Geant4 Milagro Geant4 Milagro SimulationSimulation

Vlasios VasileiouVlasios VasileiouU. Maryland 5/5/2006U. Maryland 5/5/2006

Simulation componentsSimulation components

The softwareThe software Corsika – Air shower simulation Corsika – Air shower simulation Geant4 – Simulation of the detector’s Geant4 – Simulation of the detector’s

response to the EAS particles reaching response to the EAS particles reaching the groundthe ground

Milinda – analysis software (among Milinda – analysis software (among others: adds noise, performs PMT others: adds noise, performs PMT corrections, add time jitter and smears corrections, add time jitter and smears number of pes)number of pes)

CorsikaCorsika Simulates the air showersSimulates the air showers Variety of hadronic physics models Variety of hadronic physics models

availableavailable We useWe use

Low energy hadronic model Low energy hadronic model Fluka 2005 (previously used Gheisha)Fluka 2005 (previously used Gheisha)

High energy hadronic model High energy hadronic model Nexus 3 Nexus 3 (started using it in the latest Corsika v6.500 with (started using it in the latest Corsika v6.500 with

g4sim v2.0). Previously used Venusg4sim v2.0). Previously used VenusneXus (NEXt generation of Unied Scattering approach) is a common effort of the authors of VENUS and QGSJET with extensions enabling a safe extrapolation up to higher energies, using the universality hypothesis to treat the high energy interactions . It handles nucleus-nucleus collisions with an up to date theoretical approach.

CorsikaCorsika Curved atmosphere Curved atmosphere (started using it after g4sim (started using it after g4sim

v1.2)v1.2)

In the default version of corsika a plane atmosphere is used and the In the default version of corsika a plane atmosphere is used and the thickness increases as 1/cos(thickness increases as 1/cos(θθ). The accuracy of this approximation ). The accuracy of this approximation decreases with decreases with θθ (for (for θ=θ=9090 degdeg the thickness of the planar atmosphere the thickness of the planar atmosphere reaches infinity). reaches infinity).

Corsika doesn’t allow using Corsika doesn’t allow using θ>θ>70 deg unless a curved atmosphere is 70 deg unless a curved atmosphere is selected. selected.

How it works: In the curved version the Cartesian coordinate system is How it works: In the curved version the Cartesian coordinate system is kept but the horizontal step size is limited to <20 km.kept but the horizontal step size is limited to <20 km. Longer transport Longer transport distances are divided into appropriate segments to be treated in a local distances are divided into appropriate segments to be treated in a local flat atmosphere. After each traversed segment the particle coordinates flat atmosphere. After each traversed segment the particle coordinates are transferred into the next local Cartesian coordinate system with its are transferred into the next local Cartesian coordinate system with its vertical axis pointing to the middle of Earth. vertical axis pointing to the middle of Earth. Thus the curved Earth's Thus the curved Earth's surface is approximated piece by piece by flat segments with limited surface is approximated piece by piece by flat segments with limited horizontal extensions.horizontal extensions.

Corsika settingsCorsika settingsParticleParticle Max Max θθ Spectral Spectral

indexindexEnergy RangeEnergy Range Available Available

nShowers nShowers (mil)(mil)

pp 7070 -2.75-2.75 30GeV to 100 30GeV to 100 TeVTeV

~500~500

HeHe 7070 -2.75 -2.75 30GeV to 100 30GeV to 100 TeVTeV

~250~250

γγ 4545 -2.4-2.4 30GeV to 100 30GeV to 100 TeVTeV

~315~315

Geant4Geant4 Geant4 is a simulation toolkit from Geant4 is a simulation toolkit from

CERN created for the simulation CERN created for the simulation needs of the LHC. needs of the LHC.

Biggest advantagesBiggest advantages Written in C++ and thus easy to expand Written in C++ and thus easy to expand

& debug.& debug. Wide international collaborationWide international collaboration

Far greater functionality and number of Far greater functionality and number of physics models than geant3physics models than geant3

New and more accurate physics modelsNew and more accurate physics models Fewer bugsFewer bugs

g4simg4sim Milagro simulation code based on Geant4Milagro simulation code based on Geant4 Main code written by V. Vasileiou Main code written by V. Vasileiou (it’s all my fault...)(it’s all my fault...)

C. Lansdell and A. Smith helped with debugging C. Lansdell and A. Smith helped with debugging First beta version out in Summer 2004.First beta version out in Summer 2004.

The detector model in the first version of The detector model in the first version of g4sim was almost identical to the one in the g3 g4sim was almost identical to the one in the g3 version.version.

