It depends on the semiconductor material in question (silicon, III-V (GaAs InP), perovskites(?), exotics etc.) material quality, defect density/radiation damage, etc. but basically at a low rate of carrier (electron-hole pairs) production there are competing processes that allow them to recombine locally inside the material such that they can't make it out and be used to produce a current.
Mathematical models used for terrestrial PV applications are going to focus on sunny to cloudy/overcast days and:
Source: Low Efficiency of the Photovoltaic Cells: Causes and Impacts
See also Thermal Management and Efficiency Enhancement of Solar Systems
...and may not be very accurate at Saturn (10 AU) and beyond, and experimental data for cells optimized for terrestrial applications may be worse at Saturn than those optimizes for very low light levels, but with those caveats, let's take a look at some data anyway.
Mathematical models used for terrestrial PV applications are going to focus on sunny to cloudy/overcast days and may not be very accurate at Saturn (10 AU) and beyond, and experimental data for cells optimized for terrestrial applications may be worse at Saturn than those optimizes for very low light levels, but with those caveats, let's take a look at some data anyway.
Mathematical models used for terrestrial PV applications are going to focus on sunny to cloudy/overcast days:
Source: Low Efficiency of the Photovoltaic Cells: Causes and Impacts
See also Thermal Management and Efficiency Enhancement of Solar Systems
...and may not be very accurate at Saturn (10 AU) and beyond, and experimental data for cells optimized for terrestrial applications may be worse at Saturn than those optimizes for very low light levels, but with those caveats, let's take a look at some data anyway.