Skip to end of metadata
Go to start of metadata
Table of Contents

The SPiDeR Research programme

The research goal of SPiDeR is to further develop monolithic silicon active pixel sensors for future particle physics experiments and to demonstrate their viability for vertexing, tracking and calorimetry applications.

Monolithic silicon pixel sensors for particle physics - the scientific case

As particle physics colliders reach higher energies and luminosities, finely segmented pixel detectors will be needed for vertexing, tracking and even for electromagnetic calorimetry where high granularity has been shown to be important for jet resolution using particle flow algorithms (PFA). These algorithms require efficient separation of the jet into its constituent particles, with each particle being measured accurately in the detector with the best resolution. Specifically, charged particles are measured in a tracker, photons and electrons in an electromagnetic calorimeter and neutral hadrons in a hadron calorimeter. The critical issue for the PFA approach is then to correctly assign all the energy in the calorimeters to avoid double counting; this so-called "confusion" term can dominate the jet resolution obtained. The best approach to reduce such confusion is to design the detector specifically with PFA in mind. These requirements are nicely illustrated in Figure 1 which shows a simulated event. The challenge is to provide sufficient measurements along the path of a track so that it can be reconstructed and its momentum can be measured precisely. The high density of tracks close to the interaction region and again in the outer calorimeter region demand detectors with high granularity. The innermost tracking layers need few micron precision, for reconstruction of displaced vertices. In the intermediate tracking region the emphasis is on minimizing the total amount of material. This is quantified in the following paragraphs and summarised in Table 1.

Figure 1 Simulated event in a future Detector Design


The function of a vertex detector is to provide precision spatial measurements of tracks close to the primary interaction vertex without introducing too much material. The innermost layer is placed as close as possible to the interaction vertex (typically ~1 cm). To distinguish the tracks from displaced vertices the resolution needs to be a few microns, implying a pixel size of 20 μm or smaller and analogue readout. Typical occupancies through the bunch train are on the order of 100 hits per square mm or higher. Some form of time-stamping is required to get the occupancy per time-slice below ~ 1% as required for pattern recognition.
The material per measurement layer is a crucial parameter if multiple scattering is not to spoil the measurement of track angles. Thin layers of silicon plus supports can be realised with a material budget of 0.1% radiation lengths (X0) provided the required power can be restricted to sufficiently low levels so that liquid cooling is not required.

The radiation level expected at

LaTeX Markup problem

Show/Hide latex markup

Show/Hide latex error

machines is typically 10 krads/year or
LaTeX Markup problem

Show/Hide latex markup

Show/Hide latex error

. This is 4-5 orders of magnitude lower than expected at hadron colliders.
Silicon technology is the only proven technique for realising the small pixel sizes required. In our proposal we investigate the use of a silicon sensor with in situ charge storage for time slicing, which has the potential to decrease the power and hence material budget for burst applications such as the cold superconducting linear collider. While not able to subdivide the much shorter bunch trains at CLIC , this is not needed due to the far lower luminosity per train. Readout spanning a number of CLIC bunch trains will deliver adequate background rejection.


The function of a tracking detector is to provide sufficient granularity and measurement precision to be able to find and reconstruct all charged particles, again without introducing too much material and to provide a precise measurement of the curvature. Achieving the required momentum resolution requires an rφ precision of ~15  μm. Pattern recognition for track reconstruction demands the detector occupancy to be less than 1%. This leads to two possible approaches - either long strips with time-stamping capability or higher granularity with no or limited time stamping. It is this second approach which we want to explore as it has the advantage that it can be used both in the barrel and endcap regions. The material per measurement layer is again a crucial parameter if multiple scattering and conversions are not to spoil the measurement which demands a minimal number of I/O lines and minimal power dissipation. The overall goal here is to minimise the material in front of the electromagnetic calorimeter. The radiation levels in the tracker are expected to be an order of magnitude lower than for the vertex region .

