Geant4 examples which describe or include photodiodes

Getting Started
1

Hi, What Geant4 examples describe or include photodiodes?

Geant4 Version: 11.2.1
Operating System: Windows
Compiler/Version: Visual Studio 2022
CMake Version:3.29.0-rc2

2

@alfredo10110, I am not familiar with all Geant4 examples, but referring to another post of yours [ Geant4 toolkit examples (basic, extended or advanced) which describe or include photomultiplier tubes ], I’m pretty sure some of them deal with photomultipliers. Not sure about photodiodes. Nevertheless, all the examples have a README file describing them; you can take a look.

3

Ok, thanks. Your answer has helped. So, I´m going to take a look at the README file of the Geant4 examples

4

I usually model a photodiode as a sensitive silicon die plus passive structures as usual volumes. This way, you account for direct hits and can count photons passing the volume surface.

5

Ok, thanks for adding more information. I have 5 questions:

1.- What is a silicon die?

2.- What would be the passive structures of the simulated photodiode?

3.- How long can I simulate a photodiode in Geant4?

4.- Can I simulate a photodiode in Geant4 with a basic knowledge of Geant4 and C++?, Or what knowledge of Geant4 do I need to be able to simulate a photodiode in Geant4?

5.- How difficult is it to simulate a photodiode in Geant4?, is it easy?, is it medium difficulty level?, or is it hard?

Thanks in advance

6
  1. It is a piece of silicon that senses the light. Usually, it is a 100-300 um thick plate with a chemically created or ion-implanted p-n junction, which is reverse-biased and connected to some electronics that can transform charge or current into a voltage you measure.
  2. The die cannot float in the air, so it is mounted on a dielectric pad that may or may not be mounted inside another case or epoxy mould. These structures may affect how the device behaves in a particle beam or an optical photon shower. Moulds may have different transparency depending on the wavelength. Also, the refraction indices of the parts of the system may severely affect how much light you get.
  3. I can not get the question. If you mean virtual time, it is practically unlimited. The limit for radioactive decay may be set to 1E60 years. If you mean the real time, the wall clock, it is up to you to decide. The more you need statistics, the better is the result, and the longer is the simulation. It could be a second or a week, depending on the requirements. Mind that if you plan to use scintillation or Cherenkov photon generation, the time may run out of hand pretty quickly, as a single energetic proton can generate millions of photons to track.
  4. Define basic. I deem my knowledge basic, yet I still simulate and publish results in reputable journals.
  5. Define easy. I would set the medium as the baseline, as I have little idea of the requirements. From my perspective, in a world with LLMs available, practical tasks that have been around for many years should not be hard.
1 Like
7

Ok, thank you so much for all the information. I have 3 more questions:

a) In your answer number 1, which says “100-300 um“, What units are um?, Are micrometers?

b) My question number 3 would be: How long does it take someone to simulate a photodiode in Geant4 (for someone who has a basic knowledge of Geant4 and C++)?; based on your answers number 4 and 5 (Define basic and Define easy) I suppose I can simulate a photodiode in Geant4 after a day or at most after a week, is that right?, in other words, I can simulate a photodiode in Geant4 in a short time, is that right? (considering that my knowledge of Geant4 and C++ is basic)

c) Could you please tell me the steps (as a numbered list) to simulate a photodiode in Geant4?

Thanks

8

This is a bit of an ask since so much of the answer depends on what you are trying to do/simulate/observe. Saying you want to “simulate a photodiode” is vague and unfocused. What are your trying to simulate with your photodiode? What are your metrics? What are the variables you want to control? What variables don’t matter as much (to you)?

If the questions are about its signal performance characteristics then you are better off not even using G4 in the first place but either solid state circuit simulators/codes or finite element solvers if you want to understand the fields in your use case. If it is about the “input” to your transducer element, then Geant4 can help you. If you want the entire chain then you will have to likely feed the output of Geant4 sims as an input to an entirely separate simulation to go from light to signal.