Slowly, I rewrote all the code from scratchSlowly, I rewrote all the code from scratch Using my own sources for the properties of the Using my own sources for the properties of the

detector elements (instead of the numbers in g3).detector elements (instead of the numbers in g3). Rechecking the sizes of the detector elementsRechecking the sizes of the detector elements Using a different PMT model Using a different PMT model

g4sim: Geometryg4sim: GeometryThe pondThe pond

Source of dimensions a Source of dimensions a pond schematic from Peter pond schematic from Peter Nemethy.Nemethy.

Support for water over the Support for water over the cover & air under the cover.cover & air under the cover. Currently using 1cm of Currently using 1cm of

water over the cover and no water over the cover and no air under the cover.air under the cover.

The layer of air under the The layer of air under the cover is uniform all over cover is uniform all over the water the water not a good not a good approximation.approximation.

In the g3 sim there is no In the g3 sim there is no physical cover included in physical cover included in the simulation the simulation

g4sim: Geometryg4sim: GeometryThe pondThe pond

Cover & pond linerCover & pond liner PolypropylenePolypropylene Diffuse reflector with incidence angle dependent reflectivity.Diffuse reflector with incidence angle dependent reflectivity. Calculated from Fresnel’s equations by Geant4Calculated from Fresnel’s equations by Geant4 Have to provide refractive index: 1.49Have to provide refractive index: 1.49 Increased reflectivity from the white pipes at the bottom of the pond not Increased reflectivity from the white pipes at the bottom of the pond not

included in the simulation.included in the simulation.

Reflectivity of a flat Water Polypropyleneinterface for non polarized light.

g4sim: Geometryg4sim: GeometryThe pond coverThe pond cover

Possible problem: Reflectivity from Fresnel’s equations and the results from ’94 Possible problem: Reflectivity from Fresnel’s equations and the results from ’94 David Schmidt’s Milagro memo disagree. UV reflectivity of polypropylene is very low.David Schmidt’s Milagro memo disagree. UV reflectivity of polypropylene is very low.

More reflections from the cover More reflections from the cover We trigger more easily by weaker showers. Nhit, We trigger more easily by weaker showers. Nhit, npe, mxpe, nfit distributions slightly move to lower values. npe, mxpe, nfit distributions slightly move to lower values.

Reflectivity of a flat Air Polypropyleneinterface for non polarized light from Fresnel’s equations.

Reflectivity of polypropylene (by D. Schmidt) versus the reflectivity of a material assumed to be a perfect reflector. Measurements correspond to some unknown incidence angle.

g4sim: Geometryg4sim: GeometryThe bafflesThe baffles

Baffle dimensions taken from a schematic from M. SchneiderBaffle dimensions taken from a schematic from M. Schneider Bug in g4sim v1.2Bug in g4sim v1.2

Inside part of baffles had the properties of tyvekInside part of baffles had the properties of tyvek True inside part of baffles is made out of the of unknown True inside part of baffles is made out of the of unknown

composition material called “Liner” in D. Schmidt’s memo. composition material called “Liner” in D. Schmidt’s memo. Currently in g4sim v2.0Currently in g4sim v2.0

Liner is a diffuse reflector with the reflectivity measured by D. Schmidt.Liner is a diffuse reflector with the reflectivity measured by D. Schmidt. Outside part of baffles: Diffuse reflector with the properties of Outside part of baffles: Diffuse reflector with the properties of

polypropylenepolypropylene

g4sim: Geometryg4sim: GeometryThe outriggersThe outriggers

Dimensions of outriggers taken from Dimensions of outriggers taken from a schematic by Tony Shoup.a schematic by Tony Shoup.

They are not all on the same plane They are not all on the same plane (in g3 they are)(in g3 they are)

They don’t all contain the same They don’t all contain the same amount of water. amount of water. In g4 sim their water height takes In g4 sim their water height takes

random values around a mean value. random values around a mean value. 76cm ± 10cm76cm ± 10cm ((Thanks to Scott for measuring the water Thanks to Scott for measuring the water

levels of several outriggers.) levels of several outriggers.)

g4sim: Geometryg4sim: GeometryThe outriggersThe outriggers

Inside part of outriggers lined with tyvekInside part of outriggers lined with tyvek Reflections from tyvek a mixture of specular and Reflections from tyvek a mixture of specular and

diffusediffuse The values of Geant4’s parameters that describe The values of Geant4’s parameters that describe

the reflections from Tyvek (sigma_alpha and the reflections from Tyvek (sigma_alpha and specular lobe constant) were taken from Auger specular lobe constant) were taken from Auger notes.notes.

