Hi, Bruce
I followed your guidance and did some modification. Also I tired different test for system correlation length form 2~200ps(I chose the range from reference
Thermal conductivity of Carbon nanotube Nanotechnology 11 (2000) 65–69.). And calculated the thermal conduction across the z axis. The timestep is 0.5fs (1fs systems are not stable at calculation)I did the analysis of the HACF graph. And I attached 200ps correlation length decay graph in the end. However, the HACF seems decay at zero really quick, and other part look like noise part.The thermal conductivity of cnt in here is around 0.88 w/mk. I have no idea why the thermal result is still lower than the expected result. One thing I didn’t change is the forcefiled. Does forcefield have really high influence on calculation if I used default Tersoff in lammps?
Step Temp CPU v_Jx v_Jy v_Jz v_k11 v_k22 v_k33 Density Press TotEng
8800000 333.21859 1391.903 -0.012969303 0.0086867249 0.0030157897 0.35525841 0.12557744 2.4888669 0.0028339864 -4.0776998 -3412.7176
9000000 339.90761 1423.2239 0.0039806426 0.0079442151 -0.0033400739 0.34771324 0.062181796 2.4302487 0.0028339864 2.279101 -3412.547
9200000 345.07623 1454.9235 -0.0098667929 0.0055496629 0.003617017 0.33910505 0.049894602 2.4914547 0.0028339864 1.3449785 -3412.4656
9400000 322.35507 1486.9732 -0.0044881889 0.011710868 -0.015938799 0.34873044 0.052399708 2.3833902 0.0028339864 -8.361111 -3412.5095
9600000 339.20825 1518.5993 0.0029662846 0.020995441 0.0066569117 0.34403628 0.064964102 2.3270703 0.0028339864 1.1840954 -3412.5268
9800000 316.75566 1549.9844 0.00381416 -0.0035722026 0.0049736712 0.34738361 0.060911928 2.3341339 0.0028339864 -17.329494 -3412.4839
10000000 318.70774 1581.1593 -0.0039801661 -0.0052334822 -0.010786312 0.34289347 0.054857603 2.2642135 0.0028339864 -2.9174357 -3412.4316
Thanks for the comments.They are really helpful.
Hello,
I have some comments on some of your input statements (the Tersoff version) below.
variable dt equal 0.0002 # ps
A time step of 0.5 fs (or even 1 fs) is ok for Tersoff carbon.
variable p equal 1000 # correlation length
variable s equal 10 # sample interval
variable d equal $p*$s # dump interval
So your correlation time is only 1010000.2fs = 2 ps. For Tersoff CNT, it needs a correlation time of about 1 ns (1000 ps) to achieve a converged thermal conductivity. Also, it is better to make $d larger because it corresponds to the production time for each correlation function. Good parameters for Tersoff carbon:
timestep=0.001 # time step = 1 fs
s=10 # sampling interval = 10 fs
p=10^5 # correlation time = 1 ns
d=10^7 # production time = 10 ns
Also you need to do at least 10 independent runs. Otherwise, you cannot expect to get statistically meaningful results. Thermal conductivity of CNT is very difficult to calculate. It takes a total production of about 1000 ns to obtain an accurate estimation of the thermal conductivity using Green-Kubo.
pair_coeff * * /home/tntech.edu/czhan42/lammps-16Mar18/potentials/SiC.tersoff C
As far as I know, the version by Lindsay and Broido is currently the best potential for thermal transport: https://doi.org/10.1103/PhysRevB.81.205441
compute flux all heat/flux myKE myPE myStress
variable Jx equal c_flux[1]/$V
variable Jy equal c_flux[2]/$V
variable Jz equal c_flux[3]/$V
*fix JJ all ave/correlate $s $p d &**
**c_flux[1] c_flux[2] c_flux[3] type auto file profile.dat ave running**
__variable scale equal {convert}/${kB}/$T/$T/$Vs*{dt}__
Here, you have put the volume V in the wrong position. It should be in the numerator instead of the denominator, because you have already divided J by V. If you have not divided J by V, it is correct to put V in the denominator.
variable k equal (v_k11+v_k22+v_k33)/3.0
You cannot use this formula for quasi-one-dimensional CNT. This is only for 3D isotropic systems.
print "average conductivity: $k[W/mK] @ $T K"
Did you report your thermal conductivity value based on this printed number? You need to analyse the heat current autocorrelation function and the corresponding running thermal conductivity carefully. This single number is not very meaningful.
Best,
Zheyong
