DVCS with Recoil Analysis Page

Aus HERMESwiki
Zur Navigation springen Zur Suche springen

Page maintainer: Caro, Sergey, Jennifer, Irina, Maru, Aram, Jonathan, Morgan

Workingpage.jpg This is a working page which will continuously change. The content is not reviewed and is intended for people working on this specific topic.

About this page

This page was created in May 2010 for all people working on a DVCS analysis including Recoil Detector information. Feel free to add your private section and to add, correct or update items in the common sections.

  • The chapter head line goes between == ==
  • Add more items
  • My first own item
  • I like this
  • Add a weblink like this
  • Cross out items

BCA with Recoil

Corrected

Combined fitting is working!

.

To Do list for the analysis

  • understand xml differences (Contacted Sergey)
  • learn qplot (Contacted Aram)
  • Do MC data selection


Done

  • Changed infrasturcture to root files, to accelarate plotting
  • Corrected DIS/DVCS numbers after the Talk on Wednesday
  • New data production with changed xml started.

Comparison between method 3 and method 7

Drafts of the Release Report

Changes:

  • See Caros coments
  • Section2: Data Selection, Gaps for MC
  • Section3: Statistics part filled, data qualitiy, time period cuts
  • Section3: Crosscheck plot ALU with forward redone
  • Section3: AC with recoil plot added
  • Spellchecked

Comments:

  • Still fighting the bibtex in lyx. It hates me, but i am stronger...

Comments on the Release Report

The report has to be submitted to the management latest by May 20. I suggest you send a version to the other authors of this report (as included on the title page) in advance.

Caro's comments on version 0.8, written April 29, 2013

Bibliography

There are many mistakes in the library and citations are often not complete / clearly to identify. Please check bibliography carefully and in particular take into account these comments:

  • General: document titel either in italic OR in "". Don't write all letters in capital as in [Ji04].

* [Bur10], [Zei05]: add that these documents are their PhD theses!

  • [DZ08] What is this? A release report or an internal note or... Specify! Comma after Lu, compton --> Compton
  • [Etzl12] add exact date

* [Hil99] Gernators --> Generators

  • [JB10] and [JB11]: author lists not complete. Check and add.
  • [Mae10] Rebecca's last name is Truty. Mae is her middle name.
  • [Rie12a] and [Rie12b] my last name is spelled Riedl. Add exact dates.
  • [THC09] journal citation is not complete. First page number is missing.
  • add DC92 publication
  • [Zei05] There are a couple of mistakes in the title. Check and correct.
Title page and abstract

PTitle page: add spaces before Hrachya and Caroline

Please rephrase abstract entirely: The data set of the years 2006 and 2007 collected using the recoil detector is analyzed. The first part of the 2006 runs was previously excluded from the analysis since one part of the recoil detector, the Silicon Strip Detector (SSD), was not commissioned, not operational or even dismounted. Now these data are available with a track-reconstruction method that only uses the Scintillating Fiber Tracker (SFT). This method allows for the first time the extraction of the beam-charge asymmetry in Deeply Virtual Compton Scattering (DVCS) using the recoil detector. This report summarizes the studies using kinematically complete event reconstruction.

Do not use new paragraphs within the abstract.

