Journal of the Indian Chemical Society 100 (2023) 101030

Contents lists available at ScienceDirect

Journal of the Indian Chemical Society

journal homepage: www.journals.elsevier.com/journal-of-the-indian-chemical-society

Nitrogen as a probable problematic factor of computational chemistry: A

benchmarking study
Mert Metin a, *, Tomonori Kawano a, Tadashi Okobira b
a
Graduate School of Environmental Engineering, The University of Kitakyushu, 1-1 Hibikino, Wakamatsu, Kitakyushu, Fukuoka, 808 0135, Japan
b
Department of Creative Engineering, National Institute of Technology, Ariake College, 150 Higashihagio, Omuta, Fukuoka, 836-8585, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: In the present study, theoretical IR frequencies and geometric parameters of triamterene, furosemide, and
Amine triamterene-furosemide salt were investigated. ORCA software was handled for commonly used BLYP, B3LYP,
Benchmark and HF (combined with various basis sets including 6-31G**, def2-SVP, cc-pVDZ, and pcseg-1 basis sets) cal­
Computational chemistry failure
culations. Theoretical IR frequencies and geometric parameters of triamterene and furosemide were compared
IR
Molecular geometry
with the corresponding experimental data. In addition, theoretical geometric parameters of triamterene-
Nitrogen furosemide salt were, as well, compared with the corresponding experimental data. Based on these compari­
sons, the performances of computational methods/basis sets were measured. As a result of this study, it was
discovered that the studied computational approaches were generally successful in the prediction of molecular
geometries and IR frequencies of the target molecules. When IR and geometry data were more deeply investi­
gated, it was observed that there were problems specific to nitrogen atoms. The computed geometry and IR of
amine and sulphonamide nitrogen atoms’ bonds strongly disagreed with the experimental data. Similar dis­
agreements were also observed for sulphur, oxygen, and cyclic hydrogen atoms but the level of disagreements
was lower.

1. Introduction
Computational chemistry is a field of science that uses computational
methods to study chemical phenomena [1]. It is a powerful tool that
complements experimental approaches in many areas of chemistry [2],
such as drug discovery [3–8], materials science [9–12], and environ­
mental chemistry [9,13–16]. Computational methods can provide in­
sights into the behaviour of molecules, atoms, and ions at a microscopic
level, which can be difficult to obtain experimentally [1]. One of the
main advantages of computational chemistry is that it allows re­
searchers to explore a wide range of chemical systems and conditions,
including those that are difficult or impossible to study experimentally
[2]. Additionally, computational chemistry can be used to simulate
complex systems, such as biological membranes, and study their
behaviour under different conditions [17]. As computational methods
continue to improve, their use in science is likely to become even more
widespread [1].
Advancements in spectroscopy rendered the detection of structures
and properties of chemicals possible [18]. Particularly, IR and Raman
spectroscopy are frequently utilized approaches for the investigation of
structures of large chemical systems [19,20]. Results from such in­
vestigations are required to be supported by theoretical studies. Quan­
tum chemical studies are generally highly reliable for a meaningful
comparison with the related experimental data [21]. The requirement of
huge computational resources renders quantum chemical studies on
large chemical systems highly problematic. Especially, the computation
of the IR/Raman spectra of large chemicals by the employment of
correlated ab initio approaches is a herculean mission [22].
The choice of method/basis set pair can critically act on the accuracy
and the speed of QM calculations. In general, stand-alone DFT methods
are computationally very cheap and can provide reasonable accuracy.
Differently, MP2, MP3, and MP4 methods (some enhanced versions of
HF) are computationally expensive and known to provide high-level
accuracy. On the other hand, hybrid functionals like B3LYP [3,23–26]
and M06-2X [27–29] are very popular for their good performance in
terms of accuracy and their reasonable computational cost. HF is a very
old method with known deficiencies, however, it is even in use and can
provide desired results depending on the case [30–33].
Similar to method choice, basis set preference can change the results
significantly. Various basis set families are available in many QM

* Corresponding author.
E-mail addresses: mertmetin3335@gmail.com (M. Metin), kawanotom@gmail.com (T. Kawano), okobira@ariake-nct.ac.jp (T. Okobira).

