Dear all,
Dear all,
HelloI simulated a water model with 216 by TIP3P method.
1) I used “lj/charmm/coul/long” pair style and assigned 9 and 12 angestrom
as inner and outer cutoff. Does long is long range interaction out of outer
cut off? if yes, why has it sizable value? I haven’t good sense about
using long and pppm. Could you explain more?
having the switching of the LJ part between 9 and 12 /AA is a bit agressive
(i.e. what people use in benchmarks to make their software look good).
12 / 14 is more conservative.
for the long-range coulomb you have to realize that changing the cutoff does
not mean that you change the interaction. all it does is change how much is
computed in real space vs. k-space. thus the choice of it is fairly
arbitrary
and the coulomb cutoff is a tunable parameter to optimize the speed of the
calculation. you can make some experiments with changing the cutoff and
monitoring the performance of the simulation and particularly comparing how
much time is spent on Pair, Neighbor, Comm, and Kspace; there will be an
optimum value. usually, it is the the same as or close to the outer LJ
cutoff.
to have a better understanding, i highly recommend to read up on ewald
summation
in an MD text book. even though pppm is not exactly ewald, the basic idea
is the same as you can see from using the same real space part for both.
Step Temp TotEng PotEng(E-bond+E_angle+E_vdwl+ E_coul+E_long) KinEng
E_bond E_angle E_vdwl E_coul E_long Volume Press 0 300 1322.735
744.15983 578.5752 3.56E-08 1.76E-11 789.3945 10946.56 -10991.8 6434.856
55513.42 100 520.5242 -359.721 -1363.5955 1003.875 163.248 70.55287
210.5649 9175.491 -10983.5 7312.104 -1483.59 200 320.1546 -1037.99
-1655.4341 617.4451 141.2671 129.0164 164.6259 8901.306 -10991.6 7817.167
-1399.96 300 294.1952 -1288.64 -1856.0243 567.3802 117.8539 113.0059
216.9975 8690.162 -10994 7795.684 1423.027 400 277.813 -1357.63
-1893.4186 535.7857 125.098 87.31868 250.8251 8638.334 -10995 7632.06
443.229 500 320.0567 -1258.73 -1875.99 617.2562 147.0789 86.24853
269.5639 8612.096 -10991 7480.103 -756.037 600 284.7481 -1310.52
-1859.6805 549.1606 141.1022 91.3506 216.4485 8685.086 -10993.7 7410.888
-1525.95 700 303.893 -1337.88 -1923.965 586.0832 138.6307 99.62891
271.7902 8562.786 -10996.8 7156.184 463.0154 800 303.1035 -1341.23
-1925.7912 584.5606 148.4814 99.40114 256.5872 8566.853 -10997.1 7026.405
-107.868 900 298.1551 -1326.28 -1901.2947 575.0172 150.3081 101.532
243.6802 8594.859 -10991.7 6990.484 -189.749 1000 292.0394 -1341.55
-1904.77 563.2225 149.379 116.2303 276.6879 8550.855 -10997.9 6909.445
112.0679 1100 294.0773 -1387.66 -1954.8102 567.1529 151.8573 100.9565
246.8535 8542.611 -10997.1 6815.29 -253.682 1200 287.2972 -1364.17
-1918.2418 554.0768 150.2203 105.8643 259.879 8562.868 -10997.1 6663.33
1359.726 1300 314.0518 -1339.32 -1944.993 605.6753 162.4818 106.5709
241.7463 8539.087 -10994.9 6538.435 -1764.66 1400 301.3014 -1411.49
-1992.5794 581.0851 175.2392 101.2389 240.1528 8487.847 -10997.1 6463.563
-2823.47 1500 285.4777 -1448.39 -1998.9546 550.5678 139.737 99.83439
256.6026 8501.949 -10997.1 6424.515 2370.853 1600 298.3965 -1363.14
-1938.6227 575.4827 154.2457 119.1773 240.0159 8544.681 -10996.7 6459.254
471.9212 1700 303.1832 -1357.25 -1941.9639 584.7144 162.527 97.442
226.2667 8569.988 -10998.2 6502.019 444.4899 1800 286.8965 -1372.68
-1925.9821 553.3041 181.547 103.9637 264.4572 8522.103 -10998.1 6355.337
-1092.122) I know liquid water box has equable volume and equable pressure, but I
don't know which one is better to use?
a:fix 1 all nvt temp 300 300 10
b:fix 1 all npt temp 300 300 10 iso 0.98 0.98 100depends on what you need to do with the simulation data.
if you don't need npt, use nvt. it is faster and easier to manage.
3) in both of them,I got same result.
in NVT , volume is constant exactly.pressure has large
fluctuations(+_5000).
in NPT, volume has fluctuation(+-200 angestrom^3) and pressure has large
fluctuations(+_5000) like NVT !!!!!. is it logical result?yes. as has been discussed on this many, many times. the pressure
fluctuation
are a result of the general incompressability of dense systems and dependent
on the size of the system. temperature fluctuates as well and behaves
similar
with system size, only the fluctuations are less pronounced. in both cases
you
are looking at macroscopic properties that are translated to microscopic
observables
through statistical mechanical relations. again, a suitably thorough look
into a
text book, statistical mechanics in this case, would provide you with the
details
and theoretical foundation of these properties.
4)I have found an example that used fix shake instead of fix in its water
simulation. Is it necessary to use shake algorithm for simulation water?
when do we use shake?
fix shake instead of what fix? fix rigid??
and one more time, the recommendation is to read up on things in
an MD text book. a mailing list cannot replace studying the fundamentals
of the methods your are using. the usual reason to use the SHAKE
(and RATTLE) algorithm over a rigid integrator for small molecules is
that you can use a much longer time step. the integration of the
equation of motions for the rotational degrees of freedom is dependent
of the moment of intertia and for the "pitch" rotations it is pretty small
and hence requires a small time step.
Sorry,I'm beginner in LAMMPS.
please note, that the majority of your questions don't touch on LAMMPS
specifically, but rather on the basics of doing MD simulations and using
statistical mechanics as the tool to translate between microscopic
observables and macroscopic properties. spending some time on
getting a better understanding in the theoretical foundations will help
you massively to produce meaningful simulations. one of the biggest
problems with MD simulations is, that being able to complete a trajectory
without crashing doesn't say anything about its correctness. it is
a prerequisite, nothing more.
ciao,
axel.
And see section 6.7 of the manual.
Steve