1
  • everywhere Hermes --> HERMES, Hera --> HERA
  • line 4, delete "taken", was --> were (data is plural)
  • line 5, "the Hera ring" --> "HERA-e"
  • line 6, filled --> operated
  • 2nd paragraph: add after (BCA): "in Deeply Virtual Compton Scattering (DVCS)" and "with" --> from"
  • 3rd paragraph, 1st line: input --> constraints
  • 3rd paragraph, 2nd line: "[...] Factor (CFF) \mathcal{H}" and "Distribution (GPD) H". The BCA is sensitive only to CFF/GPD H. Please read about this in the literature if you are not familiar with this.
  • "Burkardt" is actually correct... please correct once more
2
  • 2.1 (pg = paragraph)
    • 1st pg: "in a pure way" --> "with negligible contamination by associated processes, such as ep-->e\Delta^+\gamma"
    • 2nd pg: sample --> set; "data points from" --> "space points in"
    • 5th pg: was performed --> was first performed. This sentence contains almost too much information. I would have made a short itemization, showing the flow of your analysis.
  • 2.2
    • 1st pg: I would not cite [JB10] here, I would cite the DC92 publication (together with the release reports & analysis updates, as done already.) The reason: the DC90 analysis is older and I recall some differences to your analysis, e.g. in the treatment of the photon energy.
    • Table 2.1: make figure larger (over entire page width) and center figure
    • g1T rack.iSelect&... remove space between T and r
    • Caption of Table 2.1: ALU --> BSA
  • 2.1
    • Figure 2.1: there is no (0011 = without SSD) M3 track in this sketch, which I think is unfortunate. Can you add one and and indicate that the red, blue and green tracks are M7 tracks, while the newly added one (dotted; different color, ..) is M3.
  • 2.4
    • TODO: Mx2 plot with MC and experimental data; M3, M7, pure, tc, thetagammagamma; yield fractions
    • you don't call figure 2.3 in the text
    • middle of page 13: "corrected momentum values": I would more precisely write "3-momentum" here since the correction involves also the angles. Also, please add a sentence that the corrected kinematic values are NOT used in the analysis for the calculation for composed kinematic variables like thetagammagamma, tc, etc. This is not clear from your description and has lead to confusion in the (DC92-)past.
  • 2.5
    • You don't call figures 2.4 and 2.6 in the text
    • Caption of Fig. 2.4: above --> top; below--> bottom
    • Caption of Fig. 2.5: add "(top)" after method 3 and "(bottom)" after method 7
    • Caption of Fig. 2.6: left --> top (?), right --> bottom
    • line 2: "loss of statistics" --> "inevitable loss of statistics"
    • line 3: "the a" --> "a"
    • line 4: "less exact" is not very precise. Can you write a little more why? Less exact in which measure?
    • lines 4, 7: then --> than
3
  • 3.1
    • Section 3.3. is referred to twice. 3 lines later: both section 3.4. and 3.2. are referred to. Is this correct? If yes, put an "and" in between. Be consistent with section callings; since you use "Fig.", use "Sec."
    • Table 3.1: did Maru not provide pure numbers for M7? If not, put in my or Sergey's numbers at least, but fill the gap. Who did the cross check for pure M3? We need these numbers!
    • Figure 3.1
      • Caption: add which data sample this is. Pure? (I.e. which cut is on event level)
      • Call figure in the running text and shortly describe what is shows. Which conclusion(s) do you draw from the figure?
    • Table 3.3
      • Which cut is on event level? It makes no sense to quote number of recoil tracks without at least requiring a DIS lepton, since the recoil-track multiplicity highly depends on the background conditions.
      • N_c confused me. Please specify in the caption that N_C><0 refers to the beam and not the track charge.
      • Either complete the lumi numbers, or skip (my proposal)
    • Text
      • "The tables..." --> "Tables..."
      • "event number" --> "event numbers"
      • line 3: "using" --> "and"
      • add after "missing mass method" in line 2: "(unresolved sample)" and after "recoil detector" in line 3: "(pure sample)"
  • 3.2
    • " SSD unmounted": identify and quote run numbers (I would use the Delta E plots for the SSD)
    • "SFT not working properly": this is not enough. Latest for your thesis you will need more details.
    • line 4: [ref]: replace by correct reference
    • "Table sec 3.2" cite table properly
    • line 4: "pure and unresolved data" --> "pure and unresolved data samples"
    • "Captions of figures 3.2, 3.3, 3.6, 3.7: explain in the first caption what is shown: Energy deposition of space points in the SFT used for M3 tracking. Three different momentum bins are shown. Top panels mean of energy deposition, bottom panel its width. The figure shows the time dependence in bins of 500 or 1000 runs (I don't recall, check on the webpage). In the other captions you can write "Refer to caption of Fig. 3.2 for more details.""