https://doi.org/10.1016/j.jics.2023.101030
Received 22 January 2023; Received in revised form 19 April 2023; Accepted 20 May 2023
Available online 20 May 2023
0019-4522/© 2023 Indian Chemical Society. Published by Elsevier B.V. All rights reserved.
M. Metin et al.
programs. Some basis sets are parameterized for some special functions.
For instance, Jensen basis sets (i.e. pcseg-1) are specifically designed for
the computation of NMR spectra [34]. Some basis sets are limited to
some elements such as Pople basis sets (i.e. 6-31G*) [35,36]. Pople basis
sets are very old basis sets but they are even now frequently referred to
Refs. [37–40]. In addition to the family of basis sets, polarization and
diffusion/augmentation functions can own significant influences on the
computations [41].
Another notable but less discussed issue of quantum chemistry could
be the choice of software. There is various QM software, like ORCA [42],
MOPAC [43], Psi4 [44–46], MOLPRO [47,48], Gamess US [49],
Gaussian [50], CP2K [51], QUICK [52], Quantum Espresso [53–55],
Turbomole [56,57], Q-Chem [58], VASP [59–61], and NWChem [62].
Each software has its own advantageous sides. While some are
open-source (such as CP2K and Psi4), some (ORCA) are freely available
for academic purposes as others (Turbomole and Q-Chem) can provide
some discounts & promotions for academic users. MOPAC software is
limited to semi-empirical methods while ORCA can handle both DFT and
semi-empirical calculations. Similar to ORCA, Gaussian can handle a
wide range of calculation methods. Gaussian owns its special graphical
user interface (GUI) called GaussView [63]. GaussView has some prac­
ticalities. This GUI has its database where the user can access the initial
atomic coordinates of some compounds, like furosemide [64].
Furosemide is a diuretic drug with low bioavailability due to its poor
solubility and poor membrane permeability. Triamterene is a potassium-
sparing diuretic with low bioavailability due to its poor solubility. Peng
et al. successfully synthesized a novel drug-drug salt composed of tri­
amterene and furosemide. The apparent equilibrium solubility (in pH
2.0 buffer) of triamterene–furosemide salt was improved compared with
furosemide and a physical mixture of triamterene with equimolar
furosemide by 15.3-fold and 12.2-fold enhancements, respectively [65].
In this study, the authors computationally investigated furosemide,
triamterene, and furosemide-triamterene salt. The study benchmarks
commonly handled computational approaches on two chemicals
together with their salt. It is important to note that, in general, the
molecular sizes of the benchmark molecules are not large. Three of the
studied molecules, especially the salt, are very large chemicals to run a
benchmark study on them. Therefore, the large sizes of the target
chemicals separate this study from the canonical benchmark studies of
computational chemistry.
As a result of our investigation, we observed that computed data
related to amine and amide nitrogen atoms strongly disagree with the
experimental data (both geometries and IR frequencies), regardless of
the method/basis set choice. We evaluated these results by referring to
our previous study on iminodiacetic acid (IDA) [41] and to studies from
the literature [66–69].
2. Methodology
2.1. Triamterene and furosemide
Triamterene and furosemide crystal structures were obtained in cif
file format [70,71]. The cif file was opened in PyMOL [72–74] and the
molecule was saved as a pdb file. The pdb file was used by Gabedit [75]
to create input files for ORCA quantum mechanics software [76,77].
Shortly, the computational chemical calculations were done by the
following methods/basis sets;
* BLYP method with 6-31G [35,36,78], 6-31* [35,36,78], 6-31G**
[35,36,78], 6-31+G** [35,36], 6–31++G** [35,36], 6-311G**
[78–82], def2-SVP [83], cc-pVDZ [78,84,85], and pcseg-1 [34,78]
basis sets
* HF method with 6-31G, 6-31*, 6-31G**, 6-31+G**, 6-31++G**, 6-
311G**, def2-SVP, cc-pVDZ, and pcseg-1 basis sets
* B3LYP method with 6-31G, 6-31*, 6-31G**, 6-31+G**, 6-31++G**,
6-311G**, def2-SVP, cc-pVDZ, and pcseg-1 basis sets
2
Journal of the Indian Chemical Society 100 (2023) 101030
All DFT calculations were dispersion corrected by D3BJ [86,87].
Firstly, the structures were processed by BLYP/def2-SVP method/
basis set. The BLYP/def2-SVP optimized structures were verified to be
minima (apparent by the absence of negative frequencies). Thus, the
BLYP/def2-SVP optimized structures were handled as inputs for B3LYP/
def2-SVP and other DFT calculations. The B3LYP/def2-SVP optimized
structures were verified to be minima (apparent by the absence of
negative frequencies) and so they were used as inputs for HF/def2-SVP
and for other B3LYP calculations. The HF/def2-SVP optimized struc­
tures were verified to be minima (apparent by the absence of negative
frequencies) and so they were used as inputs for other HF calculations.
Some calculations ended up with negative frequencies, which signs
that the optimized structures were not minima. However, some of these
negative frequencies were not lower than − 100 cm− 1 [88]. Thus, it was
concluded that such structures were not minima but they were very
close to minima. This means that they were sufficiently well-optimized.
However, some calculations ended up with negative frequencies which
were lower than − 100 cm− 1. This drove us to use a different input
structure for these calculations. Instead of def2-SVP optimized struc­
tures, the starting pdb files were used as inputs for such calculations.
Unfortunately, BLYP/6-31++G** calculation for furosemide again had
remarkable negative frequencies. So, BLYP/6-31++G** calculation for
furosemide was removed from the study.
When ORCA starts a calculation, it starts with the calculation of the
bond lengths and angles of starting geometry. So, in the output file of
BLYP/def2-SVP calculation, geometrical parameters of the pdb file were
also available. In this way, the authors had easy access to the experi­
mentally detected bond lengths and angles of triamterene and furose­
mide. Theoretical bond lengths and triangular angles were also
extracted from the output files. Later, theoretical bond lengths and an­
gles (by various methods/basis sets) were compared to experimental
bond lengths and angles by the calculation of correlation coefficients
(R2) and Root Mean Square Error (RMSE). Based on R2 (Tables 2 and 3)
and RMSE (Tables 5 and 6) values, the performances of methods/basis
sets were assigned and compared.
2.2. Triamterene–furosemide salt
Triamterene–furosemide salt crystal structure was obtained in cif file
format [65]. The cif file was opened in PyMOL and the molecule was
saved as a pdb file. The pdb file was used by Gabedit to create input files
for ORCA quantum mechanics software. Shortly, the computational
chemical calculations were done by the following methods/basis sets;
* BLYP method with 6-31G, 6-31G*, 6-31G**, 6-31+G**, 6-31++G**,
6-311G**, def2-SVP, cc-pVDZ, pcseg-1, def2-SVPD and aug-pcseg-1
basis sets
* HF method with def2-SVP basis set
* B3LYP method with def2-SVP basis set
All DFT calculations were dispersion corrected by D3BJ.
Firstly, the structure was processed by BLYP/def2-SVP approach.
The BLYP/def2-SVP optimized structure was verified to be a minimum
(apparent by the absence of negative frequencies). Thus, the BLYP/def2-
SVP optimized structure was handled as input for B3LYP/def2-SVP and
other DFT calculations. The B3LYP/def2-SVP optimized structure was
verified to be a minimum (apparent by the absence of negative fre­
quencies) and so it was used as an input for HF/def2-SVP calculation.
Among BLYP calculations, BLYP/6-31G** and BLYP/6-311G** cal­
culations ended up with negligible negative frequencies (one for each
calculation, nearly − 4 cm− 1). Because the values of negative frequencies
were very close to 0 cm− 1, these structures were thought to be very close
to their minimum. Thus, such structures were counted as well optimized.
Differently, BLYP/6-31+G** and BLYP/6-31++G** calculations ended
up with so many negative frequencies (some of them with high negative
values). This drove authors to use a different input structure instead of
M. Metin et al.
Table 1
Correlation Coefficients for The Geometric Parameters of Triamterene-Furosemide Salt.
In this table, the green colour was used for highlighting the best-performing approach
while the red colour was utilized for the worst-performing approach.
BLYP optimized one. So, it was tried to handle the starting pdb file as an
input for BLYP/6-31+G** and BLYP/6-31++G** calculations. Howev­
er, those calculations did not appear normal to proceed (they had
excessive iterations), therefore, they were terminated. Following that, it
was decided to give a try on other families of basis sets to have at least
one diffusion function-including calculation. Thus, additional BLYP/
def2-SVPD and BLYP/aug-pcseg-1 calculations were carried out. Un­
fortunately, they also had so many negative frequencies (some of them
with high negative values), similar to other diffusion function-including
calculations. So, a diffusion function-including calculation for tri­
amterene–furosemide salt could not be included in the study.
When ORCA starts a calculation, it starts with the calculation of the
bond lengths and angles of starting geometry. So, in the output file of
BLYP/def2-SVP calculation, geometrical parameters of the pdb file were
also available. In this way, there was easy access to the experimentally
detected bond lengths and angles of triamterene–furosemide salt.
Theoretical bond lengths and triangular angles were also extracted from
the output files. Later, theoretical bond lengths and angles (by various
methods/basis sets) were compared to experimental bond lengths and
angles by R2 and RMSE values. Based on R2 (Table 1) and RMSE
(Table 4), the performances of methods/basis sets were assigned and
compared.
In addition, the bond lengths and angles from B3LYP/def2-SVP cal­
culations were individually investigated for the details about disagree­
ment sites between experimental and theoretical data (Tables 7, 8, 9, 10,
11, and 12).
2.3. IR frequencies
In this study, theoretical IR peaks of triamterene and furosemide
were investigated and compared to corresponding experimental data.
All calculations were carried out by analytical Hessians. Basis sets do not
have a strong impact on scaling factors. Therefore, in all of the calcu­
lations, the use of scaling factors for the 6-311G* basis set was preferred.
HF calculations were scaled by 0.904, BLYP calculations were scaled by
0.998, and B3LYP calculations were scaled by 0.966 [89,90].
Experimentally detected IR peaks of triamterene and furosemide
were obtained from Spectral Database for Organic Compounds (SDBS)
[91]. For IR peaks, the authors had two options, peaks measured by the
Nujol mull or KBr disk method. Both of the options were selected. It was
afforded to match experimental IR peaks with the computationally
produced ones with the help of the Vibrational Mode Automatic Rele­
vance Determination (VMARD) outputs from vibAnalysis software (Ta­
bles 16 and 17) [92]. B3LYP/def2-SVP results were used for VMARD
analysis (Table_VMARD.docx).
3
Journal of the Indian Chemical Society 100 (2023) 101030
The computationally calculated IR peaks were compared to the
experimental peaks by the calculation of correlation coefficients (Ta­
bles 13 and 14). Correlation coefficients calculated both with Nujol mull
and KBr disk IR frequencies were handled to decide which approaches
were the best and the worst performers. Here, RMSE values were not
computed because the trend similarity of the compared data was the
interest, not numerical closeness.
For triamterene-furosemide salt, no vibrational frequency data were
available to us. So, it was not possible to measure the performances of
methods/basis sets in terms of IR frequency prediction. As a solution, it
was tried to index all computations to one of the calculations. BLYP/
def2-SVP calculation was chosen as the target calculation to index all
others. Correlation coefficients were calculated to measure the similar­
ity between BLYP/def2-SVP and other calculations in terms of IR peaks
(Table 15). By the comparison of their relative similarity to BLYP/def2-
SVP computed IR frequencies, methods/basis sets were compared to
each other.
2.4. NPA charges, reactivity indices, and dipole moment and
thermodynamic data
NPA charges were calculated via NBO analysis with the help of
JANPA software [93,94]. Only the charges of heteroatoms were detailly
investigated. Heteroatom charges from different calculations were
compared (Table_NPA.docx).
Reactivity indices (eHOMO, eLUMO, electronegativity, and hard­
ness) were extracted from the result files of IR calculations, with the help
of Gabedit software. Reactivity indices from different calculations were
compared (Table_RI.docx).
Dipole moment and thermodynamic data (entropy, inner energy,
enthalpy, and Gibbs free energy) were easily available in the result files
of IR calculations. Dipole moment and thermodynamic data from
different calculations were compared (Table_TD.docx).
2.5. Supplementary materials
Some data was very long to present in this report. So, just the sum­
mary for some parts, like the NPA charges section, was presented. As a
result, there was a rich supplementary material and the authors needed
to assign a dictionary for it. There are various supplementary files,
• Supplementary_Material.xlsx includes nearly all the raw and pro­
cessed data of this study.
• Table_VMARD.docx includes VMARD analysis of B3LYP/def2-SVP
frequencies together with the experimental data comparison.
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 2
Correlation Coefficients for The Geometric Parameters of Triamterene. In this table, the orange colour was
utilized to highlight HF calculations, the blue colour was used for BLYP calculations, and the purple colour was
utilized for B3LYP calculations, on the left side of the table. On the right side of the table, it was recognized that
6-31G and def2-SVP calculations were grouped and isolated from others. Thus, the red colour was utilized to
highlight grouped 6-31G calculations while the brown colour was used for def2-SVP calculations.

• Table_RI.docx includes reactivity indices data together with the
comments on the issue.
• Table_NPA.docx includes heteroatom NPA charges data together
with the comments on the issue.
• Table_TD.docx includes the dipole moment and thermodynamic
data together with the comments on the issue.
Supplementary_Material.xlsx is a crowded file which needs a sepa­
rate legend and here it is,
• Atom_Labels_1 includes the atom number labels for the atoms of the
salt molecule.
• TFS_Geo includes the raw and processed data for the geometry pa­
rameters of triamterene-furosemide salt.
• TFS_Geo_Difference includes the error calculation for the B3LYP/
SVP geometry data of salt molecule.
• TFS_Geo_RMSE includes RMSE calculations for the geometry data of
the salt.
• TFS_NPA includes the raw and processed data for the NPA charges of
triamterene-furosemide salt.
• TFS_RI includes the raw and processed data for the reactivity indices
of triamterene-furosemide salt.
• TFS_TD includes the raw and processed data for the dipole moment
and thermodynamic parameters of the salt.
• TFS_IR includes the raw and processed data for the IR frequencies of
the salt molecule.
• Atom_Labels_2 includes the atom number labels for the atoms of
triamterene.

4
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 3
Correlation Coefficients for The Geometric Parameters of Furosemide. In this table, the green colour was
used for highlighting the best-performing approach while the red colour was utilized for the worst-
performing approach.

Table 4
RMSE Values for The Geometry Data of Triamterene-Furosemide Salt. In this table, the orange colour
was utilized to highlight HF calculations, the blue colour was used for BLYP calculations, and the purple
colour was utilized for B3LYP calculations.

5
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 5
RMSE Values for The Geometry Data of Triamterene. In this table, the orange colour was utilized to highlight HF
calculations, the blue colour was used for BLYP calculations, and the purple colour was utilized for B3LYP calculations.

