Do negative USTEP's affect my results?

FAQ AUTHOR: Alex F Bielajew, PhD (Stanford)
Professor and Ungergraduate Program Advisor
Department of Nuclear Engineering and Radiological Sciences
The University of Michigan
2355 Bonisteel Boulevard
Ann Arbor, Michigan 48109-2104
USA

The quick answer is:

If you get a few negative USTEP warnings and you are using an NRCC-supplied code like DOSRZ, you can safely ignore them. This exception gets correctly handled. The warning is to let you know that this occurred with a magnitude greater than 1e-4, a completely arbitrary limit.

If you get a lot of negative USTEP warnings and you are using an NRCC-supplied code, you probably have an error in the geometry input, plane boundaries out of order, that kind of thing. NRCC codes will abort after 1000 of these warnings. It can also happen if you are operating in non-PRESTA mode in geometries with characteristic sizes of the order of 100's of cm's or greater. There is no easy fix-up for this yet but I'd be happy to offer advice on this topic if you really need it. (Please contact me privately.)

If you get a few negative USTEP's with your own code you are *probably* OK if your geometry coding complies with EGS4's expectations, that is, if you supply a new region number even if the distance to the surface is negative. If you get a lot of negative USTEP warnings with your own code, you probably have a serious problem with your geometry coding. You can consult:

author = {W. R. Nelson and T. M. Jenkins},
title = {{ PWriting subroutine HOWFAR for EGS4 }},
journal = {Stanford Linear Accelerator Center Report SLAC-TN-87-4
(Stanford Calif)},
year = {1987},
keywords = {HOWFAR Manual},
or the report

author = {A. F. Bielajew},
title = {{ HOWFAR and HOWNEAR: Geometry Modeling for Monte Carlo
Particle Transport}},
journal = {National Research Council of Canada Report PIRS-0341},
year = {1995},
notes = {Describes geometry modeling for EGS, howfar and hownear
solutions for arbitrary quadric surfaces},
location ={http://www-personal.engin.umich.edu/~bielajew/NewStuff/pubs/reports/}


More details (Geeks only past this point!):

The negative USTEP arises because EGS4 and EGS4/PRESTA apply a multiple-scattering angular deflection at the end of the step. (ITS and MCNP do it this way as well, PENELOPE does not. EGS4/PRESTAII when it finally gets released [Please contact Dave Rogers dave@irs.phy.nrc.ca for more info] will have solved this problem as well.) Applying the multiple-scattering angular deflection at the end of the step can cause an electron or positron to scatter away from a surface that it is assumed to have crossed, leading to negative USTEP's.

There is a more complete discussion of this in the HOWFAR/HOWNEAR report mentioned above.

Applying the multiple-scattering angular deflection at the end of the step is also falling out of favor for a number of reasons.

1) This technique leads to a fairly strong step-size dependence in many applications. See:

author = {E. W. Larsen},
title = {{ A theoretical derivation of the condensed history algorithm}},
journal = {Ann. Nucl. Energy},
year = {1992},
volume = 19,
pages = {701 -- 714},
author = {I. Kawrakow and A. F. Bielajew},
title = {{ On the condensed history technique for electron transport}},
journal = {Nuclear Instruments and Methods},
year = {1998},
volume = {B142},
pages = {253 -- 280},
notes = {longitudinal/lateral distribution theory}

2) It can produce fluence singularities/anomalies that can be significant for certain applications. See:

author = {B. J. Foote and V. G. Smyth},
title = {{ The modeling of electron multiple-scattering in EGS4/PRESTA
and its effect on ionization-chamber response}},
journal = {Nucl. Inst. and Meth.},
year = {1995},
volume = {B100},
pages = {22 -- 30},
notes = {Detailed description of problems in EGS4/PRESTA.},

3) The negative USTEP problem

These difficulties are either completely or almost completely overcome by using a better electron transport algorithm such a PENELOPE, EGS4/PRESTAII or one of several others that are under development.


return to main EGS web page    return to main EGS FAQ
Webmasters

last updated 10/04/01