    • In caption of Fig. 3.3 you have to add that the top panels show protons, the bottom panels pi+.
    • In caption of figure 3.6, why do you 3x show pi+ and 1x protons?
    • Caption of figure 3.7: add "left for protons, right for pi+."
    • Call all the figures in the text to illustrate your findings!
    • Line 2 on page 22: this period --> these periods (it's more than 1, right?) No?
    • last word on p23: necessary --> necessarily
    • Top of p24: this runs --> these runs. I don't understand the argumentation. If you extract the BCA in a non-combined fit, which isn't sensitive to the beam polarization, why do you have exclude the runs?
    • "Statistics of the left data" : add content or delete headline. Better write "data after extra quality cuts" rather than "left data".
  • 3.4
    • Make all amplitude plots full width. They are too small.
    • bottom of p28: units are missing. Either add everywhere or write that all numbers in this section are in units of [Gev^2]
    • bottom of p29: "selected data sample" --> "unresolved data sample without using recoil-detector information"
    • Last 2 lines of p29: the forward spectrometer is a complex system and it is not calibrated as one piece. What has been recalibrated is the calorimeter, which is important for a DVCS analysis and in particular the stability of the Mx2 peak. Also, a problem with the pedestals in a certain time period has been fixed after the DC90 publication .
    • middle of text on p30: explain and / or refer to (Sec. 3.3): what is the 4-parameter, what is the 13-parameter fit?
    • Captions of figure 3.10 and 3.11: add "unresolved data sample" somewhere
  • 3.5
    • CAgain, a figure is not part of the running text. You have to call all figures in the beginning: "The following cross checks were performed between different analyzers. In Figs. X and Y, the results of Irina and Erik are compared ..."
    • page 33: the two analyzers are shown in Figs. 3.12 and 3.13; 3.14 shows the comparison with the published analysis.
    • Figure 3.15:
      • Add that this is the pure sample. "method 3 and method 7" alone are not enough.
      • I would find it useful a) to extend all three panels over the full width of the page and b) to normalize each blue / red histo area to 1 and then make a ratio. Only then you will have a real message. Ideally you do not want to observe any trend in the ratio plot.
    • Figure 3.16: add that both sets of data points are extracted from pure data samples
      • p35: "... in good agreement for the pure samples based on method 3 and 7, as shown in Fig. 3.16"
      • Really? Good agreement? You have to add more discussion here. This is an important point. You cannot NOT mention the possible difference seen in the lowest t-bin of the leading amplitude. Please do the following:
        • the average <t> for the lowest t-bin for a) M7 and b) M3.
        • in your event list, remove for the M7 sample all events which would not make it to the SFT (cut should be around p=250 MeV/c). Then perform the ML fit again and add it, in green or any other color, to Fig. 3.16.
        • alternatively, extract the M7 amplitude using ONLY the events you find with M3. The problem is that we are presently not looking at fully correlated event lists.
  • 3.6
    • Bottom of p35, add after "amplitudes": "extracted from the pure sample based on kinematic event fitting using M3 recoil tracking. Also shown in the figure are the results from publication [ref] extracted from the 2006/2007 data set (?? or is it the 1996-2007 result? please check) with 13-parameter fit without the use of recoil-detector data."
    • Caption of figure 3.17
      • Beginning, add: "Red Squares: "
      • add "of the pure sample" after after "BCA"
      • Add: "The blue circles represent the results extracted from the 2006/2007 data set with 13-parameter fit without the use of recoil-detector data.
    • Discuss figure 3.17 here in the text. Describe the different behaviors of the two leading amplitudes, in particular the t-dependence.
4
  • 4.1 (pg = paragraph)
    • 1st pg: The combination of "background process systematic" makes no sense; is this maybe an orphaned copy and paste?
    • 2nd pg: after "alignment file", add "of the forward spectrometer"
    • 6th pg: replace "each fitted bin from the gmcDVCS simulation" by "each bin the asymmetry amplitudes are extracted from reconstructed MC data." (gmcDVCS has not been introduced at this place, and "fitted" is not relevant here and probably rather confusing.)
    • Last pg on p37: add after "generated": "with gmcDVCS" (gmcDVCS is the name of the MC generator, gmc stands for "generator Monte Carlo")
    • Caption of figure 4.1: delete "elastic"
    • p40: the text needs some polishing. I offer to take your Latex file at some point and do this (here and at other places, after the next iteration)

