Not a scientific paper, but Q&A from Health Physics Society describes experiment with disassembled CCD and CMOS sensors.
[...] it is true that both the charge-coupled silicon devices and the
metal-oxide semiconductors may produce visible light scintillations in
response to ionizing radiation. Just as visible light will set free
electrons through photoelectric interactions in silicon, higher-energy
radiations such as alpha, beta, and gamma radiation also have the
ability to free electrons in the material. The subsequent light
emission normally occurs when free electrons combine with the holes
that represent the positive charge carriers; this recombination
process sometimes results in the emission of energy as visible light
photons. In silicon the process is relatively inefficient so that
relatively few of the ionization events that set free electrons lead
to light emission. If the frequency and density of recombination
events is sufficiently high, however, some of the light pulses may be
sufficiently intense to be visible to the naked eye, especially in a
But before going any further, reasons why in the case of Tianjin explosion footage it cannot be alpha radiation: Detecting it requires disassembled camera, because optics in front of sensor would block alpha radiation. On top of that, alpha has very low penetration, even few centimeters of air is enough to stop it, so it wouldn't even reach camera in the first place.
That being said, there is still possibility for very strong beta & gamma radiations (as nuclear explosion would produce) to interact with CCD/CMOS sensor enclosed in fully assembled camera:
The situation is quite different for beta and gamma radiation. The
maximum energy beta radiation from 40K is about 1.3 MeV, and the
approximate distance such a beta particle could travel in silicon is
about 3 mm, a dimension much greater than the thickness of the active
material in a CCD or CMOS. The average beta particle traversing the
same 20 micron path length as the alpha particle would produce
approximately 5,000 to 10,000 ionization events. Thus, the ionization
density, as well as the total ionization per particle, would be much
less for the beta radiation than for the alpha particle. This makes it
much less likely that one would be able to see the light emission with
the naked eye. There are optical microscopy enhancement techniques
that might make these weaker scintillations visible.
Gamma rays will also interact in silicon to set free electrons that
may lead to recombination and light emission, but the situation with
respect to the gamma radiation is even more restrictive as regards the
possibility of viewing the light emission. The probability that a 60
keV gamma ray will interact within a 20 micron thickness of silicon is
about 1.4 x 10-3. This compares to the probability of 1.0 for the
alpha particle (and the beta particle). Additionally, the photon
energy is much less than the alpha or beta energy and only about half
of the photon energy, on average, would be deposited in the silicon
per interaction. Thus, one would have an even more difficult time
trying to visualize light emissions from gamma-ray interactions.
On the other hand, keep in mind that such experiments talk about detecting scintillations on images, that are very dark.