Studies on signal formation processes in Micromegas - heading towards the ATLAS NSW Upgrade - Fabian Kuger 1,2 under supervision of Thomas Trefzger 1, Raimund Ströhmer 1, Rob Veenhof 2 and Christoph Rembser 2 1 Julius-Maximilians-Universitä ̈ t Würzburg (Germany), 2 CERN October 28 th 2015 Gentner Day Oct CERN sponsored by the

2 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Introduction In one sentence: The first stage of the ATLAS forward muon system is going to be replaced, using a fairly new technology: Micromegas…

Micromegas detectors 3 Studies on signal formation in MM Gentner Day CERN October 28th 2015 ❶ Primary ionization / conversion ❶ ❷ Electron drift (diffusion, attachment…) ❷ ❸ ❸ Electron losses at the mesh ❹ ❹ Electron amplification + ❺ signal capture, readout, amplification… What is a micromesh gaseous structure (Micromegas) ? How does a Micromegas detector work? Which processes contribute to the signal formation in a Micromegas?

Processes in a Micromegas - Which one to talk about? - 4 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Hurray ! Damn ! ❸ Electron losses at the mesh 10 4 – 10 5 electrons ? 1 electron ❹ Electron amplification Problem: 20” Presentation can never cover all these aspects in satisfying detail. Make your choice now!  Solution: YOU choose the topic to be discussed in detail:

Option II Understanding avalanche formation in Micromegas 5 Studies on signal formation in MM Gentner Day CERN October 28th 2015

6 Studies on signal formation in MM Gentner Day CERN October 28th 2015 ❶ Primary ionization / conversion ❷ Electron drift (diffusion, attachment…) ❸ Electron losses at the mesh ❶ ❷ ❸ ❹ ❹ Electron amplification + ❺ signal capture, readout, amplification etc. Avalanche formation - What it’s about!? - 1 electron 10 4 – 10 5 electrons ? E amp Within each gaseous detector a small number of initial electrons has to be amplified to produce a recognizable signal. In a Micromegas single electrons enter the strong E-field (≈ 40kV/cm) in a the very thin (128µm) amplification gap, where they cause an electron avalanche.

7 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Amplification of electrons into electron avalanches is a basic process in gaseous detectors. However, the details of avalanche development are not fully understood / quantified. Simulated electron avalanche, only every 1000 th collision is color tagged: elastic scattering ionization excitation - electron path e1-e1- e1-e1- Ar elastic scatteringexcitation e1-e1- e1-e1- Ar* iC 4 H 10 ionization e1-e1- e1-e1- Ar + e2-e2- ionization rate (first Townsend coefficient α) + Penning transfer e1-e1- e1-e1- Ar iC 4 H 10+ e2-e2- excitation rate(s) and transfer probability r p. rprp E Ar* > E iC 4 H 10 + ✓✓✓✗ Avalanche formation - Electron scattering processes -

8 Studies on signal formation in MM Gentner Day CERN October 28th 2015 The occurrence of secondary avalanches - feedback - influence mean gain and variance of the gain distribution. Primary avalanche β Ar + Ion feedbackPhoton feedback Ar* Avalanche formation - Secondary avalanches - Often primary and secondary avalanches are indistinguishable (space- & time-wise) and thus recognized as one avalanche!

Avalanche formation - How to describe avalanches? - 9 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Analytical Calculations (Macroscopic Model) Exponential avalanche growth is parameterized by the first Townsend coefficient α. New processes can be included by additional parameters.  Only mean gain is described by This macroscopic model Experimental Measurements Avalanche formation is a statistical process. Experimental data can be best described by a Polya-distribution, parameterized with a mean gain G and a relative variance f Montecarlo Simulations (Microscopic Model) Avalanche is grown step-by-step on microscopic-electron-tracking and - scattering level New processes can be modeled for each scattering process  Mean Gain and relative variance can be accessed in microscopic MC