Tyvek reflectivity taken from D. Schmidt’s memoTyvek reflectivity taken from D. Schmidt’s memo

Water propertiesWater properties Water absorption and scattering lengthsWater absorption and scattering lengths

In g3 sim (v3.2 pro?) 18m Absorption length and no In g3 sim (v3.2 pro?) 18m Absorption length and no scatteringscattering

In g4 sim (v0.99 – v1.3), 18m abs length and Rayleigh In g4 sim (v0.99 – v1.3), 18m abs length and Rayleigh scattering were mainly used. Simulated other abs lengths scattering were mainly used. Simulated other abs lengths too but 18m was always the pro.too but 18m was always the pro.

g4sim v2.0g4sim v2.0 Has both Mie and Rayleigh scatteringHas both Mie and Rayleigh scattering Two configurations were mainly used:Two configurations were mainly used:

Absorption Length 27.4m and Scattering Length 56.8m Absorption Length 27.4m and Scattering Length 56.8m (Att. Length 19m). Source was an email from Don Coyne to (Att. Length 19m). Source was an email from Don Coyne to the milagro mailing list dated 5/4/2004.the milagro mailing list dated 5/4/2004.

Absorption length 30m and Scattering length 50m.Absorption length 30m and Scattering length 50m.

Water propertiesWater properties Rayleigh Scattering – caused be scattering centres smaller than Rayleigh Scattering – caused be scattering centres smaller than λ/20λ/20

Rayleigh scattering length increases with Rayleigh scattering length increases with λλ44 and scattering angular distribution goes like ~(1+cos and scattering angular distribution goes like ~(1+cosθθ22)). . Forward and backward scattering probability equal.Forward and backward scattering probability equal. Rayleigh scattering very longRayleigh scattering very long Geant4 has code to simulate Rayleigh scattering. Calculates the lengths just for water using the Einstein-Smoluchowski Geant4 has code to simulate Rayleigh scattering. Calculates the lengths just for water using the Einstein-Smoluchowski

formulaformula

Water propertiesWater propertiesMie scatteringMie scattering

Mie ScatteringMie Scattering Geant4 doesn’t include code for Mie scattering, so I wrote some code to Geant4 doesn’t include code for Mie scattering, so I wrote some code to

simulate this processsimulate this process Mie’s angular distribution function is hard to solveMie’s angular distribution function is hard to solve If your scattering centers have similar properties (rindex and size) or if If your scattering centers have similar properties (rindex and size) or if

your scattering is dominated by a single species of scatterers your scattering is dominated by a single species of scatterers you can approximate the Mie complicated functions that give the scattering you can approximate the Mie complicated functions that give the scattering

angle distribution with a simple function: the Henyey-Greenstein functionangle distribution with a simple function: the Henyey-Greenstein function Needs only one parameter; the asymmetry factor (g=<cosNeeds only one parameter; the asymmetry factor (g=<cosθθ>)>)

Mie scattering in g4sim Mie scattering in g4sim v2.0 MCv2.0 MC

In the March 2004 Los In the March 2004 Los Alamos meeting Don Alamos meeting Don Coyne presented results Coyne presented results of his water scattering of his water scattering measurements. measurements.

Showed a plot of relative Showed a plot of relative intensity vs scattering intensity vs scattering angleangle

His results would agree His results would agree with a with a Henyey-Greenstein Henyey-Greenstein function of <cosfunction of <cosθθ>=0.9999. >=0.9999. Extremely forward scattering. Extremely forward scattering.

For the g4sim v2.0 MC, I used For the g4sim v2.0 MC, I used <cos<cosθθ>=0.99>=0.99

Differential (top) and integral (bottom) plots of the Mie scattering angle distributions for different asymmetry factors.

The asymmetry factor can be calculated if the refractive index The asymmetry factor can be calculated if the refractive index and the size of the scatterers is known. and the size of the scatterers is known.

There is a program called MiePlot that performs these There is a program called MiePlot that performs these calculations (if only we knew what scatters our light in the calculations (if only we knew what scatters our light in the pond).pond). Air bubbles, Al2O3, other stuff?Air bubbles, Al2O3, other stuff?

My impressionMy impression We will never know what exactly is in our waterWe will never know what exactly is in our water For now, just use a conventional very forward asymmetry factor and For now, just use a conventional very forward asymmetry factor and

if any one performs a reliable measurement of the water scattering if any one performs a reliable measurement of the water scattering angular distribution in the future, then try to change things.angular distribution in the future, then try to change things.

