Radiation-hardness requirements for future silicon detectors

Silicon detectors (sensors) play an essential role in high-energy physics and space experiments.¹ In fact, detector development follows Moore's law:² the total area of silicon strip detectors used in experiments exhibits an exponential increase over time. At the same time, the number of readout channels used increases and the cost per unit area decreases exponentially.¹ On the other hand, requirements have become more challenging because the detectors must be operated in ever-increasing particle fluxes.

The ability to pattern the sensor area to very small sizes makes them ideal for the inner tracking layers in the very large, all-purpose experiments of the Large Hadron Collider (LHC),³ where large fluxes of hadrons flow outwards from the collision point. Figure 1 gives a sense of the particle fluence (i.e., the flux integrated over the 10 years of operation) expected for the inner tracking detector ATLAS (A Toroidal LHC Apparatus) experiment⁴ following a future upgrade. The very strong radial dependence necessitates tailoring of the detector geometry to pixel sensors (inner radii) and short- and long-microstrip detectors (mid-radius and outer regions, respectively).

Figure 1. Expected radial distribution of the particle fluences for the ATLAS (A Toroidal LHC Apparatus) detector at the Large Hadron Collider upgrade for an integrated luminosity of 3000fb⁻¹(inverse femtobarns), for pions, protons, neutrons (n), and their sum, all expressed in terms of 1MeV neutron-equivalent fluences.⁵ Vertical lines indicate sensor locations. Pixels: R = 4–20cm, short strips: R = 38–60cm, long strips: R = 85–100cm.

These fluences—and the associated total ionizing dose of up to hundreds of Mrad—are much larger than those encountered in space, where we are much more concerned about single-event effects (SEEs) because of heavy ions. In accelerator-based research, secondary spallation products generated by protons can also cause SEEs, although at a fluence reduced⁶ by a factor of approximately 1.5×10⁶. Silicon sensors exhibit a variety of sensitivities to radiation,⁵ which makes them less than ideal. Yet the quest for an alternative to silicon sensors has come up dry, with the exception of diamonds for small areas of very high radiation (although even diamond has radiation issues).⁷

Irradiation by hadrons will lead to damage of the silicon bulk through creation of defects (‘traps’) in the forbidden-energy gap. They have three consequences, including increased leakage current through trap-assisted emission, increase of the full depletion voltage through an increase in the effective doping concentration in the bulk (even causing inversion of n- to p-type bulk, where ‘n’ and ‘p’ refer to electron- and hole-dominated compositions), and increased charge trapping. Combined with surface detoriation from ionizing radiation, the result is a decrease in the signal-to-noise ratio. In addition, the operational conditions are made more challenging, requiring cooling to suppress the leakage current and higher bias voltages to increase charge collection. Moreover, most radiation effects are not stable after the exposure stops. They continue to anneal, which is temperature sensitive and dictates the operating temperature even during beam-off periods.

SEEs have not been observed in silicon sensors,⁸ but sensors operating in accelerators need to survive radiation accidents because dumping very large amounts of ionizing radiation onto silicon sensors shorts the high bias voltage to the readout strips.⁹ Since present detectors have been developed for particle-fluence levels of one tenth the levels shown in Figure 1, there is obviously a need to prove that silicon sensors can function adequately after such heavy irradiation by charged hadrons and neutrons. The RD50 collaboration (sponsored by CERN, the European Organization for Nuclear Research) has been exploring development of radiation-hard semiconductor devices since 2002.¹⁰ RD50 investigates radiation hardening from many angles, including defect/material characterization, search for alternative semiconductors to replace silicon, improvement of the intrinsic tolerance of substrate material (p- versus n-type, initial doping concentration, oxygen concentration, float-zone versus magnetic Czochralski substrates), optimization of sensor geometry (collection of holes versus electrons, surface treatment), and novel (3D) detector designs.⁵

The experiments operating at the LHC have started programs¹¹ to develop semiconductor detectors that can function after an upgrade to the heavy fluences shown in Figure 1. The experiments work on both sensors and readout application-specific integrated circuits, and on characterization of their post-radiation performance. One can gauge the progress made in mitigating radiation damage by comparing the new sensors with those discussed in a review article 12 years ago.¹² The basic lesson learned is to use n-in-p sensors instead of the old p-in-n detectors. The former collect electrons instead of holes, do not exhibit type inversion, and are characterized by better charge collection after high fluence (see Figure 2) with little (and favorable) post-radiation annealing.

Figure 2. Collected charge as a function of 1MeV neutron-equivalent fluence (Φ_eq) for 23GeV and 26MeV protons and reactor neutron-irradiated float-zone (FZ, Fz) silicon-microstrip sensors at the bias voltage indicated in brackets.¹³ The advantage of the new n-in-p sensors over the older p-in-n sensors and the improvement with higher bias voltage are clear.

Further research is continuing in developing ‘punch-through’ protection structures to guard against beam accidents, optimizing surface treatment to prevent post-radiation noise increase and/or shorting out of strips, investigating operation at higher bias voltage, including charge multiplication as long as the associated noise is acceptable, investigating nonplanar (‘3D’) structures where charges are collected on alternating columns of n- and p-type material (which affords lower depletion voltages and shorter drift distances with reduced trapping),¹⁴ and optimizing the pattern of the sensors, including reduction of the dead area. This shows that although the goals of the development program look achievable, there is much work to be done before we will be able to place the sensors inside the upgrade experiment at locations indicated by the vertical lines in Figure 1.