10 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Single Electron Response – Experiment Single electrons are produced using a 337nm laser, targeting a 100µm spot on the NiCr layer through a quartz window (drift side of the MM). Gas: 95% Ar/He/Ne + 5% iC 4 H 10, 748 torr UV laser Drift region 3,2mm - 0.9kV/cm Amplification region 160µm kV/cm Segmented readout Ni-Mesh 333lpi-70% Quartz window NiCr layer 0.5nm  SER spectra: mean and variance G =  = 2.3 ± 0.1 f = 0.30 ± 0.01 SER spectrum in Ne mixture with noise peak (RMS: 380e-) and signal region and Polya fit. Experimental Measurements

Avalanche formation - Calculations - 11 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Analytical Calculations (Macroscopic Model) Exponential avalanche growth is parameterized by the first Townsend coefficient α. Penning Transfer can be included by modifying α  α Penning excitation e1-e1- e1-e1- Ar* iC 4 H 10 + Penning transfer e1-e1- e1-e1- ArAr iC 4 H 10+ e2-e2- E Ar* > E iC 4 H 10 + Secondary avalanches are described with the second Townsend coefficient β Ar *

Comparison Exp - Calc 12 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Penning transfer propabilities and the second Townsend coefficient are extracted as fitting parameters to the experimental data. Only mean gain can be compared, gain variance can not be accessed.

13 Studies on signal formation in MM Gentner Day CERN October 28th 2015 SER Simulation in Garfield++ avalanche ending plane Drift region 3,2mm - 0.9kV/cm Amplification region 160µm kV/cm electron starting point SER events are simulated in a microscopic avalanche model, using a two volume simulation for the electron drift and the avalanche development / amplification. All electrical fields are assume to be homogeneous and adopted as analytic fields. (Parallel plate approximation) Simulation of the drift process respects the starting conditions of the electron initiating the avalanche. Penning transfer rate(s) are an input parameter to the simulation, feedback can not (jet) be taken into account  Microscopic approach yields mean amplification and variance of the gain, spectra are available.  The individual treatment of each electron requires huge calculation time, sufficient statistic can not be simulated for high fields. ↯ ✓ Montecarlo Simulation (Microscopic Model)

Avalanche extrapolation How to access mean and variance for huge avalanches (> 10 5 e - ) with significant statistics? Approach: Avalanche development is a statistical process. As soon as the avalanche has sufficiently grown, single fluctuations in further amplification processes are averaged out. full (physics) simulation mathematical extrapolation  First amplification step dominates the size of the avalanche, and has to be fully simulated until avalanche reaches sufficient size (~100 e - ).  The further (CPU-consuming) amplification steps are dominated by statistics and can be handled with an mathematical extrapolation. Assumptions: - During the first step, a single starting electron is amplified into n electrons, with a probability p 1 (n). - This probability depends (mostly) on the spatial step length and the electric field the electron senses within and the gas mixture.  same conditions during different steps lead to equal p 1 (n) within each step. 14 Studies on signal formation in MM Gentner Day CERN October 28th 2015

Avalanche extrapolation  The probability of the avalanche to contain n electrons after the second step p 2 (n) is  Induction yields for the third and the l th step: Mathematical derivation of the mean, RMS and RMS/mean after the n th step yields: (All comparison between full MC and ‘first-step’ MC + extrapolation are in agreement. The model is consistent with Poisson and Legler models on avalanche development.) 15 Studies on signal formation in MM Gentner Day CERN October 28th 2015

Simulation studies Penning transfer rate (gas dependent) have been derived, using one experimental data point (mean gain).  r P NeIso95:5 =  For all simulations the full avalanche has been computed for low e-fields (≤ 27kV/cm) and ‘first-step-simulations’ were performed in the full range (20-35kV/cm).  The step-number-exponent n has been determined in a control region: 25-27kV/cm.  Mean, RMS and relative width could be calculated. Two kinds of simulations have been performed: ‘Direct-ionization-only’ simulations and ‘Penning-transfer included’ simulations.  Only combination of both yields the impact of penning transfer on the avalanche size. 16 Studies on signal formation in MM Gentner Day CERN October 28th 2015

