Development of Gossip: a gaseous pixel detector

CONCEPTATLAS Upgrade Document No: Institute Document No. Created: 24/07/2008 Page: 1 of 19

Modified: 25/08/2008 Rev. No.: 1.10

Abstract

We propose to develop Gossip, a gaseous pixel detector, as replacement for the present pixel vertex detector. This detector is expected to function well after a dose of 1016 hadrons/cm2. Another advantage of Gossip is the very low power consumption (< 0.1 W/cm2) of the pixel sensor electronics, and the wide range of possible operating temperatures. These properties greatly reduce the mass of the cooling circuit and power lines. Gossip also gives some direction information about the traversing particle: Gossip measures a track segment rather than a single hit point. Finally the costs of a Gossip tracker is expected to be lower than a tracker based on the planar silicon technology, especially because of the absence of bump bondings.

A small segment of a Gossip vertex detector, based on the CMS pixel FE chip (PSI-46) is planned to be realized in 2009 as demonstrator. In 2010, a large scale prototype vertex detector will be constructed based on the TimePix-2 CMOS pixel chip. Since TimePix-2 does not provide the aimed granularity and time resolution, we develop in parallel small prototype detectors using the Gossipo2 and the proposed Gossipo3 chip that do not have these limitations.

Gossip is a specialty of the GridPix TPC where the width of the drift gap is reduced as far as possible (to about 1.2 mm). In the following we shall treat first the principle of GridPix.

Contact Person: Harry van der Graaf (vdgraaf@nikhef.nl)

Prepared by:

H. van der Graaf (Nikhef)

F. Hartjes (Nikhef)

Checked by: Approved by:

Distribution List

ATLAS High Luminosity Steering Group

1 Introduction

1.1 Principle of the GridPix TPC and Gossip

Gaseous detectors are widely used in particle physics in the form of wire chambers since the beginning of the seventies in the last century. While the primary ionisation is mostly too low to get sufficient signal to noise ratio, amplification by avalanche is generally used to overcome this problem. Advantages of gaseous wire chambers are the low costs per detecting surface and the low material budget. However the limited position resolution, rate capability, ageing performance and channel density makes them basically unsuited for the ATLAS inner detector at the sLHC.

A first breakthrough of this situation occurred around 1990 by the development of the Micro Strip Gas Chamber (MSGC). Here the alternating anode and cathode wires were replaced by metal strips on an insulating surface, thus solving the problem of electrostatic instability that prevents wire chambers to have a pitch much smaller than 2 mm. The gas amplification occurred in the narrow region between anode and cathode strip. The MSGC was a competitor of the planar silicon strip detector offering a comparable performance for a lower price. However, the MSGC suffered from serious ageing problems which depended strongly on the cleanliness of the chamber gas. The counting rate was limited by ion drift towards the drift cathode.

Thinned (Slimmed) CMOS pixel chip

MIP track

Micromegas

insulatingspacers

Gas seal foil & cathode

50 um

~1.5 mm

Chamber gas

Fig. 1. Principle of the Gossip detector as a GridPix TPC with a narrow drift volume. **** we should make a drawing with a drafting program rather than Word (in 3D?)****

A second breakthrough came with the invention of Micromegas (Charpak & Giomataris 1995) and GEM (Sauli 1996), enabling a pixeled anode geometry. This line is continued in the GridPix detector where the fine granularity is given by the front-end pixel chip which serves as active anode. As such the GridPix concept can be used for all kind of TPC applications.

From GridPix we devised a speciality with a drift gap in the millimetre region: Gossip (Gas On Slimmed Silicon Pixels), to get a geometry comparable to planar silicon pixel or strip detectors. Its gap width is tuned to be just wide enough to detect minimum-ionizing particles at 99.5% of the hits, and thin enough to have a maximum drift time smaller than 25 ns, to collect all primary ionization within the (supposed) SLHC bunch distance. In Gossip a pixel is formed by the assembly of the detection volume: a short gas column in the drift gap, the associated hole in the Micromegas, and the under laying pad and preamp circuit of the pixel chip. As such an individual stand-alone detector is created with a pitch in X-Y direction that may be as small as 60 m. For this detector we expect at a particle rate of 0.4 GHz/cm2 an occupancy of the order of 0.15% and a position resolution for a traversing MIP track of 50 µm, values that match well the requirements of the next generation experiments at CLIC, ILC and Super LHC.

