meaningless real part spectral heat current while imaginary part has some meaningful peaks. #1
Description
I’m reproducing the spectral heat current calculation from PHYSICAL REVIEW E 93,052141 (2016). According to Eq. (5) and Eq. (A1) in the paper, the final result should be the real part of the computed spectrum.
I implemented a Fortran program to compute spectral heat flux using:
Method 1 – Direct FFT of the atomic velocity and force time series, followed by spectral projection via conjugate dot product.
Method 2 – Compute the time-domain correlation function ⟨v·f⟩ first, then apply FFT.
The inputs are atomic velocities and forces. The output files contain both real and imaginary parts of the spectral heat flux.
I observed that the imaginary part consistently shows clear peaks, while the real part is several orders smaller and noisy, which contradicts the claim in the referenced paper that the real part is physically meaningful.
I post my code later, please rename the .f90.txt file to .f90 and compile like this：
ifort -o transmission_cal_compare.out calculate_transmission_compare.f90
-I/public1/home/jingchao/fftw/1/include -L/public1/home/jingchao/fftw/1/lib -lfftw3
(Note: Change the FFTW path if your installation differs.)
Also, I would like to mention that in my approach, I did not compute the force constant matrix. Instead, I extracted the atomic forces directly from LAMMPS simulations and used those force time series as input for the spectral analysis (via FFT or correlation function), as shown in the attached code.