Comparison Exp – Calc – Sim 17 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Difference in gain estimation between Simulation and calculation  resulting from slightly different e - starting conditions, closer to reality in Sim  Sim yields a slightly higher penning transfer probability Simulation describes the mean gain and gain variance. (although with descrepancies)

18 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Results: Experiment and Simulation Experimental data Simulation Gauss fit (exp. data – noise) Gauss + Polya fit (exp. data)

Summary An extrapolation method to deal with large avalanches in microscopic simulation has been developed, tested and applied. SER – experiments and simulations delivered data to:  … understand and quantify the impact of penning transfer;  r P NeIsob95:5 =  … observe feedback and its increasing influence in high e-fields;  … compare SER spectra, where agreement is at an astonishing level. Studies including Ar + iC 4 H 10 and He + iC 4 H 10 lead to a deeper under- standing of the fundamental differences between the three noble gases.  ‘T. Zerguerras, et al: Understanding avalanches in a Micromegas from single- electron-response; NIM A 772, p.76-82, 2015 [doi: /j.nima ] 19 Studies on signal formation in MM Gentner Day CERN October 28th 2015

Acknowledgements  Thomas Zerguerras and the Orsay research group for the experimental setup, measurements and result, making this work possible  Özkan Sahin for the analytic calculations and the improvements on the macroscopic model  Rob Veenhof for fruitful discussions on simulation details  the Wolfgang-Gentner Program (BMBF) and the ATLAS MUON group for funding 20 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Sincere thanks to… You - for your attention -

Backup for Option II Understanding avalanche formation in Micromegas 21 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Backup - Experiment MM layout 22 Studies on signal formation in MM Gentner Day CERN April 29th 2015

23 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Option I Understanding electron transparency and electron losses in Micromegas 24 Studies on signal formation in MM Gentner Day CERN October 28th 2015

25 Studies on signal formation in MM Gentner Day CERN October 28th 2015 ❶ Primary ionization / conversion ❷ Electron drift (diffusion, attachment…) ❸ Electron losses at the mesh ❶ ❷ ❸ ❹ ❹ Electron amplification + ❺ signal capture, readout, amplification etc. Electron transparency - What it’s about!? - Within each gaseous detector the initial electrons may be lost before being amplified into a recognizable signal. In a Micromegas the electrons have to pass a fine metallic mesh dividing the conversion-/drift- region from the amplification gap. If they end up on the mesh surface, they are lost. Hurray ! Damn !

Electron transparency - How to describe electron transparency? - 26 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Electron transparency T of the mesh has a microscopic definition Experimental Measurements - ‘Standard procedure’ for T measurements -plenty of results from single configurations, but no systematic studies Montecarlo Simulations (Microscopic Model) Allow direct access to counting variables.  Prediction of T possible, if the simulation / model can be validated! How to get information about the electron transparency T? Analytical Calculations (Macroscopic Model) No (satisfying) analytic model exists for calculation of electron transparency!  ‘ultimate goal’ of the ongoing studies  mesh geometry (wire-diameter, mesh-opening -> open area, mesh thickness & structure)  e - approaching behavior gas (mixture, T, p) E-field (geometry, voltages) T mainly depends on

Concept of an Exchangeable Mesh Micromegas (ExMe) 27 Studies on signal formation in MM Gentner Day CERN October 28th 2015 To study different meshes under ideally comparable conditions we designed and build a Micromegas with an exchangeable mesh: - Following the floating mesh concept (as for seen for ATLAS NSW Micromegas) - Independent mesh frames allow easy mesh exchange (ATLAS NSW: mesh will be fixed on the drift panel)  Most detector inherent parameters (amplification- and drift gap thickness, readout surface etc.) are kept constant and allow direct mesh comparison ❸ ❶ ❷ ❹ ❸ Mesh frame ❶ Drift panel (stiff-back, internal gas lines, drift cathode, springs) ❷ O-Ring ❹ Readout panel (readout strips, Kapton foil with sputtered resistive pattern, connectors…) Schematic view of the ExMe1 components

