Solid Forward Calorimeter
- 1 general design ideas
- 2 which one to choose
- 3 Task list
- 4 Shashlyk Ecal
- 5 Fe/SciFi Ecal
- 6 Useful links
general design ideas
from Eugene Chudakov
a brief explanation is given in the proposal to PAC34, page 57-58.
A preshower I simulated gave a factor of 3 in e/pi rejection, which is important.
The resolution I used was 10/sqrt(E). With the shashlyk technology one can get 10% at 1 Gev easily, with a rad. hard scintillator. A better resolution of about 3-4%/sqrt(E) was obtained (KOPIO) with a high Sc/Pb ratio, but also with a rad. soft scintillator (there is a brief discussion in the proposal).
In order to make sure which resolution is enough one has to consider all the aspects including the pattern reconstruction. A poor reconstruction may lead to a more stringent requirement on the calorimeter resolution. The coordinate resolution is also very important, since tracking should start with the hit in the calorimeter.
I would suggest to assume 10%/sqrt(E) with a preshower, take some cell size to provide about 1 cm window for the shower center, and develop an algorithm for track reconstruction (Richard Holmes made some preliminary calculations). This is a critical issue, since the background in GEMs will be very high, and the calorimeter tagging is important. I used some assumptions on the GEM X-Y matching and resolutions, based on the SBB proposal in Hall A, which might be optimistic (Bogdan told me that they were concerned about it).
from Paul Souder
The Ecal has the following functions:
1. Provide a trigger.
2. Provide an energy for tracking.
3. Reject pions.
4. Study systematic errors.
Effect of resolution:
Pion rejection: the rejection factor is inversely proportional to the resolution. The asymmetry of the pions cn be measured, so the correction due to pion contamination can be made. However, if it is too big, complex systematic errors enter and the statistics is degraded.
Energy for tracking: The idea is that given a hit and an Ecal energy, one can compute where other possible hist are. This is the key to the tracking algorithm. We need a tracking Monte Carlo to determine which resolution dominates: multiple scattering, detector resolution, or Ecal resolution.
Trigger: Poor resolution may increase trigger rate.
Systematic errors: If the tracking and Ecal both have good resolution, the comparison is useful for minimizing systematic errors such as Q2.
Summary: The excellent resolution available with the sashlyk Ecal is useful. How much poorer resolution we can tolerate requires the Monte Carlo.
which one to choose
comment from Eugene
I though about using a SciFibLead calorimeter (let us call is SPACAL after their inventors at CERN), similar to the Hall D (or KLOE) one.
If the showers mostly go perpendicular to the fibers, as in Hall D, we end up with a 1-dimensional readout. The longitudinal component can be reconstructed by the time measurements from both sides. Advantage: few channels and an easy-to-make separation along the shower. Disadvantage - too high occupancy, many false combinations.
One can make it as the original SPACAL was done - the shower goes along the fibers. It will be similar to shashlyk - 2 dimensional structure, many channels. Disadvantage: the preshower must be a different detector. In general, I do not see any advantage to shashlyk.
SHASHLYK: In my opinion it was the most promising technology for SoLID. However, it requires some R&D: in order to make the preshower one should use some of the WLS fibers for the preshower only. Either the preshower photodetector is located in front of the calorimeter (not an appealing option for many reasons), or these fibers come through the main "shower" part, optically insulated, say with black plastic tubes.
From Paul Reimer:
As I see it, the problem is how do we read out a segmented calorimeter if it is located completely inside the iron of the endcap flux return of the solenoid. There are three answers that I can think of:
- Use rather expensive APD's and if we need to service them, we need to open up the entire detector. This is the solution from our proposal, I think.
- Figure some way to get the light outside of the flux return iron, using wave shifters and fibers. The problem here is that you loose a lot of light in the process.
- Make the calorimeter part of the flux return iron so that the light naturally ends up at the downstream end of the flux return iron.