• Tri_Geo includes the raw and processed data for the geometry pa­
rameters of triamterene.
• Tri_Geo_Difference includes the error calculation for the B3LYP/
SVP geometry data of triamterene.
• Tri_Geo_RMSE includes RMSE calculations for the geometry data of
triamterene.
• Tri_NPA includes the raw and processed data for the NPA charges of
triamterene.
• Tri_RI includes the raw and processed data for the reactivity indices
of triamterene.
• Tri_TD includes the raw and processed data for the dipole moment
and thermodynamic parameters of triamterene.
• Tri_IR includes the raw and processed data for the IR frequencies of
furosemide.
• Atom_Labels_3 includes the atom number labels for the atoms of
furosemide.
• Fur_Geo includes the raw and processed data for the geometry pa­
rameters of furosemide.
• Fur_Geo_Difference includes the error calculation for the B3LYP/
SVP geometry data of furosemide.
• Fur_Geo_RMSE includes RMSE calculations for the geometry data of
furosemide.
• Fur_NPA includes the raw and processed data for the NPA charges of
furosemide.
• Fur_RI includes the raw and processed data for the reactivity indices
of furosemide.
• Fur_TD includes the raw and processed data for the dipole moment
and thermodynamic parameters of furosemide.
• Fur_IR includes the raw and processed data for the IR frequencies of
furosemide.
2.6. Atom labels
In the article, some tables and descriptions need information about
atom numbers/labels of the studied compounds. So, just below, atom
number labels for triamterene, furosemide, and their salt were provided.
Figures were obtained with the help of Avogadro and Gabedit soft­
ware. Readers should note that Avogadro and Gabedit programs differ
from ORCA in their way of labelling atoms of target molecules. They
start counting atoms from 1 while ORCA starts from 0. JANPA and
vibAnalysis handle the same numbering mechanism. Please investigate

6
M. Metin et al.
Table 6
RMSE Values for The Geometry Data of Furosemide. In this table, the orange colour was utilized to highlight HF
calculations, the blue colour was used for BLYP calculations, and the purple colour was utilized for B3LYP
calculations.
this document according to this knowledge.
3. Results
3.1. Bond lengths and triangular angles
3.1.1. Comparison of the geometries via R2
According to correlation coefficients presented in Tables 1 and in the
prediction of bond lengths of triamterene-furosemide salt (Table 1, left
side), BLYP/6-31G** performed the worst. Oppositely, HF/def2-SVP
performed the best. Except for BLYP/6-31G** and BLYP/6-31G, all
other methods/basis sets performed somehow well. The performances of
HF/def2-SVP, B3LYP/def2-SVP, and BLYP/def2-SVP were similar.
In the case of triangular angles (Table 1, right side), all methods
seem to fail in the prediction of the bond angles of triamterene-
furosemide salt. HF/def2-SVP was more successful than other
methods/basis sets and BLYP/6-31G** performed the worst. HF/def2-
SVP and B3LYP/def2-SVP outperformed BLYP/def2-SVP.
Correlation coefficients suggest that HF/def2-SVP was the most
successful method/basis set pair and BLYP/6-31G** performed the
worst, in the computation of the geometry parameters of triamterene-
furosemide salt.
According to correlation coefficients presented in Tables 2 and in the
7
Journal of the Indian Chemical Society 100 (2023) 101030
prediction of bond lengths of triamterene (Table 2, left side), all
methods/basis sets performed well. The best-performing approach was
BLYP/6-31G while the performance of HF/def2-SVP was lower than
others. On the left side of Table 2, BLYP calculations dominate the up­
side of the list, B3LYP dominates in the middle, and HF dominates the
downside. This means that the choice of the method matters for the
computation of triamterene bond lengths. Additionally, HF/6-31G per­
formed better than other HF calculations, BLYP/6-31G performed better
than other BLYP calculations, and B3LYP/6-31G performed better than
other B3LYP calculations. HF/def2-SVP performed worse than other HF
calculations, BLYP/def2-SVP performed worse than other BLYP calcu­
lations, and B3LYP/def2-SVP performed worse than other B3LYP cal­
culations. This shows that the computation of bond lengths of
triamterene was affected by the polarization function and by the use of
different basis set families.
In the prediction of triangular angles of triamterene (Table 2, right
side), all methods/basis sets had low-level success. However, BLYP/6-
31G* performed better than others. It is easily recognizable that, on
the right side of Table 2, def2-SVP and 6-31G calculations were grouped
at the downside of the list. So, def2-SVP and 6-31G calculations were the
least successful approaches. This demonstrates that the computation of
triangular angles of triamterene was strongly affected by the polariza­
tion function and by the use of different basis set families.
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 7
Triamterene-Furosemide Salt Bond Lengths Displaying Remarkable Differences Between the B3LYP/def2-SVP
and Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds and yellow for S
bonds. For atom labels of triamterene-furosemide salt, see Fig. 1.

Correlation coefficients suggest that the computation of triamterene
geometry was strongly affected by the polarization function and the use
of different basis set families. BLYP/6-31G* and/or BLYP/6-31G** can
be evaluated as the most successful methods/basis sets among studied
approaches, in the prediction of triamterene geometry.
According to correlation coefficients presented in Tables 3 and in the
prediction of bond lengths of furosemide (Table 3, left side), all ap­
proaches except HF/6-31G and BLYP/6-31G performed somehow well.
HF/6-31+G** and HF/6-31++G** performed the best while BLYP/6-
31G performed the worst.
In the prediction of triangular angles of furosemide (Table 3, right
side), all of the studied approaches failed. HF/6-31G performed the best
while BLYP/6-31G performed the worst.
Correlation coefficients suggest that the prediction of furosemide
geometry was best performed by B3LYP/6-31++G** approach and
worst performed by BLYP/6-31G as both bond lengths and triangular
angles were investigated.
3.1.2. Comparison of the geometries via RMSE
According to Table 4, all RMSE values for the bond lengths were
greater than 0.05 and all RMSE values for the bond angles were less than
5. This could be interpreted as a sign of the better functioning of all
calculations in the case of the bond angles, as compared to the bond
lengths of triamterene-furosemide salt. Additionally, the method choice
seems to strongly act on the RMSE values, and so on the success ratio, for
the bond lengths and angles of triamterene-furosemide salt.

8
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 8
Triamterene-Furosemide Salt Triangular Angles Displaying Remarkable Differences Between the B3LYP/
def2-SVP and Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds,
yellow for S bonds, and red for O bonds. For atom labels of triamterene-furosemide salt, see Fig. 1.

Table 9
Triamterene Bond Lengths Displaying Remarkable Differences Between the B3LYP/def2-SVP and
Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds. For atom labels of
triamterene, see Fig. 2.

Table 10
Triamterene Triangular Angles Displaying Remarkable Differences Between the B3LYP/def2-SVP and
Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds. For atom labels of
triamterene, see Fig. 2.

RMSE values for the bond lengths and angles propose that HF/def2-
SVP was the best-performing approach while BLYP/6-31G** was the
worst performer for the molecular geometry of triamterene-furosemide
salt.
According to Table 5, all RMSE values for bond lengths were greater
than 0.05 and all RMSE values for the bond angles were highly less than
5. This could be interpreted as a sign of the better functioning of all
calculations in the case of the bond angles, as compared to the bond
lengths of triamterene. Additionally, the method choice seems to
strongly act on the RMSE values for the bond lengths of triamterene as
the absence of polarization function and the utilization of def2-SVP basis
set caused worse performance in RMSE values for the bond angles.
RMSE values for the bond lengths and angles propose that HF/6-
311G** was the best-performing approach while BLYP/6-31G was the

9
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 11
Furosemide Bond Lengths Displaying Remarkable Differences Between the B3LYP/def2-SVP and
Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds, yellow for S bonds,
and red for O bonds. For atom labels of furosemide, see Fig. 3.

Table 12
Furosemide Triangular Angles Displaying Remarkable Differences Between the B3LYP/def2-SVP and
Experimental Results (≥2%, in Ǻ). In the table, blue colour was handled for N bonds, yellow for S bonds,
and red for O bonds. For atom labels of furosemide, see Fig. 3.

worst performer for the molecular geometry of triamterene.
According to Table 6, all RMSE values for bond lengths were greater
than 0.05 and all RMSE values for bond angles were less than 5. This
could be interpreted as a sign of the better functioning of all calculations
in the case of the bond angles, as compared to the bond lengths of
furosemide. Additionally, the method choice seems to strongly act on
the RMSE values for bond lengths and angles of furosemide.
RMSE values for the bond lengths and angles propose that HF/6-
31+G** was the best-performing approach while BLYP/6-31G* was the
worst performer for the molecular geometry of furosemide..
Table 7 indicates that B3LYP/def2-SVP bond lengths for triamterene-
furosemide salt strongly disagree with the experimental data in the case
of amine and sulphonamide N-H bonds and cyclic C-H bonds. The level
of disagreement was numerically very larger for amine N-H bonds as
compared to the C-H bonds. The table does not include the hydrogen
bonds of O9-H40 and O2-H53. These are the bonds binding triamterene
with furosemide. So, the hydrogen bond lengths are truly predicted.
Similarly, the bond angles associated with the O9-H40 and O2-H53
bonds were not in Table 8. Thus the bond angles for O9-H40 and O2-
H53 bonds were also well-predicted. These observations are remark­
able and positive regarding the performance of the B3LYP/def2-SVP
approach.
Table 8 demonstrates that the sulphonamide N-H bond angle and
pentagonal bond angles of triamterene-furosemide salt have relatively
large errors in comparison with the other bonds. The error was larger for
the N-H bond angle. The errors in bond angles (less than 10%) were
smaller and less abundant in comparison with the errors in bond lengths
(which can easily exceed 10%)..
Table 9 indicates that B3LYP/def2-SVP bond lengths for triamterene
strongly disagree with the experimental data in the case of amine N-H

