[Halld-pid] light guide simulation

[Alexander Ostrovidov]

On Tuesday, March 22, 2011, Beni Zihlmann wrote:
> Hi All,
> I have simulated a light guide using GEANT4 and looked at how much light
> we lose.
> You can find the results at the following link.
> http://www.jlab.org/Hall-D/software/wiki/index.php/Light_Guide_Design
> 
> Because it is very hard to model a bend in the tapered section where by
> "tapered"
> I mean the part that converts a rectangular shape into a round shape I
> left this
> part strait. I introduced a bend afterwards in the round section. That
> is much easier
> to model. It turns out that most light gets lost in the tapered section
> anyway even
> without a bend. Secondly as expected a larger bend causes more light to
> be lost.
> In general we can expect a light loss of about 25% to 30% through the
> full light guide
> including interfaces also to the PMT.
> 
> any thoughts and ideas are welcome,
> 

Beni,

An interesting result. What I'm puzzled about is that more than
half of all photons (15% out of 25%)  are lost in the first very short
rectangular section of the light guide. Obviously, this is not due to 
photon absorption in lucite because this section is only about 
one-tenth of the total length. I doubt that this is due  to rectangular
shape of this section either. TOF paddle has the  same shape and is 
about 100 times longer, which means that amount of light from hits 
at the far end of TOF paddle would be equal to (1-0.15)^100=nothing
if such shape is a culprit.

I suspect that the reason for this large loss is  2pi solid angle of
photons generated near the light-guide entrance. A significant
fraction of these photons will have incident angle higher than
the angle of total internal reflection. Many of them will escape
when they hit light guide wall for the first time. That, of course,
is likely to happen in its first section for large-angle photons.

Loss of light in light guides is most important for the smallest 
signals coming from hits at the far end of a TOF paddle. My
feeling is that mostly photons which are nearly parallel to
the paddle will reach the light guide at the opposite end.
So, my suggestion would be to run your Monte Carlo 
with photons at 0 (or nearly 0) angle instead of 2pi soild angle 
to see how the conclusion about light loss in the light guide
depends on the assumption about initial angular distribution
of the generated photons.

Sasha