Function and Design

A detailed description of the geometry in not appropriate for this manual but an overview will help you to understand how the code works and how it is organised. The geometry has to service the following basic requests:-

REGION

Where am I? i.e. given X,Y,Z return region.

NEAR

Distance to nearest boundary from given X,Y,Z.

NEXT

Distance to boundary in given direction from given X,Y,Z.

PICK

Generate point randomly in region.

Callers may optionally pass the particle type (so that different particles could ignore certain elements of the geometry). Within the PMT support structure the geometry is full 3D. However the PMT panels and tubes are currently represented by flat surfaces so requests such as PICK are invalid in these regions.

The geometry design all flows from two guiding principles:-

• Each homogeneous region (called a detector element) has its own code to handle requests. This is done to keep execution time down for then the code can exploit any time saving feature specific to the region. For example each region knows exactly which other regions border it. This approach is not data driven; adding a new region means adding new (although small) routines to deal with it.

• There is only one copy of the code that deals with each boundary. This ensures a self consistency; regardless of which region a boundary is viewed from it has to look the same.

These principles appear to be in conflict: each boundary has to be dealt with by one routine yet each is shared by two regions each with their own routine. The solution is to assign priorities to each region. The only requirement is that two neighboring regions have different priorities. The boundary is owned by the region of higher priority; code dealing with the lower priority region must call the higher priority one to deal with the surface. In this scheme each region is defined by 2 types of boundary:-

EXTERNAL BOUNDARY

to regions of lower priority.

INTERNAL BOUNDARY

to regions of higher priority.

Code for a given detector element has only to deal with the boundary it owns i.e. the external boundary.

Figure 12.1: Region and Boundaries

As an illustration consider diagram 12.1 where the numbers represent both the regions and their priorities. The table below shows which boundaries are internal to each region.

Region	Internal Boundary
1	2 3 4
2	3
3	4
4	None

So the geometry is an embedding geometry, regions of higher priority are embedded in regions of lower priority. As a specific example, the acrylic vessel is considered to be a solid sphere of acrylic in which is embedded a sphere of D$_2$O , a set of belly plates and two concentric cylinders that represent the chimney. Of course, the concept of embedding can be nested. For example, the detector starts as a solid ball of rock in which are embedded a counting room and the inner darkness. Within these regions further regions are embedded, and the process repeats until a complete model of the detector has been assembled. So, strictly speaking, all other regions are internal to the rock. However, the only internal boundaries that have to be dealt with here, or in any other region, are those that are directly visible: the rock cannot `see' individual PMTs!

For panel zones and PMT hexagons, it is the `array of' that embeds. Internally these regions are divided up into a set of tessellating sub regions that all have the same priority. The concept of boundary consistency is not compromised however, as the consistent handling of internal boundaries is part of the `array of' concept. For example, a panel is viewed as a honeycomb of hexagonal prisms, each having its own serial number but all having the same priority. The boundary of every hexagon is handled by the same code, thus ensuring consistency.