10
M. Metin et al.
Table 13
Correlation Coefficients for IR Frequencies of Triamterene-Furosemide Salt. Here, the red
colour was utilized to sign HF/def2-SVP calculation which was least similar to BLYP/def2-
SVP.
bonds and benzene C-H bonds. The level of disagreement was numeri­
cally a little bit larger for the C-H bonds as compared to the amine N-H
bonds.
Table 10 shows that amine N-H bond angles of triamterene have
relatively large errors in comparison with the other bonds. The errors in
bond angles (less than 6%) were smaller and less abundant in compar­
ison with the errors in bond lengths (which can easily exceed 10%).
Table 11 shows that B3LYP/def2-SVP bond lengths for furosemide
strongly disagree with the experimental data in the case of amine and
sulphonamide N-H bonds, carboxylic acid oxygen bonds, and cyclic C-H
bonds. The level of disagreement was numerically larger for amine N-H
bonds as compared to the C-H bonds.
Table 12 demonstrates that carboxylic acid oxygen bond angles of
furosemide have relatively large errors in comparison with the other
bonds. The errors in bond angles (less than 10%) were smaller and less
abundant in comparison with the errors in bond lengths (which can
exceed easily exceeds 10%)..
3.2. IR frequencies
According to Tables 13 and in IR frequencies, all calculations were
similar to BLYP/def2-SVP and so to each other. Among all approaches,
HF/def2-SVP was the least similar to BLYP/def2-SVP. Thus, HF differed
also from other approaches. Therefore, the preference of method can
influence the IR results for the salt molecule.
In the prediction of IR frequencies of triamterene (Table 14), all
methods/basis sets performed well. The best-performing approach was
B3LYP/6-31+G** while the performance of BLYP/6-31G was lower than
others.
Here, on the left side of Table 14, BLYP calculations accumulated at
the bottom of the table. This could be a sign of the remarkable impact of
the method preference. However, no such order was observed for the
right side of the table. Nevertheless, the method choice could be the
primary factor acting on the computation of IR frequencies of
triamterene.
In the prediction of IR frequencies of furosemide (Table 15), all
methods/basis sets performed well. The best-performing approach was
B3LYP/6-31G** while the performance of HF/6-31G was lower than
others. Moreover, HF calculations accumulated at the bottom of
Table 15 (both right and left sides). Therefore, the method choice ap­
pears as the major factor influencing the computation of IR frequencies
of furosemide.
In case all the observations done for IR frequencies of three of the
studied chemicals were summed up, it can be suggested that BLYP/6-
31G** could be the optimal approach, in terms of computational cost
and satisfactory accuracy. However, B3LYP/6-31G* appears to be the
most accurate one. Furthermore, the method preference seems to be the
11
Journal of the Indian Chemical Society 100 (2023) 101030
frontier factor influencing the IR frequencies of the studied molecules.
3.2.1. VMARD analysis
Table 16 shows that B3LYP/def2-SVP IR frequencies strongly
disagree with the experimental data in the case of amine N-H bonds,
benzene C-H bonds, cyclic N-H bonds, and cyclic C-N bonds. The level of
disagreement was numerically very large for amine N-H bonds as
compared to others.
Table 17 shows that B3LYP/def2-SVP IR frequencies strongly
disagree with the experimental data in the case of amine N-H bonds,
benzene C-H bonds, sulphonamide N-S bonds, and S-O bonds of sul­
phonamide. The level of disagreement was numerically very large for
amine N-H bonds, as compared to others.
3.2. NPA charges, reactivity indices, and dipole moment and
thermodynamic data
The method preference was the primary factor acting on the
computation of NPA charges of heteroatoms, reactivity indices, ther­
modynamic data, and dipole moment of three of the studied molecules.
For details, see Table_NPA.docx, Table_RI.docx, and Table_TD.
4. Discussion
Here, in this study, the performances of computational chemical
approaches on the prediction of molecular properties of triamterene,
furosemide, and triamterene-furosemide salt were investigated. Exper­
imental data for molecular geometries and IR frequencies were
compared to the theoretical ones. Both R2 and RMSE can be utilized as
tools to measure the similarity between two data sets. However, RMSE
concentrates on the numerical closeness of the compared data sets while
R2 reflects the similarity in terms of the trend. For IR frequencies the
trend of the data matters more than the numerical closeness so R2 is
handled to compare IR frequencies. Differently, it was observed that
both R2 [64] and RMSE [66] have been handled by chemists to compare
molecular geometry data (bond lengths and angles). Here, both R2 and
RMSE calculations were applied to the molecular geometry data.
According to R2 data, in the prediction of bond lengths of triamterene
and furosemide, all methods/basis sets performed well. The method
preference was the major factor acting on the accuracy of the compu­
tations. In the prediction of bond lengths of triamterene-furosemide salt,
except BLYP/6-31G** and BLYP/6-31G, all other methods/basis sets
performed somehow well. R2 data suggests that, in the case of triangular
angles, all methods failed in the successful prediction of the experi­
mental angles of three of the studied molecules. Based on the R2 data of
three of the studied chemicals, it can be suggested that BLYP/6-31G*
could be chosen as the optimal approach for the prediction of the
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 14
Correlation Coefficients for IR Frequencies of Triamterene. In this table, the green colour was used for high­
lighting the best-performing approach while the red colour was utilized for the worst-performing approach.

molecular geometries of the studied compounds, in terms of computa­
tional cost and satisfactory accuracy. However, B3LYP/6-31++G** was
the one correlating best with the experimental data.
According to RMSE data, in the prediction of bond lengths of tri­
amterene and furosemide, all methods/basis sets performed well. The
method preference appeared to be the major factor acting on the accu­
racy of the computations. In general, both HF was superior to B3LYP and
BLYP and HF/6-31G* emerged as the most accurate approach. All RMSE
values for the bond lengths of the studied compounds were greater than
0.05 and all RMSE values for the bond angles were less than 5. This could
be interpreted as a sign of the better functioning of all calculations in the
case of the bond angles, as compared to the bond lengths of the inves­
tigated compounds. Such a proposal is the just opposite of what it was
said for R2 data which suggests that the performance of the utilized
methods/basis sets were very bad for bond angles and they were very
good for the bond lengths. It was found the answers for the conflict, as
bonds and bond angles from B3LYP/def2-SVP calculation were indi­
vidually investigated (Tables 7, 8, 9, 10, 11, and 12). The errors in bond
angles were smaller (less than 10%) and less abundant in comparison
with the errors in bond lengths (which can exceed easily exceeds 10%) of

12
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 15
Correlation Coefficients for IR Frequencies of Furosemide. In this table, the green colour was used for high­
lighting the best-performing approach while the red colour was utilized for the worst-performing approach.

the three of the studied molecules. So, this indicates that the predictions
of the bond angles were more successful than those of bond lengths.
Therefore, it was concluded that R2 is not a good tool for the comparison
of molecular geometry data and its use should be avoided in the case of
molecular geometry data.
In the prediction of IR frequencies of triamterene and furosemide, all
methods/basis sets seem to perform well. Based on our observations of
IR frequencies of three of the studied chemicals, it can be suggested that
BLYP/6-31G** could be the optimal approach among the studied ap­
proaches, in terms of computational cost and satisfactory accuracy.
However, B3LYP/6-31G* provided the best correlation with the
experimental frequencies.
In this report, significant sites of the VMARD analysis for B3LYP/
def2-SVP frequencies were provided. R2 comparison of IR frequencies
suggested that nearly all methods performed well on the IR prediction
for the studied molecules. However, the VMARD analysis demonstrated
that there are significant site-specific problems with the computational
IR frequencies, in terms of their association with the experimental data.
Especially, amine/amide N-H bonds both from triamterene and furose­
mide were problematic in this sense. These problems were, as well,
shared by other methods/basis sets (for details, please visit Tri_IR and
Fur_IR Sheets in Supplementary_Material.xlsx), in addition to

13
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 16
Triamterene Frequencies Displaying Remarkable Differences Between the B3LYP/def2-SVP and Experimental Results (Frequencies in cm− 1). For the full version of the
table, see Table_VMARD.docx. For atom labels of triamterene, see Fig. 2.
KBr Disk Frequencies Nujol Mull Frequencies B3LYP/def2-SVP Frequencies Scaled Frequencies (0.966) Composition

3473 3473 3730.98 3604 -0.6952 (54.2%) BOND N11H26

+0.5728 (44.6%) BOND N11H25
3390 3389 3692.87 3567 -0.6898 (54.2%) BOND N13H29
+0.5828 (45.8%) BOND N13H30
3370 3370 3614.23 3491 +0.8199 (47.5%) BOND N12H27
+0.7902 (45.8%) BOND N12H28
3291 3286 3588.62 3467 -0.8702 (49.4%) BOND N11H25
-0.7280 (41.3%) BOND N11H26
3137 3127 3565.54 3444 +0.8522 (49.3%) BOND N13H30
+0.7264 (42.0%) BOND N13H29
2955 3201.44 3093 -0.8397 (46.8%) BOND C15H20
-0.3927 (21.9%) BOND C16H21
+0.2252 (12.5%) BOND C19H24
-0.1678 (9.3%) BOND C17H22
2925 3191.08 3083 +0.7286 (34.4%) BOND C17H22
-0.3671 (17.3%) BOND C15H20
+0.3349 (15.8%) BOND C18H23
+0.3134 (14.8%) BOND C16H21
-0.2441 (11.5%) BOND C19H24
2855 3179.05 3071 -0.6758 (35.8%) BOND C16H21
+0.6451 (34.2%) BOND C18H23
+0.2400 (12.7%) BOND C15H20
-0.1891 (10.0%) BOND C19H24
1679 1680 1682.45 1625 -0.4795 (17.1%) BOND C1 N2
+0.3159 (11.3%) BOND C1 N11
-0.2670 (9.5%) BOND N4 C5
1495 1495 1589.36 1535 -0.2394 (6.5%) BOND C3 N4
-0.2284 (6.2%) ANGLE H29 N13H30
-0.2233 (6.0%) BOND C1 C10
+0.2118 (5.7%) BOND C1 N2
-0.2111 (5.7%) BOND C3 N12
1287 1287 1364.48 1318 -0.3047 (8.7%) BOND C14C19
+0.2983 (8.5%) BOND C16C17
-0.2803 (8.0%) BOND C17C18
-0.2778 (7.9%) BOND C15C16
+0.2671 (7.6%) BOND C14C15
+0.2537 (7.2%) BOND C18C19