** p40: add a summary that the BCA amplitudes were not corrected for any background. This conclusion is somewhat missing here.

  • 4.3 Summary of systematics: add text!
5
  • Discussion of results and plots proposed for Release: add plots
    • I would discuss the results in section 3, not here. Just put the plots here.
    • Which plots do you want to release? How many?
Tasks left

Bibliography

  • Use Sergeys bibtex file
  • Check bibliography carefully

3

  • 3.1
    • Table 3.1: did Maru not provide pure numbers for M7? If not, put in my or Sergey's numbers at least, but fill the gap. Who did the cross check for pure M3? We need these numbers!
  • 3.2
    • I don't understand the argumentation. If you extract the BCA in a non-combined fit, which isn't sensitive to the beam polarization, why do you have exclude the runs?-> we only loose e+ statistics here. Need to decide if we want to have a time period here which is only suspicious by the polarization.
    • "Statistics of the left data" : add content or delete headline. This sub chapter only contains a table. put some text here.
  • 3.5
    • Figure 3.16: add that both sets of data points are extracted from pure data samples
      • p35: "... in good agreement for the pure samples based on method 3 and 7, as shown in Fig. 3.16"
      • Really? Good agreement? You have to add more discussion here. This is an important point. You cannot NOT mention the possible difference seen in the lowest t-bin of the leading amplitude. Please do the following:
        • the average <t> for the lowest t-bin for a) M7 and b) M3.
        • in your event list, remove for the M7 sample all events which would not make it to the SFT (cut should be around p=250 MeV/c). Then perform the ML fit again and add it, in green or any other color, to Fig. 3.16.
        • alternatively, extract the M7 amplitude using ONLY the events you find with M3. The problem is that we are presently not looking at fully correlated event lists.

4

  • 4.1
    • p40: the text needs some polishing. I offer to take your Latex file at some point and do this (here and at other places, after the next iteration)
  • 4.3 Summary of systematics: add text!' TIME DEPENDENCE?

Associated asymmetry (DC93)

Drafting

Release


Pure sample (DC92)

Publication October 2012

Versions - Popular summary - Directory for supporting plots - Drafting working directory - Mailing list archive of DC92

Release April 2011

Studies related to Photon Detector

2010

List as of May 2010:

  • Master comparison (single-charge beam-helicity asymmetry) to 2006/2007 Feb.2010 released result (2 beam charges): use identical cuts, but allow only those events for those the expected kinematics of the recoiling proton is within the acceptance of the Recoil, but without requiring Recoil tracking information. Do this also on Monte Carlo, maybe even start with MC.
  • Pure elastic amplitude (BSA) from kinematic fitting, by applying chi2<CUT cut. Involve as much statistics as possible.
  • Background amplitude from kinematic fitting, by applying chi2>CUT cut (+ Mx2 cut in order to be in exclusive window). Use all traditional DVCS cuts in order to be able to compare to the Feb.2010 released result.
  • Associated asymmetry.
  • Beam-charge asymmetry. Involves SFT-only tracking.

Some notes after the exclusive meeting July 20, 2010. (CR)

"DVCS" means in particular single-charge beam-helicity asymmetry amplitude, but it can also be extended to beam-charge asymmetry amplitude once SFT-only tracking information is involved. In the latter case, in addition charge-difference and charge-average beam-helicity amplitudes can be extracted.

0. Standard DVCS amplitude: traditional DVCS analysis with spectrometer-only.

1. DVCS + proton expected in RD (spectrometer-only information). Corresponds to the Master comparison in the list above, in case 2 beam charges are used.

2. DVCS + proton detected in RD (only up to PID, without kinematic constraints. I.e. if there exists any proton in the event).

3. DVCS + proton reconstructed in RD (including kinematic constraints, e.g. kinematic fitting, and possibly using the Photon Detector as veto). Corresponds to the clean elastic amplitude in the list above.

  • 0. minus 1. gives the information how much the DVCS amplitude changes only due to the modified acceptance (modified phase space)
  • 1. minus 2. gives the information how the amplitude is affected due to the inefficiency of the RD.
  • 2. minus 3. gives the background amplitude (associated plus some semi-inklusive).
    • need to properly propagate statistical uncertainties of asymmetries
    • contamination by elastic process can be corrected for - both fraction and asymmetry are known
  • In the future, there is the plan to directly select the associated process by kinematic fitting using the pi0 and the Delta+ as constraints. In addition: PD

Caro

To be edited only by Caro.
Last update: 17.12.2010.

