*
* $Id: gedoc,v 1.1 1996/03/06 15:31:14 mclareni Exp $
*
* $Log: gedoc,v $
* Revision 1.1  1996/03/06 15:31:14  mclareni
* Add geane321 history, CMZ and doc files
*
*
*CMZ :  3.21/02 29/03/94  15.41.49  by  S.Giani
*-- Author :
 
*-- Author :
   1. Introduction
 
   2. Definitions
 
   3. Description of the User Routines and COMMONs
 
   4. Examples of application
 
   5. Interface with GEANT
 
   6. Acknowledgements
 
 
 
   1. Introduction
      ============
 
      The present Package allows the user to calculate the
average trajectories of particles and to calculate the
transport matrix as well as the propagated error covariance
matrix. It makes use of a set of routines worked out
by the European Muon Collaboration [1] and it is integrated
to the GEANT3 system [2] with expected applications in both
simulation and reconstruction context.
 
      The package is available as a PAM-file. It contains
two basic PATCHes, one with the original EMC routines,
the other with new GEANT-type tracking routines and interface
between them and the EMC routines. This second PATCH
contains those routines the user should invoke
to carry out the tracking (routines ERTRAK, EUFILL, EUFILP, EUFILV).
In addition to this a series of utilities are available for
the user (e.g. to transform the track representation from one
system to another or to carry out 5 X 5 matrix multiplication
in an optimum way).
 
     In Section 2 we give the definitions of the mathematical
quantities to be dealt with. In Section 3 the description of
the user routines are given. In Section 4 we illustrate the
application of the program by several examples. Finally in
Section 5 the GEANT part is described.
 
      Further development will concern the improvement of the
error matrix by taking into account the Landau tail in the
fluctuation of the energy loss, the bremsstrahlung and direct
pair production of the muons.
 
 
      2. Definitions
         ===========
 
         2.1 Track variables, Representations
             --------------------------------
 
      The particle trajectory is characterized by 5 independent
variables as a function of one parameter (e.g. the pathlength).
Among the 5 variables 1 is related to the curvature (to the absolute
value of the momentum, p), 2 are related to the direction of the
particle and the other 2 are related to the spatial location.
The most usual representation of these 5 parameters are:
 
       I.
 
       1/p, lambda, phi, y_perp, z_perp
 
 
where lambda and phi are the dip and azimuthal angles related
to the momentum components in the following way:
 
            p_x = p cos(lambda) cos(phi)
            p_y = p cos(lambda) sin(phi)
            p_z = p sin(lambda)
 
 
y_perp and z_perp are the coordinates of the trajectory in a
local orthonormal reference frame with the x_perp axis along the
particle direction, the y_perp being parallel to the x-y plane.
This representation is usually applied in the overall reference
frame. (In the EMC code this reference frame is labelled by 'SC'
since the overall system was identified with that of the Streamer
Chamber.)
 
       II.
 
       1/p, y', z', y, z
 
 
where y'=dy/dx and z'=dz/dx. This representation is particularly
useful in fixed target experiments, where the trajectory is evaluated
on successive parallel planes (which are perpendicular to the x-axis).
(In the EMC code this representation is labelled by 'SP' since a
convenient mathematical description of a trajectory being approxima-
tely parallel to the x-axis is a 'spline'.)
 
       III.
 
       1/p, v', w', v, w
 
 
where v'=dv/du and w'=dw/du in an orthonormal coordinate system with
axis u, v and w. This representation is paricularly useful when the
trajectory has to be evaluated on different detector planes
in a colliding beam experiment, where the planes can take a great
variety of directions.(In the EMC code this representation is
labelled by 'SD' as System of Detection.)
 