B3LYP/def2-SVP.
Actually, our group have heard rumours (Acknowledgements)
about the failure of computational chemistry approaches in the case of
nitrogen (specifically its bond with hydrogen). Moreover, we have
experienced problems with this issue. In our previous study on IDA, HF,
B3LYP, and semi-empirical calculations failed in the prediction of IR
frequencies. IDA is a small molecule with a secondary amine structure.
The disagreement between experimental and computational data was
very strong specifically in the frequencies higher than 2200 cm− 1. Now,
we guess that amine nitrogen (specifically its bond with hydrogen)
played a role in the problem there. In addition, we have observed some
problems with the N-H bond in the case of proton NMR chemical shifts
(unpublished, Study of benchmarking methodologies for 1H NMR
chemical shifts of thiazolidin-4-one derivative; A novel drug
candidate). In our new study, we studied 2-(Furan-2-ylmethylenehy­
drazono)-5-(3-methylbenzyl)thiazolidin-4-one with HF/6-31G, HF/6-
31G**, HF/def2-SVP, HF/def2-SVPD, HF/def2-TZVP, B3LYP/def2-SVP,
BLYP/def2-SVP, PBE0/def2-SVP, PBE/def2-SVP, TPSSh/def2-SVP,
TPSS/def2-SVP, and BP86/def2-SVP. However, none of the studied ap­
proaches could remediate the strong disagreement between the theo­
retical and experimental proton NMR chemical shift for the N-H site
(unpublished, Study of benchmarking methodologies for 1H NMR
chemical shifts of thiazolidin-4-one derivative; A novel drug
candidate). As opposed to such observations by our group, Karakaya
et al. successfully reached a good agreement between theoretical and
experimental IR and geometry data for the nitrogenous sites of the gli­
clazide molecule, by B3LYP and HF calculations [64]. However, our
group was not the single group that experiences problems with nitrogen
atoms [66,67,97].
While this study was being reviewed, our group studied more on IDA.
As a result, we suggested that the O-H and N-H bonds of an IDA molecule
could be linked to other IDA molecules’ N and O atoms via hydrogen
bonds, in the solid phase. Although probably single molecule compu­
tation significantly affected the results for IDA vibrational frequencies,
the error cannot be attributed only to this factor. Nevertheless, the failed
predictions for IDA appear to strongly associate with the performance of
computational chemistry approaches (unpublished, Benchmarking
computational chemistry approaches on iminodiacetic acid, Part
II).
N-H bonds are known to be strongly affected by hydrogen bonds
[98]. A similar situation could be observed also in the case of O-H bonds
and C-H bonds of benzene in case there is hydrogen bonding. Let’s as­
sume that N-H bond lengths in the triamterene side of the salt molecule
were wrongly predicted because of hydrogen bonding with neighbour­
ing salt molecules. Then, how the case of the N2-H41 bond (Fig. 1)
should be described? In triamterene-furosemide salt, N2-H41 and
N6-H54 bonds (Fig. 1) connect two sides of the molecule with hydrogen
bonds. According to Table 7, the N2-H41 bond was the most problematic
bond to be predicted. The error rate was 31.42%. N2-H41 bond already
owns a hydrogen bond with the neighbouring oxygens. Authors don’t
know how salt molecules interact with each other and the way they are
organized in the solid phase. Although such bonding, in the cif file of the
salt, was not observed, possibly, N2 may link to hydrogen from other salt
molecules. However, such bonding should not lead to an extreme change
causing a 31.42% prediction error. Such errors should be also related to
the performance of computations. In all B3LYP, HF, and BLYP calcula­
tions of the salt molecule, the N2-H41 bond length was predicted
wrongly (TFS_Geo Sheet in Supplementary_Material.xlsx).

14
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Table 17
Furosemide Frequencies Displaying Remarkable Differences Between the B3LYP/def2-SVP and Experimental Results (Frequencies in cm− 1). For the full version of the
table, see Table_VMARD.docx. For atom labels of furosemide, see Fig. 3.
KBr Disk Frequencies Nujol Mull Frequencies B3LYP/def2-SVP Frequencies Scaled Frequencies (0.966) Composition

3401 3401 3671.76 3547 -0.6433 (50.2%) BOND N11H31

+0.6372 (49.8%) BOND N11H32
3351 3352 3535.62 3415 +0.8091 (46.4%) BOND N11H32
+0.8038 (46.1%) BOND N11H31
3284 3287 3490.31 3372 +0.9841 (77.6%) BOND N2 H22
2955 3240.07 3130 +0.9776 (85.3%) BOND C5 H23
-0.1002 (8.7%) BOND C8 H24
2924 3091.4 2986 +0.7270 (53.5%) BOND C9 H30
-0.5824 (42.9%) BOND C9 H29
2854 3041.67 2938 -0.8344 (53.4%) BOND C9 H29
-0.6815 (43.6%) BOND C9 H30
1674 1674 1786.38 1726 +0.8360 (49.7%) BOND C18 O20
-0.1099 (6.5%) BOND C4 C18
+0.0980 (5.8%) BOND C1 N2
1509 1504 1624.69 1569 +0.3693 (13.1%) BOND C1 C4
-0.2981 (10.6%) BOND C5 C6
-0.2854 (10.1%) BOND C1 N2
+0.1856 (6.6%) ANGLE C1 N2 H22
-0.1618 (5.8%) BOND C1 C8
1353 1368.77 1322 +0.6001 (45.0%) BOND S3 O12
-0.5917 (44.4%) BOND S3 O14
1078 1149.33 1110 -0.7400 (32.4%) BOND S3 O14
-0.7366 (32.3%) BOND S3 O12
540 588.58 569 +0.5366 (10.5%) TORSION C4 C18 O19H28
+0.5355 (10.5%) OUT N2 C1 C4 C8
+0.4904 (9.6%) TORSION C5 C4 C18 O19
+0.3160 (6.2%) TORSION N2 C1 C4 C5
514 577.74 558 -0.4968 (8.8%) ANGLE C6 S3 O12
-0.4920 (8.7%) ANGLE C6 S3 O14
-0.4242 (7.5%) ANGLE O19C18 O20
+0.3445 (6.1%) BOND C4 C18

Although we heard some rumours, a report specifically emphasizing well-functioning of the approach of interest on the studied chemicals
the problems with nitrogen atoms discussed here could not be found. In and their investigated sites, the computation will be safer.
case there is, this study will be a good supplementation to that for all 5
molecules (the three being from this study + IDA + the thiazolidin-4-one 4.1. What are the possible solutions?
derivative) our group have studied in our past and current studies,
B3LYP, like other methods, did not end up with a trustable level of ac­ With inspiration from the work by Spinn et al. [67], we propose that
curacy for the nitrogenous sites. Amines cover a huge portion of organic the addition of an extra point charge nearby nitrogen atoms could be a
and biological chemistry and they play very significant roles in biolog­ good solution to the problems regarding N-H bonds. We did not have the
ical and environmental processes. So, in the case where N-H bonds play opportunity to try such a solution thus our group encourage readers to
significant roles and computations on nitrogen are problematic, there do it. If any of the readers go for it and achieves success, the authors will
could be wrong conclusions on biological and chemistry issues. In such a be happy to know it. In a such situation please communicate with our
situation, the use of computational chemistry could be a problem, rather group through the corresponding author.
than a remedy. Authors don’t know where such problems with nitrogen
originate from.
4.2. Other approaches and techniques
In the literature, some cases where computations for N-H bonds
failed can be found [68,69,99–104]. For example, the N-H bond
Authors could only study with HF, BLYP, and B3LYP methods, in this
stretching frequency for N-(2,5-dimethyl-4-nitrophenyl)-4-methylben
study. But, there is a need to measure the success of higher-level ap­
zenesulfonamide was predicted with a 5.2% error rate by
proaches like B2LYP, a double hybrid, and MP2. However, such ap­
B3LYP/6-311++G** approach while the error rate was around 39%
proaches can also fail like the methods studied, here. They may provide
when NMR shift for the nitrogen-bonded hydrogen was computed by
just fine-tuned versions of the results provided by HF and B3LYP. It is
B3LYP/6-311+G(2d,p) [103]. In support of this, Karrouchi et al. pre­
also a good idea to go for another benchmark study with three of the
dicted (by B3LYP/6-311++G**) NMR shifts of the nitrogen-linked hy­
studied molecules to test the performances of some new methods, such
drogens of (E)-N’-(4-methoxybenzylidene)-5-methyl-1H-pyrazole-
as GFN2-xTB, also known as XTB2 [106]. Further, some other ami­
3-carbohydrazide with 29.2% and 29.6% error rates [100]. Especially,
ne/amide molecules should be investigated. However, vibrational
the problem becomes more noticeable in the case of proton NMR shifts
spectrometry data for large molecules are crowded and complex. As
[97]. However, it is significant to emphasize that failure is not the case
opposed to this, computational and experimental NMR data is easier to
for every computation [105].
study. Therefore, this study could be followed by an NMR study, rather
Based on the above discussion, the authors propose that before deep
than an IR investigation.
investigations of the molecules by computational chemistry approaches,
it could be a good option to first investigate if the selected computational
5. Conclusion
approach is functioning sufficiently well on the chemical of interest. For
that, a comparison with any available experimental data, like crystal­
In this study, various commonly used methods/basis sets were
lography, vibrational spectrometry, bond lengths, bond angles, and/or
benchmarked mainly on molecular geometry and IR frequencies of tri­
NMR chemical shifts, could be useful. In the case there seems general
amterene, furosemide, and their salt. The method preference was the

15
M. Metin et al. Journal of the Indian Chemical Society 100 (2023) 101030

Fig. 1. Atom labels for triamterene-furosemide salt. The B3LYP/def2-SVP optimized structure was visualized by Avogadro software [95,96].

