NPOL Analysis Program for E93-038

Version 2.4

Purpose:

    Our version of the Hall C engine analyzes HMS data and produces an ntuple containing analyzed HMS variables plus raw NPOL variables.  The program npol_calculations.exe analyzes the NPOL data and performs kinematic calculations, producing an expanded ntuple.  This document provides brief instructions for using the program and some sample results.  A more formal report describing the calculations in more detail will be provided later.

Event Classification

    Kinematic calculations depend upon reconstruction of tracks for the events of interest.  Although the simplest events contain a single hit in the front and rear arrays, most events are more complicated.  The steps of the reconstruction process are summarized below.

1. The first step in the procedure is to evaluate mean pulse height, mean time, and position for every hit comprising an event.  A filtering procedure then discards hits which fall below software pulse-height thresholds, or reconstruct to positions significantly outside physical detectors, or fall outside time windows suitable for quasifree knockout and subsequent flight through the polarimeter.  
2. During the filtering process we attempt to ascertain which hit produced the trigger, even if that hit fails one of the selection criteria.  
3. The next step is to group the remaining hits into recognizable patterns.  A cluster in the front array is defined as a set of hits that are close in both time and position and which contain no missing layers.  We assume that the first nuclear interaction occurs within the layer closest to the target containing hits.  If that layer contains several hits, we assign the nuclear interaction either to the top or the bottom hit by determining whether the nucleon scatters upward or downward.  An error condition is set if the first layer contains a noncontiguous set of detectors because such gaps may indicate an accidental spoils the reconstruction.
4. If there is a second cluster in the front array, we determine whether the nucleon scatters upward or downward using hits in the rear array.  We then decide whether the first scattering was upward or downward by comparison with the first cluster.  Note, however, that is difficult to analyze multiple scattering of the projectile within the front array and there is considerable room for improvement of this algorithm.  Fortunately, the frequency of such events is small enough that they can be discarded without significant loss of statistics.
5. We then look for a hit in the rear tagger along the chosen track; if one is present within a specified distance the particle detected in the rear is assigned positive charge.  If both top and bottom rear subarrays contain filtered hits and it is not possible to choose the most probable track unambiguously, an error condition is set.  

    The following table summarizes the track reconstruction error conditions.

The following table describes the tracking information contained in the output ntuple.

Instructions

Requirements:

The program is presently installed in the directory /u/group/e93038/Analyzer2.4/KinematicCalculation/npol_calculations.exe.  The following environment variables must be established by the script that runs the program.

A sample script, do_npol_calculations, has been provided.  It uses the sample files and directory names listed above.  The program assumes that data are stored in ntuples with names of the form polraw#####_n.rzdat where ##### denotes the 5-digit run_number and n is an index for members of an ntuple chain belonging to a particular run.  Output ntuples are written to files named pol#####_n.paw.  In addition, a journal file, pol#####.log, is produced.

Instructions:

1. Use source setup_analyzer to establish and verify the required environment variables.
2. Examine the raw data and adjust the tracking cuts as needed.
3. The program is launched by the command: do_npol_calculations   run_number   target  calibration_number where you must supply the appropriate run number and target (deuteron or proton) values.  Time calibration parameters are read from a file named time_#####.dat where ##### is a 5-digit calibration_number for a nearby run used for calibration.
4. Examine the output ntuple.  The auxiliary program basic_histograms produces an hbook file containing various diagnostic and physics histograms.  The auxiliary program check_detectors produces a more extensive set of diagnostic histograms useful for evaluating the performance of each detector.  PAW kumacs for displaying these results are available.

Limitations:

1. Kinematic calculations do not account for the energy loss suffered by protons passing through the lead sheets or the detector elements.
2. The analysis does not account for bending of proton trajectories in Charybdis.
3. Compatibility with earlier versions requires that events be limited to 62 hits even though there are 70 detectors, but this limitation is benign because such high multiplicities should never be seen if the data acquisition is working properly.

