I am trying to reproduce the quantity referred to as the degree of charge transfer reported in this publication by Zhu et al. [Ref. 1] for the F4-TCNQ molecule. I am interested in section 3.2 in the main text about using the bond length as an estimate for the charge transfer amount.

capture of figure 3 from reference Zhu et al.]

In figure 3b, the authors show the correlation between the degree of charge transfer, CT$_{pop}$ from their natural population analysis compared to the approximation of the bond lengths of F4-TCNQ from equation 1, shown below.

$$\rho = \frac{\alpha_{CT}-\alpha_{0}}{\alpha_{-1}-\alpha_{0}}$$ with $\alpha_i = l_3 (l_2+l_4)$ where $l$ is the bond length, and subscript $i=0, -1, CT$ denote the netural molecule, the anion, and the compelx, respectively.

It's not clear to me, how the authors determined the values on the y-axis in Figure 3b, and there are no further clarification or references to a methodology for the values within the text or supplement information. Both axes in Figure 3b use the F4-TCNQ anion as an arbitrary reference.

I can calculate $\rho$ from the above equation, but I want to compare $\rho$ to the CT$_{pop}$ at the level of theory I am employing, verifying the accuracy of my calculations.

Through reading and googling, I believe the CT$_{pop}$ refers to a population analysis such as a Mulliken or natural population (AKA Lowdin population), referenced in Mendez et al. [Ref. 2] (bottom of page 9 in the Theory footnote). I do not know how to obtain the degree of charge transfer from the resulting population analysis.

I have one reference from Joo et al. [Ref. 3] who gives more details on a similar method, writing on page 2, left column:

The calculated atomic charges were then used to obtain the degree of charge transfer between the donor and acceptor molecules, which may be defined as the sum of all atomic charges on the donor part of the complex.

The above text leaves me to believe the method of determining the degree of charge transfer is as follows:

  1. Calculate a population analysis of the DA (donor and acceptor) complex to determine the atomic charge
  2. Determine the atomic charge of each of the fragments D and A by summing the atomic charge for the D and A fragments, respectively
  3. Report the atomic charge of fragments as the degree of charge transfer.

The last aspect is Joo et al. [Ref. 3] reports the value for the electron donor while Zhu et al. [Ref. 1] indicates the value for the electron acceptor, which is F4-TCNQ.

Does this procedure sound reasonable? Any thoughts or comments would be appreciated.

TL;DR How can I reproduce the values shown in Figure 3b on the y-axis, the degree of charge transfer from a natural population analysis?


  1. Zhu L; Kim E.-G; Yi Y; Bredas J.-L. Charge Transfer in Molecular Complexes with 2,3,5,6-Tetrafluoro-7,7,8,8- Tetracyanoquinodimethane (f4-Tcnq): A Density Functional Theory Study. Chemistry of Materials 2011, 23 (23), 5149–5159 DOI: 10.1021/cm201798x.
  2. Méndez Henry; Heimel, G.; Winkler, S.; Frisch, J.; Opitz, A.; Sauer, K.; Wegner, B.; Oehzelt, M.; Röthel Christian; Duhm, S.; et al. Charge-Transfer Crystallites As Molecular Electrical Dopants. Nature Communications 2015, 6 (1) DOI: 10.1038/ncomms9560.
  3. Joo B; Kim EG. Model-Independent Determination of the Degree of Charge Transfer in Molecular and Metal Complexes. Chemical Communications (Cambridge, England) 2015, 51 (81), 15071–15074 DOI: 10.1039/c5cc05274b.

I will evade answering the main question because I am not feeling expert enough, but I want to point out that Natural population analysis (NPA) is very different (and better) from the Loewdin one. Loewdin is computed from orbitals of a special basis set (although it comes from the density, it is still free-atom-derived basis set which is used to obtain the density matrix). NPA (part of NBO) is from density matrix which is adjusted to the particular molecular environment, hence better representing the "real" atomic population. NPA is available in the commercial NBO program, old version of which is also included with Gaussian package (maybe also in other packages, I don't know). You can also try AIM charges which are absolutely basis-independent (if Your density is good enough).

Personally I would think the analysis You proposed is fine with NPA or AIM charges, but I state again that I am not an expert in the topic.

See https://www.q-chem.com/qchem-website/manual/qchem43_manual/sect0061.html and also http://www.huntresearchgroup.org.uk/teaching/teaching_comp_chem_year4/L7_bonding.pdf .

| improve this answer | |
  • 2
    $\begingroup$ Related: How dependent are computed charges using the quantum theory of atoms in molecules on the used level of theory? Please note that the NBO version included in Gaussian is long deprecated; worst of all, results are not consistent with newer developments, and there are studies that point out why. I strongly recommend not using anything earlier than NBO6. $\endgroup$ – Martin - マーチン Jul 4 '19 at 9:44
  • $\begingroup$ Thank you @Igor Mihailovs, I didn't realize there was a difference between NPA and Loewdin. Does AIM stand for Atoms In Molecules by Bader? I don't have much experience with either an NPA or AIM analysis. I have been playing around in Multiwfn link to the site, and I did recognize AIM analysis implemented. $\endgroup$ – user3587374 Jul 4 '19 at 18:32
  • $\begingroup$ Yes, AIM is for "Atoms-In-Molecules". Actually, Gaussian seems to also have some AIM-related optionas computable (look at IOp(6/35) options). But I have, too, never used it, so I cannot comment further. Just running IOp(6/35=2) for AIM charges produces no result. $\endgroup$ – Igors Mihailovs Jul 5 '19 at 10:25
  • 1
    $\begingroup$ @igor I think there is a free implementation of NPA, try looking up janpa. Multiwfn can do aim charges, it's actually easy and well explained in the manual. NBO is - in my experience - well worth the money, more so than Gaussian for example. $\endgroup$ – Martin - マーチン Jul 5 '19 at 10:47
  • 1
    $\begingroup$ I am not 100% certain, but I read on the CCL archive that AIM in Gaussian was unstable and therefore deprecated/never fixed. Also on the matter at hand, I am not familiar with the literature on the subject. While AIM has a rigorous condition for charges, it is a more or less arbitrary one. It doesn't hurt to look at other decomposition schemes to find whether they agree. $\endgroup$ – Martin - マーチン Jul 5 '19 at 14:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.