Dear Dr. Axel,
I’m using the svsl method to study the phonon dos of a multi-component alloy.
After obtaining the slspring data, I used the svsl code to get the DOS of one structure.
The DOS is expected to reach 0 density around the frequency of 1e13. However, I got a long tail of noise up to 6e13. The densities from 1e13 to 6e13 are very low, and should be considered as numerical error, but not valid data. Is there a way to set a cutoff frequency? or, maybe it indicates that there is some deeper error of the calculation that I have to re-exam the whole calculation?
Thank you in advance.
Yi Wang
Nanjing University of Sci. & Tech.
The problem has been solved.
It is proven I have been wise in my own conceit.
During the fitting for the slspring parameters, I chose different orders of polynomial for different pairs. Some pairs need a order of 4 th polynomial, while 1 st order polynomial is enough for other pairs. So I got a slspring.out file with mixed orders after editing by hand.
However, the editing did not work exactly as I thought. The integer number in slspring.out for each pair indicates the order of polynomial (plus 1). My original editing gives different integers (for the order of polynomial) in slspring.out. Somehow, this kind of editing caused the long tail of noise.
The solution is that re-editing the slspring parameters, unifying the integers of orders to the highest one, and patching zeros for the parameter set. Then the noise is gone.
Sorry to the thread readers for your precious time.