Auxiliary programs

    A couple of simple programs for producing useful histograms from the output ntuple have been provided, together with sample kumacs for display purposes.  

basic_histograms

    The program basic_histograms found in subdirectory BasicHistograms produces an hbook file containing many histograms that are useful for evaluating the quality of the data.  The program is launched by the command make_histograms run_number.  The paw kumac basic_histograms.kumac can then be used to display the histograms.  The file histo_38295.ps contains a set of figures based upon run 38295.  Below we provide a brief description of the pages.

1. Mean pulse height spectra for front, rear, front tagger, and rear tagger arrays.  No cuts were applied.  Calibrations of the front and rear arrays were used to convert ADC channels to MeVee.  The spectrum for the rear array includes a structure beyond 60 MeVee arising from ADC overflow.
2. The position spectra for the front and rear arrays show sharp edges defined by the sizes of these detectors.  Constant-fraction discrimination gives these detectors have relatively good position resolution.  Positions in the taggers were calibrated by comparing tracks with the nearest hit in the front array, but the resolution with leading-edge discrimination is relatively poor giving these spectra a more Gaussian appearance.  The width of these spectra is significantly less than the size of the detectors, which are not completely illuminated.
3. The mean-time spectum for the front array shows a dominant peak produced by the primary scattering event centered at zero, followed by secondary peaks arising from multiple scattering of the recoil proton, superimposed upon an uncorrelated background.  Although no cuts were applied to this spectrum, it is clear that track reconstruction can safely disregard hits outside a (-5,+10) ns window.  The peak in the rear array corresponds to quasielastic scattering between the front and rear arrays.  Early hits can be eliminated using a window (-10, 25) ns.  The front tagger also displays a quasifree proton peak superimposed upon an uncorrelated background.  Offsets were chosen to place the proton peak in the rear array at zero also.  Thus, initial and final protons can be eliminated using a (-12 ns, +5 ns) cut.
4. The trigger time spectrum shows a 0.5 ns resolution for reconstruction of the trigger time.  The error spectrum shows that the most common tracking problem is exclusion of all rear hits, usually because the rear hit occurs too early to arise from a valid track.
5. The rveto spectrum of distances between the incoming track and the nearest hit in the front tagger contains a relatively narrow distribution superimposed upon a flatter distribution, with a significant change in slope at about 30 cm which is related to the position resolution for the tagger array.  Therefore, we identify tracks with front tagger hits within 30 cm and within the time window as charged particles and disregard front tagger hits at larger distance or inappropriate time.  The rtag spectrum is essentially featureless because recoil protons are widely distributed in angle and originate anywhere along the track.  Nevertheless, we use the same proximity conditions to identify a charged track from front to rear, arguing that similar position resolution is expected.
6. Hit frequencies are shown by detector number and by array.
7. Filtered hits shows multiplicities after application of filtering conditions.
8. The raw tof spectrum is simply ph_start converted to ns.  The upper right spectrum shows the path-length correction to the electron time of flight.  The tof spectrum includes offset and trigger-delay corrections for the neutron arm.  The ctof spectrum subtracts the calculated flight time for a d(e,e'n) reaction and, hence, should be centered at zero.  This spectrum selects neutrons without track reconstruction errors and includes HMS particle-identification cuts (see item 10).  We then obtain FWHM about 2 ns.  The tail on the right side might include neutrons from the d(e,e'p)(p,n), which are expected to be late because of proton energy loss in various materials.
9. The front-to-rear velocity spectra require neutrals in front and rear, HMS particle identification, and include a ±2 ns cut on ctof..  The velocity ratio compares the measured velocity to that for the p(n,n)p reaction at the same scattering angle and shows a strong quasifree peak.
10. Electron particle identification (PID) spectra include the Cerenkov signal converted to number of photo-electrons (npe), total energy in the show detector, velocity, and chi-square for track reconstruction.  Spectra here that include electron PID require >2 photo-electrons, >1 GeV in the shower detector, and 0.7<beta<1.3.
11. HMS focal-plane variables.  Some neutron spectra include a cut on |delta|<4.
12. Target variables reconstructed using HMS optics.  Some neutron spectra include a |z|<7 cm cut on the vertex position.
13. Momentum transfer variables. 
14. Electron scattering angles, invariant mass, and virtuality.
15. theta_nq and phi_nq are neutron polar and azimuthal angles relative to the momentum transfer, while theta_sc and phi_sc are polar and azimuthal scattering angles.  These spectra require neutrals in front and rear, electron PID and delta cuts, and a tight ctof cut.
16. Components of missing momentum, total missing momentum energy, and mass. Note that the missing mass is sharp because the neutron momentum is computed from the electron scattering kinematics and neutron angle; this calculation does not use the time of flight.