Fig. 2. Atom labels for triamterene. The B3LYP/def2-SVP optimized structure
was visualized by Avogadro software.
primary factor influencing the fate of the results for molecular geome­
tries, IR frequencies, NPA charges of heteroatoms, reactivity indices,
thermodynamic data, and dipole moment of three of the studied mole­
cules. The studied computational approaches generally performed well
in the IR frequency and molecular geometry prediction missions.
However, some site-specific problems were recognized. In the study,
amine and amide sites of the studied molecules were unsuccessfully
predicted, by the studied approaches, in terms of molecular geometry
and IR spectroscopy.
Although it was not mentioned in the article, there is one more
probability behind the observed disagreements between the experi­
mental and theoretical data. The experimental data could be the prob­
lematic side, rather than the theoretical results. Even though
experiments are commonly accepted as a reference, various natural and/
or human-sourced factors can affect the experimental results. Although
this is not the case mostly, this scenario should be also mentioned and
kept in mind.
We wish that this study will guide the scientists who computationally
investigate amines/amides and computational chemistry techniques.

Funding

Our group thanks The Ministry of Education, Culture, Sports, Sci­

ence, and Technology (MEXT) of Japan for their grant of a PhD schol­
arship to author M.M. while this study is being done.

Availability of data and material

All data generated or analysed during this study are included in this
published article and its supplementary files.

Code availability

Here, in this study, authors handled ORCA (version 4.2.1), Gabedit

(version 2.5.2), Avogadro (version 1.2.0), JANPA (version 2.02), vibA­
nalysis (version 1.2.2), and PyMOL (version 2.5.0). All of these programs
were freely and easily available from their website except ORCA. ORCA
was freely available to M.M. for academic purposes. In the enrichment
process of the introduction section ChatGPT [107] was utilized while
Grammarly [108] was handled for the improvement of the article’s
Fig. 3. Atom labels for furosemide. The B3LYP/def2-SVP optimized structure
was visualized by Avogadro software.
grammar and fluency.

