Room: AAPM ePoster Library
CdTe-based semiconductor detector for ‘correction-less’ small-field dosimetry
A recently proposed approach to design a correction-less detector for small-field dosimetry employs mass-density as the principal determinant of detector water-equivalence. We evaluated the approach for a range of geometric parameters utilizing Cadmium Telluride (CdTe) and traditional silicon (Si) diode sensitive media.
Monte Carlo simulation (MCNP5 package) was used to optimize dosimeter designs combining a semiconductor sensitive volume with an air-gap, following a mass-density matching approach. Si and CdTe of 3, 30 and 300µm thickness were modeled in combination with air gaps varying from 0.4 to 4 mm, with surrounding PMMA layer that extended at least 1 mm beyond the detector in every direction. Detectors were irradiated with 6MV and 6FFF spectra, representative of Varian TrueBeam linac sources. Energy deposition was scored with pulse-height tally in semiconductor and equivalent water with air gap volumes.
For 3, 30, and 300µm thick semiconductor layers the density-matched air gaps are 0.0145, 0.145, 1.45mm for CdTe and 0.004, 0.04, and 0.4mm for Si. These configurations resulted in close water signal matching for Si, but deviated by a factor of 2 to 3 for CdTe. Since the electron density rather than mass-density governs photon interactions under radiation therapy energy range, a more sophisticated matching strategy should be employed for high atomic number solid state detectors, such as CdTe. Details of the input spectra even as close as 6MV and 6FFF further complicate density-matching.
Combining a high-density solid state detector and air gaps to achieve mass-density matching to water typically results in a detector with low sensitivity, problematic for small-filed dosimetry. Substituting standard Si diode with CdTe offers a significant increase in detector sensitivity, but requires additional considerations for correction-less design, where Monte Carlo method is uniquely suitable for a prototype development.