The functioning of Gossip is illustrated in fig.1. In the gas filled drift volume, clusters of electron-ion pairs are created along the track of a traversing particle. Due to the electric field, the electrons drift towards the Micromegas grid and are subsequently focused into the holes of the grid. The grid is put at about -400 V with respect to the (grounded) anode pixel chip, creating a strong field in the 50 µm thick avalanche gap. As a result, each single electron entering a grid hole causes an electron avalanche of sufficient charge to be detected by the pixel front-end preamp. By recording the arrival time of the avalanche, the original position of the primary electron can be deduced. As such Gossip is a single-electron sensitive Time Projection Chamber (TPC).

Fig.2. The picture shows an opened GridPix detector. The chamber, of which the guard electrode and chip is visible, includes a TimePix chip which has been covered by a 20 µm thick layer of amorphous silicon. On top of this, an InGrid structure was deposited. The drift gap was 30 mm high.

Fig.3. The plot shows a measurement made by this detector of two tracks from a 90Sr source. The fiducial readout surface was 14 mm x 14 mm (256 x 256 pixels, with square pixel pitch of 55 µm). The detector was placed in a magnetic field of 0.2 T with the field lines running vertical.

The Gossip detector has been made possible by two recent technical innovations: InGrid [Appendix 1] and WaProt [Appendix 2]. InGrid is a Micromegas foil that is produced on the surface of the pixel chip by means of photolithographic wafer post processing. With this technology the pillars can be made much narrower such that they fit in the space between two pixels, eliminating the insensitive zone that occurs with the broader pillars of Micromegas. In addition this method, that is well suited for mass production, enables a much better geometric accuracy of the amplification area.

WaProt is a high resistivity layer that is deposited on the surface of the pixel chip as a protection against microdischarges. This protection is essential since the functioning of gas avalanche detectors is always

accompanied by certain degree of microdischarges. Starting with a 20 µm thick layer of amorphous Silicon, we recently got good results with 7 µm of Si3N4. The resistivity of this layer can be tuned with a Si dopant from 1014 to 108 Ωcm.

Equipped with WaProt and InGrid, the pixel anode chip forms the monolithic active readout anode of the gas-filled drift volume.

1.2 Properties of Gossip

Basically, the detection medium in Gossip is gas. Compared to solid state detectors, gas has the advantage of a negligible mass of the detecting medium (only the housing has some mass) but the disadvantage of too low primary ionisation for direct detection. Therefore, the charge signal has to be enhanced by an avalanche process, making operation more critical. Compared to solid state detectors, these properties result for Gossip in the following advantages and disadvantages:

Advantages of Gossip

Gas can be exchanged or refreshed: therefore there is no radiation damage of sensor material.

The (charge) signal can be made sufficiently large by applying gas amplification.

There is no bias current.

In gas ɛr=1: therefore, and for geometrical reasons, the signal source capacity is as low as ~10 fF, allowing low-power, low noise preamps.

Gossip measures, in three dimensions, the positions of all single electrons of a track, left in the gas, by a passing fast charged particle. A track segment is thus measured instead of track point position, and dE/dX information is obtained, and δ-rays can be recognized and omitted.

The technology to produce Gossip detectors is cheap, no bump bonding is required.

Gossip is little sensitive for neutrons and X-rays.

Gossip can operate in a wide temperature range between -30 deg C and room temperature.

The low electronics power dissipation and the higher operating temperature greatly reduces the demands on the cooling system. As a result the mass of the cooling system may be significantly diminished.

Disadvantages & limitations of Gossip:

Deterioration of the position resolution by diffusion of the drifting electrons.

Most hits liberate only a few primary electrons, so s are not easily rejected.

Discharges are possible between grid and pixel chip that may damage or destroy the pixel chip. This problem has been solved using WaProt (see Appendix WaProt).

Possible ageing by deposit on the anode, i.e the pixel chip leading to a rate dependent decrease of the gas gain.

More services: two high voltage lines are needed instead of one (grid + drift anode) + two thin gas lines.

The grid voltage is critical and depends on parameters like the gas composition and the detector geometry. An increase of the grid voltage by -30V gives an increase of the gas gain by a factor of two.

Fig.4. First working prototype of a Gossip detector(left). As FE readout chip the CMS Pixel FE chip (PSI-46) was applied. On top of the chip, which was protected by a layer of amorphous silicon, a Micromegas foil was placed. This unit was covered by a gas/cathode seal foil.

Fig.5. Example of a β-event from a 90Sr source. While most events activate a small cluster of one or more pixels, this event displays a track which must have been running rather

parallel to the chip surface.

