Zero superconducting gap after Padé continuation & memory

Post here questions linked with issue while running the EPW code

Moderator: stiwari

Post Reply
huantran
Posts: 1
Joined: Mon Mar 06, 2023 7:53 pm
Affiliation: Georgia Institute of Technology

Zero superconducting gap after Padé continuation & memory

Post by huantran »

Dear community,

I have two questions on running EPW. First, it looks to me that in my case, solving the Eliashberg equations in imaginary-axis is done, but then the Padé approximation return zero znorm and delta. I am not sure why, and my output is below. If the input is needed, I will post.

The second question: the required memory increases continuously (see below). Is this normal? I am not sure if I can afford this memory demand if it keep increasing.

Thank you. Any suggestions are appreciated.
Huan Tran

Ref 1. Output related to Pade approximation
temp( 2) = 4.00000 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw( 2) = 462
Cutoff frequency wscut = 1.0017


Size of allocated memory per pool: ~= 64.2616 Gb
Actual number of frequency points ( 2) = 462 for uniform sampling
Size of allocated memory per pool: ~= 64.2804 Gb
Size of allocated memory per pool: ~= 64.2851 Gb
Size of allocated memory per pool: ~= 90.1699 Gb
iter ethr znormi deltai [meV]
1 2.507534E+00 1.663165E+00 1.298018E+00
2 7.410276E-01 1.664141E+00 1.336912E+00
3 1.323488E-01 1.664234E+00 1.333875E+00
4 3.748113E-02 1.664299E+00 1.320079E+00
5 1.231918E-02 1.664405E+00 1.306483E+00
6 1.783111E-02 1.664587E+00 1.283840E+00
7 5.049641E-03 1.664641E+00 1.277238E+00
8 5.863746E-03 1.664682E+00 1.271005E+00
9 7.913872E-03 1.664620E+00 1.279887E+00
10 5.969435E-04 1.664614E+00 1.280667E+00
11 1.104081E-03 1.664624E+00 1.279403E+00
12 4.137961E-04 1.664628E+00 1.278883E+00
13 4.159026E-04 1.664631E+00 1.278384E+00
14 1.612322E-05 1.664631E+00 1.278395E+00
Convergence was reached in nsiter = 14

Chemical potential (itemp = 2) = 1.4631500743E+01 eV
Temp (itemp = 2) = 4.000 K Free energy = -0.000335 meV
Size of allocated memory per pool: ~= 90.1687 Gb
Size of allocated memory per pool: ~= 96.2839 Gb
Size of allocated memory per pool: ~= 96.2651 Gb
Size of allocated memory per pool: ~= 96.2639 Gb
Min. / Max. values of superconducting gap = 0.000000 1.871303 meV
iaxis_imag : 52204.52s CPU 52318.48s WALL ( 2 calls)

Pade approximant of anisotropic Eliashberg equations from imaginary-axis to real-axis
Cutoff frequency wscut = 1.0000

Size of allocated memory per pool: ~= 96.3151 Gb
pade Re[znorm] Re[delta] [meV]
416 NaN NaN

Convergence was reached for N = 416 Pade approximants

Size of allocated memory per pool: ~= 96.3150 Gb
Min. / Max. values of superconducting gap = 0.000000 0.000000 meV

Size of allocated memory per pool: ~= 96.3660 Gb

Size of allocated memory per pool: ~= 96.3150 Gb
raxis_pade : 16.68s CPU 17.84s WALL ( 2 calls)


Size of allocated memory per pool: ~= 96.3127 Gb
itemp = 2 total cpu time : 52336.32 secs


Size of allocated memory per pool: ~= 96.2615 Gb

Ref 2. Output related to memory

grep "Size of allocated memory per pool" epw.out
Size of allocated memory per pool: ~= 52.0221 Gb
Size of allocated memory per pool: ~= 64.3063 Gb
Size of allocated memory per pool: ~= 64.2686 Gb
Size of allocated memory per pool: ~= 64.2663 Gb
Size of allocated memory per pool: ~= 64.3174 Gb
Size of allocated memory per pool: ~= 64.3174 Gb
Size of allocated memory per pool: ~= 64.3683 Gb
Size of allocated memory per pool: ~= 64.3174 Gb
Size of allocated memory per pool: ~= 64.3127 Gb
Size of allocated memory per pool: ~= 64.2615 Gb
Size of allocated memory per pool: ~= 64.2616 Gb
Size of allocated memory per pool: ~= 64.2804 Gb
Size of allocated memory per pool: ~= 64.2851 Gb
Size of allocated memory per pool: ~= 90.1699 Gb
Size of allocated memory per pool: ~= 90.1687 Gb
Size of allocated memory per pool: ~= 96.2839 Gb
Size of allocated memory per pool: ~= 96.2651 Gb
Size of allocated memory per pool: ~= 96.2639 Gb
Size of allocated memory per pool: ~= 96.3151 Gb
Size of allocated memory per pool: ~= 96.3150 Gb
Size of allocated memory per pool: ~= 96.3660 Gb
Size of allocated memory per pool: ~= 96.3150 Gb
Size of allocated memory per pool: ~= 96.3127 Gb
Size of allocated memory per pool: ~= 96.2615 Gb
Size of allocated memory per pool: ~= 96.2616 Gb
Size of allocated memory per pool: ~= 96.2741 Gb
Size of allocated memory per pool: ~= 96.2773 Gb
Size of allocated memory per pool: ~= 113.5338 Gb
Size of allocated memory per pool: ~= 113.5330 Gb

hpaudya1
Posts: 190
Joined: Tue Mar 21, 2017 7:11 pm
Affiliation:

Re: Zero superconducting gap after Padé continuation & memory

Post by hpaudya1 »

Hi Huan Tran,

First: It looks like your Pade calculation is not converged. I recommend you to decrease the value of "npade" (the default value is 90, https://docs.epw-code.org/doc/Inputs.html#npade)

Second: The memory requirement depends mainly on number of k/q points, number of bands inside the Fermi window, and the number of Matsubara frequencies at that specific temperature. Please have a look at these numbers in your output files.

Please upload your input/output files for further investigation.

Happy EPWing,
Hari

Post Reply