check_detectors

    The program check_detectors, found in subdirectory CheckDetectors, produces an hbook file containing many histograms useful for evaluating the performance of each detector.  Sample scripts for running the program and displaying the histograms using PAW may be found in the Samples subdirectory.   Sample output is contained in check_38295.ps.  

    The program reads parameters for event selection from a file, identified as check_detectors.cuts in the sample script.  The sample input file includes comments identifying these cuts.  The table below lists the available cuts, the parameters chosen by the sample file for Q²=1.5 (GeV/c)², and the histograms affected by those cuts.  Some cuts are defined by either a minimum or maximum value, but most cuts require both.  All histograms are subject to electron PID cuts.  Spectra for the front array array require p_zfr=0; spectra for the rear array require p_zfr=0.and.p_zre=0; spectra for the front and rear tagger arrays require p_nvt>1.  These choices of charge are hardwired and were chosen to emphasize the performance for (n,n') events of greatest interest to G_En.

    Some additional cuts are inherited from the filtering performed by npol_calculations.  Only events with p_error=0 are histogrammed by check_detectors.  Therefore, there must be at least one hit in both front array and rear arrays satisfying the position and time filters.  The tracking algorithms in npol_calculations attempt to identify the hit in the front array representing the primary neutron scattering event, called idfront, and the hit in the rear array, called idrear, representing detection of the scattered neutron.  Spectra for individual detectors in the front or rear arrays are sorted by either idfront or idrear, discarding secondary hits due to recoil protons or rescattering.  Self-timing peaks are sorted by idtrig, which is the hit believed to be responsible for the trigger, which need not be the same as idfront.  Therefore, these spectra retain multiple-hit events while testing the tracking algorithms.

A brief description of some of the output pages follows.

1. Position spectra for each array include all hits subject only to the charge selection and electron PID cuts above.
2. Mean-time spectra for each array include all hits subject to charge selection and electron PID cuts.  The restriction on the front mean-time is a result of filtering.
3. The dtof spectra for the top and bottom rear arrays include a ctof cut that emphasizes the quasifree neutron scattering events.
4. Position spectra for the primary neutron scattering event in the front array are all very similar, with sharply defined edges at the borders of the physical detectors.  Some affect of attenuation is apparent in deeper layers.
5. Position spectra in the rear arrays follow a fairly simple pattern.  The outer detecters in each layer have more strength because they are wider.  The positive slopes arise from the forward-peaked angular distribution for neutron scattering (positive x is toward the back of the detector).  This slope is smaller for inner than for outer layers.  The requirement on ctof and (n,n') limit the statistics in these spectra.
6. Position spectra for the front tagger are all very similar.  Position spectra for the rear tagger are also similar to each other, with outer detectors being a little weaker.
7. Mean-time spectra for the front array shoul show a single narrow peak centered at zero, but shoulders on the late side for detectors 4 and 14 might indicate occasional errors in identifying idfront.
8. Mean-time spectra for the rear array should show a neutron scattering peak superimposed upon an uncorrelated background.  The offsets are chosen to center these peaks at zero.  The relative intensities generally follow the pattern expected from geometry, except that detectors 22 and 36 are somewhat too weak.  Both of these detectors also had poor position and pulse-height spectra throughout the January run.
9. Mean-time spectra for the tagger arrays should contain a dominant peak at zero produced by quasifree protons passing through all elements of the tagger and front array, with secondary features attributed to recoil protons, superimposed upon an uncorrelated background.  All spectra for the front tagger are very similar.  Outer elements of the rear tagger are somewhat weaker and have flatter background.
10. All self-timing peaks are centered at zero and contained within about ±1 ns.  These p_tc are cut off on the positive side because idtrig is identified as the detector with the smallest value of p_tc.
11. The ctof spectra all centered at zero with similar widths and strengths.  The uncorrelated background appears to decrease slightly with depth.
12. The dtof spectra are very similar to the mean-time spectra for the rear array.
13. Pulse-height spectra for the front array are all very similar to each other, with minor variations in threshold and leakage through the discriminators.
14. Pulse-height spectra for the rear array sometimes show ADC overflow within the plotted range, which is represented by a spike at the end of the left or right spectra and an associated ridge in the mean pulse height.  The droop in phR22 is probably related to the anomalous position spectrum in detector 22.  Similarly, the steeper than average slope in phL36 and droop in phR36 are probably responsible for its unusual position spectrum.  The thresholds are clearly visible in these spectra, but these discriminators are fairly leaky.
15. The pulse-height spectra for most tagger elements look fine, but the thresholds and gains for both 45R and 46R are too small, producing unusual structure on the low side.