1.3 Other possible applications of Gossip/GridPix in ATLAS

a. Gossip as replacement of the SCT Si strip detectors . The essential element of this detector would be a large CMOS chip (30 mm x 30 mm) containing 512 ‘strixels’ (thus 512 pixels with, as input pad, a strip with a length equal to the full width (30 mm) of the chip.

Lower costs are the main advantage of a Gossip Strixel detector with respect to a Si strip detector: compare the price of CMOS and Si sensor material per cm2, and the fact that the functionality and the costs of FE chips is included in the strixel CMOS chips.

The power dissipation of this detector would be a factor 4 less compared to equivalent Si strip detectors. *** we have to work this further out (numbers) or we should omit this statement****

b. GridPix as Transition Radiation Detector. Using a 17 mm wide drift gap and a Xe-filling, Gridpix detects the conversions from soft X-rays that are emitted by transition radiators in front of the detector. These conversions appear as clusters of tens of electrons that are superimposed onto the MIP track,. Several TRT layers could be installed at the outer radius inside the solenoid. *** also here some numbers and documentation (Anatoli?)***.

We plan to submit a separate R&D proposal on this subject.

c. Using GridPix as a LVL1 trigger . In a GridPix detector with a drift gap of ~20 mm, tracks appear as projection on the XY pixel plane. The projected length depends on the angle of incidence, and hence on the momentum of the track. The projected track length could be collected within one µs, and could thus provide LVL 1 trigger.

We plan to submit a separate R&D proposal on this subject.

d. Preshower detector : track sampling/photon discrimination, dE/dX measurement [more description required]

e. Hadron Calorimeter : digital (track counting) hadron calorimeter. The energy resolution and granularity of such a calorimeter would be very competitive, provided that there is sufficient correlation between tracks at both sides of the absorbers. With GridPix, dE/dX is measured precisely, and low-cost CMOS wafers can be applied as readout. *** I believe that in the past gaseous dE/dx detectors were not very successful since they require a very long path to get sufficient primary statistics. Would this be better in this case??***

2 Participating Institutions

****To be completed and confirmed. I suggest that we only mention here senior staff members to make the list not too long ****

Nikhef Amsterdam

o Harry van der Graaf

o Fred Hartjes

o Nigel Hessey

o Els Koffeman

University of Nijmegen

o Adriaan König

o Thei Wijnen

Technical University Twente/MESA+

o Jurriaan Schmitz

Physics Institute of the University of Bonn

o Klaus Desch

o Norbert Wermes

Moskow Physical Engineering Inst. (MePhI)

o Anatoli Romaniouk

o Serguei Morozov

o Seguei Konovalov

Saclay/CEA: Paul Colas, Yannis Giomataris

o ATLAS: Claude Guyot

University of Freiburg

o Andreas Bamberger *** still in function??***

Liverpool?

LBL?

Harvard?

o John Oliver

MIT?

o Ulrich Becker

o Joe Paradiso

Collaboration with CMS: PSI

o Roland Horisberger

o Tilman Rohe

3 Topics, goals and subprojects

3.1 Topics

a. Investigation of the optimal operational parameters. This involves both simulation and experimental verification

o Optimization of the counter gas

o Required gas gain

o Nature and resistivity of the chip protection layer

o Thickness of the drift gap

b. Measurement of the performance of Gossip for MIP tracks

o Efficiency

o Time resolution

o X-Y resolution

o Double track separation

o δ-ray identification and suppression

o Background dependence for neutrons and X-rays

o Dependence of the gas gain on temperature and pressure

c. Development of the robustness of operation in a harsh environment

o Spark protection

o Radiation tolerance for MIPs until 1016/cm2

o Rate dependence up to 1 GHz/cm2

d. Design of a generic Gossip tracker

o Mechanical support structure

o Services

o Cooling (CO2 based)

o Gossip hybridization (MCM)

o Dedicated Gossip Read Out Chip (ROC) optimized for low input capacity and negligible input current (possibly TimePix-3)

3.2 Final goal of the Gossip R&D

a. The development of a Gossip tracker as replacement for the present pixel inner tracker of ATLAS.

b. Developing a Gossip tracker for the B-layer replacement: GOAT-1 (GOssip-ATlas), prior to the full upgrade of the ATLAS ID provided that the technology is available in time. GOAT-1 has been simulated (see appendix). ****how far are we with this??***

3.3 Subprojects

Based on the R&D topics that are formulated above, we aim to pursue the following subprojects.