16
M. Metin et al.
Authors’ contributions
The conception and design of the study: M.M. and T.K.
Acquisition of data, analysis, and interpretation of data: M.M.
Drafting the article: M.M.
Proofreading: T.O.
Declaration of competing interest
The authors declare that they have no conflict of interest.
Acknowledgements
The authors also want to thank Uezu Kazuya from The University of
Kitakyushu for sharing his experiences with computational chemistry.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.jics.2023.101030.
References
[1] C.J. Cramer, Essentials of Computational Chemistry: Theories and Models, John
Wiley & Sons, 2013.
[2] F. Jensen, Introduction to Computational Chemistry, John wiley & sons, 2017.
[3] H. Boulebd, Theoretical insights into the antioxidant activity of moracin T, Free
Radic. Res. 54 (4) (2020) 221–230.
[4] G. Palermo, M. De Vivo, Computational chemistry for drug discovery, in:
B. Bhushan (Ed.), Encyclopedia of Nanotechnology, Springer Netherlands,
Dordrecht, 2014, pp. 1–15, https://doi.org/10.1007/978-94-007-6178-0_
100975-1.
[5] G. Sliwoski, S. Kothiwale, J. Meiler, E.W. Lowe, Computational methods in drug
discovery, Pharmacol. Rev. 66 (1) (Jan. 2014) 334–395, https://doi.org/
10.1124/pr.112.007336.
[6] P.A. Townsend, M.N. Grayson, Density functional theory in the prediction of
mutagenicity: a perspective, Chem. Res. Toxicol. 34 (2) (Feb. 2021) 179–188,
https://doi.org/10.1021/acs.chemrestox.0c00113.
[7] E. Köse, M. Erkan Köse, S. Güneşdoğdu Sağdınç, Principal component analysis of
quantum mechanical descriptors data to reveal the pharmacological activities of
oxindole derivatives, Results in Chemistry 5 (Jan. 2023), 100905, https://doi.
org/10.1016/j.rechem.2023.100905.
[8] I.Y. Joel, et al., Insights into features and lead optimization of novel type 1½
inhibitors of p38α mitogen-activated protein kinase using QSAR, quantum
mechanics, bioisostere replacement and ADMET studies, Results in Chemistry 2
(Jan. 2020), 100044, https://doi.org/10.1016/j.rechem.2020.100044.
[9] J. Bhawsar, P.K. Jain, P. Jain, Experimental and computational studies of
Nicotiana tabacum leaves extract as green corrosion inhibitor for mild steel in
acidic medium, Alex. Eng. J. 54 (3) (Sep. 2015) 769–775, https://doi.org/
10.1016/j.aej.2015.03.022.
[10] G. Gece, The use of quantum chemical methods in corrosion inhibitor studies,
Corrosion Sci. 50 (11) (Nov. 2008) 2981–2992, https://doi.org/10.1016/j.
corsci.2008.08.043.
[11] A. Irfan, A. Mahmood, Designing of efficient acceptors for organic solar cells:
molecular modelling at DFT level, J. Cluster Sci. 29 (2) (Mar. 2018) 359–365,
https://doi.org/10.1007/s10876-018-1338-x.
[12] A. Irfan, A. Mahmood, Computational designing of low energy gap small
molecule acceptors for organic solar cells, Journal of the Mexican Chemical
Society 61 (4) (2017), https://doi.org/10.29356/jmcs.v61i4.461.
[13] J. Stevens, Virtually going green: the role of quantum computational chemistry in
reducing pollution and toxicity in chemistry, Physical Sciences Reviews 2 (7) (Jul.
2017), https://doi.org/10.1515/psr-2017-0005.
[14] J.J. Villaverde, C. López-Goti, M. Alcamí, A.M. Lamsabhi, J.L. Alonso-Prados,
P. Sandín-España, Quantum chemistry in environmental pesticide risk
assessment, Pest Manag. Sci. 73 (11) (2017) 2199–2202, https://doi.org/
10.1002/ps.4641.
[15] D. Xia, et al., Potential application of machine-learning-based quantum chemical
methods in environmental chemistry, Environ. Sci. Technol. 56 (4) (Feb. 2022)
2115–2123, https://doi.org/10.1021/acs.est.1c05970.
[16] R. Kumar Mandal, S. Ghosh, T. Pal Majumder, Comparative study between
degradation of dyes (MB, MO) in monotonous and binary solution employing
synthesized bimetallic (Fe-CdO) NPs having antioxidant property, Results in
Chemistry 5 (Jan. 2023), 100788, https://doi.org/10.1016/j.
rechem.2023.100788.
[17] M. Tuckerman, Statistical Mechanics: Theory and Molecular Simulation, Oxford
university press, 2010.
[18] M. Hochlaf, Advances in spectroscopy and dynamics of small and medium sized
molecules and clusters, Phys. Chem. Chem. Phys. 19 (32) (2017) 21236–21261.
17
Journal of the Indian Chemical Society 100 (2023) 101030
[19] H. Kim, M. Cho, Infrared probes for studying the structure and dynamics of
biomolecules, Chem. Rev. 113 (8) (2013) 5817–5847.
[20] Y. Fan, J. Ho, R.P. Bettens, Approximating Coupled Cluster level vibrational
frequencies with composite methods, J. Phys. Chem. 110 (8) (2006) 2796–2800.
[21] R.A. Friesner, Ab initio quantum chemistry: methodology and applications, Proc.
Natl. Acad. Sci. USA 102 (19) (2005) 6648–6653.
[22] S.S. Khire, N. Sahu, S.R. Gadre, Harnessing desktop computers for ab initio
calculation of vibrational IR/Raman spectra of large molecules, J. Chem. Sci. 130
(11) (2018) 1–7.
[23] T.M.A. Al-Shboul, et al., Synthesis, characterization, computational and
biological activity of some schiff bases and their Fe, Cu and Zn complexes,
INORGA 10 (8) (Aug. 2022), https://doi.org/10.3390/inorganics10080112. Art.
no. 8.
[24] I.B. Isik, N. Tekin, S.G. Sagdinc, The analyses of solvent effects on infrared spectra
and thermodynamic parameters, Hirshfeld surface, reduced density gradient and
molecular docking of ketoprofen as a member of nonsteroidal anti-inflammatory
drugs, J. Mol. Struct. 1250 (Feb. 2022), 131861, https://doi.org/10.1016/j.
molstruc.2021.131861.
[25] D. Rajaraman, L.A. Anthony, P. Nethaji, R. Vallangi, One-pot synthesis, NMR,
quantum chemical approach, molecular docking studies, drug-likeness and in-
silico ADMET prediction of novel 1-(2,3-dihydrobenzo[b][1,4]dioxin-6-yl)-2-
(furan-2-yl)-4,5-diphenyl-1H-imidazole derivatives, J. Mol. Struct. 1273 (Feb.
2023), 134314, https://doi.org/10.1016/j.molstruc.2022.134314.
[26] A. Mili, et al., Tautomerization of diazene<=>hydrazine via single proton
tautomerization, spectral, XRD/HSA-interactions, optical and DFT/TD-DFT of
new hydrazine ligand, J. Mol. Struct. 1272 (Jan. 2023), 134113, https://doi.org/
10.1016/j.molstruc.2022.134113.
[27] H. Zhang, H. Zhou, H. Qin, J. Liu, W. Fan, The mechanism for the oxidation of
benzoquinoline: a theoretical study, Combust. Flame 248 (Feb. 2023), 112553,
https://doi.org/10.1016/j.combustflame.2022.112553.
[28] A. Briš, Y. Murata, Y. Hashikawa, D. Margetić, Utilization of sym-tetrazines as
guanidine delivery cycloaddition reagents. An experimental and computational
study, J. Mol. Struct. 1272 (Jan. 2023), 134207, https://doi.org/10.1016/j.
molstruc.2022.134207.
[29] M.Z. Sadvakassova, et al., Rotational barriers in N-benzhydrylformamides: an
NMR and DFT study, Molecules 28 (2) (Jan. 2023), https://doi.org/10.3390/
molecules28020535. Art. no. 2.
[30] J.E.M. McGeoch, M.W. McGeoch, Chiral 480 nm absorption in the hemoglycin
space polymer: a possible link to replication, Sci. Rep. 12 (1) (Sep. 2022), https://
doi.org/10.1038/s41598-022-21043-4. Art. no. 1.
[31] S. K. Mishra, A. Parikh, K. Rangan, and A. K. Sah, “Crystal structure of N-(2-
Hydroxynapthylidene)-L-isoleucinyl-4,6-O-ethylidene-β-D-glucopyranosylamine
and an insight from experimental and theoretical calculations,” Cryst. Res.
Technol., vol. n/a, no. n/a, p. 2200209, doi: 10.1002/crat.202200209.
[32] C. Sudhakar, et al., Pharmacological and quantum chemical studies of 2-amino­
benzo[d]thiazol-3-ium 4-chlorobenzenesulphonate: synthesis, spectral, thermal
analysis and structural elucidation, Results in Chemistry 4 (Jan. 2022), 100442,
https://doi.org/10.1016/j.rechem.2022.100442.
[33] P.S. Idante, G.C. Apebende, H. Louis, I. Benjamin, U.J. Undiandeye, I.J. Ikot,
Spectroscopic, DFT study, and molecular docking investigation of N-(3-
methylcyclohexyl)-2-phenylcyclopropane-1-carbohydrazide as a potential
antimicrobial drug, J. Indian Chem. Soc. 100 (2) (Feb. 2023), 100806, https://
doi.org/10.1016/j.jics.2022.100806.
[34] F. Jensen, Unifying general and segmented contracted basis sets. Segmented
polarization consistent basis sets, J. Chem. Theor. Comput. 10 (3) (2014)
1074–1085.
[35] M.M. Francl, et al., Self-consistent molecular orbital methods. XXIII. A
polarization-type basis set for second-row elements, J. Chem. Phys. 77 (7) (1982)
3654–3665.
[36] W.J. Hehre, R. Ditchfield, J.A. Pople, Self—consistent molecular orbital methods.
XII. Further extensions of Gaussian—type basis sets for use in molecular orbital
studies of organic molecules, J. Chem. Phys. 56 (5) (1972) 2257–2261.
[37] A.M. Alsubaiyel, et al., Adsorption and electronic properties of pristine and Al-
doped C60 fullerenes using N2O molecule: a theoretical study, J. Mol. Liq. 369
(Jan. 2023), 120855, https://doi.org/10.1016/j.molliq.2022.120855.
[38] V. Sathya, V. Srinivasadesikan, L. Ming-Chang, V. Padmini, Highly sensitive and
selective detection of tryptophan by antipyrine based fluorimetric sensor, J. Mol.
Struct. 1272 (Jan. 2023), 134241, https://doi.org/10.1016/j.
molstruc.2022.134241.
[39] I.A. Udofia, et al., Experimental and theoretical calculation of pKa values of
substituted-2,4,6-trinitrodiphenylamines, J. Mol. Liq. 371 (Feb. 2023), 120926,
https://doi.org/10.1016/j.molliq.2022.120926.
[40] R. kerkour, N. Chafai, O. Moumeni, S. Chafaa, Novel α-aminophosphonate
derivates synthesis, theoretical calculation, Molecular docking, and in silico
prediction of potential inhibition of SARS-CoV-2, J. Mol. Struct. 1272 (Jan.
2023), 134196, https://doi.org/10.1016/j.molstruc.2022.134196.
[41] M. Metin, T. Kawano, T. Okobira, Benchmarking computational chemistry
approaches on iminodiacetic acid, J. Indian Chem. Soc. 100 (2) (Jan. 2023),
100895, https://doi.org/10.1016/j.jics.2023.100895.
[42] F. Neese, Software update: the ORCA program system—version 5.0, Wiley
Interdiscip. Rev. Comput. Mol. Sci. 12 (5) (2022) e1606.
[43] J.J. Stewart, MOPAC: a semiempirical molecular orbital program, J. Comput.
Aided Mol. Des. 4 (1) (1990) 1–103.
[44] J.M. Turney, et al., Psi4: an open-source ab initio electronic structure program,
WIREs Computational Molecular Science 2 (4) (2012) 556–565, https://doi.org/
10.1002/wcms.93.
M. Metin et al.
[45] R.M. Parrish, et al., Psi4 1.1: an open-source electronic structure program
emphasizing automation, advanced libraries, and interoperability, J. Chem.
Theor. Comput. 13 (7) (Jul. 2017) 3185–3197, https://doi.org/10.1021/acs.
jctc.7b00174.
[46] D.G.A. Smith, et al., PSI4 1.4: open-source software for high-throughput quantum
chemistry, J. Chem. Phys. 152 (18) (May 2020), 184108, https://doi.org/
10.1063/5.0006002.
[47] H.-J. Werner, et al., The Molpro quantum chemistry package, J. Chem. Phys. 152
(14) (Apr. 2020), 144107, https://doi.org/10.1063/5.0005081.
[48] H.-J. Werner, P.J. Knowles, G. Knizia, F.R. Manby, M. Schütz, Molpro: a general-
purpose quantum chemistry program package, WIREs Computational Molecular
Science 2 (2) (2012) 242–253, https://doi.org/10.1002/wcms.82.
[49] G.M.J. Barca, et al., Recent developments in the general atomic and molecular
electronic structure system, J. Chem. Phys. 152 (15) (Apr. 2020), 154102,
https://doi.org/10.1063/5.0005188.
[50] M.J. Frisch, et al., Gaussian 16, Rev. C.01, Wallingford, CT, 2016.
[51] T.D. Kühne, et al., CP2K: an electronic structure and molecular dynamics software
package - quickstep: Efficient and accurate electronic structure calculations,