Summary of recent changes

Version 2.4

• A new variable, p_rtof, is defined as the difference between the measured and predicted time of flight from front to rear.  The measured time, p_dtof, already includes the time offsets for all detectors.  The predicted time is based upon free NN kinematics using the incident momentum calculated from electron kinematics and the hit positions in front and rear arrays.  This quantity was available in earlier versions of the time calibration program, but for for historical reasons (largely compatibility with the earlier ntuple definitions) was not available in the second-pass ntuple.  It has the advantages of better resolution and being centered upon zero for all kinematics.
• The definition of p_td has been changed for the rear array so that quasielastic events should appear near zero.  Hence, the filtering of hits in the rear array can now use a narrower window, typically (-10, 25) ns.  
• The identification of the charge detected in the front and rear arrays has been changed.  The previous versions tried to identify a neutron and basically said that if it's not a neutron then it must be a proton.  The new logic requires a more definite identification of a proton.  An incident proton (more precisely, charged particle) must have hits in both front tagger arrays consistent in both time and space with the front hit.  Similarly, when both rear tagger arrays are available, both must agree with the charge identification.  If the conditions for a proton are only partially satisfied, the charge is classified as ambiguous.  Therefore, the charge variables are now interpreted as -1:ambiguous, 0:neutron, or 1:proton.  A significant number of incident particles (about 15% for Q2 = 1) are now in the ambiguous category, most because there is one hit in the front tagger within the appropriate time window.  Such events may reflect tagger inefficiency or noise, but can alo be caused by (p,n) reactions in the lead that are accompanied by a photon or a low-energy secondary that fires just one tagger element in coincidence.  Whatever their origin, we cannot then uniquely determine the charge of the primary track.
• The algorithm for selecting the hit for protons that represents a large-angle nuclear scattering event has been improved also.  Instead of always choosing a hit in the first layer, code provided by Brad Plaster looks for a deviation from a straight trajectory and uses the appropriate hit in the preceding layer.  If no deviation is found, the final hit is chosen.  The analyzing power can be improved by using a cut on scattering angle to eliminate the events with only Coulomb multiple scattering and no large-angle nuclear scattering.
• Times for hits in the front and rear array relative to the scattering time in the target, reconstructed from the HMS optics and the trigger time, are stored in the ntuple within the array p_t.
• For convenience, hs_q2 was added to the npol block of the ntuple.  Although logically it belongs to the first block, consistency with the present version of engine precludes that placement. 