In this proposal we wish to investigate a novel solution which is an evolution of the ISIS and TPAC developments. Pixels will be realised using the 4T imaging technology which has recently become available to us and will provide better noise performance than previous 3T architectures like TPAC. It provides efficient charge collection and storage with low power. The readout will be realised as a conventional rolling shutter, with the columns subdivided to give the required time slicing. It is an important question for our R&D programme to determine whether this approach using a pinned photodiode or the ISIS approach using a photogate will provide full detection efficiency for minimum-ionising particles (MIPs) with compact clusters and adequate precision. This could be more challenging with the larger pixels (~25-50  μm) needed for tracking. The readout and memory buffers will be distributed over the pixel area with no dead space using the deep p-well INMAPS technology, which was pioneered for TPAC. The time-stamp resolution will be adjustable from the full 1 ms in the barrel down to 0.1 ms in the very forward region. Power dissipation even in the forward region is expected to be compatible with gas cooling. The radiation tolerance of the structure will be investigated to understand if the principle can be developed for other applications.

Digital electromagnetic calorimetry

Any foreseen future collider detector will need to reconstruct quark final states through hadronic jets. There has been an intense study of techniques to give good jet energy resolution over the last few years and the consensus is that Particle Flow Algorithms (PFA) are most likely to give the best performance. For electromagnetic calorimeters, silicon-tungsten (Si-W) sampling calorimeters are the favoured technology for a PFA-based calorimeter, as the tungsten absorber has a small radiation length (3.5mm) and Molière radius (9mm) while the silicon sensitive layers can be made very thin. Studies so far have used silicon detectors with diode pads with dimensions of several mm and have reconstructed the shower energy through an analogue measurement of the energy deposited in the silicon.

Figure 2 Comparison of ideal analogue and digital ECAL resolutions without any detector effects.

An alternative approach is instead to attempt to count the charged particles crossing the silicon detectors and use this as the shower energy estimator. Both the number of charged particles and the energy deposited in the silicon are on average proportional to the total energy of the electromagnetic shower. However, the energy deposited arises because of the passage of the charged particles through the silicon. The actual energy deposited is not a constant for each particle but depends on the path length and velocity of the particle, as well as having a fluctuation described by the Landau function. This means the energy deposited has extra effects which smear the value over and above the fundamental shower fluctuations which contribute to variations in the number of charged particles. This is illustrated in Figure 2, where an idealised simulation shows the fundamental shower resolution obtained when using the analogue energy deposited or the digital number of particles. Here, idealised means no instrumentation effects (electronics, etc.) is added to the analogue value and the number of particles is counted perfectly, so showing the fundamental differences of the two approaches. The basic ideal resolution of the analogue case is seen to be around 50% worse than for the digital case. Resolutions close to the ideal analogue value have been achieved with low noise systems. However it is not clear how close it will be possible to get to the ideal digital value in practise; to determine this is the major aim of this Workpackage.
A sensor for a digital electromagnetic calorimeter (DECAL) detector must provide a measurement of the number of tracks created in a shower. This can be achieved by binary readout of pixels with a threshold set to be efficient for MIPs at normal incidence. This requires the probability of two charged particles crossing the same pixel to be low. Although the average incident track density is much lower than is the case for vertexing, the local density within a shower is very high and can be up to about 100 particles/mm2 in the core of high energy electromagnetic showers. The CALICE-UK collaboration has shown using simulation that the required performance can be achieved using 50  μm pixels  but it is highly sensitive to the way pixel hits are clustered when counting tracks. The simulation cannot be trusted at this level of detail and so such results must be verified with real data. The material per measurement layer is not an issue per se for calorimetry but it is important to minimise the thickness of the sensitive layers so that the lateral spread of the shower is minimised. This can be mostly achieved if the cooling, and the number of power cables and I/O lines is minimised, hence pushing for low power. Cost is a critical issue as any Si-W calorimeter will require silicon sensors to cover very large areas of about 1000-2000 m2, an order of magnitude larger than e.g. the CMS tracker silicon. This requires sensors which can be manufactured using a standard (and hence affordable) process supplied by multiple vendors.
We have developed a small scale sensor, TPAC1, with 50  μm pixels using four charge collection diodes per pixel to optimise charge collection. As a next step we would aim for a larger, reticle-size device suitable for constructing a DECAL stack. This would be used to test experimentally the technique of digital electromagnetic calorimetry.

