Recent progresses in the development of Monolithic Active Pixel Sensors

Introduction The ambitious physics goals of the CBM experiment call for unprecedented performances of its vertex detector, as well in terms of spatial resolution and material budget as for read-out speed and radiation hardness. Monolithic Active Pixel Sensors (MAPS) are considered as candidates for this application as they already demonstrated an excellent spatial resolution (~1.5 μm) and a low material budget (< 120 μm silicon). However, the high luminosities envisaged for CBM set stringent requirements in radiation hardness and readout speed, which impose further investigations in the detector design. Moreover, a model allowing to simulate MAPS within the CBM simulation framework is needed for feasibility studies. The corresponding R&D project, involving IReS and GSI, was started in 2003 [1]. The progresses achieved in 2004 are summarised hereafter. 1. Investigation of a new fabrication process The performances of MAPS depend quite dramatically from a few fabrication parameters. The thickness of the epitaxial layer (EL) is of prime importance, as it determines the active volume and therefore the signal magnitude. The modest signal charge (a few 100 e ENC) delivered by a standard, <10 μm thick EL, makes it difficult to drive the signal conditioning micro-circuits necessary for the fast, massively parallel read-out required for CBM. By the end of 2003, the manufacturer AMS announced a new 0.35 μm fabrication process optimised for MAPS imaging applications. It was supposed to provide an EL of ~20μm thickness and very low leakage currents from the sensing diodes. MIMOSA-9 was designed, fabricated and tested on a 120 GeV/c pion beam at the CERN SPS in order to explore this process. Excellent performances (e.g. 1.5 μm spatial resolution, 99.5% detection efficiency) were achieved. However, in contrast to the foundry announcement, the total cluster charge observed (~800-900 e ENC) and visual checks by microscope suggest an EL thickness of ~10 μm 11 μm. The leakage current was measured to be about one order of magnitude below the typical values of previous chips but increased by up to two orders of magnitude already after 20kRad integrated ionising radiation dose. Nevertheless, the new fabrication process can be considered as the most effective tested up to now and solutions to its radiation sensitivity are very likely to be found soon. 2. Fast read-out architecture The strategy of a massively, column parallel read-out architecture requires integrating advanced functionalities, like leakage current and pedestal subtraction, inside each pixel. Every column has to be equipped with a single discriminator handling the signal coming out sequentially from the pixels. Three different sensors with column parallel read-out (MIMOSA 6-8) were prototyped and tested since 2003. The noise of MIMOSA-6 came out to be enhanced by large pixel-to-pixel signal dispersions. In contrast, preliminary test results on MIMOSA-8 show an individual pixel noise of 1318 e ENC and a charge-to-voltage conversion gain of 50110 μV/e, depending on architecture details [2]. Those performances, which are comparable to conventional MAPS, come together with a modest pixel-to-pixel dispersion. This architecture, which is still being characterised in detail, provides the path to follow in 2005 for the next R&D steps.