For this version of the Mie scattering codeFor this version of the Mie scattering code There is no energy dependence of the asymmetry factorThere is no energy dependence of the asymmetry factor No back-scattering (easy to add, thought about it after I generated No back-scattering (easy to add, thought about it after I generated

all this data)... all this data)...

Water propertiesWater propertiesMie scatteringMie scattering

PMT propertiesPMT properties

GeometryGeometry g4sim v>1.2 uses a full optical model of the g4sim v>1.2 uses a full optical model of the

PMTPMT Taken from GLG4SIM generic Geant4 Taken from GLG4SIM generic Geant4

application (G. Horton, D. Motta et al) and application (G. Horton, D. Motta et al) and modified and bug fixed for Milagro.modified and bug fixed for Milagro.

PMT modelPMT model The PMT model in the G4 Milagro

simulation tracks the photons until they are converted to photoelectrons or until they are absorbed upon incidence on a surface or until they exit the PMT.

It simulates reection/refraction/absorption at the glass, photocathode, dynodes and silvered surface of the PMT.

The absorption probability of a photon transversing the photocathode material is calculated using its complex refractive index and thickness.

The probability of a photon detection is a product of the probabilities of the following steps:

The absorption of the photon in the photocathode material and the resulting creation of a photoelectron

The liberation of the photoelectron in the PMT vacuum

The collection of the photoelectron at the 1st dynode

PMT modelPMT model The absorption probability is be calculated from the

complex refractive index of the photocathode material and depends on the energy and the incidence angle.

Some of the photoelectrons produced will be liberated into the vacuum and later collected by the 1st dynode.

The probability for a liberated into the vacuum photoelectron to be properly collected by the 1st dynode is called the Collection Efficiency of a PMT (CE).

For the PMT model in the simulation the CE is 1; all liberated in the vacuum photoelectrons are detected. And collection efficiency effects are applied later in milinda.

PMT modelPMT model Let the liberation probability of a

photoelectron into the vacuum be called LP. The Quantum Efficiency (QE) of a

photocathode is defined as the ratio of liberated into the vacuum photoelectrons over the incident on the photocathode photons.

If the probability of a photon being absorbed in the photocathode is called A then

QE = A*LP (1)

PMT modelPMT model The LP is only energy dependent, while the A and

QE are both energy and incidence angle dependent. The QE data from the spec sheets corresponds to

normal incidence, let's call it QEn(E). The absorption probability at normal incidence,

An(E), can be calculated using the complex refractive index and the thickness of the photocathode material.

So the LB(E) can be calculated as LB(E) = An(E)/QEn(E) (2) Deriving LB(E) from (2) and by calculating A(E,θ)

we can derive, using (2), the angular dependence of the quantum efficiency.

PMT testsPMT tests

I tested some Milagro PMTs I tested some Milagro PMTs Mainly, made pulse height distributions Mainly, made pulse height distributions

for different conditions and measured the for different conditions and measured the relative detection efficiency on the relative detection efficiency on the surface of the PMTsurface of the PMT

PMT gain and detection efficiency PMT gain and detection efficiency decrease as we illuminate points far from decrease as we illuminate points far from the top of the photocathode.the top of the photocathode.

I sent a memo out (5/1/2006) about these I sent a memo out (5/1/2006) about these tests.tests.

PMT testsPMT tests

Pulse height distribution for illumination at different positions of the photocathode .

PMT #1024, HV 1800V, PMT vertical

PMT testsPMT tests

PMT testsPMT tests

PMT testsPMT tests

Magnetic field effects PMT illuminated at the top

PMT testsPMT testsRelative detection efficiencyNormal illumination at different points of the photocathode. PMT #1024, 1800V, vertical position

PMT testsPMT tests

Relative efficiency vs incidence angle. Illumination of the top of the photocathode under different angles

Milinda’sMilinda’s initial processing of the MC initial processing of the MC eventevent

Noise additionNoise addition The MC event is read by DataRead_MCASCII5The MC event is read by DataRead_MCASCII5 AddNoise code superimposes noise to the eventsAddNoise code superimposes noise to the events

Cosmic ray noiseCosmic ray noise Simulated 5 GeV – 100 TeV , 90deg zenith angle protons Simulated 5 GeV – 100 TeV , 90deg zenith angle protons

thrown uniformly on the pond with cores distributed to +-thrown uniformly on the pond with cores distributed to +-5km. Kept events with any pmts hit and at most 10AS 5km. Kept events with any pmts hit and at most 10AS PMTs hit. Saved in the config_milagro/noise.dat filePMTs hit. Saved in the config_milagro/noise.dat file

Milinda adds, randomly in time, events from this file until Milinda adds, randomly in time, events from this file until the hit rate of the AS tubes becomes 20KHz. the hit rate of the AS tubes becomes 20KHz.