      Of course, all the above representations of the trajectory
are equivalent and one can go from one representation to the
other by calculating the corresponding Jacobian. These Jacobians
are provided by the following EMC routines:
 
     S/R TRSCSP    from I   to II
     S/R TRSPSC    from II  to I
     S/R TRSCSD    from I   to III
     S/R TRSDSC    from III to I
 
 
 
     2.2 Error Propagation
         -----------------
 
 
     Let us denote in the following the 5 independent variables at
a given value of parameter l_0 (e.g. pathlength) by x_i(l_0),
(i=1,...,5). In many applications we are interested in the evolution
of the average value of x_i for l>l_0: E(x_i). This is calculated
by GEANT as will be outlined in Section 5.
 
     The knowledge on the avarage trajectory is characterized by the
5 X 5 covariance matrix of the variables:
 
   sigma_$ij(l_0) = E(x_i(l_0).x_j(l_0)) - E(x_i(l_0)).E(x_j(l_0))
 
We are also interested in the evolution of sigma_$ij for l>l_0,
which we call error propagation. If the particle is propagating in a
deterministic way, i.e. without any random process involved (
e.g. in vacuum) then the propagation of sigma is simply described by
the so called transport matrix in the following way:
 
   sigma_$ij(l) = T_$jm(l,l_0).sigma_$mn(l_0)T_$in(l,l_0)
 
where the transport matrix expresses the infinitesimal change
in the parameters at l with respect to the change of parameters
at l_0:
 
        T_$ij(l,l_0) = delta (x_i(l))/ delta (x_j(l_0)).
 
    In a realistic detector, however, the particles undergo random
processes as well, like Multiple Coulomb scattering, energy loss
due to delta ray production, etc. therefore the error propagation
should contain an additional term:
 
    sigma_$ij(l) = T_$jm(l,l_0).sigma_$mn(l_0).T_$in(l,l_0) +
                  + sigma_$ij:$random(l).
 
   The program calculates sigma_$ij(l) step by step using the above
recursive formula, where T_$ij and sigma_$ij:$random refers to the
actual step and sigma_$mn is the cumulative error for all previous
steps. For the mathematical formulae to calculate T_$ij for a finite
step the reader is referred to Ref [1].
 
   By invoking the subroutine ERTRAK (see next Section) the user
will have access to the average trajectory, to the full error matrix
represented and in addition to this the program makes available also
the transport matrix given by which can be useful in several
applications (see Section 4.)
 
 
 
   3. Description of the User Routines and COMMONs
      ============================================
 
   To run the program the user should first initialize GEANT,
(set-up the geometry and initialize the appropriate physics
processes). This procedure will be described in Section 5.
The tracking with error propagation is carried out by invoking
subr. ERTRAK. However, before calling ERTRAK the user should
provide informations to the program in two commons:
/ERTRIO/ and the pair /EROPTS/ and /EROPTC/. For this purpose
a series of user routines
are forseen (routines EUFILL, EUFILP, EUFILV), which should be
called by the user. The result of the tracking is partly
returned in the arguments of the routine ERTRAK and partly can be
accessed through the common /ERTRIO/. In the following we give
a description of the user routines (subr. ERTRAK, EUFIL,L,P,V)
and that of the commons /ERTRIO/, /EROPTS/ and /EROPTC/.
 
 
       3.1 User Routines
           -------------
 
    The output parameters are denoted by asterisk in the calling
sequence.
 
      SUBROUTINE ERTRAK (X1, P1, X2*, P2*, IPA, CHOPT)
      ================================================
 
 
   Performs the tracking of the track from point X1 to
             point X2
   (Before calling this routine the user should also provide
             the input informations in /EROPTS/, /EROPTC/ and /ERTRIO/
             using subr. EUFIL(L/P/V)
 
          X1       - Starting coordinates (Cartesian)
          P1       - Starting 3-momentum  (Cartesian)
          X2       - Final coordinates    (Cartesian)
          P2       - Final 3-momentum     (Cartesian)
          IPA      - Particle code (a la GEANT) of the track
 
