Potential energy curve of O2

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
jesper
Posts: 3
Joined: 22 May 2017, 15:52
First name(s): Jesper
Last name(s): Janssen
Affiliation: Eindhoven University of Technology
Country: Netherlands

Potential energy curve of O2

Post by jesper » 23 May 2017, 15:08

My goal is to calculate transport properties like the thermal conductivity and viscosity of various gases. In order to obtain these properties I need the interaction potential between the particles. I am currently testing DALTON for interactions involving oxygen atoms. I used the following input files
O2.mol
(134 Bytes) Downloaded 45 times
O2.dal
(261 Bytes) Downloaded 52 times
Using these files I obtained the energies -150.0973834488 Hartree for the singlet state; -150.1468234657 Hartree for ^3B1g and -149.9202076234 Hartree for ^3B1u.
By moving the O atoms further apart
O2_far.mol
(134 Bytes) Downloaded 47 times
I obtain -149.8815131593 Hartree for the singlet state; -149.8601469889 for ^3B1g and -150.0244409923 Hartree for ^3B1u. Thus the ^3B1u state ends up below the ^3B1g state. According to Bytautas http://dx.doi.org/10.1063/1.3298376 (MRCISD+Q calculation) this should not happen. Additionally, a potential energy difference of about 7.8eV is obtained for the ^3B1g state which should be around 5.21eV (vibrational offset is also included). This shows that I make an error of about 2.6eV.

My questions:
- Is it possible to improve my result using a CCSD calculation? Or is it fundamentally not possible to obtain an accurate representation of the two triplet O atoms with the excitation operator being limited to T = T1 + T2 in a CCSD calculation?
- Should I somehow use a different initialization for the CCSD calculation? Open shell initializations are not supported. Is there a smarter way to initialize the system by using something other than .HF?
- Can I only avoid the difficulties above by using a CI calculation? For example using a GASCI calculation?

xiongyan21
Posts: 125
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

Re: Potential energy curve of O2

Post by xiongyan21 » 24 May 2017, 04:16

Dear Dr. Jessen
Based on your input, the use of aug-cc-pV5Z makes the energy of ^3B1g lower and that of ^3B1u higher, which displays the correct order.
What still remains is whether CCSD can be suitable to obtain the correct energies for this geometry.
Last edited by xiongyan21 on 30 May 2017, 12:52, edited 3 times in total.

jesper
Posts: 3
Joined: 22 May 2017, 15:52
First name(s): Jesper
Last name(s): Janssen
Affiliation: Eindhoven University of Technology
Country: Netherlands

Re: Potential energy curve of O2

Post by jesper » 29 May 2017, 12:39

Dear xiongyan21,

I tested your suggestion by running calculations at multiple basis sets. At a separation of 121pm I obtain

basis set 1Ag 3B1g 3B1u 1Ag -3B1g 3B1u-3B1g
aug-cc-pVDZ -149.9588425 -150.0098259 -149.7776001 0.050983449 0.232225792
aug-cc-pVTZ -150.0973834 -150.1468235 -149.9202076 0.049440017 0.226615842
aug-cc-pVQZ -150.1668218 -150.2156908 -149.9901849 0.04886895 0.225505832
aug-cc-pV5Z -150.1920768 -150.2406963 -150.0156604 0.048619511 0.225035883
aug-cc-pV6Z -150.207449 -150.2559531 -150.0311067 0.048504047 0.224846358

which indeed ensures that 3B1g is lower than 3B1u. However, at a separation of 351pm this should still be the case.
I cannot reproduce the order that you obtained. Therefore I would like to see your input file. The results at 351pm

basis set 1Ag 3B1g 3B1u 1Ag -3B1g 3B1u-3B1g
aug-cc-pVDZ -149.783203 -149.7611533 -149.9203879 -0.022049645 -0.159234641
aug-cc-pVTZ -149.8815132 -149.860147 -150.024441 -0.02136617 -0.164294003
aug-cc-pVQZ -149.9415116 -149.9202939 -150.0861467 -0.021217718 -0.165852777
aug-cc-pV5Z -149.9623077 -149.941161 -150.1076189 -0.021146708 -0.16645795
aug-cc-pV6Z -149.9758815 -149.9547774 -150.1214645 -0.021104025 -0.166687101

which indicates that the basis set has a relatively small impact when going to aug-cc-pVQZ or higher.

xiongyan21
Posts: 125
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

Re: Potential energy curve of O2

Post by xiongyan21 » 30 May 2017, 01:56

Dear Prof. Janssen
Sorry, I didn't try the separation that large.
Thanks a lot for the investigation. Anyway, CCSD may not be suitable for this case.
This reference illustrates normalized and completely renormalized CCSD(T) and CCSD(TQ) methods can provide significantly improved results at large N-N separations.
Renormalized CCSD(T) and CCSD(TQ) approaches: Dissociation of the N2 triple bond
《Journal of Chemical Physics》 , 2000 , 113 (14) :5644-5652
The following illustrates excited-state potential energy curves.
Excited-state potential energy curves of CH + : a comparison of the EOMCCSDt and full EOMCCSDT results
《Chemical Physics Letters》 , 2001 , 347 (1–3) :237-246
You can find other references.

Diverse, efficient normalized and completely renormalized coupled cluster methods are the exciting and now mature features in GAMESS and NWCHEM. Some multi-reference coupled cluster methods are implemented in NWCHEM and some ROHF coupled cluster ones, e.g., CCSD, EOMCCSD and CR-CC(2,3) in GAMESS.

I think but I am not very sure (It may fail) there may be , but may not limited to, one kind of energy among various kinds of ground states and excited states energies obtained to fit each in this case, which may be a turn-key solution.

By the way, there are wonderful non-RHF DFT and TDDFT methods in GAMESS, which should be much more economic.
Last edited by xiongyan21 on 12 Mar 2018, 02:27, edited 14 times in total.

jesper
Posts: 3
Joined: 22 May 2017, 15:52
First name(s): Jesper
Last name(s): Janssen
Affiliation: Eindhoven University of Technology
Country: Netherlands

Re: Potential energy curve of O2

Post by jesper » 19 Jun 2017, 16:55

Dear xiongyan21,

These papers refer to very interesting methods. I was not aware of these possibilities. It seems that DALTON cannot provide such corrections so I will consider using other packages like NWCHEM or GAMESS. Thank you for suggesting these packages!

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests