Sciencemadness Discussion Board

Most reactive combination?

zoomer - 22-5-2005 at 19:25

I’m interested in modeling a few reactions in software as a personal project, and I want to start by defining the minimum reaction unit of space. This unit will be determined by the largest distance that any two atoms, in any ionic state, in any phase, are going to be able to cover to react with each other. Another way of defining that is to find the “most reactive” combination (having the greatest affinity) of elements and determining how far apart they can be and still be guaranteed to react. What would those two elements be? I’m guessing a halogen + alkali metal? Based on eV potentials, maybe cesium & chlorine? I am just guessing, so if anyone has better info, or a better idea, I would greatly appreciate your enlightening me. Thanks!

Z

Marvin - 22-5-2005 at 23:00

More information please, how are you planning to model them?

Just because a reaction yeilds a large amount of energy does not mean it will occur at the longest distance. How about a system based on force? If ions form, your distance based estimations go out the window.

darkflame89 - 23-5-2005 at 00:07

OOoo, I would like to modify your suggested reactants a bit: use fluorine instead of chlorine? Cesium reacting with fluorine would be nasty...
Or another suggestion would be to use a powerful oxidiser and a powerful reducer in a reaction. Such as using perchloric acid...

zoomer - 23-5-2005 at 08:30

I’ve become interested in difficult-to-achieve reactions, and would like to know what’s going on at the atomic level, and in slow motion. My intent is to build a “virtual beaker” in SW where materials can be “added,” and their activity modeled in very small time slices, maybe as small as 1 nsec. Using an object-oriented language (probably Java), I can model each material, including and accounting for every known characteristic, even relativistic ones. An “environment engine” would model the spatial relationships, accounting for movements as modulated by the characteristics of the medium (water, air, vacuum, etc plus heat, gravity, etc). As materials move about in the medium, they may eventually encounter others, and their interaction would then be calculated for each time slice.

To finish this amount of calculation before the end of the millennium, I need to work on a very small scale. I also want to start very, very simply, so one requirement is define the optimal calc space (a “cell”), being a sphere exactly the size of the largest distance possible. Any larger means exponentially more calculations; any smaller might mean missing some possible reactions at the furthest extreme. However, when I get my next supercomputer :) I’ll start adding cells together for additional possibilities.

I’m also limiting my initial efforts to a small handful of elemental atoms and simple compounds such as H2O, HCl, etc. So while more complex compounds would be more interesting, right now they are way beyond the model’s capability.

I’m sure every college and major chem lab already has this SW, but I don’t have the $100K or so to buy it. I suppose my first question should have been, “Does anyone know of freeware source code that does this?” But then, this is the DIY zone, right?

I had not considered using force to define the cell size, can you give an example of how that might be used, and what values might result?

Thanks!

Z

BromicAcid - 23-5-2005 at 10:26

I've seen a movie of cesium reacting with fluorine, it was kind of ill shot though. It would be better to see lithium react with fluorine though, being that both are very small molecules and completely at opposite ends in terms of electronegativity.

Polverone - 23-5-2005 at 11:20

There is a huge body of literature and code already in existence for computational chemistry, and very little of it is as simple as the program you've thought to build yourself. 1 ns is actually a long time in computational chemistry; the standard length for a molecular dynamics simulation of solvated DNA that I run is 2 ns, and that takes roughly 80 hours of real time to calculate using 16 processors on a MPP computer.

Several well-known computational chemistry suites are free for academic use but completely non-graphical. You will have to learn their special command languages and use separate analysis/visualization programs to make sense of large outputs. If you already know how to program this shouldn't bother you too much, but if you want a point and click interface, almost all of your options are commercial and expensive. Some packages are free to academics but you will have a hard time getting ahold of them if you don't have a university to go through.

Here's a <A HREF="http://www.netsci.org/Resources/Software/">listing</A> of software from NetSci.

<A HREF="http://zeus.polsl.gliwice.pl/~nikodem/linux4chemistry.html">Linux4Chemistry</A> has a large listing of chemistry software that runs under Linux, and tells you right away whether the product is free for academic use, shareware, or commercial. Most of the software shown there runs under other Unix systems too, and possibly even Windows.

