Google Calculator

James Rogers jamesr at best.com
Mon Aug 18 15:20:47 PDT 2003


Russell Turpin wrote:
> 
[...fine ideas elided...]
 
> (3) Once a set of related units, constants,
> and formulae have been selected, the
> visitor should be able to generate a file
> in C, C++, Python or other likely language
> that contains these in language-specific
> form. In other words, if you're writing
> code for some automated quantitative
> analysis device, based on reaction rates
> as a function of pH, temperature, etc.,
> you should be able to download all your
> constants and conversion formulae from
> this website.


This would only work for toy chemistry, not "real" chemistry.

As far as the chemical process modeling goes, many (most?) of the non-trivial
non-ideal mathematical conversion functions for a given chemical reaction are
trade secrets and/or considered very valuable intellectual property.  Anybody
able to provide the data/functions would have to charge a lot of money for them
to recover cost of getting that data in the first place.

I'm pretty sure most people don't realize how profoundly UGLY chemistry is from
a mathematical standpoint.  They give you neat and pretty equations in basic
chemistry because fairly large errors are tolerable, but for industrial purposes
where a couple percent deviation in efficiency is the difference between gross
profitability and bankruptcy, everyone uses an entirely different set of
equations for these things.

The problem is that all the industrial equations are created by curve-fitting
empirical data.  The resulting equations are very ugly, but much more accurate
along a broader parameter range.  Consequently, there are many different
equations for the same set of conditions, some of which are better or have more
dimensions than others depending on the parameters of the measurement for the
data points to fit the curve to.  For many various common chemical systems there
are companies that own and license models for particular simple industrial
chemical system.  In theory anybody could generate their own models just by
setting up a high-precision lab and doing the measurements, but it is generally
cheaper to just buy reputable chemical models from another source to plug into
ASPEN (industrial chemical process modeling software).  In fact, I got my start
in the software business writing FORTRAN code on mainframes to brute force
unsolvable systems of differential equations to generate derivative chemical
systems equations.

Off-hand it would seem natural to assume that nobody owns the equations that are
used for real-world quantitative chemistry, but once you realize how difficult
and expensive it is to obtain even a reasonably good approximation and that
everybody's version of it will be somewhat different (with some better than
others), it is more obvious why many of these systems of chemical equations are
not passed around willy-nilly but are actually purchased from companies that
expend dedicated effort to discover an improved approximation.  If quantitative
chemistry was not computationally intractable even for trivial cases, this might
be different.


-James Rogers
 jamesr at best.com



More information about the FoRK mailing list