Dark noiseDark noise 2KHz for Pond PMTs and 20KHz for Outrigger PMTs. 2KHz for Pond PMTs and 20KHz for Outrigger PMTs.

There maybe another source of dark noise in our pond.There maybe another source of dark noise in our pond. Relative rate of muons to single pe hits doesn’t seem to Relative rate of muons to single pe hits doesn’t seem to

agree between MC and data for these noise rates.agree between MC and data for these noise rates.

Muon PeakMuon Peak Number of pes a muon produces on the MU layer Number of pes a muon produces on the MU layer

PMTsPMTs

Muon peak from dataMove the time window of the edge-finder off time, where the pond hits come from other unrelated showers.

Muon peak from the MCAnalyze non triggering data

Milinda’s initial processing of the Milinda’s initial processing of the MC eventMC event

PMT gain and efficiency PMT gain and efficiency correctionscorrections

Milinda code was written to apply the PMT Milinda code was written to apply the PMT efficiency and gain corrections.efficiency and gain corrections.

In g4sim v2.0 the position (distance from PMT In g4sim v2.0 the position (distance from PMT axis) of each pe detection is saved in the axis) of each pe detection is saved in the output file. output file.

Milinda’s code GetPEAmp.cMilinda’s code GetPEAmp.c Discards some of the MC pes based on the position Discards some of the MC pes based on the position

dependent PMT efficiencydependent PMT efficiency Assigns a pulse height to the surviving pes using Assigns a pulse height to the surviving pes using

the pulse height distributions as a PDF. (Does the the pulse height distributions as a PDF. (Does the same for the dark noise hits, using a dark noise same for the dark noise hits, using a dark noise pulse height spectrum)pulse height spectrum)

Milinda’s initial processing of the Milinda’s initial processing of the MC eventMC event

Some notesSome notes If g4sim v1.2 or earlier is usedIf g4sim v1.2 or earlier is used

Can’t have PMT corrections applied. No positions of Can’t have PMT corrections applied. No positions of the pe detections are saved in the MC file.the pe detections are saved in the MC file.

If you were using code with the eventdata bugIf you were using code with the eventdata bug Using Milinda.Reco.Event.EventData->CalData Using Milinda.Reco.Event.EventData->CalData

instead of Milinda.Reco.Event->CalData instead of Milinda.Reco.Event->CalData You bypass the calibration processYou bypass the calibration process

For g3 and g4sim v1.2-v1.3, you get pe smearing, time For g3 and g4sim v1.2-v1.3, you get pe smearing, time jittering with the old g3 method. No noise or calibration on jittering with the old g3 method. No noise or calibration on the eventsthe events

In g4sim v2.0 I stopped saving all the pieces of information In g4sim v2.0 I stopped saving all the pieces of information that milinda would calculate (time jitter and pe smearing that milinda would calculate (time jitter and pe smearing info). info).

So for g4sim v2.0, you would get no time jittering, no pe So for g4sim v2.0, you would get no time jittering, no pe smearing (integer number of pes – nb2&X2 definitely wrong), smearing (integer number of pes – nb2&X2 definitely wrong), no noise, no PMT corrections, no calibrations, no noise, no PMT corrections, no calibrations, no 2.6 spectrum no 2.6 spectrum on the crabon the crab

Further readingFurther reading ““Results from the Geant4 Milagro SimulationResults from the Geant4 Milagro Simulation”, Milagro ”, Milagro

Memo, V. Vasileiou, 04/15/2006Memo, V. Vasileiou, 04/15/2006 ““PMT Tests at UMDPMT Tests at UMD”, Milagro Memo, V. Vasileiou, ”, Milagro Memo, V. Vasileiou,

05/1/200605/1/2006 ““An electronics simulation and improved noise model for An electronics simulation and improved noise model for

MilagroMilagro”, Milagro Memo, A. Smith, 2/4/2004”, Milagro Memo, A. Smith, 2/4/2004 g4sim webpage g4sim webpage http://umdgrb.umd.edu/vlasisva/g4simhttp://umdgrb.umd.edu/vlasisva/g4sim Milinda manual Milinda manual http://http://

umdgrb.umd.edu/milinda/doc/manual/MilindaManual.htmlumdgrb.umd.edu/milinda/doc/manual/MilindaManual.html