Ionizing-Radiation Detectors & Counting Statistics

Ionizing radiation deposits energy in matter through ionization or excitation. A detector converts that interaction into collected charge, scintillation light, or electron-hole pairs. Gas-filled ionization chambers, proportional counters and Geiger–Müller tubes; scintillators; and semiconductor detectors deliberately trade sensitivity, energy resolution, speed, robustness, and cost.

Counts are not activity #

Observed rate depends on source emission, solid-angle geometry, attenuation in air/windows/shields, intrinsic detector efficiency, energy threshold, dead time, and background. Moving a source or changing its container invalidates an efficiency calibration even if the detector is unchanged.

For independent arrivals with expected count $N$, Poisson standard deviation is $\sqrt N$, so relative counting uncertainty is approximately

$$\frac{u_N}{N}=\frac{1}{\sqrt N}.$$

One hundred counts carry about 10% relative counting uncertainty; 10,000 counts about 1%, before efficiency and background uncertainty. Longer counting improves statistics only as square root of time.

If sample and background are counted independently, net counts are subtracted but variances add. A small difference between two large counts can therefore be statistically weak.

Dead time and pile-up #

After an event, pulse shaping and detector recovery make the channel temporarily insensitive or distort overlapping pulses. At high rate, measured rate becomes nonlinear and spectra suffer pile-up. The correction depends on whether the system is adequately modeled as paralyzable or non-paralyzable and must be validated for the actual electronics.

Selection logic #

• Presence/rate monitoring needs efficiency and count-rate capacity.
• Spectrometry needs pulse height related to deposited energy and sufficient energy resolution.
• Dosimetry needs calibrated response to radiation type and energy, not merely counts.
• Safety surveys need background, geometry, response-time, saturation, and alarm behavior verified.

Record detector type, bias, threshold, live time, dead time, geometry, background, calibration source, energy range, and uncertainty. Check sources and blanks reveal drift; reproducible fixtures control geometry.