28 Studies on signal formation in MM Gentner Day CERN October 28th 2015  Single processes can not be measured directly  Spectrum of the signals induced by γ from Cu-X-Ray  Position of the K α -Peak (and Ar-Escape-Peak) Experimental Setup ❸ mesh losses ❶ primary ionization ❷ electron drift ❹ amplification  Disentangling a single process ( ❸ ) requires control and stabilization of all other S = Signal (measured in ADC counts), A = fraction of electrons lost to attachment, G = Gain per electron ❶ ❷❸❹❺ On the experimental side… underlying assumption

Experimental data analysis 29 Studies on signal formation in MM Gentner Day CERN October 28th 2015  Variation of the drift voltage causes a systematic shift of the spectrum, and accordingly the K α -Peak  signal loss corresponds to a loss of amplified e -  caused by decreased Transparency with increasing U drift (increase of mean energy for electrons approaching the mesh) K α -Peak Position over U Drift Normalization to Maximum = 1  Transparency  Normalizing the curves corrects for all proportional factors

Results – Experiment only 30 Studies on signal formation in MM Gentner Day CERN October 28th 2015 MeshT | U drift =300V ,5% ,3% ,6% ,0% Comparing the experimental results for different meshes: Experiment `stand alone` interpretation mesh-opening & wire-diameter [μm]

Modeling and simulation approach 31 Studies on signal formation in MM Gentner Day CERN October 28th 2015 On the simulation side…  Full simulation of e.g. a single 8keV-γ-event (by chaining the simulation of all processes) is extremely CPU intensive  Single process simulation is more reasonable (statistics) To simulate the mesh transition ❸ :  the geometry has been modeled and the electrical field calculated using FEM (Ansys) ❸ mesh losses ❶ primary ionization ❷ electron drift ❹ amplification  electron microscopic tracking performed in Garfield++  Electron end points yield T

Comparing Exp & Sim for mesh Studies on signal formation in MM Gentner Day CERN October 28th 2015 Comparing simulation and experimental results reveals: Simulation Experiment ✗ Systematic deviations at the low and high drift field region EXP vs. SIM Mesh Good agreement in the overall T estimation

Comparing Exp & Sim for mesh Studies on signal formation in MM Gentner Day CERN October 28th 2015 More pronounced discrepancies with other Exp data: ✗ Agreement looks less convincing EXP vs. SIM Mesh Simulation Experiment ✗ Systematic deviations at the low drift field are much more pronounced ✗ Same ‘crossing’ of simulated and experimental data is observable at high U Drift

Comparing Experiment and Simulation 34 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Reviewing the underlying assumption ❸ mesh losses ❶ primary ionization ❷ electron drift ❹ amplification Simulation of the attachment losses during electron drift ( ❷ ) :  crucial effect of small gas impurities (> % O 2 ) at low U Drift +O 2 +CO 2  non-negligible attachment to CO 2 at high e - energy T only - Norm = 1 T *(1-A) - Norm = Max sim = 1 T *(1-A) - Norm = Max sim < 1 rejectedrevised assumption ❷❸

Results – combined interpretation: Experiment & Simulation 35 Studies on signal formation in MM Gentner Day CERN October 28th 2015 MeshT | U drift =300V ,5% ,3% ,6% ,0% Comparing the experimental data under interpretation of simulation results: MeshT *(1-A) | U drift =300V ,9% ,4% ,3% ,4% Experiment `stand alone` interpretation

Summary  Simulation-Experiment-Comparison enlarged our understanding of the process of signal formation in Micromegas  Importance to double-checking experimental assumptions (and interpretation of data) by correct modeling & simulation!  Experimental measured curves are effected by transparency- and attachment losses, both effects can be separated using the simulation  good agreement between the data measured and the simulation results  To exploit the full data taken a careful analysis is required (and in progress) 36 ExMe Transparency studies DPG Frühjahrstagung Wuppertal October 28th 2015 MeshEXP: T | U drift =300V EXP & SIM: T *(1-A) | U drift =300V ,5%86,9% ,3%98,4% ,6%61,3% ,0%93,4%