J. Chem. Phys. 152 (19) (May 2020), 194103, https://doi.org/10.1063/
5.0007045.
[52] A. Shajan, M. Manathunga, A. Goetz, K. Merz, Geometry Optimization: A
Comparison of Different Open-Source Geometry Optimizers, 2023.
[53] P. Giannozzi, et al., Quantum espresso: a modular and open-source software
project for quantum simulations of materials, J. Phys. Condens. Matter 21 (39)
(Sep. 2009), 395502, https://doi.org/10.1088/0953-8984/21/39/395502.
[54] P. Giannozzi, et al., Advanced capabilities for materials modelling with Quantum
ESPRESSO, J. Phys. Condens. Matter 29 (46) (Oct. 2017), 465901, https://doi.
org/10.1088/1361-648X/aa8f79.
[55] P. Giannozzi, et al., Quantum ESPRESSO toward the exascale, J. Chem. Phys. 152
(15) (Apr. 2020), 154105, https://doi.org/10.1063/5.0005082.
[56] R. Ahlrichs, M. Bär, M. Häser, H. Horn, C. Kölmel, Electronic structure
calculations on workstation computers: the program system turbomole, Chem.
Phys. Lett. 162 (3) (Oct. 1989) 165–169, https://doi.org/10.1016/0009-2614
(89)85118-8.
[57] F. Furche, R. Ahlrichs, C. Hättig, W. Klopper, M. Sierka, F. Weigend, Turbomole,”
WIREs Computational Molecular Science 4 (2) (2014) 91–100, https://doi.org/
10.1002/wcms.1162.
[58] Y. Shao, et al., Advances in molecular quantum chemistry contained in the Q-
Chem 4 program package, Mol. Phys. 113 (2) (Jan. 2015) 184–215, https://doi.
org/10.1080/00268976.2014.952696.
[59] G. Kresse, J. Hafner, Ab initio molecular dynamics for liquid metals, Phys. Rev. B
47 (1) (Jan. 1993) 558–561, https://doi.org/10.1103/PhysRevB.47.558.
[60] G. Kresse, J. Furthmüller, Efficiency of ab-initio total energy calculations for
metals and semiconductors using a plane-wave basis set, Comput. Mater. Sci. 6
(1) (Jul. 1996) 15–50, https://doi.org/10.1016/0927-0256(96)00008-0.
[61] G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy
calculations using a plane-wave basis set, Phys. Rev. B 54 (16) (Oct. 1996)
11169–11186, https://doi.org/10.1103/PhysRevB.54.11169.
[62] E. Aprà, et al., NWChem: past, present, and future, J. Chem. Phys. 152 (18) (May
2020), 184102, https://doi.org/10.1063/5.0004997.
[63] R. Dennington, T. Keith, J. Millam, GaussView, version 5 (2009).
[64] M. Karakaya, M. Kürekçi, B. Eskiyurt, Y. Sert, Ç. Çırak, Experimental and
computational study on molecular structure and vibrational analysis of an
antihyperglycemic biomolecule: gliclazide, Spectrochim. Acta Mol. Biomol.
Spectrosc. 135 (Jan. 2015) 137–146, https://doi.org/10.1016/j.saa.2014.06.152.
[65] B. Peng, J.-R. Wang, X. Mei, Triamterene–furosemide salt: structural aspects and
physicochemical evaluation, Acta Crystallogr. B 74 (6) (Dec. 2018), https://doi.
org/10.1107/S2052520618013185. Art. no. 6.
[66] C. Kramer, A. Spinn, K.R. Liedl, Charge anisotropy: where atomic multipoles
matter most, J. Chem. Theor. Comput. 10 (10) (2014) 4488–4496, Oct, https://
doi.org/10.1021/ct5005565.
[67] A. Spinn, P.H. Handle, J. Kraml, T.S. Hofer, K.R. Liedl, Charge anisotropy of
nitrogen: where chemical intuition fails, J. Chem. Theor. Comput. 16 (7) (Jul.
2020) 4443–4453, https://doi.org/10.1021/acs.jctc.0c00204.
[68] M. Macit, H. Tanak, M. Orbay, N. Özdemir, “Synthesis, crystal structure,
spectroscopic characterization and DFT studies of bis[(1Z,2E)-N-(2,6-
diethylphenyl)-N′ -hydroxy-2-(hydroxyimino)acetimidamidato]nickel(II),”, Inorg.
Chim. Acta. 459 (Apr. 2017) 36–44, https://doi.org/10.1016/j.ica.2017.01.013.
[69] A. Fatima, et al., Quantum computational, spectroscopic, Hirshfeld surface,
electronic state and molecular docking studies on sulfanilic acid: an anti-bacterial
drug, J. Mol. Liq. 346 (Jan. 2022), 117150, https://doi.org/10.1016/j.
molliq.2021.117150.
[70] B.K. Sarojini, H.S. Yathirajan, B. Narayana, K. Sunil, M. Bolte, CCDC 657920:
Experimental Crystal Structure Determination, Cambridge Crystallographic Data
Centre, 2008, https://doi.org/10.5517/CCQ2M7M.
[71] M. Tutughamiarso, M. Bolte, CCDC 668589: Experimental Crystal Structure
Determination, Cambridge Crystallographic Data Centre, 2008, https://doi.org/
10.5517/CCQFQD7.
[72] L.L.C. Schrödinger, “The PyMOL molecular graphics system, version 1.8, https://
pymol.org, November, 2015.
[73] L.L.C. Schrödinger, The JyMOL Molecular Graphics Development Component,
Nov, 2015, Version 1.8.
[74] L. Schrodinger, The AxPyMOL Molecular Graphics Plugin for Microsoft
PowerPoint, Schrödinger, LLC, New York, NY, 2015 version 1.8.
[75] A.-R. Allouche, Gabedit—a graphical user interface for computational chemistry
softwares, J. Comput. Chem. 32 (1) (2011) 174–182.
18
Journal of the Indian Chemical Society 100 (2023) 101030
[76] F. Neese, The ORCA program system, WIREs Computational Molecular Science 2
(1) (2012) 73–78, https://doi.org/10.1002/wcms.81.
[77] F. Neese, Software update: the ORCA program system, version 4.0, WIREs
Computational Molecular Science 8 (1) (2018) e1327, https://doi.org/10.1002/
wcms.1327.
[78] F. Weigend, Accurate Coulomb-fitting basis sets for H to Rn, Phys. Chem. Chem.
Phys. 8 (9) (2006) 1057–1065.
[79] T. Clark, J. Chandrasekhar, G.W. Spitznagel, P.V.R. Schleyer, Efficient diffuse
function-augmented basis sets for anion calculations. III. The 3-21+ G basis set
for first-row elements, Li–F, J. Comput. Chem. 4 (3) (1983) 294–301.
[80] M.J. Frisch, J.A. Pople, J.S. Binkley, Self-consistent molecular orbital methods 25.
Supplementary functions for Gaussian basis sets, J. Chem. Phys. 80 (7) (1984)
3265–3269.
[81] R. Krishnan, J.S. Binkley, R. Seeger, J.A. Pople, Self-consistent molecular orbital
methods. XX. A basis set for correlated wave functions, J. Chem. Phys. 72 (1)
(1980) 650–654.
[82] A. McLean, G. Chandler, Contracted Gaussian basis sets for molecular
calculations. I. Second row atoms, Z= 11–18, J. Chem. Phys. 72 (10) (1980)
5639–5648.
[83] F. Weigend, R. Ahlrichs, Balanced basis sets of split valence, triple zeta valence
and quadruple zeta valence quality for H to Rn: design and assessment of
accuracy, Phys. Chem. Chem. Phys. 7 (18) (2005) 3297–3305.
[84] T.H. Dunning Jr., Gaussian basis sets for use in correlated molecular calculations.
I. The atoms boron through neon and hydrogen, J. Chem. Phys. 90 (2) (1989)
1007–1023.
[85] D.E. Woon, T.H. Dunning Jr., Gaussian basis sets for use in correlated molecular
calculations. III. The atoms aluminum through argon, J. Chem. Phys. 98 (2)
(1993) 1358–1371.
[86] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, A consistent and accurate ab initio
parametrization of density functional dispersion correction (DFT-D) for the 94
elements H-Pu, J. Chem. Phys. 132 (15) (2010), 154104.
[87] S. Grimme, S. Ehrlich, L. Goerigk, Effect of the damping function in dispersion
corrected density functional theory, J. Comput. Chem. 32 (7) (2011) 1456–1465,
https://doi.org/10.1002/jcc.21759.
[88] ORCA input library - ORCA common errors and problems. https://sites.google.
com/site/orcainputlibrary/orca-common-errors-and-problems. (Accessed 24
December 2022).
[89] CCCBDB Vibrational frequency scaling factors. https://cccbdb.nist.gov/vibnotes.
asp. (Accessed 16 January 2023).
[90] CCCBDB vibrational frequency scaling factors. https://cccbdb.nist.gov/vsfx.asp.
(Accessed 16 January 2023).
[91] AIST:Spectral database for organic compounds,SDBS. https://sdbs.db.aist.go.jp
/sdbs/cgi-bin/cre_index.cgi. (Accessed 24 December 2022).
[92] F. Teixeira, M.N.D.S. Cordeiro, “Improving vibrational Mode interpretation using
bayesian regression,”, J. Chem. Theor. Comput. 15 (1) (Jan. 2019) 456–470,
https://doi.org/10.1021/acs.jctc.8b00439.
[93] T.Y. Nikolaienko, L.A. Bulavin, D.M. Hovorun, JANPA: an open source cross-
platform implementation of the Natural Population Analysis on the Java
platform, Computational and Theoretical Chemistry 1050 (2014) 15–22.
[94] T.Y. Nikolaienko, L.A. Bulavin, Localized orbitals for optimal decomposition of
molecular properties, Int. J. Quant. Chem. 119 (3) (2019), e25798.
[95] M.D. Hanwell, D.E. Curtis, D.C. Lonie, T. Vandermeersch, E. Zurek, G.
R. Hutchison, Avogadro: an advanced semantic chemical editor, visualization,
and analysis platform, J. Cheminf. 4 (1) (2012) 1–17.
[96] “Avogadro: an open-source molecular builder and visualization tool.” [Online].
Available: http://avogadro.cc/.
[97] H.C. Da Silva, W.B. De Almeida, Theoretical calculations of 1H NMR chemical
shifts for nitrogenated compounds in chloroform solution, Chem. Phys. 528 (Jan.
2020), 110479, https://doi.org/10.1016/j.chemphys.2019.110479.
[98] G. Żuchowski, K. Zborowski, The influence of solvent molecules on NMR
spectrum of barbituric acid in the DMSO solution, Cent. Eur. J. Chem. 4 (2006)
523–532.
[99] Z. Afroz, M. Faizan, M.J. Alam, V.H.N. Rodrigues, S. Ahmad, A. Ahmad,
Synthesis, structural, hirshfeld surface, spectroscopic studies and quantum
chemical calculation of the proton transfer complex between 2-amino-4-hydroxy-
6-methylpyrimidine and salicylic acid, J. Mol. Struct. 1171 (Nov. 2018) 438–448,
https://doi.org/10.1016/j.molstruc.2018.06.020.
[100] K. Karrouchi, et al., Synthesis, structural, molecular docking and spectroscopic
studies of (E)-N’-(4-methoxybenzylidene)-5-methyl-1H-pyrazole-3-
carbohydrazide, J. Mol. Struct. 1225 (Feb. 2021), 129072, https://doi.org/
10.1016/j.molstruc.2020.129072.
[101] N. Dege, et al., Quantum computational, spectroscopic investigations on N-(2-((2-
chloro-4,5-dicyanophenyl)amino)ethyl)-4-methylbenzenesulfonamide by DFT/
TD-DFT with different solvents, molecular docking and drug-likeness researches,
Colloids Surf. A Physicochem. Eng. Asp. 638 (Apr. 2022), 128311, https://doi.
org/10.1016/j.colsurfa.2022.128311.
[102] K. Karrouchi, et al., Synthesis, X-ray structure, vibrational spectroscopy, DFT,
biological evaluation and molecular docking studies of (E)-N’-(4-(dimethylamino)
benzylidene)-5-methyl-1H-pyrazole-3-carbohydrazide, J. Mol. Struct. 1219 (Nov.
2020), 128541, https://doi.org/10.1016/j.molstruc.2020.128541.
[103] P.K. Murthy, et al., Synthesis, characterization and computational study of the
newly synthetized sulfonamide molecule, J. Mol. Struct. 1153 (Feb. 2018)
212–229, https://doi.org/10.1016/j.molstruc.2017.10.028.
[104] N. Dege, et al., The synthesis, characterization and theoretical study on nicotinic
acid [1-(2,3-dihydroxyphenyl)methylidene]hydrazide, Spectrochim. Acta Mol.
M. Metin et al.
Biomol. Spectrosc. 120 (Feb. 2014) 323–331, https://doi.org/10.1016/j.
saa.2013.10.030.
[105] N. Boukabcha, et al., Synthesis, crystal structure, spectroscopic characterization
and nonlinear optical properties of (Z)-N’-(2,4-dinitrobenzylidene)-2-(quinolin-8-
yloxy) acetohydrazide, J. Mol. Struct. 1194 (Oct. 2019) 112–123, https://doi.
org/10.1016/j.molstruc.2019.05.074.
[106] C. Bannwarth, S. Ehlert, S. Grimme, GFN2-xTB—an accurate and broadly
parametrized self-consistent tight-binding quantum chemical method with
19
Journal of the Indian Chemical Society 100 (2023) 101030
multipole electrostatics and density-dependent dispersion contributions, J. Chem.
Theor. Comput. 15 (3) (2019) 1652–1671.
[107] T. Brown, et al., Language models are few-shot learners, Adv. Neural Inf. Process.
Syst. 33 (2020) 1877–1901.
[108] My Grammarly - Grammarly. https://app.grammarly.com/. (Accessed 9 April
2023).