Flavio

The most likely explanation is that the equilibrium pressure is not the one that you set. I know that the value you use seem small, but it may very well be too high. What is the pressure converging to? Also, make sure you understand the pressure units. The MC barostat is much more stable. Perhaps you could run a short MC simulation in the NPT ensemble and see if it goes in the same direction. Just pointing out possible causes of error...

In general, less coupled thermostats favour strand diffusion. So if your system involves more than one strand, and their diffusion speed is important (i.e., hybridization events) using less coupled thermostat speeds things up. On the other hand, more coupled thermostats produce more stable trajectories and are better suited to reach equilibrium from highly strained configurations (i.e., configuration coming from file converters or from the cadnano interface). So to your specific points: 1. The rationale...

Hi Derek, the code does reset only when going forward, it is intrinsic in the introduction of the "-1" interface that it should do so. The reset interface is there to make sure that the true initial states "i=0" are decorrelated as much as possible from one another. Your conclusion sounds reasonable, the more you pull the more the duplex frays and the longer it will take between i=0 forward crossings. The link is broken and I cannot see it unfortunately Flavio

First of all, I don't know whether it is intentional or not, but Martin is using normal traps, not mutual traps. One of them has also stiffness=10 and the other one has stiffness=1, perhaps that's intentional and perhaps not, but regardless it should be running. So, why did fix_diffusion=0 do the trick? The reason is that with fix_diffusion=1, the code periodically shifts all the particle positions so that they are in the main simulation box. This is because if the code runs for a long time not doing...

Hi Martin, try fix_diffusion=0 in your input file. If it helps, I will tell you why. Best Flavio

By "can't add more interfaces" I meant that you are allowed to, although I am not sure it really helps. Interfaces only help if the average time it takes to fail or succeed is longer than some correlation time in your system. More interfaces help if they do "contain information", i.e., two consecutive interfaces n and n+1 must be "far apart" enough so that the fate of one starting configuration at n is to some degree uncertain: some attempts will fail, some will succeed. If all attempts succeed starting...