5
$\begingroup$

I am a 10+1 student, currently preparing for JEE. It requires me to prepare for Maths, Physics and Chemistry. I particularly enjoy logic related subjects and application, so I find maths and physics enjoyable, however, I am troubled by the way chemistry currently exists (or is taught to an high schooler).

From my current understanding of parts of concepts of Chemical Bonding (Ionic bonds, VSEPR Theory, Molecular Orbital Theory, ...), I understand that bonds are formed on the basis of approach of atoms, achieving lower potential energy, and thus more stability. But from the POV of solving problems, we are expected to learn rules, most of which come with a variety of exceptions. Also, my teachers have told me that all of these are theories, and match the real results to a fairly good extent.

I hope the intro wasn't too long, and here I come to my main question. If I haven't bored you till here, here is my question:-

If chemistry is mainly related to achieving lower potential energy on account of various forces, why can't we simulate all of chemistry?

To elaborate, being a computer sciences enthusiast, I wondered why we can't/don't have perfect knowledge about how things work on an atomic level. My solution/thoughts were that we could have an AI model which is trained and simulates approach of charged entities, achieving stability, then slowly formation of atoms, then molecules by making the model have atoms approach each other, in various conditions; all of this done while scripting the physics laws mainly governing these, i.e., electrostatic attraction and the others involved which I may not be aware of. The thought that made me think of this was that in VSEPR and hybridization theory, we were taught the alignment of hybridized orbitals, and had to memorize the bond angles, when I realized that those were the angles just formed when charged particles were left to revolve in a sphere and would align themselves in a symmetrical position to achieve no net forces acting on each of them. Is it that such an idea has not been thought of(which I highly doubt) or some other more logical reasons?

If you have read my question till here, I am really thankful to you, since I know I can be very annoying and hard to speak to. If you can answer my queries, I am even more grateful and look forward to interacting.

Thanks,

Ravi Arora

$\endgroup$
13
  • 14
    $\begingroup$ "Give me a computer big enough and a memory large enough and I can simulate the world." - misattributed to Archimedes... $\endgroup$
    – Jon Custer
    Sep 24 at 19:44
  • 2
    $\begingroup$ What makes chemistry great is its complexity. Chemical reactions are much more complicated than just rolling marbles on a potential energy surface. Typically, higher excited states (difficult to calculate) play a role as well and many particles interact in a complex sequence with short lived intermediate states. Even the simplest reactions are the topic of current research. Machine learning is used in chemistry as well, but this is of course pretty much a black box and does not really provide insight (although it is useful for solving some problems). $\endgroup$
    – Paul
    Sep 24 at 20:02
  • 2
    $\begingroup$ An exhaustive list of why quantum mechanics is hard to model can be made, but in short, one big reason is if we wanted to know everything about the system, we must then know the wave function for that system. This equation contains all the information of the system. Additionally, when we operate on these equations, (e.g. take the derivative or something like that) we cant always find closed form (finite) solutions, leading us to use approximation or numerical methods to find the solution to the wave equation. $\endgroup$
    – user98623
    Sep 24 at 20:19
  • 4
    $\begingroup$ As an indication of the problem, an accurate calculation is truly vast, for example in benzene with 42 electrons and a minimum of 72 orbitals there will be $\approx 10^{20}$ integrals to solve. So severe approximation has to be made, even a small fraction of this number say $10^9$ integrals is still an enormous challenge, and benzene is a small molecule. $\endgroup$
    – porphyrin
    Sep 25 at 8:06
  • 1
    $\begingroup$ I have no idea why this has been closed as a question. It has already generated two good answers so doesn't seem to be suffering from a lack of clarity in what is being asked. $\endgroup$
    – matt_black
    Sep 26 at 14:13
9
$\begingroup$

I wondered why we can't/don't have perfect knowledge about how things work on an atomic level

Well... we do. The things you have suggested (like VSEPR theory) are very crude, though, and not suitable for such tasks. Obviously, if it was a matter of just coding in Coulombic forces, somebody would have done it already.

More appropriately, everything should be simulated using quantum mechanics. There are lots of software that will do that.

The question then becomes why can't we just simulate everything using quantum mechanics. Part of the answer is that the Schrodinger equation cannot be analytically solved for anything more complicated than a one-electron system. Thus, there are some approximations involved; and the results are then not guaranteed to be accurate. Going beyond that, the answers here: Why is chemistry unpredictable? will likely explain that; and I'd argue that apart from the misconception that VSEPR is the theory that underpins all of chemistry, this is basically a duplicate question.

That said, quantum chemistry is used a lot in order to predict (and more often, rationalise) chemical reactions. So it's not impossible to do that. It's just impossible to do it in a total fashion, where literally everything in chemistry is simulated.