Acknowledgements  Rui de Oliveira and the CERN PCB workshop for assistance in design and production of the Exchangeable Mesh Micromegas prototype  the CERN RD51 Laboratory, in particular Patrik Thuiner & Eraldo Oliveri for their support during the experimental set-up, data-acquisition and -analysis  Rob Veenhof for fruitful discussions on simulation details  the Wolfgang-Gentner Program (BMBF) and the ATLAS MUON group for funding 37 Studies on signal formation in MM Gentner Day CERN October 28th 2015 Sincere thanks to… You - for your attention -

Backup for Option I Understanding electron transparency and electron losses in Micromegas 38 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Backup – ExMe detailed layer description 39 Studies on signal formation in MM Gentner Day CERN April 29th 2015 Drift panel - with internal gas distribution and HV conduct - mounted on honeycomb + FR4 stiff-back - carrying springs pressing down the mesh frame O-ring placed between external FR4 frame (5mm) and mesh frame (4mm+springs) Mesh frame Mesh glued on lower side, aligned with r/o board via pins in the corner. Ground contact to copper ground on r/o plane. (! Non-flatness of the frame due to mesh tension ~500µm!) Readout panel - copper readout strips routed to Panasonic connectors - Kapton ™ foil with sputtered - resistive pattern - cover lay (128µm pyralux) with pillar structure and ‘frame’ to define mesh boarder height - glued outer FR4 frame - connectors for HV, r/o (Panasonic) and grounding

Backup ExMe Study subjects 40 Studies on signal formation in MM Gentner Day CERN April 29th 2015 A variety of mesh specification details can be studied: - different wire diameter - different openings with same wires - no/ soft / strong calendared meshes - different types of weaving (plain vs. twill weave) - alternative mesh material (metalized synthetics)  First measurements show the severe impact of mesh geometry on gain behavior The ExMe readout is divided in four sectors, covered by different spaced pillar patterns. (Pillar-arrangement in regular triangles with different side-length between 5-10mm)  Impact on gain behavior is observed Second ExMe chamber is available, where the sputtered resistive layer is replaced by a screen-printed one. Sector C 8,5 mm Sector D 10 mm Sector A 5 mm Sector B 7 mm

Matching Data to impurities concentrations Problem: Oxygen-Impurities in <0,1% concentration varies during hours / days  Different curves / data taken must be individually matched to the right concentrations Simulation Attachment losses over 5mm drift 41 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Matching Data to impurities concentrations Follow-up question: Does this tiny (<0.1%) O 2 impurity effect the gas properties sufficiently to influence transparency simulations (simulated with pure ArCO 2 93:7) ? Deviation in v D only in T=100%-range  No impact on T simulation no systematic change in d T (deviations are statistical fluctuations)  No ‘adjusted’ Transparency simulation (due to tiny O2 impurities) necessary! 42 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Matching Data to impurities concentrations Experiment Simulation Convincing overall agreement for all meshes ✗ Still systematic deviations  other gas impurities(H 2 O …) ? Exp? Sim? 43 Studies on signal formation in MM Gentner Day CERN April 29th 2015

Matching Data to impurities concentrations Simulating impact of H 2 O contamination (very recent results!): Higher H 2 O concentrations (<1%) increase attachment at high U drift …  Tiny H 2 O impurities are unlikely to cause the visible systematic effect!  Further possible contaminants to be studies Almost no extra attachment for tiny (<0.1%) H 2 O impurities No significant effect on v D or d T … and have significant impact on v D and d T  Adjusted Simulation of T necessary?  Verify/ reject H2O concentration in experiment 44 Studies on signal formation in MM Gentner Day CERN April 29th 2015