[Edited on 5-24-2005 by Polverone]

12AX7 - 23-5-2005 at 12:07

Wouldn't you have to model the molecules, ions and atoms as points in space with a distinct charge (or combination of charge, as polar molecules) and consider each one in relation to the next by inverse square law? Rather inconvienient of course, so you could count up particles within a certain number of radii (those outside of which have little effect anyway and can be ignored).

It seems to me your finite elements are already defined as quantized atoms, rather than a random density in a volume of space?

Tim

zoomer - 23-5-2005 at 18:31

Polverone, thanks for the URL! I hadn’t thought to look in the linux index. At first glance I think there are some things there I can use, at least to get started. I’ll check it in depth this weekend

I’m not looking for GUI input or graphical output, I’m OK with a command line in and tables out. I’m a SW engineer, so programming scripts and such should be fairly straightforward.

12AX7, I hadn’t gone very far, but I was leaning away from a point-space relationship (kinda like the grid in the Battleship game) to a probability-based environment, which seems to be a common approach. And if I understand your question, yes, I am thinking of each entity as an individually represented quantum in the model.

Z

Marvin - 23-5-2005 at 21:55

I think getting something down that 'proves the concept' should not be that difficult, but I'm sure it will find no general use.

I would suggest basing it on a precompiled MOPAC binary. I'm unsure if you can get it to spit out forces directly, when I had to do something similar we ended up iterating to a minimum enthalpy of formation.

Without clever and rather unsubtle methods you will be restricted to a very small subset of reactions. You also have to decide what to do with energy, if you conserve many reactions will simply not happen eg cesium and fluorine atoms hitting eachother will just bounce off, whereas if you remove all the energy any two atoms will form a 'molecule'. You could use room temperature as a guide but this brings other problems.

Madandcrazy - 24-5-2005 at 06:18

Nice URL Polverone.

A couple of computational chemistry tools
designing own models and diagrams with molecules, atoms and ions.

unionised - 24-5-2005 at 08:54

"This unit will be determined by the largest distance that any two atoms, in any ionic state, in any phase, are going to be able to cover to react with each other. "
Over what timescale?
Given time, a hydroxide ion will attract a proton from as far as you like unless something gets in the way.

zoomer - 24-5-2005 at 17:59

Marvin, I’m not looking to make something that’s going to solve every equation, or for complex molecules. I’m thinking more of a framework such that when I come across an interesting but relatively uncomplicated reaction, I can add some new entries in a database, turn them on and see what happens, and maybe validate what’s actually in the beaker. Useful? Probably not, but making H2SO4 from CuSO4 may not be all that useful, either, but look at how many threads there are on the subject. :D Thanks for the mention of MOPAC, that could be a very helpful lib.

unionized, no time span, and with the atoms “at rest.” True, eventually any two atoms/molecules at any initialization will eventually interact, but only because they are moving and will eventually cross a threshold of attraction. It’s kinda like when you were a kid and had two magnets on the table and you very slowly moved one until the other snapped onto it. I’m trying to find what that threshold distance is. I know that two atoms moving a very high rate in different directions will have a different threshold, but I’m trying to keep things simple at first.

Z

unionised - 29-5-2005 at 02:39

Even if they start at rest, because they attract each other, they will move towards one another. There is no "threshold" for the Coulomb law.
The 2 magnets on a table only has a threshold because of static friction. If you could do the experiment in zero gravity that threshold wouldn't be there.

There is probably some effective limit where the energy of interaction is much less than the thermal energy, but that's a bit wishy-washy.

zoomer - 29-5-2005 at 17:02

Understood, the attractive forces diminish with the square of the distance but never go to zero. However, even in the hypothetical case of perfectly at rest and zero friction, wouldn't the attractive force(s) need to be a certain level to overcome the inertia of the particles? My physics knowledge is as rusty as my chem so I could be way off on this.

12AX7 - 29-5-2005 at 17:06

No, inertia is overcome by time.

Over great distances (think exponential decay) however, the forces might as well be zero, something which could optimize your program.

On the other hand, you really have to take distance as propability over time of transmission of a momentum photon. But I know very very little of Feynman diagrams and such...