3.3.1 Gossip demo based on the PSI-46 ROC

A Gossip made from a single PSI-46 chip equipped with a HV-protection layer and InGrid is at present being investigated. A test beam experiment is foreseen where a stack of these Gossip detectors will be placed in a silicon telescope. This will give us experimental numbers for the efficiency, position resolution and double track separation. However, this chip is made by an older technology and is optimized for silicon sensors. As a result, both the noise level and the granularity are worse than aimed for at Gossip and there is practically no drift time measurement.

3.3.2 Gossip demo based on the Gossipo chip

Basically, the existing Gossipo-2 chip, designed in 130 nm technology, has completely the aimed performance in terms of pixel size (55 x 55 µm), time resolution (1.8 ns) and input noise **NEC. However, the minor dimensions of the readout pad matrix (16 x 16 pixels => a square of 0.88 mm) make it hard, but not impossible, to measure the position resolution in the XY plane. For Gossipo-2 the variations in the pixel settings vary such that at best 85% of the pixels are operational. Therefore we will use in a later stage the anticipated Gossipo-3 chip that will have more uniform pixel settings.

Fig.6. A Gossipo-2 chip equipped with an Ingrid structure. The SU-8 insulator carrying the (black) aluminium mesh is visible as a transparent square.

3.3.3 Gossip demo based on the TimePix-2 ROC

The planned TimePix-2, to be designed in 130 nm technology, will also have a 1.8 ns resolution in the arrival time measurement of the hit, enabling track reconstruction in the 1.2 mm wide drift gap. Since the chip is also designed for use with moderately irradiated silicon, the front-end performance will not be optimal for Gossip.

3.3.4 Upgrade of Gossipo-2 to Gossipo-3 and Gossipo-4

We plan to design an improved pixel front-end to overcome the range problems that were experienced with Gossipo-2. To facilitate tracking, Gossipo-3 will possibly have more pixels than Gossipo-2. For Gossipo-4 we will try a readout architecture comparable to Timepix-2.

3.3.5 Development of TimePix-2

As a follow-up of the existing TimePix (in 250 nm technology) we plan the following modifications:

1. Convert the design to 130 nm technology

2. Implement the Gossipo frontend

3. Adapt the digitizing parameters to the Gossipo design

a. 55-60 µm pixel size

b. 560 MHz TDC clock

To enable a larger scale testing, the chip will be produced in an engineering run

3.3.6 Development of the Gossip Multi-Chip Module (MCM)

We will design a structure supporting a number of ROCs (16?) that includes the cooling circuitry, HV and LV distribution structure, data communication lines, and the gas circuitry. To minimise the dead inter-chip zone, we will use the seamless technology that is developed by the ReLAXD collaboration. Prototyping will include thermo-mechanical simulations and testing.

3.3.7 Development of a track fitting routine

Basically, with Gossip the track can be reconstructed from the individually arriving electrons. However, the performance of this concept may be greatly spoilt by gas diffusion and time slewing of the front-end electronics. Using the data from future test beam experiments, we have to find out if track fitting is useful at all or whether we may better average the XY locus of the recorded hit points.

3.3.8 Engineering study for a stave structure using Gossip MCMs

Various options like a string concept will be simulated using finite element calculations and thermal modelling, and prototyped. Finally this study should end in the construction of a stave for the B-layer. Topics are the implementation of cooling, distribution of electrical power, and design of a minimal mass data transport.

3.3.9 Industrializing of the dedicated wafer post processing

Presently InGrid and WaProt are only applied on a lab scale by processes with limited reliability. Studies will be done to come to a reliable process that is applicable in industry. At present we are in contact with IZM-Berlin about this.

3.3.10 Radiation tolerance studies

To study the performance of Gossip in the hottest part of the SLHC where the radiation environment is dominated by MIPs, we use a strong 90Sr source with radiation levels up to 1.3 GHz/cm2. At this rate a monthly dose of 3.3 * 1015 MIPs/cm2 can be obtained. We aim for a radiation tolerance for MIPs of at least 1016 cm-2. In addition the test also provides very useful information on the reliability of operation at such a high rate. Samples that pass this severe test will be tested at a later stage at the CERN mixed source facility and irradiated with a high dose of neutrons.

Parallel to this study we plan to use a generic ageing set-up where electrons are liberated from the drift cathode by means of UV light. For this set-up only non-suspect materials will be used like stainless steel and ceramics. This opens a way to easily trace possible ageing compounds in the chamber gas.