          CHOPT
              'B'   'Backward tracking' - i.e. energy loss
                                 added to the current energy
              'E'   'Exact' calculation of errors assuming
                                 helix (i.e. pathlength not
                                 assumed as infinitesimal)
              'L'   Tracking upto prescribed Lengths reached
              'M'   'Mixed' prediction (not yet coded)
              'O'   Tracking 'Only' without calculating errors
              'P'   Tracking upto prescribed Planes reached
              'V'   Tracking upto prescribed Volumes reached
              'X'   Tracking upto prescribed Point approached
 
 
       SUBROUTINE EUFILL (N, EIN, XLF)
       ===============================
 
 
    User routine to fill the input values of the commons :
               /EROPTS/, /EROPTC/ and /ERTRIO/ for CHOPT = 'L'
 
         N     Number of predictions where to store results
         EIN   Input error matrix
         XLF   Defines the tracklengths which if passed the
                      result should be stored
 
        SUBROUTINE EUFILP (N, EIN, PLI, PLF)
        ====================================
 
 
    User routine to fill the input values of the commons :
               /EROPTS/, /EROPTC/ and /ERTRIO/ for CHOPT = 'P'
         N     Number of predictions where to store results
         EIN   Input error matrix (in the 'Plane' system )
         PLI   Defines the start plane
                      PLI(3,1) - and
                      PLI(3,2) - 2 unit vectors in the plane
         PLF   Defines the end plane
                      PLF(3,1,I) - and
                      PLF(3,2,I) - 2 unit vectors in the plane
                      PLF(3,3,I) - point on the plane
                                   at intermediate point I
 
 
        SUBROUTINE EUFILV (N, EIN, CNAMV, NUMV, IOVL)
        ============================================
 
 
    User routine to fill the input values of the commons :
               /EROPTS/, /EROPTC/ and /ERTRIO/ for CHOPT = 'V'
         N     Number of predictions where to store results
         EIN   Input error matrix
       CNAMV   Volume name of the prediction
        NUMV   Volume number (if 0 = all volumes)
        IOVL   = 1  prediction when entering in the volume
               = 2  prediction when leaving the volume
 
 
         2.2 User COMMONs
             ------------
 
      CHARACTER*8      CHOPTI
      PARAMETER        (MXPRED = 10)
      DOUBLE PRECISION ERDTRP
      LOGICAL          LEEXAC, LELENG, LEONLY, LEPLAN, LEPOIN, LEVOLU
 
      COMMON /EROPTS/ ERPLI(3,2), ERPLO(3,4,MXPRED), ERLENG(MXPRED),
     ,                NAMEER(MXPRED), NUMVER(MXPRED), IOVLER(MXPRED),
     ,                LEEXAC, LELENG, LEONLY, LEPLAN, LEPOIN, LEVOLU
      COMMON /EROPTC/CHOPTI
 
      COMMON /ERTRIO/  ERDTRP(5,5,MXPRED), ERRIN(15), ERROUT(15,MXPRED),
     ,                 ERTRSP(5,5,MXPRED), ERXIN( 3), ERXOUT( 3,MXPRED),
     ,                 ERPIN(3),ERPOUT(3,MXPRED),NEPRED,INLIST, ILPRED,
     ,                 IEPRED(MXPRED)
 
 
LEEXAC           = .TRUE. if CHOPT = 'E' in  ERTRAK
LELENG           = .TRUE. if CHOPT = 'L' in  ERTRAK
LEONLY           = .TRUE. if CHOPT = 'O' in  ERTRAK
LEPLAN           = .TRUE. if CHOPT = 'P' in  ERTRAK
LEPOIN           = .TRUE. if CHOPT = 'X' in  ERTRAK
LEVOLU           = .TRUE. if CHOPT = 'V' in  ERTRAK
IOPTER(I)        = 1 if the Ith letter of the alphabet is
                 occuring in CHOPT in ERTRAK (else = 0)
 
