new application of interval computations

[Vladik Kreinovich]

An Application of Interval Computations to Gravity 
Featured as One of the Major Scientific Results

The latest issue of "Discover", the popular science magazine,
features, among other stories about the major recent scientific results, 
a story about measuring the gravitation constant G (according to 
Newton's law, the attraction force of a body with mass M at a distance R 
is GM/R^2). Due to the fact that on Earth, the gravitational interaction 
between bodies is much weaker than any other interaction, this constant is
the worst known among the fundamental physical constants. What is even worse, 
different known measurements of G seem to be inconsistent: there are 
several measurement results with accuracy estimates; each gives an 
interval of possible values of G, so, ideally, the actual value of G
must be in all of them, but ... these intervals have no common points.

Physicists and applied mathematicians from Wuppertal, Germany, 
led by Prof. Dr. H. Mayer (Physics) and Prof. Dr. B. Lang (Math), 
analyzed this situation and discovered that this seeming inconsistency 
is caused, partially, by neglecting certain physical sources of error, 
but mainly, by using approximate error estimation techniques for data 
processing algorithms, techniques that often underestimate the 
resulting error. Instead, they propose to use computations with 
automatic result verification (in particular, interval methods). 

The paper by B. Lang and co-authors will appear in No. 3 (1996) of 
"Reliable Computing"; the author's email is langamath.uni-wuppertal.de.
A brief description of this application is placed on the interval Website
http://cs.utep.edu/interval-comp/main.html

This is the second time in half a year that a result using interval 
computations is featured as one of the major scientific breakthroughs:
the previous was the result about the double bubble featured last November.