$\endgroup$
4
  • $\begingroup$ I am aware that VSEPR theory is not the theory which truly generalizes chemistry. In fact, our teachers share that at high school level, most of what we are learning is not 100% accurate and, at times, contradicts what someone pursuing sciences would find. And they also instruct us that we have accomplished a lot more in chemistry, but is beyond the scope of school studies. Coming to my main question, I want to ask, why can't it just be simulated using Coulombic forces and various approaches? What is the need for Quantum Mechanics to be introduced? $\endgroup$
    – Ravi Arora
    Sep 25 at 3:23
  • 5
    $\begingroup$ You may ask yourself a similar question as to why you can’t simulate the entire universe using classical mechanics. What is the need to bring in relativity and other more advanced theories. Answer is, classical mechanics just isn’t a good enough theory to explain every experimental observation. Same with electrostatics and quantum mechanics in chemistry. In fact you need relativity too in chemistry. $\endgroup$
    – orthocresol
    Sep 25 at 7:59
  • 3
    $\begingroup$ @RaviArora Chemistry is not (only) governed by classical coulombic forces. In fact, chemistry only emerged when the first atoms (helium) became neutral after the big bang. You need quantum mechanics to form chemical bonds and, as pointed out by orthocresol, this is not trivial and requires approximations and numerical approximations. One needs to introduce quantum mechanics because atoms and molecules behave quantum mechanically and not classically. This is just a consequence of many observations. $\endgroup$
    – Paul
    Sep 25 at 8:31
  • $\begingroup$ @orthocresol Paul thanks a bunch for taking the time to write these answers. I just have one last question in mind, which isn't directly related, maybe. That would be:- Will we ever be able to photograph an atom? I remember from one of my lectures that we can't even shine light on an electron (if that was practically possible) since it would jump to another energy level. So what I am trying to ask more clearly is, do we know if all this is going to be theoretical, or do we not know yet if we can achieve visual view of atoms/molecules? $\endgroup$
    – Ravi Arora
    Sep 25 at 16:14
7
$\begingroup$

We can't simulate everything because there isn't enough computing power to do it

The fundamental problem of making chemical predictions is that they are extremely hard to do because of the combinatorial complexity of what needs to be simulated.

Chemists can and do simulate, though. Big computational models can approximate molecular structures and reactions. But it takes a metric shit tonne of computer power even for simple molecules.

Start with the basic equation about the electronic structure of atoms, the Schrodinger equation. We can derive analytical solutions only for systems with a single electron (this is, by the way, also true of the equations of gravity for systems with more than two large objects). So the best possible answers are derived from simulations of the more complex systems.

The big problem is that the complexity of the equations explaining systems grows extremely rapidly with the number of parts and rapidly exceeds all the computing power on the planet.

Consider the supercomputer-level of computer power needed to create AlphaZero, the system that can play extremely good chess. Chess is a very small system: there are only 8 by 8 positions on the board and a small finite number of pieces. The list of rules is short. The system is simpler than all but the most basic molecule. But it still took enormous computational power for a computer to learn good strategies to play the game.

Molecules and their possible reactions are far more complex than that. chemistry is played on an infinite board and the rules about interactions are far more complex than the rules of how chess pieces interact.

Even at a different level of analysis, like finding all the possible proteins with just 52 amino acids in them (never mind determining their structure), is overwhelmingly complex. There are more possible combinations of 20 amino acids than there are atoms in our galaxy.

So, while simulation can sometimes but used in chemistry in limited circumstances, there is not enough computer power in the known universe to turn chemistry into a subject where we simulate everything.

$\endgroup$
1
  • $\begingroup$ That explains it really well, thanks a lot. Well, I wish there's a way to mark all answers as right since all of them were very explanatory. $\endgroup$
    – Ravi Arora
    Sep 25 at 16:08
6
$\begingroup$

What you have in mind sounds like Molecular Mechanics. Atoms are treated there like classical particles and are simulated using classical physics, including Coulombic forces. But we know that this is not sufficient. Effects like protons tunneling can not be explained within that model, just to name one.

Including all known laws of physics into simulations of chemistry would lead to simulations that would require thousands of years( and that is probably a massive underestimation of the time required). That is obviously impractical. So we have to find approximations that are on one hand accurate enough to model the current problem of interest and on the other hand fast enough to be done on a human time scale(days, weeks, months, years at most). Methods that are accurate and fast are tailored to solve particular problems and up to this point, there is no approximate method that is well suited to describe all of chemistry in a fast and accurate manner. Indeed, picking the right method for the chemical problem is by itself a non trivial task and current researchers often have to decide which method to pick for the problem that they are investigating. This is a point where researchers from the theoretical field and experimental field collaborate in many cases.

The development of new methods is also an ongoing process, so its understandable that they aren't taught at school. Why teach a method that might be outdated in a few years ? And learning/teaching all the physics and math required to understand the methods is also beyond school. School only teaches you some basic concepts that allow you to rationalize about chemistry on a shallow level.

VSEPR is an approximate model that allows us to predict equilibrium geometries, and like all approximate methods, there are cases where it works well and cases where it fails. The strongest point of it is its simplicity, we don't even need a computer to do it. Most approximate methods are themselves non trivial and require a lot of computational power. The approximations taught at school are typically methods that are doable by hand, which is probably also more satisfying when you are new to chemistry. Imagine you were taught methods that you could only apply with a large computer cluster, which is most likely not available to the common school student. I guess that most people wouldn't find that helpful at all and would only be frustrated with chemistry.

$\endgroup$
1
  • $\begingroup$ I understand your point of the need of a large computer cluster to solve a small problem, but I hope we can reach a stage where this could be done on a more compact system in much lesser time. With the advent of quantum computing, I hope we can get to a point where everything can be simulated. From my current understanding and perspective, that would be very helpful in the fields of medicine, to study adverse reactions of chemicals, and a lot of other fields. I hope we reach such a point someday soon:) $\endgroup$
    – Ravi Arora
    Sep 25 at 16:05

Not the answer you're looking for? Browse other questions tagged or ask your own question.