Sciencemadness Discussion Board

An extremely ambitious pipe dream (Reality simulation)

Swinfi2 - 22-10-2018 at 19:51

I have to start off with a disclaimer, I'm not even sure how to start this project. I'm not qualified for a start, I have a 3° BSc in environmental chemistry and I'm at the height of the Dunning–Kruger effect with my skills in C#, I've made 2-3 non-functional games with some working features and that's about it. But that's exactly what I want to do! (except without the failing part)

What kind of video-game? A simulation of reality of course, but to what level of detail? Enough level of detail that people set up a forum to look up cheat codes for it and call it Science-Madness.

Assuming that it isn't physically impossible because of the Bekenstein limit, I need to simulate almost every aspect of chemistry to the point you can collect periodic table elements and most organic chem with minimal errors.

But this is to be a game not a boring simulation, you have a fixed amount of time before natural disaster strikes. Start Now or Prepare Carefully?

So where do I start simulating;? The atoms? The surfaces that bind each substance? The potential energy of each substance? The way things crystallise and phase changes? Why some explosives are super sensitive and others can be hit with planes without exploding? (I'm looking at you Fox-7) How electricity/EM/magnetism interacts with matter? (fml) Anywhere else?

And I'm also imagining the utility, I remember the old Crocodile Chem they used to use at my high school and how it failed to extrapolate past basic reactions, but it was still useful, imagine that but good? (and first person?)(and zombies are chasing you)(and the last thing you have is a Molotov cocktail)(that you made yourself in-game with perchlorate)

phlogiston - 23-10-2018 at 01:11

Consider this analogy:
if you want to simulate a waterfall, you can do so at many different levels.

At the very lowest level, you could simulate individual quarks, electrons, and bosons, and gravity. Theoretically, that should work. However, no known computer is able to simulate an entire macroscopic waterfall at that level.

Alternatively, you could simulate molecules of water. You'd only need to consider the much simpler forces between the molecules. You can go full quantum, but even an approximation will do. Current computers may simulate a very small droplet of water this way. Still, no known computer can accomplish this for an entire watefall.

You could also simulate water, based on its macroscopic properties of water. Surface tension, viscosity, density, compresability, etc. This will simplify the calcualtions tremendously, to the point where maybe even a powerful desktop computer will be able to simulate a waterfall.

You could also design a particle system and play with it enough make it look convincingly like water, even if the underlying 'physics' have no realation with real water.

In my opinion the last option is best for games.

Going back to your chemistry game: if you are going for a game, just put enough Chemistry in it to be educational and fun. You could allow for a lot of possible reactions and Chemistry in a high-level format (A+B --> C).
It would hopeless to try to achieve a simulation of reactions, recrystalizations, distillation, etc, etc at a quantum-chemistry level. If you'd achieve that, it would not be a mere game but the best tool in Chemistry ever.

Swinfi2 - 23-10-2018 at 03:13

I'm not saying I personally have the ability to predict if any reaction is going to occur(given the specificity of all possible conditions) but as chemists we all get a sense for it, we know when we mix an oxidising agent and a reducing agent that there is potential energy there *(if you know what you're doing).

The problem faced by the computer is exactly as you put it, if you simulate the quarks, it's too fine grained to see the whole picture, same with atoms. But we take our knowledge (and research) of these systems and apply it with "enough detail" to predict the outcome, we don't always need to calculate the stoichiometry. Then on the surface of it, the game has to model the bulk properties, then decide methodically (or Monte-Carlo, seems like the done thing lol) if any higher level of detail is required. The game needs a 'built in chemist' that says "ignore those rocks they won't change at this time scale" and "yep, that just exploded in your face".

My first thoughts were how is Q-Chem done today with stuff like gromacs in tiny folded up box universes, but that uses lots of computation for the accurate calculations. How many samples of a wave equation do you really need to hold molecules together? or would it be better to integrate the sum of available states? and then I see research papers using this stuff and atoms ghosting each other :mad:

My ideal goal here is as you put it "the best tool in chemistry ever." After that the "game part" is simple, darken the sky, irradiate/contaminate the surface, add zombies/looters.

But at the same time, at least in the beginning, I don't mind if my approximations are wrong. They just have to go in the right direction and get better over time as computing power goes up and the approximation gets less approximate. I'm pretty sure an AI could do it, but it would put us all out of a job *facepalm*

JJay - 23-10-2018 at 18:06

IBM has an AI that predicts the products of chemical reactions: https://rxn.res.ibm.com/

They're really only scratching the surface of what's possible, but it's only a matter of time before a pharmaceutical company puts together an AI that can write working syntheses of novel compounds. I wouldn't be surprised if that's happened already.

MichaelBijanAfghani - 24-11-2021 at 17:59

I think the closest thing to what you're talking about is ab initio molecular dynamics, or quantum chemistry. It simulates atoms based on how their electron clouds interact with each other, allowing simulation of chemical reactions. I wish I could do that, but I don't have a good enough understanding of molecular dynamics, let alone taken a computational chemistry or pchem 2 course.

Michael Bijan Afghani

[Edited on 25-11-2021 by MichaelBijanAfghani]