_nnlo_as_01180
b      Grid       FkTable      rel. diff
--+------------+------------+-------------
0   2.7400418e0  2.7371347e0 -1.0609672e-3
1   2.7347464e0  2.7319391e0 -1.0265585e-3
2   2.7298912e0  2.7269958e0 -1.0606106e-3
3   2.7085291e0  2.7056548e0 -1.0612065e-3
4   2.6923000e0  2.6897803e0 -9.3589734e-4
5   2.6496577e0  2.6472263e0 -9.1763037e-4
6   2.5742308e0  2.5722173e0 -7.8215776e-4
7   2.3675544e0  2.3656406e0 -8.0836710e-4
8   1.9878036e0  1.9862178e0 -7.9778839e-4
9   1.5272002e0  1.5261000e0 -7.2043371e-4
10  1.0084506e0  1.0077076e0 -7.3677997e-4
11 4.8781048e-1 4.8746251e-1 -7.1332031e-4

This looks OK.