Goal: analyse 06e1 and 07c2 hydrogen data. Separate elastic and non-elastic.

Analysis Code

  • For new analysis framework, see here
  • My analysis C-code is located in
/user02/criedl/recoil

All my plotting kumacs (PAW) are located in 

/user02/criedl/recoil/kumacs
  • Code log, table of analyzed data samples, printed out event list
  • Remarks.
    • As it is programmed now (Aug.2010), the correction formula for leptons is applied to all photons - be careful when making a pi0 analysis!
    • For hrc-format data (i.e. not uDST), the code is not yet upgraded to apply the correction formula only for --newcalo, also, the E_uncorr is not assigned!!
    • If I want to study Q2<1 events, I have to release the condition when my small ntuples are filled!

Kinematic Fitting

  • Technical description
    • The source of Sergey's kinematic fitting routine (C-based) is stored on geordi: /group02/rcoilgrp/Software/KinFit/
      • The user needs to add the path /group02/rcoilgrp/Software/KinFit/include/ to his Makefile and compile the code together with /group02/rcoilgrp/Software/KinFit/src/*.c. This is one way of doing it; another would be to work with the library /group02/rcoilgrp/Software/KinFit/lib/libKinFit.a
      • test.c is an example how to implement it into a user code. It includes some documentation.
    • kinfit (iprocess, DoingMC, pbeam, mom_meas, mom_fit, &chi, &ndf, pull)
      • In every DVCS candidate event, I am calling this function for each M7 track.
      • iprocess = 1 for DVCS
      • DoingMC = 0 (real data), 1 (MC)
      • pbeam: beam momentum
      • mom_*: 4x4 array holding the four-momentum vectors of the measured kinematics (*=meas = input) and the fitted kinematics (*=fit = output). The four-momentum vectors are written as (px,py,pz,E). For DVCS, the order of particles is: (scattered lepton, real photon, recoiling proton, unused).
      • chi: Fehler beim Parsen (Konvertierungsfehler. Der Server („https://wikimedia.org/api/rest_“) hat berichtet: „Cannot get mml. Server problem.“): {\displaystyle \left.\chi ^{2}\right.}
      • ndf: number of degrees of freedom = 4 for DVCS.
      • pull: 12 array holding the pull distributions (in the same order as the four-momentum vectors in meas_*).
        pull(i) = (k_fit(i)-k_meas(i))/sqrt(sigma2(k_fit(i))-sigma2(k_meas(i)))
  • chi2_kinefit
    • My code
      • rdtrack.chi2_kinefit_dvcs[irdtrack] = chi2_kinefit of individual rdtrack
      • event.chi2_kinefit_dvcs = 1 if there exists an M7 track in the event with chi_kinefit<10 (cut as of 1.9.2010), 0 otherwise.
      • rdtrack.issmallestchi2_kinefit_dvcs[irdtrack] = 1 if this track is the M7 track with the smallest chi2_kinefit of all other M7 tracks in the event, 0 otherwise.
    • Best chi2 kinefit cut?
      • chi2=13.4% corresponds to a fit probability of 1%.
      • MC: what does the chi2 look like for recoiling proton that are reconstructed? Elastic, SIDIS, associated separated?
  • Fitted parameters handed back from kinfit()
    • kinfit() introduces some correlations between the lepton, photon and proton kinematics.
    • The DVCS missing mass using the fitted instead of the measured paremeters should be the squared proton mass - this is one of the input constraints / requirements.
    • Pull=1: the scale of the measurement error
    • The open question is what we will do in the end - will we use the 'measured' kinematics (HTC) or the 'fitted' kinematics (from kinfit())for the extracted DVCS amplitudes? For some parameters (e.g. theta), there is a bias.

