Step-wise shift of total energy along with lattice constants

Posted on : February 14, 2017 (Tue) 15:26:41

by 澤田　亮人

Dear researchers,

I calculated bcc-Pt metal/alloy have step-wise shift of total energy at a critical lattice costants.

bcc-Pt metal is known to be ubstable, however the total energy after the shift is sometimes lower than its stable (fcc-Pt) energy.

I am afraid this will prevent from estimating stable crystal structure in alloy. How can I avoid such a kind of trouble?

Best wishes,

[Same question in Japanese]
bccのPt合金などで特定の格子定数において、階段状のtotal energyのシフトが発生するようです。Ptは本来fccであるものですが、bcc-Ptのエネルギーのシフト後のtotal energyはfccのものより少なくなってしまいます。

こうしたエネルギーシフトは合金の結晶構造などを予想する際に問題となる（複数の結晶構造で計算し、最もエネルギーの低いものを採用するという手が使いにくくなる）のではと心配しています。

問題の回避方法について、アドバイスをいただけると幸いです。よろしくお願い申し上げます。

[Re:05] Step-wise shift of total energy along with lattice constants

Posted on : February 27, 2017 (Mon) 09:42:41

by Hitoshi GOMI

Dear KKR Administrator,

Thank you for your reply.
I calculated without the SIC, and it works fine.

Hitoshi

[Re:04] Step-wise shift of total energy along with lattice constants

Posted on : February 26, 2017 (Sun) 23:35:08

by Administrator

Sorry, I forgot to give one of important notes:
In source/cstate.f, the values of 'sic' and 'eoff' are set in data statements:

data istop/50/, tol/1d-8/, eb/-20d0/, sic/.true./
& ,eoff/ 1d3, 1d3, 1d3,-4d0/

These data control if the self-interaction corrections be made or not. In the present case of Pt, for the lattice constant smaller than around 5.54, the SIC correction will be made. For the lattice constant larger than 5.58, this does not occur. For a intermediate value of the lattice constant 5.54~5.58, actually there are two solutions, SIC being included for one of them, not for another. This happens because of the fixed valued of 'eoff' specified in source/cstate.f. To avoid this, you may either switch off the SIC procedure by using the data

data istop/50/, tol/1d-8/, eb/-20d0/, sic/.false./
& ,eoff/ 1d3, 1d3, 1d3,-4d0/

or, if you would like to use SIC, modify data statement as follows:

data istop/50/, tol/1d-8/, eb/-20d0/, sic/.true./
& ,eoff/ 1d3, 1d3, 1d3,-20d0/

The values '-20d0' is rather arbitrary. In this case SIC is applied for f core states
whose energy is higher than -20Ry. If SIC is not applies, the total energy should be around -36804 Ry, while it should be around -36824Ry with SIC.

KKR Administrator

[Re:03] Step-wise shift of total energy along with lattice constants

Posted on : February 26, 2017 (Sun) 19:56:16

by Hitoshi GOMI

Dear KKR administrator,

I tried the following input with and without your suggestion, but both calculations show the same energy jump at between a=5.5 and 5.6 Bohr.

c------------------------------------------------------------
go data/bccPt
c------------------------------------------------------------
c brvtyp a c/a b/a alpha beta gamma
bcc 5.60, , , , , ,
c------------------------------------------------------------
c edelt ewidth reltyp sdftyp magtyp record
0.001 1.2 sra mjw nmag 2nd
c------------------------------------------------------------
c outtyp bzqlty maxitr pmix
update 4 200 0.01
c------------------------------------------------------------
c ntyp
1
c------------------------------------------------------------
c type ncmp rmt field mxl anclr conc
Pt 1 1 0.0 2
78 100
c------------------------------------------------------------
c natm
1
c------------------------------------------------------------
c atmicx atmtyp
0 0 0 Pt
c------------------------------------------------------------



Following the total energy with decreasing the lttice constant.
# a(Bohr) Energy(Ry)
6.60 -36804.667463414
6.50 -36804.690033507
6.40 -36804.712063384
6.30 -36804.733114209
6.20 -36804.752647273
6.10 -36804.769324504
6.00 -36804.779290960
5.90 -36804.781558124
5.80 -36804.775368583
5.70 -36804.759709762
5.60 -36804.733832340
5.50 -36824.482094153
5.40 -36824.426626461
5.30 -36824.352951349
5.20 -36824.260076372



I did not find any change in parameters, even without the modification of source/spmain.f.

out/bccPt_5.20.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.30.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.40.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.50.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.60.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.70.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.80.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_5.90.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.00.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.10.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.20.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.30.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.40.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.50.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1
out/bccPt_6.60.out: last= 243 np= 19 nt= 273 nrpt= 169 nk= 29 nd= 1

Hitoshi

[Re:02] Step-wise shift of total energy along with lattice constants

Posted on : February 16, 2017 (Thu) 20:42:58

by Administrator

It is crucial to keep the condition of the calculation during the series of runs to avoid such jumps that may happen when continuously changing the lattice constant. In most cases the calculational condition changes when the number of the real space lattice vectors and that of the reciprocal space vectors, which are used for generating structural Green's function, change. These number can be seen from the output:

last= 243 np= 15 nt= 331 nrpt= 177 nk= 554 nd= 1

In this case, 331 G-vectors and 177 R-vectors are used. These numbers depends on the lattice constant as well as the energy range specifies by input parameter 'ewidth'.
To keep these values unchanged, you have to fix the value of some control parameters:
Look into the program 'spmain.f'. At lines # 277 and 278 (if you are using the latest version of cpa2002v009c), you will find lines that seem like

275 c------- fix ew and ez to some special values
276 c write(*,'(1x,a)')' Fixed ew and ez are used'
277 c ew=-0.45563
278 c ez=1.2
279 c---------------------------------------

You may uncomment the lines from 276 through 278, giving some values to ew and ez. You have to choose the values from a typical calculations of your system. In the output, you will find the values for 'ew' and 'ez' that were used in that calculation. If you specify those values at these lines, the program is forced to use them for ew and ez. Try this and see. Does the total energy still jump? Never forget to comment them out when you use the code for usual ways.

KKR administrator

[Re:01] Step-wise shift of total energy along with lattice constants

Posted on : February 14, 2017 (Tue) 23:16:04

by Hitoshi GOMI

Hello 澤田 亮人,

Would you please let me know the input file and the range of the lattice constant?

Hitoshi

澤田さま

インプットファイルと格子定数の範囲を教えてえいただけますか？

五味