NEPRED           Number of predictions (c.f. N in EUFILL,P,V)
ERPLI            Initial plane descriptor (c.f. PLI in EUFILP)
ERPLO(,,INLIST)  Final plane descriptor - first 3 vectors are identic
                 with PLF in EUFILP, the 4th vector is the cross-product
                 of the first two vectors (plane normal)
ERLENG(INLIST)   Lengths to store results (c.f. XLF in EUFILL)
NAMEER(INLIST)   Volume names   to store results (c.f. CNAMV in EUFILV)
NUMVER(INLIST)   Volume numbers to store results (c.f. NUMV in EUFILV)
IOVLER(INLIST)   (c.f. IOVL in EUFILV)
 
ILPRED           Current number of prediction
IEPRED(ILPRED) = INLIST if the ILPREDth prediction reached (else = 0)
ERDTRP(,,ILPRED) Double precision value of the Transport Matrix
                 at the prediction ILPRED
ERRIN            Input Error Matrix in Triangular form
ERROUT(,ILPRED)  Output Error Matrix in Triangular form
                 at the prediction ILPRED
ERTRSP(,,ILPRED) Single precision value of the Transport Matrix
                 at the prediction ILPRED
ERXIN            Starting coordinates (c.f. X1 in ERTRAK)
ERXOUT(,ILPRED)  Output coordinates at the prediction ILPRED
ERPIN            Starting momentum
ERPOUT(,ILPRED)  Output momentum at the prediction ILPRED
 
 
 
    Note that ERRIN, ERROUT, ERPIN, ERPOUT, ERTRSP and ERDTSP are
given by the program in the representation which is requested by
CHOPT in subr. ERTRAK. (E.g. if CHOPT='P', all the above quantites
are given in the representation III.)
 
 
 
   4. Examples of Application
      =======================
 
 
    4.1 The simplest case: Representing the trajectory at another point
        ---------------------------------------------------------------
 
    Usually the particle trajectory is not measured at the  point of
production where its physical parameters are of interest. Therefore
the measurement has to be extrapolated back close to the origin. This
can be achieved by the simple call:
 
        CALL ERTRAK(X1,P1,X2,P2,IT,CHOPT).
 
Since this extrapolation is opposite to the particle direction,
CHOPT should contain the letter 'B'. If the tracking should be
stopped on a prescribed plane, CHOPT should also contain 'P',
and before invoking ERTRAK the user should call subr. EUFILP.
This extrapolation can be carried out simultaneousely onto several
planes, in this case the 1st argument of EUFILP is greater than 1.
The result can be retrieved from the common /ERTRIO/ as described
in Section 3.
 
     4.2 Joining track elements in different parts of the detector
         ---------------------------------------------------------
 
    It happens frequently that one measures a part of a trajectory
in a downstream detector and would like to join this information
to another one obtained in a detector close to the interaction
point. Since there are usually several trajectories which could be
a priori joined the first task is to find the one which  matches
the best. The next task is to improve the trajectory parameters.
 
    One chooses a plane near to the interaction point and extrapolates
onto this plane all candidate trajectories as described in the
preceeding section. For the i-th trajectory one obtains an
avarage value  x_i and a covariance: sigma_i. (In this discussion
the indices will represent the trajectory numbers.) Next one
extrapolates back the trajectory from the downstream detector
to the same plane and obtains x_d and  sigma_d. One can then construct
a chi:2 for each track i:
 
       chi:2_i = (x_i-x_d)(sigma_i+sigma_d):$-1(x_i-x_d)
 
    The matching condition can be defined as:
 
                   chi:2_i.leq.chi:2_0,
 
where chi:2_0 is some prescribed value.
 
     Having chosen trajectory i for the matching the improved track
parameters can be obtained by minimizing
 
   chi:2 = (x-x_d).sigma_d:$-1.(x-x_d)+(x-x_i).sigma_i:$-1.(x-x_i)
 
w.r.t. x resulting in:
 
  x_$impr = (sigma_d:$-1+sigma_i:$-1):$-1.
                      (sigma_d:$-1.x_d+sigma_i:$-1.x_i)
 
The covariance of x_$impr
 
       sigma_$impr =  (sigma_d:$-1+sigma_i:$-1):$-1
 
shows explicitely the improvement of the trajectory parameters.
 