9

a. Yes, it is a colloquial term for micrometres, as not everyone can type µm without effort.

b. Yes. Seems so.

c. You need to

  1. Define the requirements for the simulation

  2. Define the geometry in Geant4, likely in GDML if you need some real-life replica and not an abstract photodiode in vacuum

  3. Design the simulation and the readout.

  4. Process the data and make a report.

  5. Repeat if needed.

10

Ok, thank you for all the information. Regarding your questions:

What are you trying to simulate with your photodiode?, Answer: I want to simulate a silicon PIN photodiode; My goal is to get an energy spectrum (a plot of Counts versus channel number / energy) with Geant4 (this spectrum is a Pulse Height Spectrum); I already got the energy spectrum experimentally (in the Laboratory) using a photomultiplier tube, the detector material, a NIM bin and a Multichannel analyzer; What I still need to do in the Lab is get the energy spectrum using a PIN photodiode; Regarding the simulation in Geant4, I want to get the energy spectra using first a photomultiplier tube and after a PIN photodiode; This would be done to compare the experimental results and the Geant4 simulation results; In the Laboratory, my experiment consists of a primary particle source, which is a gammas radioactive source very close to the detector material (thallium activated sodium iodide); the detector material is optically coupled to a silicon PIN photodiode (a light detecting semiconductor device with a layered structure; a P-type (positive), I-type (intrinsic) and N-type (negative) layer); this experimental setup is connected to a NIM (Nuclear Instrumentation Module) bin that provides high voltage; the NIM bin is connected to a Multichannel Analyzer, which consists of hardware (an electonic board and a desktop computer) and software that provides energy spectra (histograms). So, What would I need to do to get the energy spectra with Geant4?. I only need to add that the type of detector I am using in the Lab is a gamma ray detector.

What are your metrics? Answer: With this question, are you asking, What do I want to measure?; if so, what i want to get with Geant4 are the energy spectra using a silicon PIN photodiode and a multichannel analyzer; I think i could simulate the silicon PIN photodiode in Geant4; however, I don´t know how i could include the multichannel analyzer in the simulation with Geant4; Would I need to include a Geant4 digitizer module to do so?

What are the variables you want to control? Answer: Since I want to get the energy spectra, I think the variables i want to control are the counts and the energy, would this be okay?, Would this be enough?

What variables don´t matter as much ? Answer: I think the variables that wouldn´t matter so much to me would be all those that are neither counts nor energy, would this be okay?, or, would this be enough?

Thanks

11

Ok, thanks. I´ll keep that in mind. Regarding the geometry of the simulated photodiode, what do you recommend most?: should the geometry be done in GDML format?, Or in a non-GDML format?.

12

There are two layers to this.

  1. The “spectrum” of light that reaches your PIN Diode. Each gamma event will create many photons and some subset of these photons will reach your PIN Diode sensor. Geant4 can help with this by giving you the time of arrival of photons and the number of photons that arrive for each gamma event. This is not a spectrum you will ever be able to “see” in a real setup. Let’s call this L(x).
  2. The “spectrum” resulting from the conversion of light to an electronic signal, complete with electronic noise, amplification, shaping (if done), integration, etc. This is a function that maps f(x): L(x) |→ S(x) where S(x) is what you are actually measuring. Geant4 is not really appropriate for this part. Other codes are better suited for this part. However, Geant4 is useful in that your simulation can give you event histories of photons and their time of arrival which can be used as an input to these other codes. Geant4 handles the radiation transport part, not the electronics part.

because of this Geant4 only cares about the physical properties of your system including the optical interfaces between them until your silicon PIN photodiode front surface. At that point, you will likely have to use or do something else to continue forward.

Would I need to include a Geant4 digitizer module to do so?

You can certainly write your own but that will be a lot of work for such a simple system. Such a thing does not exist in Geant4 out of the box. There isGATE, which is a wrapper for Geant4, that does but any questions you would have would have to be directed towards that independent collab instead of here.

1 Like
13

Ok, thank you for all the information. In your paragraph number 2 (The “spectrum“ resulting…), What is f(x)?.

So, based on everything you have told, the best thing I can do is to include Sensitive detectors and Hits in the Geant4 simulation, right?