Zhiwen is looking at option 3. The first question to answer is what does the magnetic field do to the shower? To do this, I wanted to start with a simulation code that has been tested against measurements in a test beam. Thus we started with Hertzog et al's code which agrees with their test beam results (the code is called */SciFi, what a cool name!). My thought is that we need to do the following:
- Get the W/SciFi code from Hertzog running and show that we get similar results to what they got in their paper (complete)
- Add a magnetic field and study the changes in shower shape, transverse and longitudinal (in progress, near done)
- Switch from tungsten to iron and repeat step 2 (in progress, near complete)
- Change geometry from "fiber ribbons" to fibers on equilateral triangles and repeat 2 (starting, could be done in a week)
- Do 3d magnetic field simulation of calorimeter's Fe/SciFi geometry to get field correct. I don't have access to TOSCA easily at Argonne, but I do have access to another 3d field program. Here Zhiwen could start working with Juliette and learn TOSCA. It is a very marketable skill (many university professors do not know how to do it).
- Repeat step 4 (will be completed a week or less after step 5)
- Design calorimeter based on results of step 6. This is optimizing resolution as a function of iron to scintillating fiber ratio, determining the number of radiation lengths needed to contain the shower to the extend that we feel it needs to be contained and optimizing segmentation.
- Cost calorimeter
- Build prototype
- . . .
Work through step 6 will completely the question of will it work. From what Zhiwen's shown me, the answer will clearly be yes, but I would like him at the end of step 6 to write a tech note on this to document it, and that will take some time. Then we move on to designing and building. At some point around step 7/8 we need to present this to the collaboration and get their approval. There is also an issue of getting the preshower to work (to get the light out). This would be an issue with any choice of calorimeter. When I was at JLab last week (week of March 7), I think that Zhiwen and I figured out a cost effective solution, but it depends on the segmentation required by the experiment.
In proposal, PVDIS covers 22-35 deg, the area is 3.14*(280**2-120**2)=2e5 cm2, but with baffle only about half 1e5 cm2 is needed
In proposal, SIDIS covers 7-12 deg, the area is 3.14*(220**2-120**2)=1.1e5 cm2
so both have similar area if we can rearrange the modules.
|module size (cm2)||module cost ($)||number of module||total cost ($)|
|estimation based on 1.1e5 cm2 area|
For SIDIS largeangle EC, the area is 3.14*(140**2-75**2)=0.5e5 cm2, the cost is half if use same module.
4x4cm2 module has 16 fibers, so the total number of fiber is 0.1M.
the position resolution is a couple cm.
If we only need forwardangle EC to has 30 segmentation to match GEM for trigger, what's largest module size we can have?
- pricing info
Now we can produce modules with the following parameters: 1) transverse size - 38.2x38.2mm**2 or 76.4x76.4mm**2 2) lengh - no limitations 3) scintillator thickness - 1.5mm (the thickness of 1mm scintillator close to the limit, a 0.5mm has bad light yield and attenuation) 4) lead - from 0.2mm upto 2mm 5) production speed - 200 modules per month; can be increased by 2-3 times; For the production of these modules have molds for stamping of lead plates and casting scintillators. The price of modules with 0.8mm lead plates/1.5mm scintilator and 160 layers with Bicron fibers is 370USD. The price of modules with thinner sampling and more layers will be higher by 100 or 200USD, real price depend on the number of layers and the total number of modules. a) the price of raw material is around 20% b) the production of lead plates and scintillators should be the same or a little more; c) the assemble also should be the same or a little more; so a 76.4x76.4mm**2 module is about $700, 110x110mm**2 module is $1k If you want modules with different transverse size, it's not a big problem. It will be necessary to produce a new stamp for the lead and mold for casting scintillators. The cost of new stamp and new molds around 50kUSD and will take approximately 6 months to design, manufacture and testing. The new equipment ensures production of lead plates to 150K and 500K scintillators.
