Problem 444

Summary:	OpBoundaryProcess doesn't behave as claimed in the manual
Product:	Geant4	Reporter:	ob214
Component:	processes/optical	Assignee:	gum
Status:	CLOSED FIXED
Severity:	normal
Priority:	P2
Version:	other
Hardware:	PC
OS:	Linux

Description ob214 2003-01-04 19:57:53 CET

I believe the Optical Boundary Process defined in G4OpBoundaryProcess doesn't
do what is said it does in the manual.
Namely, if you define an optical surface with an index of refraction, the
manual says that there will be 2 boundary effects: "This allows the simulation
of boundary effects at the intersection between the medium and the surface
layer, as well as at the far side of the thin layer".
This doesn't work.
The index of the thin layer is used only if the index of the second material
has not been defined (line 203-208) and in any case I believe only 1 boundary
effect is simulated (I'm not 100% sure of that fact though but after a quick
look at the code it looks like that).
Also, at the moment if the second material doesn't have an index of
refraction, the only thing checked to decide whether or not to carry on is the
existence of an optical surface (line 180-191). If the user didn't define an
index for the surface, Rindex2 is undefined and the boundary process shouldn't
carry on. This is not checked at the moment.

At last, could I make a suggestion.
At the moment, it not possible to simulate a boundary effect between 2 volumes
made of the same material even if an optical surface between those 2 volumes
is defined (line 124-125).
You have to define the same material twice to be able to do that.
It would be nice to allow a boundary effect between 2 volumes made of the same
material to be simulated if and only if an optical surface is defined.
Therefore complex object geometry (with therefore no boundary effect between
the different volumes describing that object) could be implemented if no
optical surface is defined. But at the same time, boundary effects between
crystals (array of crystals in a calorimeter for instance) could be forced to
be simulated simply by defining an optical surface (with a different
refraction index) without having to define 2 materials. This is what has to be
done at the moment and one has to make sure 2 direct neighbour crystals are
defined with different "version" of the same material.

Comment 1 gum 2003-01-28 17:42:59 CET

(1) If you specify the UNIFIED model, define an optical surface with an index of
refraction, and the "backpainted" surface finish, two boundary effects are
simulated. The first is between the original medium and the thin surface layer.
The second is Lambertian reflection off the 'paint' at the far side of the thin
layer. It is not possible to simulate a thin film between two dielectric media
by simply specifying an optical surface. In such a case, the thin film has to be
defined as a true G4 volume and give it an index of refraction. The G4Navigator
will then track the optical photon across two boundaries: n1 | n_layer | n2 (or
again n1). For example, if you wanted to simulate the silicon oil between the
PMT glass and your scintillator, you'd have to define the silicon oil film as a
true G4 volume. I agree, the documentation should be more precise and clearly
state that 'backpainted' finish assumes that a metallic paint (Lambertian
Reflection) is the second boundary.

(2) It is true that the thin layer index is only used when the second material
has no index define, else the second material's index is used. I agree, the
optical surface index could probably take precedence instead without ill effect.

(3) I also agree that by requiring the index of refraction be different on both
sides of an intersection, an unnecessary complication arises when an array of
crystals touch, each of which is wrapped but the wrapping and thin layer are not
implemented as physical G4 volumes but as part of an optical surface. - The
check is there to bypass the process in cases where the index of refraction is
equal on both sides of the boundary. Unless I can think of a problem, I shall
change the code so that the process returns only if, in addition, no optical
surface was defined as you have suggested.

(4) Thanks for pointing out that Rindex2 can be undefined if the user didn't
specify an index for the second medium and didn't specify one for the thin layer.

As a general comment, another solution to your present problem would be to allow
for a thin air space between your different (wrapped) crystals by placing
slightly smaller volumes into an array. In this case, the G4Navigator would find
the air space with no index defined, and use the optical surface as expected.
Down to the allowed precision of the G4Navigator, there probably is a thin air
space between crystal cases.

Comment 2 gum 2003-03-15 18:56:59 CET

I have adopted the following logic:

Material1 -> Rindex1

if(!Rindex1) kill the photon - return (as before)

Defaults (as before)
--------------------

theModel = glisur;
theFinish = polished;

G4OpticalSurfaceType type = dielectric_dielectric;

if(OpticalSurface){

    type      = OpticalSurface->GetType();
    theModel  = OpticalSurface->GetModel();
    theFinish = OpticalSurface->GetFinish();

    if(theFinish == ...backpainted){
      OpticalSurface -> Rindex2
      if(!Rindex2) kill the photon - return
    }
}

if (type == dielectric_dielectric){
   if(theFinish == 'polished' || 'ground') {

// e.g. no 'paint' and not 'dielectric_metal'

   if(Material1 == Material2)return

   Material2 -> Rindex2
   if(!Rindex2)kill photon - return

   }
}