The photomultiplier:  Dark background
 

If a photomultiplier is made to see no light at all, a small output current persists. This is known as the dark background or dark signal and its source is chiefly from electrons which have sufficient thermal energy to allow them to be spontaneously emitted from the photocathode. When photon-counting techniques are used, this background signal from the detector is referred to as the dark count. The magnitude of the dark background can, therefore, be reduced by enclosing the photomultiplier in a cooled housing: dry ice (solid CO₂) is frequently used as a coolant. A recorded signal is, therefore, the combination of the detector’s response to light plus the contribution from the dark background. When observations are made of faint objects, the dark signal may be sufficiently significant to require separate measurement and then subtraction from the measurement of the combined signal. If the magnitude of the dark signal were known perfectly, its effect would not cause any deterioration in the measurement obtained from the faint object. However, as in the case of the photoelectrons, the emission of thermal electrons is a statistical process and the dark background shows shot noise fluctuations. Any measurement of the background carries with it an uncertainty and this, in turn, affects the accuracy of any determination of the weak light-level.