Version 2.3

• The makefiles have been substantially rewritten for use with the gmake utility and will produce executables for either Linux or Sun computers.  The script setup_analyzer establishes pointers to the executables for the current platform and environment variables for the files containing parameters and cuts for the desired kinematical conditions.  Although these scripts were written specifically for the JLab environment, modification for use at other institutions should be straightforward.  Details may be found at installation notes. 
• The kumacs have been modified to produce postscript output quickly without the HIGZ screen.  Use pawps kumac_name [run_number].  This procedure is much faster than before and is suitable for batch processing also. 
• Numerous small changes were made for compatibility with g77, including the addition of trigonometric functions with arguments in degrees.
• Tests for unphysical kinematics in the calculation of neutron time of flight expected for the d(e,e'n)p reaction have been improved, which eliminates some runtime errors but increases the frequency of p_error=8. 
• The consistency of detector numbers and multiplicities are tested.  Inconsistent events are tagged with p_error=9.
• Scattering calculations are skipped if there were any track reconstruction errors.

Version 2.2

• Installed in /u/group/e93038/Analyzer2.2.
• A new program check_detectors.f produces an hbook file suitable for evaluating the performance of each detector and the PAW script check_detectors.kumac produces a postscript file with plots organized in a convenient arrangement.
• The basic_histograms program reads cuts from an input file.

Version 2.0

• Installed in /u/group/e93038/Analyzer2.0.
• Data for the new fourth tagger layer is included, but the programs will still work for runs replayed with earlier versions of engine before implementation of the fourth tagger.  Note that if any rear tagger hit is found within the specified time and distance from the front-rear track, the particle detected in the rear is tagged as charged.  We do not require hits in both rear tagger layers or consistency between hits in the two layers.

Version 1.6

• Installed in /u/group/e93038/Analyzer1.6.
• We returned to the use of velocity calculated from electron kinematics and npol angle for calculations of missing energy, momentum, and mass; the change to measured flight time in Version 1.5 turned out to be a bad idea and has been reversed.  The problem is that this experiment provides no independent measurement of momentum for comparison with time of flight, so that cuts upon missing mass or momentum are correlated with ctof.
• Filtering for hits in the rear array has been modified.  If there are hits that satisfy the selection criteria, only those are retained.  If no such hits are present, all hits are retained.  Thus, dtof and vratio can be calculated for every event with an acceptable hit in the front whether or not valid hits are present in the rear.  This method allows more stringent time filtering for the rear array while retaining early dtof events for evaluation of accidentals.

Version 1.5

• Calculations of missing momentum, energy, and mass now use the measured time of flight instead of that predicted from electron kinematics and hadron angle.  Although this approach sacrifices some resolution, it should help to exclude events from other reactions, such as pion production.  Note, however, that these quantities can be calculated only when the flight time gives a velocity less than lightspeed.  Furthermore, accidental coincidences on the late side of the coincidence peak tend to accumulate within an artificial missing mass peak which resembles, but does not correspond to, nucleon resonances.
• Bugs in the reconstruction algorithm for multiple-hit events in the front array that led to some spurious error conditions in earlier versions have been corrected.  The fraction of events that can be reconstructed completely has increased.
• A third block has been added to the ntuple containing track reconstruction diagnostics.  The p_ntype variable is now longer used, having been superceded by the more detailed information in the tracking block.
• Walk corrections for the tagger times are included.
• Event initialization problems have been corrected.

Version 1.4

• File names are now based upon environment variables, which facilitates use of the program in batch mode for production processing.  A sample script is provided as do_npol_calculations.
• A new time-of-flight variable, p_tof, includes the timing offsets and trigger correction, giving better resolution than the raw tof variable, ph_start.
• The definition of the coincidence time of flight, p_ctof, is now ctof = tof - ntof where tof is the corrected time of flight and ntof is predicted from the electron kinematics and nucleon angle.  The result should be zero, within resolution, if the event arises from the d(e,e'n)p reaction.
• RF information is no longer used in the ctof algorithm.
• Ntuple chains are now processed, producing a corresponding output chain.

Maintained by: James J. Kelly

Last revised: January 3, 2002