Summary of the requirements for a pixellated silicon sensor at a linear collider



Tracking Barrel/Forward


Pixel size (square)

20  μm

25 - 50  μm

50  μm

Spatial occupancy
(EM shower core)

100 / mm2

0.1 - 2/mm2

100 / mm2

Material per layer (X0)

< ~0.1%



Output Signal

>5 bit analogue

Analogue or binary


Noise (MIP)


< 0.1


Table 1 Summary of the requirements for a pixellated CMOS sensor

Monolithic silicon pixel sensors - the technical case

The use of silicon sensors, including charge-coupled devices (CCDs) and hybrid silicon pixel detectors for vertexing, hybrid silicon strip detectors for tracking and silicon pad detectors for luminosity monitoring is by now well established in particle physics. Silicon is the technology of choice where high precision measurements are required and its use is likely to continue into the future, which is evident from the current proposals for sLHC upgrades and the recent linear collider detector concept studies. The urgent requirement is to develop novel solutions which have faster readout and/or lower power density so that the material per sense layer can be reduced by factors of 2-5. In addition the cost per unit area is a limiting factor.
Our research proposal is focused on developing monolithic silicon pixel sensors which provide the required granularity for future high energy and high luminosity experiments but also addresses the major issues such as power consumption, connectivity and cost reduction. Because a monolithic silicon device integrates both the sensor and the front end (FE) electronics, it immediately decreases the material per measurement layer and simplifies assembly. The use of CMOS processes available from multiple vendors, compared with silicon detectors requiring specialised processing, provides a cost benefit and opens up the use of monolithic silicon pixel sensors for tracking and calorimetry.

Figure 3 Schematic cross-section of an INMAPS wafer
with the deep P-well implant. Not drawn to scale

Figure 4 Principle of ISIS operation

In the early '90s CMOS Monolithic Active Pixel Sensors (MAPS) were proposed as imaging sensors. Charge-Coupled Devices (CCD) were then the dominant imaging technology but it was immediately recognised that an image sensor in CMOS has significant cost advantages as it allows the integration of the sensor with sophisticated readout electronics including analogue-to-digital conversion and signal processing. It also allows very small pixel sizes building on the reduction in feature size of CMOS. Since then the technology has continued to improve by the introduction of features such as low dark current diodes and charge transfer, which were previously found only in CCD technology. It was quickly shown that the devices can act as particle detectors and that they have the following attractive features:

  • High granularity: pixel sizes down to about 1 micron are possible
  • Simplicity: CMOS integration lowers the need for high density connection to external electronics, which requires complicated flip-chip technologies in hybrid pixel detector systems
  • Low cost: as CMOS is a mature, industrial process with many vendors. (but particle detectors might require additional processing)
  • Low power: because of the absence of standing current in many designs, e.g. the rolling shutter architecture, as well as the low voltage levels of CMOS
  • The potential radiation hardness of the CMOS circuitry

However, there are a number of limitations of the technology which have delayed the take-up in particle physics experiments:

