[Clascomment] OPT-IN: Measurement of Unpolarized Cross Sections and Polarized Cross Section Differences for Deeply Virtual Compton Scattering

[Sebastian Kuhn]

This is an extensive paper with important results, which I will need some time to digest. However, I wanted to quickly convey a couple issues that caught my eye right away:

1) The binning is introduced early on (p. 4) and referenced frequently later on when discussing results. It is shown in Fig. 5, yet the text (line 270) refers to Table II which only appears at the very end and shows bins in x and Theta, NOT x and Q2.

2) The style of some figures is quite unconventional. In particular Figs. 9-16 all have an ugly grey box with little useful information on the top right - I bet the journal editors won't like those. Also, not all panels in Figs. 13-16 show the promised fit functions. Finally, I would prefer having all systematic error bands as straight horizontal bands, as opposed jumping around with the data points.

3) I am worried about the (lack of) description of the elastic normalization. First of all, you don't compare the overlap (or lack thereof) of the proton and electron phase space for the 2 channels, so it's not clear if the "loss" for DVCS and ep should be expected to be equal. You also don't mention whether you apply radiative corrections to the elastic data. You don't discuss the event selection cuts to define elastic events. You also don't mention whether you simulated elastic events as well and used the same procedure to correct them for acceptance and detector efficiency (I presume so but it should be said). Finally, you use the average (over sectors) normalization for the DVCS data (btw, the difference between 0.926 for the "global average" and the value 0.906 for µ in Eq. 22 isn't well motivated), and then you apply a 5% uncertainty due to sector-to-sector variations. However, you do not comment on or include the rather large fluctuation from Q2 bin to Q2 bin (+/-20% in Fig. 
 20) or the reason why the first data point is off even further (and can be neglected). Overall, the reader may well doubt the 5% value for the uncertainty.