 So now we have both a potential function and a mechanism to update coordinates. That's all we need This is a small example system of water just to give you a feeling of the time scales I think this entire movie covers a few picoseconds or so That means that the individual time steps to also to integrate the fastest motions correct here They're gonna need to be in the ballpark of femto seconds In fact, if I have individual bonds that vibrate here I need to stick to one femtosecond time steps 10 to the minus 15 I would need to do 10 to the power of 15 steps to reach a second that would never happen We would like to extend that and I'll get back to how in a second But in principle we can set up a flimellation flow chart now You're gonna need some sort of state to start from if it's water I can throw out water molecules randomly or form an ice crystal or if it's a protein I can get one of the protein data bank or maybe you can create a small helix structure yourself based on how a helix should look like Then we're gonna need to calculate the forces and that's goes from this potential definition to tell the truth We never calculate the potential we calculate the forces directly because we typically don't need the potential I might calculate the potential every 100 steps to write out what the energy is But I don't need V to get to the next time step just V just F The second part I'm gonna need I'm gonna need to apply Newton's equations of motion to update my positions And once I've done that I probably want to calculate some properties such as energy and temperature I might want to save the coordinates to a trajectory so that I can look at them That's the last round box there and then it's time to go back and keep doing this millions of times and that's the problem because On the one hand, we have some very costly interactions here as I told you but be we're gonna need to Reproduce this every sorry repeat this every one femtosecond Now there are a couple of tricks we can do to take this faster if I want to do this faster Remember the limitation the leap frog integrator needed roughly five points per period So if I want to take longer time steps, I should remove the fastest motion in the system And one of those fast motions was an individual bond vibrating And that's why as I mentioned already lecture one It turns out to be very convenient to replace those springs that are anyway crappy representation of an harmonic bond and just have these Bonds be constants. You might not be able to see it here, but trust me. They are not changing in length In this particular water model even the angle here is fixed That also turns out to be a perfectly adequate representation because I'm not I'm interested in water here As a solvent to my proteins not having a perfect water model so With a few of those simplifications I might be able to get to two or maybe four femtoseconds if I'm lucky But somewhere there I'm starting having torsions and other things involved that I simply can't remove anymore So no matter how you slice this We're gonna end up doing billions of time steps to get to interesting time scales So you're gonna need a fast computer here