   This procedure can be easily generalized to join more than 2
measurements on the particle trajectory. If e.g. between the two
above planes there is another detection plane, one can first
merge the informations of the downstream and intermediate
plane and continue the backtracking from the intermediate
plane to the plane close to the interaction point with the
improved trajectory parameters.
 
   The procedure can be used in principle also if not all
the five parameters are measured (e.g. if only the
coordinate informations are available). In this case one
starts the back-tracking with some initial values of the
unmeasured parameters and assigns an error to these
parameters which is much larger than the difference between
the true and the initial value. The user is however has to
ensure that the result is stable against the choice of
the starting value of parameters and errors (e.g. by performing
several iterations). These problems can be overcome
by a fitting procedure which is described in Section 4.4.
 
      4.3 Prediction of the trajectory
          ----------------------------
 
      It is often needed to predict the particle trajectory in a
detector plane at a certain confidence level in order to
perform pattern recognition. An example is to find hits from
a penetrating particle inside a segmented calorimeter when the
particle trajectory is well measured at the two extrems of the
calorimeter.
 
      In the case of 1 intermediate plane inside the calorimeter
the solution can be obtained by combining the methods outlined
in the previous two sections. One extrapolates the measured track
parameters from the two endplanes of the calorimeter onto the plane
set up inside the calorimeter (Section 3.1) and one joins the two
 informations on that plane (Section 3.2). This procedure of course
can be carried out on any number of intermediate planes. However,
if there is a large number of planes, it is advantageous to carry
out the tracking in one direction and in one go, for which case a
method is outlined below.
 
      Let's start the tracking from one end of the calorimeter
and denote by x_i and by x_e the average track parameters on the
intermediate plane i and on the other end-plane e, respectively.
 Let's denote the true track parameters on the same planes by x
and by y, respectively. The corresponding chi:2, which we should
minimize w.r.t. x is:
 
         chi:2 = (x-x_i).sigma_i:$-1.(x-x_i) +
                 + (y(x)-x_m).sigma_$em:$-1.(y(x)-x_m)
 
where x_m is the measured trajectory at the end-plane with
covariance matrix sigma_m, sigma_i is the propagated error
matrix from the starting plane to plane i and
 
            sigma_$em = sigma_m + sigma_$ei
 
where sigma_$ei is the propagated error from plane i to plane e
(n o t  including the error on plane i itself).
 
   The minimization results in the following equation:
 
   sigma_i:$-1.(x-x_i) + dy/dx.sigma_$em:$-1.(y(x)-x_m) = 0,
 
which we solve by linearization:
 
                  x = x_i + Delta x
 
      y = y(x_i)+dy(x_i)/dx.Delta x = x_e+T(e,i).Delta x.
 
   The result is:
 
        Delta x = sigma_x.T:T(e,i).sigma_$em:$-1.(x_m-x_e)
 
where:
 
   sigma_x = (sigma_i:$-1+T:T(e,i).sigma_$em:$-1.T(e,i)):$-1
 
is the covariance matrix of the trajectory prediction at the
plane i (T:T means the transpose of T).
 
    The following glossary gives the correspondance between the
mathematical quantities used in the above equations and the
varibales in the user common /ERTRIO/:
 
x_i        ERXOUT(,I), ERPOUT(,I) (I standing for prediction i)
x_         ERXOUT(,IE), ERPOUT(,IE) (IE standing for prediction e)
sigma_i    ERROUT(,I)
T(e,i)     T(e,1).T:$-1(i,1)
T(e,1)     ERTRSP(,,IE)
T(i,1)     ERTRSP(,,I)
sigma_$ei sigma_e-T(e,i).sigma_iT(e,i)
sigma_e    ERROUT(,IE)
 
 
 
 
     4.4 Fitting trajectory parameters
         -----------------------------
 
 
      In the above examples all of the 5 variables of the trajectory
have been known at least in one space point. However, in most of the
cases direct mesurements yield only the coordinate informations, from
which one should reconstruct the curvature and the direction. The
following example shows how to use the program package for this pur-
pose. This tool can be applied in the most general case: in inhomoge-
neous magnetic field and even if the particle passes through a great
amount of material.
 