- spiral fiber
Fibers are not straight, it can be seen in the photo. Fibers pass in the module for volumetric spiral with 2mm diameter and 80mm step. Slide 9 explains how organized spiral inside the module, of course only a picture is unclear. I will tryed to explain this slide. Right on the slide plate shows one of 4 types of scintillators, 9.5 mm is the distance between the centers of spirals. Centers of the holes place on a circle with a diameter of 2 mm centered at the center of the spiral. On the left slide shows an enlarged view of one spiral. There are 4 types of scintillator and lead plates in which holes are arranged with a slight shift. On the slide, they are shown 1 - pink, 2 - blue, 3 -red, 4 - gold. These 4 types taken together make up 1 / 4 helix, the following 4 layers are rotated 90 degrees and provides the following quarter of a spiral, etc. Thus, 16 layers provide a full turn of the spiral. In Compass case, I've been collecting double instead of single layers, i.e. 2 layers of lead-scintillator of the same type. This is to ensure the diameter of curvature of the fiber more 70mm, the diameter of light losses are insignificant and there is no damage to the fibers.
Each fiber glued by epoxy an individual hole in a special black plastic flange. This is guaranteed to kill all scintillation light, which is distributed under the cladding, and also makes it easy to replace broken or inefficient fiber. Of course the fibers may be connected with light guide or directly from the photomultiplier. In the COMPASS we use a photomultiplier FEU-84, Russia made, for readout.
Shashlyk Ecal was used by E865 at Brookhaven, the PHENIX RHIC detector, the HERA-B detector at DESY, and the LHCb detector at CERN.
Information from the 2008 KOPIO NIM paper:
- cross section 110x110 mm^2
- 300 layers, each layer consisting of a 0.275-mm lead plate and a 1.5-mm scintillator plate. Spacing between sci tiles is 0.350mm. The lateral size of the plates is 109.7x109.7mm^2;
- The scintillator used was BASF143E-based. In the 2004 NIM paper it was PSM115 polystyrene. The radiation stability of BASF143E-based scintillator is a dose level 120 krad, and the recovery time is about 80 days.
- Sci tiles have "lego" type locks that maintain the position of the Sci layers and the 350-um gap, providing sufficient room for the 275-um lead tiles without optical contacts between lead and scintillator.
- Each plate has 144 holes equidistantly arranged in a 12x12 matrix, in which the spacing between the holes is 9.3mm. The diameter of the holes is 1.3 mm, both in the lead and the sci plates.
- Inserted into the holes are 72 1.5-m long, 1.0-mm diameter WLS fibers Y11-200MS, with each fiber looped at the front of the module. The loop has a radius of about 2.5cm, which may be approximated by a mirror with a reflection coefficient of about 95%.
- Light from the fibers are collected by Avalanche Photo Diodes and Wave-Form Digitizer (WFD) readout.
- The module has 4 1-mm stainless steel wires to hole all plates together in compression (with longitudinal loading up to 800kg).
- The module is wrapped with 150-um TYVEK paper which has light reflection efficiency of about 80%.
KOPIO test results using LEGS photon beam:
- The APDs wre keped in cooling units to remove from data analysis temperature dependent effects in APD gain stability.
- Operation in a magnetic field of up to 500 Gauss
- energy resolution is (1.96 + 2.74)%/sqrt(E) for 50-1000MeV photons, time resolution is (72 + 14)/sqrt(E) (ps), where "+" means a quadratic summation.
how to run code
checkout from svn https://jlabsvn.jlab.org/svnroot/solid/CaloSim/ and there's a README file
old instruction is somewhat relevant.
- 0. get all files from http://hallaweb.jlab.org/12GeV/SoLID/download/ec/ec_SciFi/
- 1. have a standard build of geant4 and setup its environment. (I used version 4.9.3.p02 with instruction http://geant4.slac.stanford.edu/installation/)
- 2. have compat-gcc-*-c++* package installed (I use compat-gcc-34-c++-3.4.6-20.fc14.x86_64 on my fedora 14 x64 system)
- 3. use my version GNUmakefile to replace GNUmakefile in the individual packages because the old ones are not clean and won't work.
