djboersma
commented
4 years ago Thanks for bringing this up! I agree that the current implementation of convergence/divergence not ideal. One problem that I see is that for TPS, it's not possible to model a beam that is e.g. divergent at low energies and convergent at high energies. Using the sign of one of the optical parameters would possibly help to remove this limitation.
Question: I think that in your question you use "divergence" to mean the standard deviation of the angular distribution of the source particles, denoted by mSigmaTheta and mSigmaPhi in the code. Right?
It's a pity that "emittance" is defined in such a weird way. Instead of emittance I'd like to use the correlation coefficient rho between position and angle of a particle, which would be -1 for a perfectly convergent beam, zero for a beam that is neither convergent nor divergent (which happens at the "waist" of an initially convergent beam) and +1 for a perfectly divergent beam. The non-diagonal element in the Gaussian (https://github.com/OpenGATE/Gate/blob/develop/source/physics/src/GateSourcePencilBeam.cc#L181) would then be equal to -rho*mSigmaX*mSigmaTheta and similar for the Y and Phi Gaussian. If the "source properties files" would provide a polynomial (or maybe we should use tables instead, like we discussed a while ago) of the correlation coefficient instead of the emittance, then your wish to encode the convergence/divergence as a sign in one of the optical parameters is perfectly implemented, I think.
Your proposal to allow the standard deviation of the angular coordinate to have a sign, positive referring to convergence and negative to divergence, is in my opinion less optimal. Because in the example above, in which a beam is divergent at low energy (mSigmaTheta positive) and convergent at high energy (mSigmaTheta negative), so then there is an energy in between at which mSigmaTheta will be zero (regardless of whether we implement this with polynomials like now or with tables in the near future), while actually the standard deviation of the angles is not zero at all (for this energy the particles happen to be generated at the waist of the beam).
In short:
- let's combine this change with the plan to move to tabulated (instead of polynomial) source properties
- in the tabulated source properties, the user specifies the correlation coefficient instead of the "emittance"; this allows a natural encoding of the convergence and divergence without any unphysical artifacts.
andiresch
GrevillotL
Current status
For beam optics (in the PBS and TPS source), the user needs to provide, sigma, divergence, emittance as well as the beam convergence (e.g. converging or diverging beam). Due to the internal code, the sign of the divergence can not be used (squaring). Consequently, an additional parameter, beam convergence is used.
Proposal
I propose to implement the possibility to use the sign of the divergence (negative for convergence, positive for divergence) as a default. The current convergence flags could be used as a backup to keep the macros compatible. If one of the convergence flags is set, these should take precedence.
I believe this could be implemented with minimal changes, while simplifying macros and the overal use of GATE.