14

Unless you want to write your own digitizer class, yes. I guarantee that all the information you need is some combination of :

  1. Number of total photons for each event
  2. Arrival time of photons for each event
  3. (x, y, z) and (px, py, pz) of the photons entering the “photodiode” volume
  4. Wavelength of the photons

Each one of these becomes a train of delta “inputs” to some other simulator to convert from light pulses to electronic signals outside of Geant4. But what you will also find is that each event might generate (depending on material and incident energy) upwards of 10-100s of thousands of photons giving you that many (2, 3, 4 in list above) data points making these files absolutely enormous and take a long time to run if it doesn’t crash first. So likely you will need to make some compromise that is entirely dependent on what you really need for your setup. There is no checklist here.

As an example you could simplify things as:

  1. Only record 2, 3, 4 for the first X photons that arrive where X is twice the detection limit of your electronics.
  2. Calculate just the “center of mass” and “center of momentum” of the photons. This would just be a weighted average of the (x, y, z) and (px, py, pz) of photons that reach the photodiode volume with a “timestamp” corresponding to your detection limit for example, the 5th photon. Just drop 4 entirely.

A non-linear arbitrary function that converts a train of light pulses into an aggregated electronic signal. My claim is that you need to use another simulation framework geared for charge injection of electronics such as your pin diode to get it. Photons convert to electrons inside the diode and there is an entire ecosystem of tools in solid state physics to do this.

1 Like
15

Thanks for all the information. I´ll keep that in mind. However, I have the following questions:

1. Everything you mention in your first numbered list can be represented as “Geant4 Hits”, right?; 2. Regarding delta “inputs“, do you mean that they are inputs that have a shape like a Dirac delta?;

3. Regarding your first numbered list, Could it be said that a light pulse can be represented by one of the things you mention in your first numbered list or a combination of them?

4. What does the detection limit of electronics mean?

5. Is “upwards of” equivalent / synonim to “more than”?

6. Could you please explain me paragraph 2 of your second numbered list in more detail (I didn´t understand what timestamp means; What would be a synonym for timestamp?, or how would you define timestamp?) (I didn´t understand either “corresponding to your detection limit for example, the 5th photon, Just drop 4 entirely“; Could you please explain this in other words / “in a simpler way”) ?

7. Which simulators would you recommend for simulating the electronics of my experimental setup?.

16

Not really following this forum thread, but…in my experience, for most applications, simply the number of photons hitting the photodiode is a good estimator of the electrical signal magnitude.

2 Likes
17

Ok, thank you for providing more information. Regarding the above, what would be an equation / formula that includes / describe the number of photons hitting the photodiode as well as the electrical signal magnitude?

18

Yes

Regarding delta “inputs“, do you mean that they are inputs that have a shape like a Dirac delta?

Yes

Could it be said that a light pulse can be represented by one of the things you mention in your first numbered list or a combination of them

Depends on what you are trying to measure and the time resolution of your photodiode. There is usually an impedance intrinsic to your device that will become a decay constant when read out of, for example, a 50 ohm cable. But that is entirely not something G4 can model.

The spec sheet of your transimpedance amplifier will let you calculate or estimate the single photon amplitude and be dependent on your circuit. A typical value might be like 0.01 mV. If the RMS noise of your circuit is 2 mV that translates to 2/0.01 = 200 photons as the ballpark limit of detection. Until you have the 200th photon arrive from the incident pulse, it will be indistinguishable from noise.

Could you please explain me paragraph 2 of your second numbered list in more detail

Timestamp or “arrival time” at the surface of your pin diode. Your pin diode signal will be the superposition of many, many photons arriving at slightly different times. As in the example of the previous section, you actually cannot tell that any incident radiation has hit your detector until you get the 200th photon arriving at the pin photodiode because it will be blended into the noise. For the purposes of calculating your signal you can average the momentum and position of all photons for the event and record a very compact event.

Yes, it is and probably a good first step. It just depends on the precision desired for the the actual final pulse shape with regards to filtering schemes. It is a very common problem tackled with SiPMs for example.

1 Like
19

Ok, thanks for your answers. I’ll keep them in mind.