problem in polarizability calculations

Find answers or ask questions regarding Dalton calculations.
Please upload an output file showing the problem, if applicable.
(It is not necessary to upload input files, they can be found in the output file.)

Post Reply
ikea1984
Posts: 4
Joined: 21 Mar 2016, 12:57
First name(s): Haitao
Last name(s): Liu
Affiliation: Institute of Applied Physics & Computational Mathematics
Country: China

problem in polarizability calculations

Post by ikea1984 » 21 Mar 2016, 14:28

Dear All,

I am a totally newbie for DALTON program. I want to calculate the polarizability of some atomic clusters. But I met a problem when I tried the test calculation with a Xe atom. The keyword .POLARI and .ALPHA will give quite different results. I hope someone could tell me the answer or some hint, thanks a lot. My input file is as follows

DALTON.INP:
**DALTON INPUT
.RUN PROPERTIES
.DIRECT
**WAVE FUNCTIONS
.DFT
CAMB3LYP
**PROPERTIES
.POLARI
.ALPHA
*ABALNR
.FREQUE
1
0.0
**END OF DALTON INPUT

MOLECULE.INP
BASIS
ANO-RCC
Calculation of the polarizability
------------------------------------------
1
54. 1
Xe 0.0 0.0 0.0


The POLARI and ALPHA results are 24.80 and 26.59 a.u. respectively.

hjaaj
Posts: 371
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: problem in polarizability calculations

Post by hjaaj » 21 Mar 2016, 15:59

Dear Haitao

You have found a bug. The correct number is the 26.59 au (I calculated it with finite difference, to be 100% sure), which is the value from the .POLARI option.

I also tried Hartree-Fock, and as I expected .ALPHA and .POLARI then give the same value. Note that two completely different modules in Dalton are called with .ALPHA and .POLARI. The Xenon polarizability results tell us that (some?) DFT functionals are not properly implemented in the routines activated with .ALPHA :(

I am convinced that you will get the correct results in your research project by continuing with the .POLARI option only.

-- Hans Jørgen

ikea1984
Posts: 4
Joined: 21 Mar 2016, 12:57
First name(s): Haitao
Last name(s): Liu
Affiliation: Institute of Applied Physics & Computational Mathematics
Country: China

Re: problem in polarizability calculations

Post by ikea1984 » 21 Mar 2016, 16:12

Dear Hans,

Thanks for quick and detailed reply. I agree that the 26.59 au is the correct result which I verified with some other program. Note that this value is calculated by the .ALPHA (the linear response module I thought) instead of the .POLARI keyword. I did another test with the Xe atom without symmetry. Both .POLARI and .ALPHA will give the same correct value. So I guess the problem occured when the .POLARI and symmetry are adopted simultaneously. Hope my update will be useful for you.

There is one more question I did not know, what is the difference between the .POLARI and .ALPHA, in the theoretical aspect?

Haitao

hjaaj
Posts: 371
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: problem in polarizability calculations

Post by hjaaj » 21 Mar 2016, 19:44

ikea1984 wrote:Thanks for quick and detailed reply. I agree that the 26.59 au is the correct result which I verified with some other program. Note that this value is calculated by the .ALPHA (the linear response module I thought) instead of the .POLARI keyword. I did another test with the Xe atom without symmetry. Both .POLARI and .ALPHA will give the same correct value. So I guess the problem occured when the .POLARI and symmetry are adopted simultaneously. Hope my update will be useful for you.
No. The .POLARI option gives the correct value, the .ALPHA option gives the wrong value for Xe and CAMB3LYP.
From your "another test" it seems as if the bug is related to the DFT contrinbutions to non totally-symmetric response vectors...
ikea1984 wrote:There is one more question I did not know, what is the difference between the .POLARI and .ALPHA, in the theoretical aspect?
For static polarizabilities the results should be the same, so in the theoretical aspect there should be no difference. However, there is a fairly big difference in the implementation. With .ALPHA Dalton calls the same linear response module as used for wave function response contributions to the molecular Hessian, and that module can only calculate static singlet responses. With .POLARI Dalton calls the more general RESPONS module to calculate the polarizability, and you could get the same results by calling the response module with .RUN RESPONSE (but **RESPONS input is somewhat different from the **PROPERTIES input, and the outputs are also different - usually nicer structured when **PROPERTIES).

hjaaj
Posts: 371
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: problem in polarizability calculations

Post by hjaaj » 21 Mar 2016, 20:11

Oops. SORRY, you are right. I had switched what is .ALPHA and what is .POLARI. So they have to be switched in my answer above.


.POLARI is most useful when you also calculate the molecular Hessian and do a vibrational analysis (.VIBANA) - then you will get the static electronic polarizability from .POLARI, the so-called vibrational polarizability from .VIBANA, and the total (the sum) when both are specified. .POLARI is the oldest of the two options .POLARI and .ALPHA, .ALPHA was introduced to do frequency-dependent polarizabilities.

xiongyan21
Posts: 262
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

Re: problem in polarizability calculations

Post by xiongyan21 » 24 Mar 2016, 03:50

Dear All
I can repeat the calculation from Dr. Liu by the change of the format of .mol file into one which can be recognized by Dalton2016.0 and obtain the same results.
Can we carry out the same calculation through coupled cluster methods, e.g., CCSD,CC2,CC3,etc, using the keywords .POLARI and .ALPHA?
Thanks very much for your advice.
Very Best Regards!
Last edited by xiongyan21 on 25 Mar 2016, 05:07, edited 1 time in total.

lyzhao
Posts: 74
Joined: 11 Nov 2013, 00:36
First name(s): Youzhao
Last name(s): Lan
Affiliation: Institute of Physical Chemistry
Country: China

Re: problem in polarizability calculations

Post by lyzhao » 24 Mar 2016, 03:59

Dear xiong,
To calculate response properties such as polarizability using CC* method, please refer to CC section in manual.

Lan

xiongyan21
Posts: 262
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

Re: problem in polarizability calculations

Post by xiongyan21 » 24 Mar 2016, 05:13

Dear Dr. Lan
Does it mean it is impossible for us to include the keywords .POLARI and .ALPHA in a coupled cluster polarizability calculation?
Thanks very much!
Very Best Regards!
Last edited by xiongyan21 on 27 Mar 2016, 12:10, edited 3 times in total.

lyzhao
Posts: 74
Joined: 11 Nov 2013, 00:36
First name(s): Youzhao
Last name(s): Lan
Affiliation: Institute of Physical Chemistry
Country: China

Re: problem in polarizability calculations

Post by lyzhao » 24 Mar 2016, 11:28

Dear Xiong,
Why do you use .POLARI and .ALPHA in CC?
Do you want to test CC module?
According to the manual, .POLARI belongs to the directives for evaluation of HF, SOPPA, and MCSCF molecular properties.

Lan

Post Reply

Who is online

Users browsing this forum: No registered users and 16 guests