      Suppose we would like to reconstruct the particle trajectory
x_0 at plane 0 by measuring the coordinate informations x_i:m
at N different detector planes (i=1,...,N). If in the formation
of the trajectory random processes can be neglected, then
the average trajectory can be obtained by
minimizing the following chi:2 w.r.t. x_0:
 
 chi:2 = Sum_$i=1:N[(x_i(x_0)-x_i:m).sigma_i:m:$-1.(x_i(x_0)-x_i:m)]
 
where x_i are the true track parameters at plane i, and sigma_i:m is
the 2 X 2 covariance matrix of the measurement on plane i. This results
in the following equation:
 
    Sum_$i=1:NT:T(i,0).sigma_i:m:$-1.(x_i(x_0)-x_i:m) = 0
 
where T(i,0) is the transport matrix between plane 0 and plane i.
This equation is again solved by linearization. In the first
approximation we calculate the trial trajectory x_i:t on plane i
starting with a value x_0:t. The true value is then obtained by:
 
              x_0 = x_0:t + Delta x_0
with
 
Delta x_0 = sigma_$x_0.Sum_$i=1:NT:T(i,0).sigma_i:m:$-1.(x_i:m-x_i:t),
 
where the covariance matrix of x_0 is given by:
 
  sigma_$x_0 = Sum_$i=1:NT:T(i,0).sigma_i:m:$-1.T(i,0)]:$-1.
 
     If in the formation of the particle trajectory random processes,
like multiple Coulomb scattering, cannot be neglected then obviousely
there are correlations in the error matrix between different planes
and therefore the above chi:2 should be written as
 
          chi:2 = (x(x_0)-x:m).sigma:$-1.(x(x_0)-x:m)
 
where x is a vector of length 2 X N containing the coordinate values
(xi,eta) of the average trajectory plane by plane:
 
     (xi_1,eta_1,xi_2,eta_2,...,xi_N,eta_N),
 
x:m is the corresponding vector of the measured coordinates.
sigma is a 2N X 2N matrix, whose 2 X 2 diagonal submatrices can be
written as
 
            sigma_$ii = sigma_i:m + sigma(2)_i:r
 
where sigma(2)_i:r is the 2 X 2 part of the covariance matrix sigma_i:r
due to random processes. The off-diagonal 2 X 2 matrices give the the
correlations between planes:
 
          sigma_$ij = T(j,i).sigma_i:r      (i<j)
 
where T(j,i) is the 5 X 5 transport matrix between plane i and j and
only the 2 X 2 part of sigma_$ij is considered.
 
   The minimization procedure is formally the same as before:
 
                 x_0 = x_0:t + Delta x_0
 
with
 
     Delta x_0 = sigma_$x_0.tau:T.sigma:$-1.(x:m-x:t),
 
where the 5 X 5 covariance matrix of x_0 is given by:
 
      sigma_$x_0 = (tau:T.sigma:$-1.tau):$-1.
 
Here tau is the joint transport matrix of dimension 5 X 2N
containing the transport matrices T(i,0) (5 variables for
plane 0 and 2 coordinate variables for plane i).
 