- 4. go to each package and compile by "make", you should have the executable in bin directory.
- 5. use my vis.mac to replace vis.mac in the individual packages to use OpenGL viewer instead of the author suggested HepRepXML
- 6. read the CaloSim.pdf to know more about the program.
Note from the author, Noah Schroeder (email@example.com, firstname.lastname@example.org)
"I'm pretty sure Calosim1 was just a copy of calosim that i used to experiment with new things, so you should start from the basic CaloSim program. As for the lightguide and fiber sims, optimization of the lightguides was largely a separate issue from the calorimeter itself, so it was easiest to deal with that on its own. In order to do that, we needed a good simulation for the distribution of the photons coming out of each fiber, hence the fiber sim. How we did the alorimeter was just measuring the energy deposit of a shower in a given calorimeter chunk, rather than simulating the actual transit of the photons through a fiber, then to a lightguide, then to a PMT. As for putting them together, we never got that far, so i'm not sure the easiest way to go."
"As far as the geometry goes, I think those dimensions weren't of any particular significance, we weren't that close to doing full scale sims yet. As far as the 5 degree angle, The calorimeter face will be angled 5 degrees away from directly at the incoming high energy incident particles to prevent channeling, where a particle travels down a single fiber"
To understand shower size for Fe and W
formula refer to Eugene's calorimeter talk (p18-22) http://www.jlab.org/div_dept/consortium/08series/calor_lect.pdf * Critical Energy Ec = 670MeV/(Z+1.24) * Shower width R = 2*X0*21MeV/Ec/d * Shower peak Dmax = x0*(ln(E/Ec)-0.5)/d * Shower depth D = Dmax+X0*(0.08*Z+9.6)/d
Z X0(g/cm2) d(g/cm3) Ec(MeV) R(cm) Dmax(cm)(1.5,2.0,2.5,3.0,3.5,11GeV) D(cm)(1.5,2.0,2.5,3.0,3.5,11GeV) Fe 26 13.84 7.87 23 3.24 6.48 6.95 7.36 7.67 7.95 9.97 27.01 27.50 27.90 28.22 28.50 30.51 W 74 6.76 19.3 8.3 1.75 1.65 1.75 1.82 1.89 1.94 2.34 7.08 7.18 7.26 7.32 7.38 7.78
For pure material, shower R_Fe/R_W = 1.85 , Dmax_Fe/Dmax_W = about 4, D_Fe/D_W = about 3.6 For absorber/fiber sandwich, shower size should be enlarged somewhat and the ratio will shifted toward 1
resolution is taken as σE/E in fiber by gaus fit.
e- beam, 1mm wide, 5 degree in XZ plane, energy 1.5,2.0,2.5,3.0,3.5GeV Calorimeter 20cm in X (width), 10cm in Y(height), 40cm in Z(depth) 0.5mm absorber/0.5mm diameter fiber sandwich, the rest is glue. plane along Y, fiber along Z. fiber and glue are both assumed to be scintillator BC404 field is uniform 2T along Y
if preshower has no radio segmentation, only azimuthal segmentation, we can have lightguide on side. So what's rate?
According to rate estimated in PVDIS proposal (p40 http://hallaweb.jlab.org/collab/PAC/PAC34/PR-09-012-pvdis.pdf), For PVDIS with baffle, pion rate is 140MHz. this gives 5MHz(200ns) in 10deg. If beam bucket interval is 2ns, we have one pion every 100 buckets.
According to rate estimted in SIDIS proposal (p49 http://hallaweb.jlab.org/collab/PAC/PAC35/PR-10-006-SoLID-Transversity.pdf), For SIDIS, pion rate is below 3Mz, this gives 100KHz(10us)
(are they correct?)
Fiber sampling Tungsten powder calorimeter (Generic Detector R&D) for an Electron-Ion Collider
- Generic Detector R&D Program for an EIC (managed by BNL)