Im using Abinit Code for band structure calculations. To calculate transport properties using Boltztrap code, i have printed input files for Boltztrap code using following input in Abinit:
Code: Select all
#defination of k -point grid
kptopt1 4
ngkpt1 6 6 2
prtden1 1
toldfe1 5.0d-7
#Dataset 2 :
prtbltztrp2 1 # boltztrap output
iscf2 -2
getden2 -1
kptopt2 4
ngkpt2 12 12 4
nband2 80
tolwfr2 1.0d-12
After this four files have been generated, viz. .ns_energyso, .struct, .def and .intrans. (attached here)
when i run Boltztrap code. using
Code: Select all
../../../../src/x_trans BoltzTraP -so
then on terminal display:
Code: Select all
================ BoltzTraP vs 1.2.5 =============
STOP error in factorization
11.2u 0.0s 0:11.40 98.6% 0+0k 10720+16io 16pf+0w
When i checked .outputtrans file, the calculation stopped at
Code: Select all
Allocating engre...
engre allocated
About to enter fite4...
fite4 matrix setup: 0.192
inf = 1
From file fite4.F90 in src folder of Boltztrap source code,
Code: Select all
WRITE(6,'(A,f10.3)') 'fite4 matrix setup:',t2-t1
! print*,'xx',SUM(hmat)
! LU factorization of hmat matrix, kept in 2 triangular blocks of the matrix
CALL DGETRF(bs%nkpt-1,bs%nkpt-1,hmat,bs%nkpt-1,ipiv,inf)
! print*,'xx',SUM(hmat)
write (6,*) 'inf = ', inf
IF(inf/=0) STOP 'error in factorization'
Boltztrap's output file .outputtrans also attached.
There is a problem in LU factorization of matrix. In my case, the value inf is 1, it means U(1,1) element in Upper matrix (U) is zero. Now further I am unable to solve this problem.
- 1. How the U(1,1) element can be non-zero?
- 2. For spin-orbit coupling calculation kptopt = 4 is used. Is it the reason for error?
Janpreet Sidhu
Panjab University
Chandigarh
India