    Again the correspondance between the above symbols and variables
calculated by the program package is given by the following glossary:
 
     In case of tracking from plane 0 to plane i in one go:
 
 
x_i:t           ERXOUT(,I), ERPOUT(,I) (I standing for prediction i)
sigma_i:r       ERROUT(,I) (starting with 0 error)
T(j,i)          T(j,0).T:$-1(i,0)
T(i,0)          ERTRSP(,,I)
 
 
 
 
    5. Interface with GEANT
       ====================
 
 
     The following diagram shows the program flow-chart of ERTRAK.
Two running examples show how to build up a complete application.
(see patch EREXAM1 and EREXAM2)
 
     As can be seen from the flow-chart the particle is propagated
through the experimental setup by the routine ERTRGO which is a
simplified version of the GEANT routines GTRACK and GTVOL. In fact,
the calculation of the average trajectory is independent of the basic
GEANT Tracking Package.
 
     On the other hand, the GEANT Geometry Package is invoked
by calling GMEDIA,GTMEDI,GINVOL,GTNEXT (see the flowchart).
Therefore the GEANT Data Structures: JVOLU, JMATE, JTMED, JPART
must be set up at initialization time (see patch erexample below).
 
     In fact, in the JMATE data structure, only the energy loss
tables (JLOSS and JRANG) are required. The energy loss calculation must
include the contributions coming from the undetectable random
processes: delta rays, bremsstrahlung and direct pair production
of muons. The user should check that the values of the parameters
DCUTE/M, BCUTE/M, PPCUTM have been set coherently with his
application. (In most of the cases these parameters should be set
to their maximum value (say 10 TeV).)
 
    No other GEANT facilities are required to run ERTRAK.
No GEANT data structurs are overwritten. Only the commons /GCKINE/
and /GCTRAK/ are used to keep the current values of the parameters
of the average trajectory.
 
     For debugging purposes the control is given to the user at each
tracking step via the routine EUSTEP. This is the equivalent of
the GEANT GUSTEP routine. The routine ERXYZC is a copy of GPCXYZ.
 
    After the initialization a call to ERTRAK can be done at any
place of the user's code both in a context of Reconstruction or
Simulation, without interfering with other eventual GEANT calls
in the same event.
 
 
 
               Flow-Chart for subroutine ERTRAK
               ================================
 
 
    ERTRAK
    ======
      |
      | ----> GUFLD
      | ----> TRSCSD
      | ----> TRSDSC
      | ----> ERBCER
      | ----> GEKBIN
      | ----> ERPINI ----> TRPROP
      | ----> ERTRGO
              ======
                |
                | ----> GMEDIA
                | ----> GUFLD
                | ----> EVOLIO
                | ----> ERSTOR
                |       ======
                |         |
                |         | ----> ERBCER
                |         | ----> ERBCTR
                |         | ----> TRSCSD
                |         | ----> DMMSS
                |
                |
                | ----> ERTRCH -| ----> GTNEXT
                |       ======  |
                | ----> ERTRNT -| ----> GUSWIM ----> GUFLD
                |       ======  | ----> GINVOL
                |               | ----> ERLAND
                |               | ----> GEKBIN
                |               | ----> ERPROP
                |               |       ======
                | ----> EUSTEP  |          |
                | ----> EVOLIO  |          | ----> GUFLD
                l               |          | ----> TRPROP
                l ----> GTMEDI  |          | ----> TRPRFN
                                |          | ----> SSMTST
                                |          | ----> ERMCSC
                                |
                                | ----> ERSTOR
                                        ======
 
 
 
   6. Acknowledgements
      ================
 
      The authors of the present interface benefitted numerous critical
remarks and useful suggestions from the authors of the GEANT3 Package,
especially from F. Bruyant (CERN), which are greatly acknowledged here.
 
      Complaints and suggest must be sent to one of the authors :
          Innocent@cernvm    Maire@cernvm    Nagy@cernvm
 
 
 
                References:
                ===========
 
 
[1] W.Wittek, EMC Internal Reports: EMC/80/15,   EMCSW/80/39,
                                    EMCSW/81/13, EMCSW/81/18
    A.Haas,   The EMC Utility Package: UTIL42
 
[2] R.Brun, F.Bruyant, M.Maire, A.C.McPherson, P.Zanarini
    DD/EE/84-1, May 1986
 
 
 
 