3.3.11 Simulation studies

In the framework of RD51 we will investigate by simulation the full functionning of gthe Gossip detector

o Optimization of the counter gas for Gossip application. Possibly the outcome of the radiation tolerance studies will give constraints on the gases that can be used. The following parameters will be investigated:

Ionisation density

Cluster density

Drift velocity

Diffusion

o Drift and diffusion of single electrons

o Gas amplification process

Shape of the charge signal distribution, comparing to existing experimental results from Gossip prototypes.

Determining the Pólya distribution factor from experimental results

o Electrical charge signal development

o Modelling of discharges

Best way to reduce the charge to a pixel front-end in case of a discharge

o Developing a track reconstruction algorithm. With Gossip the locus of a traversing track is reconstructed from a number of individual hits from mostly single electron events.

4 Relation to existing efforts

A number of existing R&D activities is related to the Gossip research mentioned in this proposal.

o The Gossip technology with a bigger drift distance, GridPix, is proposed for tracking at CLIC and the ILC. Related with EUDET WP2

o Next-Quad: a planar suspension of 4 Timepix chips to be applied in a TPC. Ongoing project at Nikhef

o Activities of the ReLaxeD collaboration: a planar suspension & cooling of 64 Medipix/Timepix chips forming a large area pixel

o Detector. EUREKA/SenterNovem project (FOM-Nikhef/IMEC/Panalytical/Canberra Olen)

o Activities of the RD-51 collaboration: R&D on Micro Pattern Gas Detectors in various technologies

Large-area GEM and Micromegas detectors

ATLAS Muon Upgrade

o Si-TPCs for CLIC & ILC

o Mini/Micro TPCs for Double β Decay and WIMP search experiments

o Activities of the TimePix-2 Consortium (CERN, Nikhef, Saclay, Bonn, Freiburg, Panalytical)

Development of TimePix-2

o The DICE experimentt (Delft University of Technology, Nikhef)

Measurement of Internal Conversion using a µ–TPC.

5 Time schedule

5.1.1 Gossip demo based on the PSI-46 ROC

A few prototypes already exist, lab tests and beam tests to be completed by mid 2009

5.1.2 Gossip demo based on the Gossipo chip

Gossip-2 tracker operational end 2008, beam tests finished by mid 2009. Gossipo-3 tracker test possibly by mid 2010, depending on the availability of the chip/

5.1.3 Gossip demo based on the TimePix-2 ROC

Time schedule dependent on the availability of the chip, research possibly finished by end 2010.

5.1.4 Upgrade of Gossipo-2 to Gossipo-3 and Gossipo-4

Gossipo-3 could be available by mid 2009, Gossipo-4 by end 2009.

5.1.5 Development of TimePix-2

Possibly available by end 2009/ mid 2010.

5.1.6 Development of the Gossip Multi-Chip Module (MCM)

Basically an ongoing activity that will continue until the start of mass production. Possibly a viable solution by mid 2010.

5.1.7 Development of a track fitting routine

Again an ongoing activity. It should be possible to have a viable solution by end 2009.

5.1.8 Engineering study for a stave structure using Gossip MCMs

An activity running parallel to the development of the MCM with a comparable time schedule.

5.1.9 Industrializing of the dedicated wafer post processing

We hope to have a workable process developed by mid/end 2010.

5.1.10 Radiation tolerance studies

An ongoing study where results for practical use are expected by mid 2009. To be continued until at least end 2010.

5.1.11 Simulation studies

o Optimization of the counter gas for Gossip application. Mid 2009 we should appoint a favourite.

o Drift and diffusion of single electrons, gas amplification, electrical charge signal development, modelling of discharges: to be completed by mid 2010.

6 Resources

6.1 Existing resources

Source Manpower (fte/year)

Material budget (k€/year)

Nikhef: budget for generic detector R&D 5 70

Nikhef: ATLAS upgrade budget 2 30

Medipix-3 Consortium ??? ???

Eudet: prototyping with TimePix-1 ROC pm pm

STW project “There is plenty of room at the top” 1? 20?

MarieCurie grant 2 pm

6.2 Expected resources

Source Manpower (fte/year)

Material budget (k€/year)

Nikhef/Bonn/CERN/Saclay/Freiburg/: TimePix-2 development 3? 150?

STW proposal Nikhef/Univ. Twente: “There is even more room at the top”

1 50?

7 Appendices

7.1 Appendix InGrid

7.2 Appendix WaProt

7.3 Appendix GOAT-1

8 References

[1] a backup document.