the Negative Frequency in Anharmonic Phonons

[Zhennan-Lin]

Dear Tadano Sensei,
Recently, I used the SCPH method to calculate phonon spectra with anharmonic effects. Below the phase transition temperature, the entire phonon spectrum shows no imaginary frequencies. However, above the critical temperature, imaginary frequencies appear. I tried randomly sampling structures from MD simulations at different temperatures, but the issue persists. How can I eliminate these high-temperature imaginary frequencies?
calculated details:
3 MODE = optimize
4 NAT = 96
5 NKD = 2 ; KD =X X
6 NMAXSAVE = 3
7 /
8
9 &interaction
10 NORDER = 3
11 NBODY = 2 3 3
12 /
2 &cutoff
23
24 - None None 12.0 12.0 12.0
25 /
26 &optimize
27 LMODEL = elastic-net
28 NDATA = 80
29 FC2XML = R_2nd.xml
30 DFSET= DFSET_AIMD_random
31 L1_RATIO = 1.0 # LASSO
32 L1_ALPHA = 2.51189e-06 # Add this line.
33 CV_MINALPHA = 1.0e-8
34 CV_MAXALPHA = 0.01
35 CV_NALPHA = 30
36
37 STANDARDIZE = 1
38 CONV_TOL = 1.0e-9 # was CONV_TOL = 1.0e-8 in CV
39
40
41
42 CV = 0
43
44 /
45 STANDARDIZE = 1
46 CONV_TOL = 1.0e-9
47 /
48

[Zhennan-Lin]

Dear Prof. Tadano,

I am pleased to inform you that the issue of imaginary frequencies in my SCPH calculations has been resolved. After adjusting the temperature parameters to TMAX = 1400, and DT = 100, all negative frequencies disappeared. Could you kindly help me understand the underlying reason for this improvement?

Thank you for your guidance.

Best regards,
Zhennan-Lin

[Zhennan-Lin]

Dear Prof. Tadano,

I am writing to seek your guidance regarding convergence difficulties in my SCPH calculations. Despite implementing your previous suggestions—specifically, reducing MIXALPHA to 0.05 and setting DT=1—the system fails to converge, as shown in the output below:
Temp = 7.800000e+02 : convergence achieved in 139 iterations.
Temp = 7.700000e+02 : convergence achieved in 136 iterations.
Temp = 7.600000e+02 : convergence achieved in 133 iterations.
Temp = 7.500000e+02 : convergence achieved in 129 iterations.
Temp = 7.400000e+02 : convergence achieved in 125 iterations.
Temp = 7.300000e+02 : convergence achieved in 120 iterations.
Temp = 7.200000e+02 : convergence achieved in 112 iterations.
Temp = 7.100000e+02 : convergence achieved in 98 iterations.
Temp = 7.000000e+02 : convergence achieved in 54 iterations.
Temp = 6.900000e+02 : convergence achieved in 101 iterations.
Temp = 6.800000e+02 : convergence achieved in 121 iterations.
Temp = 6.700000e+02 : convergence achieved in 135 iterations.
Temp = 6.600000e+02 : convergence achieved in 146 iterations.
Temp = 6.500000e+02 : convergence achieved in 417 iterations.
Temp = 6.400000e+02 : not converged.
Temp = 6.300000e+02 : not converged.
Temp = 6.200000e+02 : not converged.
Temp = 6.100000e+02 : not converged.
Temp = 6.000000e+02 : not converged.
Temp = 5.900000e+02 : not converged.
Temp = 5.800000e+02 : not converged.

Additionally, I suspect the persistent negative frequencies might be raised by this issue. Could you kindly advise on potential strategies to resolve this convergence problem?

Thank you very much for your time and consideration. I greatly appreciate your expertise.

Best regards,
Zhennan-Lin

[ttadano]

This may be due to inaccuracy of IFCs or inappropriate settings in the &scph field. I need more information about the calculations and target systems to suggest a solution.