The best device for that is SiPm [Silicon Photomultiplier].

It was demonstrated applied to gamma spectroscopy in sizes as large as 75x75mm e.g.

Silicon photomultipliers in gamma spectroscopy with scintillators

Grodzicka-Kobylka, M.; Moszyński, M.; Szczęśniak, T.

Publication: Nuclear Inst. and Methods in Physics Research, A, Volume 926, p. 129-147. Pub Date: May 2019 DOI: 10.1016/j.nima.2018.10.065

actually is no any problem to replace PMT with something else but you should remember that you need to compensate amplification PMT provide with using an additional amplifier to obtain the same one pulse feedback

The answer, as with most things, is complicated. Each optical detection technology has its pros and cons. The only way to properly answer the question is to perform a system noise analysis of the competing technologies to determine which technology is best, or which at least is adequate for your application. It also depends on other factors. Does the detector have to be smaller than you can achieve with a photomultiplier tube (PMT)? Is the detector going to be subjected to magnetic fields, which would disturb the operation of a PMT?

I assume you're trying to read and resolve the energy of gamma photons coming in for identification of radionuclides.

Firstly, you have to consider the quantum efficiency of the optical detection method. This noise source could be described as optical shot noise. Basically, the relative error in the detection of the optical pulse will go as 1/sqrt(N) where N is the number of optical photons detected.

NaI(Tl) has an optical yield of 38000 photons per MeV. If you detected 100% of the optical photons generated from, say, a 662 keV Cs0137 gamma line this would contribute a relative error in the detected gamma energy of a 662 keV gamma line of 1/sqrt( 38000 * 0.662) = 0.63%.

PMTs have a quantum efficiency of about 25%. That means that the relative error contributed from optical shot noise would be 1/sqrt(0.25) = 2 times higher. So, 0.63% x 2 = 1.3%.

NaI(Tl) has an intrinsic spread in its optical yield of about 8%, so at least for a PMT this isn't really the dominant source of noise, at least at 662 keV. When you start talking about lower energy gamma lines, the effect becomes more significant.

You also need to consider the dark current of the detector. If, for example, you're using a photodiode or an avalanche photodiode, you need to include the statistical variation caused by shot noise in the dark current. This is a function of your integration time for collecting the optical pulse. NaI(Tl) has a scintillation decay time of 0.23us. You'd probably want to collect for three or four decay periods to make sure you've collected most of the optical photons generated. The longer the integration window, the more dark current you collect, and hence the more statistical noise from dark current you collect.

The contribution of amplifier noise must also be factored in for photodiodes and avalanche photodiodes.

(Be warned that the noise in avalanche photodiodes is greater than the noise you'd expect just based on their leakage current. The avalanche photodiodes amplify single-electron noise, which makes the noise "clumpier" than single electron shot noise. I seem to remember this increases the effective shot noise by a factor of a few.)

Silicon photomultipliers don't have dark current in the same way, but they do have random firings. You'd need to look up the specs on the silicon photomultiplier and factor in these random firings in your noise budget.

Leakage currents in photodiodes, avalanche photodiodes, PMTs and the random firings in silicon photomultipliiers all go up with temperature. If you can cool down your detector, especially silicon semiconductor detectors, you'll reduce this noise source significantly.

PMTs are good in that they have a large surface area for collecting the optical photons from the scintillator. That's more difficult to achieve using photodiodes, avalanche photodiodes, or silicon photomultipliers. What's more, the larger the area of these devices, the higher the leakage current you'll experience and the higher the shot noise current.

Finally, if you're trying to get accurate energy on the measured a gamma spectral lines, you need to have consistency in the efficiency of the optical collection. If the collection efficiency of optical photons from a gamma photon absorbed in one part of the scintillator is 10% different than the efficiency of collection in another part of the scintillator, that's an immediate 10% difference in the measured energy of the gamma line. If you use multiple optical detectors and from this can estimate roughly where the gamma photon was absorbed, you can compensate for this effect somewhat. That makes the system more complicated. Best for a simple system to just wrap the scintillator in a highly reflective material that only allows light to exit on to your optical detector surfaces. A slight roughening of the surface of the scintillator can also help, as it randomizes the paths of the optical photons and can help to eliminate focusing effects with the internal reflections of the optical photons passage through the scintillator. (You can model these sorts of effects using Geant4. However, there is a significant learning curve to using Geant4, and you need to be a reasonably proficient C++ programmer.)

You can do this, but the matter is not that easy. Several important things must be taken into account, such as the role of amplifiers in the condensation of the signal instead dynodes. In addition, a substance that condenses electrons at a specific point must be added.

Also, a substance should be added around the photo diode to collect light as much as a substance Tio2.

It is interesting problem. You can do it. First, you can choose APD S8664-1010 and crystal CsI(Tl) or LaBr3(Ce) (areas are 1 cm**2). Second, you can use CR-110 or eV-5093 with bias resistor 200 M). Bias voltage is about 380 V. All of them must be placed in a metal box.

I think it is possible only in case of your photons are in the visible rang with high accuracy focusing is needed