  • The size of the induced signal is small because the sensitive layers are thin. Also, the time it takes the signal to develop is large because the devices rely on diffusion to collect the charge. This also affects the radiation tolerance of the sensor as charge will be lost by trapping after irradiation. We have previously proposed an approach to solving this problem which relies on the use of a deep n-doped diode to create an electric field [14]. In this proposal we intend to investigate options to increase the signal size and decrease the charge collection time through the use of high resistivity epitaxial layer material, allowing complete depletion of this layer. This should also increase the radiation tolerance.
  • In order to maintain 100% detection efficiency, historically only NMOS electronics (which sits in a p-well) could be used on its own rather than together with PMOS electronics, (which sit in an n-well), as these PMOS n-wells collect the charge, reducing the signal being collected by the sense diodes. This limited the sophistication of the on-sensor electronics and has been overcome in advanced optical sensors by introducing a shielding deep p-implant, and recently adapted by us for the TPAC sensor ]. In this process, the highly doped buried layer isolates the PMOS transistors from the sensitive volume from which the charge is collected, making it possible to integrate a complex electronics processing chain into a pixel, in the same way as is done for hybrid active pixel sensors, while preserving 100% efficiency.
  • In order to minimise power consumption, there is a need to develop robust, in-pixel analogue storage. In the ISIS architecture, the raw signal charge is stored in charge transfer structures inside the pixel, equivalent to short CCD registers, as illustrated in Figure 4. This CCD register is protected from parasitic charge collection by a deep p-well or by a deep p-implant by the abovementioned approach. The CCD register itself has been implemented using a buried n-channel and is 20 cells long. The principle of raw-charge storage results in a high immunity to noise pickup and minimises front-end electronics power. We are developing this approach with a commercial CMOS foundry and we propose to explore ways of minimising the pixel size. The ISIS devices already developed for high speed cameras have a large market, and this approach for particle detection could open the door to high speed x-ray imaging systems, for which there are many potential scientific applications.
  • Imaging APS(Active Pixel Sensors) devices can have very small pixel sizes, down to 1.2  μm, which are not directly usable for particle detection. The relatively large pixel sizes required for calorimetry and tracking applications (~25-50  μm) require custom-designed diode geometries. A multi-diode geometry has been developed by us for the TPAC1 sensor and has been shown to collect charge efficiently. In this proposal the 4T cell circuitry developed commercially for optical imaging will be optimised for charge collection in large pixels whilst still providing a low input capacitance by using a pinned photodiode. This will also allow correlated double sampling (CDS) for very low noise operation and low power.
  • There has been limited experience with large scale systems using monolithic pixel sensors. Any pixel detector will by definition have a large number of readout channels and an important aspect of our research is to prototype a large sensor and hence study issues of power distribution, signal transmission and system integration.





Charge collection & storage

Photogate + CCD register

Pinned photodiode (4T)

4 Diodes plus >100 transistors per pixel

Pixel size (square)

20 µm

25 µm (50 µm)

50 µm

Time slicing

x 20 in-situ storage cells

x10 rolling shutter

x 8192 time stamp

Noise minimisation

raw charge storage and CDS


Signal shaping and pseudo-CDS

Power minimisation

Delayed slow readout
(rolling shutter)

Rolling shutter

Asynchronous operation

Radiation tolerance

Enhanced compared to CCD by small number of transfers

High resistivity study to enhance

High resistivity study to enhance.

Yield and cost

Custom CMOS process

Standard CMOS process

Standard CMOS process

Table 2 Architecture comparison

In order to investigate these issues we propose an R&D programme which addresses the fundamental issues discussed above by developing three sensors. Two of these devices are based on work initiated within the framework of the LCFI and CALICE-UK projects. The ISIS design will demonstrate low-noise charge storage in a pixel size which is compatible with the requirements of a vertex detector. The TPAC2 sensor will be a large area sensor which will be used to validate the principle of digital electromagnetic calorimetry. This should determine the limiting factors for operating a DECAL detector and so give an indication of the required future direction for development of such sensors.
Within this proposal we will also study the promising 4T cell technique. Combined with shielding implants, this should lead to significant improvements in S/N compared with 3T devices, lower power and higher efficiency, which would be beneficial for calorimetry. In addition, the 4T technology would be a very good match to tracking sensor requirements. Taken together, this programme will therefore demonstrate the basic feasibility of monolithic silicon sensors for vertexing, tracking and calorimetry applications and allow future experiments to tailor devices for particular applications. Integrating CCD structures in a CMOS process will open new opportunities for a wide range of scientific applications.

#trackbackRdf ($trackbackUtils.getContentIdentifier($page) $page.title $trackbackUtils.getPingUrl($page))
  • No labels