What is the state of the art in particle detection and localization

I am researching methods to detect the position of radioactive materials (emitting gamma and beta particles), and would like to know what current methods are used to do this.

What type of sensors are used, and how do they process the information to detect the position of a radioactive material in 3d space?

Are there any limitations/difficulties with these methods? e.g computationally expensive, inaccurate, only works for one type of element etc.

I'm developing my own method to detect the position of radioactive material in 3d space, and I'd like to know what the current standard method for doing this is so I can research and compare the methods.

My method involves liquid scintillation counters and a neural network.

The material will be highly radioactive. e.g 100g of uranium. It is limited to (4x14x6)m area, and could be buried in a solid, surrounded by air, inside a container. However I am assuming that the matter surrounding the radioactive material is uniformly distributed with constant density.

-
If you re-word this question to ask something more along the lines of "what is the state of the art in particle detection and localization" and describe what you're trying to detect (charged particles?) the question would be vastly improved. – Brandon Enright Jan 15 '14 at 18:46
You haven't mentioned the environment. Will your source be in vacuum? In air? In some liquid? Buried in a solid? Supported in a heavy, duty steel frame with various bits and pieces but a lot of air? These things really matter as the penetration of the various particles varies. Also, can you impose an strong magnetic field without messing things up? – dmckee Jan 15 '14 at 19:14
Thank you for your suggestions. I have edited the question. – Blue7 Jan 16 '14 at 1:18
The thing is that the amount and composition matter. Less than $1 \,\mathrm{kg/cm^2}$ of almost anything will block a huge fraction of the natural decay gammas making direct detection require considerable integration time. For cases like that neutron fluorescence‎ has been proposed, but it requires a fairly complex arrangement. – dmckee Jan 16 '14 at 2:40

1 Answer

Finding a radioactive item is conceptually the same as finding a light source. You detect what it emits with a sensor that measures the angle the radiation comes from and project it back. Do this with a few detectors and find the common point. The problem comes if you can only absorb the radiation without measuring the direction.

What radiation is it emitting? High energy particles are best detected with some sort of track chamber-the track points back to the source. High energy gamma rays make an electron-positron shower that points back to the source, though maybe not as accurately. Low energy alpha particles have the problem that they don't have much range and also have their paths deflected easily.

Charged macroscopic objects are hard to sense. About all you can use are electric field probes of some sort. How close will you be?

-
Thank you for your answer, I did not realise that detectors could measure the direction of the particle. The element will be emitting beta and gamma radiation, and the furthest it can be from a detector is 8 meters. I will be using multiple detectors though, so the furthest it will be from a single detector is less than 2 meters. – Blue7 Jan 15 '14 at 16:38
Unless you can measure the direction, you can't determine the position. The particles are not emitted simultaneously, so you can't use time of arrival for anything. The first step is to understand what your sensor can do. – Ross Millikan Jan 15 '14 at 16:43
My plan is to use the count rate at each detector to determine the position. e.g in a square array of detectors with the element near the bottom right corner, the detectors closer to the element will have a higher count rate than the detectors further away. The information will be fed into a neural network that has been trained to give the position of the element. I will test using simulations if this method is feasible. First, I just wanted to know what the current methods for position detection are so I can research them and include them in the literature review. – Blue7 Jan 15 '14 at 16:56
That sounds like it should work. I doubt you need a neural network. The counting rate will give you a measure of distance. Now for a given location, you can compute the count rate expected at each detector location. You can minimize the error between the prediction and measurement using a multidimensional minimizer. They are described in any numeric analysis text. – Ross Millikan Jan 15 '14 at 17:04
You could look at chapter 10 of Numerical Recipes. Obsolete versions are free. It gives a discussion and algorithms in C. I don't know much about neural networks, but I don't think of them for accurate numerical computation. – Ross Millikan Jan 15 '14 at 17:19