Tim

budullewraagh - 29-5-2005 at 17:33

in terms of potential difference, the greatest would be lithium with fluorine. cesium with fluorine would be more violent, although hardly more so visibly speaking. fluorine with any alkali metal will react so vigorously that it is impossible to tell the difference without frame by frame analysis

zoomer - 29-5-2005 at 18:17

I now have two votes for lithium + fluorine, so I'm going to go with that as my baseline. And considering the helpful comments from other people on distance v. force, I'm going to modify my original requirement from a distance "guaranteed" to react, to "98% probability [ie. the last SD] in a vacuum." Any thoughts on that? Any guesses as to how close fluorine can get to lithium before it reacts (with a 98% probability)?

12AX7 - 29-5-2005 at 20:05

Hmm, no idea...

These ions or diatomic (gaseous) molecules? That will surely complicate issues... any idea how to model orbitals?

I would consider it reacted when they reach the ionic radius and form a salt. Until then, they release electrical potential energy, accelerating towards each other. Which begs the question of where the energy goes when they connect, thermal energy obviously but particles of this size don't simply "glow" all of a sudden. And what happens if the reaction takes place in aqueous solution where the ions react, yet are technically dispersed in the solution, more or less seperated?

Yeesh... God sure made a good particle physics "simulator", didn't he? :P

Tim

[Edited on 30-5-2005 by 12AX7]

Polverone - 29-5-2005 at 20:34

I really suggest that you find and at least skim some books on computational chemistry or maybe plain physical chemistry first unless you cannot learn any way but from painful experience. I think you can find books on axehandle's FTP site and perhaps in the "books" threads here (served from rapidshare) if you cannot get to a library.

I know that you say you're just starting out now, but can you give an example of the sort of calculations you'd actually like to do with your finished program? You may be building a framework that leads nowhere if you're not careful. Leading computational chemistry systems have dozens if not hundreds of highly skilled man-years of work invested in them, and they're often still frustratingly limited and opaque to the non-specialist chemist. Some programs still in use have been in development for more than 30 years. Since many if not most computational chemistry systems are available as source code, you could save a vast amount of time and effort by building on the work of others instead of coding from scratch.

unionised - 30-5-2005 at 02:42

The "size" you are after is called a reaction cross section. Unfortunately, almost all the references on the web seem to be to nuclear reactions.
I think you need a good book on kinetics as well as one on computational chemistry.

zoomer - 2-6-2005 at 19:57

Polverone,

Your advice is well considered, thank you. The answer to your question about learning styles is mostly “painful experience.” :)

Granted, I may have jumped in a little fast, but that tends to be most satisfying for me. In a former life I was a SW engineer, for many years specializing in modeling complex systems, especially for the insurance industry. I’m sure I could pick up an existing program somewhere and have it do all the things I wanted, and more. But for me, there is absolutely no joy in that. What I’m doing now is similar to the experimental things you tried in your cyanide synthesis thread, only my medium of choice in this case is software.

Also, from my systems experience, plus endless hours of reviewing code during the Y2K hoax, I have little faith in code that is decades old. As you correctly point out, many work-years of experience & knowledge are imbedded in such programs, but so are years of mistakes, guesses, assumptions, obfuscations and even outright sabotage (surprisingly much more often than you would think, especially in an academic environment.) And, I have reached an age where I don’t have to work in FORTRAN if I don’t want to. :D

I find it much more enjoyable to discover things for myself, even if I’m the 4 billionth person to do so. I have downloaded some of the material you mentioned, and even went to the library. But I’m still going to try to find a lot of things on my own. I have no illusions that I will discover something never found before. By someone else, that is. The simple questions I asked were just to get me started; if you are concerned that this thread goes on endlessly, I don’t plan to do that. unionized has given me the search term I was looking for (Thanks unionized!) and now I’m off on a bold new adventure, so to speak.

Promise, I’ll let you know if I find anything interesting. :)

All is well then!

Polverone - 2-6-2005 at 21:23

If you don't like FORTRAN, you'll probably not enjoy working on any of the stately, ancient codebases of the long-developed computational chemistry programs. I didn't really know what your background/interests were with this program. If you go in with your eyes wide open to the fact that you won't easily catch up with the state of the art, that's fine. I was just hoping you weren't attempting to bite off more than you can chew.

It may still serve you well to learn an existing package or two, just so you can have something to compare results with.

Francium under water ? ?

franklyn - 12-3-2007 at 16:26

with maybe a little Plutonium in the center :D

http://video.google.com/videoplay?docid=-2134266654801392897...

.