fix npt or nvt

Dear all:
  
  I have some problems about 'fix nvt' and 'fix npt', though discussion about these two fixes have been held for so many times, i still cann't completely know how to exactly use these two commands. Here are my problems.

  Firstly, i construct a bicrystal with boundary condition(bc) 'psp', then heat it to a certain temperature after minimization. What i want is to heat the bicrystal while keeping atoms fluctuation as small as possible, is it right to say like this because atoms fluctuation is unavoidedable during heating ? Here bc in
y-direction is 's', which will undoubtedly led to increase of y-dimension, no matter whether i use 'fix nvt' or 'fix npt'. According to my understanding, compared with constant dimension caused by periodic boundary, increased dimenion caused by non-periodic boundary will led to larger atoms fluctuation during heating a system. So i want to find a way to reduce increase of dimension during heating.

  According to my simulation results, i found that: If 'fix nvt' is used, then x-dimension and z-dimension will keep constant while y-dimension increasing; if 'fix npt' is used, then dimensions in x,y and z directions all will increase, but increase of x-dimension and z-dimension are very small, and increase of y-dimension is much smaller than increase of y-dimension when 'fix nvt' used.Here are part results of 'fix npt' and 'fix nvt' for pure Ni.

velocity all create 1400 2222
fix 1 all npt temp 1400.0 1400.0 0.2 x 0.0 0.0 1.0 z 0.0 0.0 1.0
Step Temp PotEng Lx Ly Lz Pxx Pyy Pzz
       0 1400 -75527.242 38.518729 139.40627 34.457144 17885.22 19476.379 17810.531
     200 657.3217 -73776.736 39.022955 140.26357 34.899363 -1089.8773 13196.074 -3105.5526
     400 759.79105 -73727.506 39.026565 140.57181 34.845843 1495.6443 13609.49 1848.6423
     600 830.41051 -73561.059 39.074984 140.9155 34.890088 -315.01296 11666.599 505.73567
     800 903.18819 -73362.67 39.109281 140.90193 34.947492 -439.21822 9807.0176 -109.60324
    1000 984.64085 -73155.919 39.136166 141.44609 34.972028 433.85512 9527.1239 527.14558
    1200 1056.8826 -72910.343 39.191075 142.31878 35.017632 259.04562 7221.4809 104.61742
    1400 1141.4766 -72684.753 39.219943 143.08067 35.050564 -320.17207 5752.8217 947.99463
    1600 1224.2348 -72461.34 39.248373 143.09356 35.094673 -597.33511 3755.0626 635.18484
    1800 1292.1665 -72228.019 39.267683 144.79429 35.157866 -999.67771 -103.91794 -307.10116
    2000 1339.3723 -71992.451 39.249359 144.99963 35.197569 522.19771 -1682.9446 783.23654
    2200 1389.5794 -71821.185 39.258769 145.64774 35.205314 -1947.7082 -6223.2918 -715.47662
    2400 1430.6614 -71719.774 39.19593 145.92665 35.193619 -336.62678 -9533.278 284.40277
      .... .... .... .... .... .... .... .... ....
   18600 1404.2031 -71746.319 39.478877 144.45835 35.277525 380.48091 4342.4721 1263.3162
   18800 1395.0332 -71739.323 39.469375 144.95422 35.300668 -1376.2202 1231.34 -1173.2426
   19000 1406.3462 -71775.302 39.386187 144.71338 35.276763 -544.56491 3518.1711 1295.0136
   19200 1391.0004 -71756.359 39.363782 145.15262 35.278777 -121.79337 1119.9905 -58.165145
   19400 1394.3011 -71750.162 39.375895 145.64053 35.230991 -286.96558 -226.15536 307.17412
   19600 1406.1116 -71749.029 39.351365 145.72665 35.215892 -221.80943 -774.53321 906.92484
   19800 1395.1934 -71704.295 39.319323 145.96507 35.251567 -25.411606 -2082.7949 -889.87345
   20000 1402.3496 -71702.223 39.288336 145.97492 35.221666 1458.3162 -2521.4996 373.40728

What i want is to heat the bicrystal while keeping atoms fluctuation as small as possible, is it right to say like this because atoms >fluctuation is unavoidedable during heating ?

no, this isn't right. When you heat the sample, the atoms move
where they want to go.

Steve

Dear Steve:
   Thanks for your response.

   I do konw atoms will move during heating, but i think there still exists difference in atoms fluctuation when heating a same sample to a same temperature by using 'fix nvt' and 'fix npt', respectively. So that is why i want to know is 'fix nvt' or 'fix npt' doing better in heating an non-periodic sample while keeping atoms fluctuation as small as possible. Steve, according to your understanding of MD and Lammps, is what i thought possible to realize?

L.Yang

hi,

why don’t you just try which of the fixes suits your needs better? you can easily monitor the displacements of your particles and compute the fluctuations.
however, it may also be a good idea too look at other parameters to finally decide what you actually want to use.

best,

nikita