Selection algorithm: rdPID, track hypotheses, topologies, suspicious tracks, multiplicities

  • rdPID. As of August 2010, I assign the rdPID in the following way.
    • First, for each M7 track the spacepoint configuration ("track topology") and charge are read from g1RDTrack.bHitP.
    • Then, for each M7 track it is checked if rdPID values are available in g1RDPID. If yes, the rdPID sum of all layers where there is a spacepoint is calculated. If the rdPID sum is smaller than PIDCUT-pion (currently 0), the track is assumed to be a pion, if the sum is larger than PIDCUT-proton (currently 0), the track is assumed to be a proton.
    • If there exists no rdPID, the track is assumed to be a negative pion if its momentum is negative, or a proton if it is a 2SP track and/or StoppedProton hypo track. Note that the rdPID assigment happens before any tracking parameters are read.
    • Remarks.
      • Throw out the tracks with 0011 && p>0.250 GeV at an early stage?
      • Momentum-dependent PID-cut?
  • Track hypotheses. After Feb. 17, 2011 - no rdPID is used anymore for selection of rdHypo
    • The Proton hypo params are read, if available AND if the chi2 of the track fit is not more than 100;
    • Otherwise, the StoppedProton hypo parameters are read if available (resulting in about 3-4% of all protons);
    • If neither StoppedProton nor Proton hypos are available, the Pion hypo params are read (in about 0.5% of the p>0 cases).
    • Remarks.
      • (Aug 2010) I haven't worked on readrecoil_hrc() for quite some time - the track hypo selection is not yet tuned there (i.e. hrc-format as opposed to uDST-format) to the latest "best way".

Cuts not related to Recoil

  • thetagaga: make existing plot for trad., exists track in RD, and deltap cut also for MC, separating elastic, asso and sidis. If the upper thetagammagamma cut (<45mrad) is removed, the azimuthal phi distribution rises for small absolute values of phi. An upper cut of <70mrad is enough to flaten the phi-dsitribution. (Didi: is due to singularity of BH propagator) Question 6.5.2010: Is this a problem for the extracted asymmetry amplitudes if the phi distribution is not flat? Probably not. The upper cut should rather be driven by accpetance.
  • t Mandelstam (constrained)
  • Impact of cuts on the photon. Of single photon events (1 DIS lepton, 1 photon) which fulfill the delta(p)<1 GeV/c cut
    • the Ecalo > 5 GeV cut cuts away 13%
    • the Epre > 1 MeV cut cuts away 18%
    • the Ecalo and Epre cut together cut away 28%
    • the fiducial volume for the photon in the calo cuts away 16%
    • all 3 cuts together cut away 39%
  • Calorimeter
    • E_calo Do we need the upper 5GeV cut, which reduces background, or not??? There exists the statement that the events with E_calo < 5 GeV are not DVCS, but here we fully trust and rely on our Monte Carlo. According to Elke, the cut with Ecalo_gamma > 5 GeV came from a mismatch between the MC missing mass distribution and the data one. We found later out it had to do with the fact that we needed to open the Q2-cut and W2-cut for lepto.
    • Use photons that do not shower in the preshower (Epre<1MeV)? Compare Mx2 for showering and non-showering photons. plot exists. Resolution is slightly worse. One needs however a different z-depth for those events. The missing mass distro E<1, E>1MeV in MC looks intersting - which correction did I apply for E<1 events?
    • Photon energy in Monte Carlo?
  • Tracking
    • Any change in ZVx cut?
  • Links

Asymmetry extraction

  • 06e1 and 07c2 hydrogen data.
  • Started with chi2 fit method and asymmetries calculated by C-code. BSA is calculated in my code binned in aphi,t,x,Q2 and (t,aphi). bsa.kumac plots everything.
  • Meanwhile, I use an EML macro (Physica adapted from Andy Miller) to extract the asymmetry amplitudes. The fitting is done in
/user02/criedl/MaximumLikelihood/2010

See README. bsa_eml.kumac plots everything.

  • I use the bblists for 06e1 and 07c2, some additional cuts on burst level (see e.g. Dietmar's release report) and SSD data quality from g1Quality. See extended analysis report for more details.

2007/2008/2009 Studies

Related Pages

Abgerufen von „https://hermes-wiki.desy.de/index.php?title=DVCS_with_Recoil_Analysis_Page&oldid=13611“