Chemical Physics Letters 599 (2014) 169174

Contents lists available at ScienceDirect

Chemical Physics Letters

journal homepage: www.elsevier.com/locate/cplett

Fourier type potential energy function for conformational change

of selected organic functional groups
Anita Rgyanszki a, Attila Surnyi a, Imre G. Csizmadia a,b, Andrs Kelemen c, Svend J. Knak Jensen d,,
Selma Yarligan Uysal e, Bla Viskolcz a
a

Department of Chemical Informatics, Faculty of Education, University of Szeged, Boldogasszony sgt. 6, H-6725 Szeged, Hungary
Department of Chemistry, University of Toronto, M5S 3H6 Toronto, Ontario, Canada
Department of Applied Informatics, Faculty of Education, University of Szeged, Boldogasszony sgt. 6, H-6725 Szeged, Hungary
d
Department of Chemistry, Langelandsgade 140, Aarhus University, DK-8000 Aarhus C, Denmark
e
Department of Chemistry, Faculty of Arts and Science, Eskisehir Osmangazi University, TR-26480 Eskisehir, Turkey
b
c

a r t i c l e

i n f o

Article history:
Received 30 January 2014
In nal form 12 March 2014
Available online 20 March 2014

a b s t r a c t
The energy changes associated with internal rotation of a functional group in a molecule depend on the
topology of the chemical environment. Energies obtained from electron structure and force-eld calculations have been analyzed by Fourier expansions. The ndings show that rotation around bonds connecting atoms without lone pairs can be described with a one term Fourier-series. In contrast, two or three
terms are needed if the connected atoms have lone pairs. The analysis inspires adoption of a simplied
Fourier expansion that reproduces the data well, suggesting that Fourier-type-series with few terms
are useful in describing any internal rotation analytically.
2014 Elsevier B.V. All rights reserved.

1. Introduction
1.1. Perspective
Protein folding is a century old problem. With the aid of conformational analysis the biological problem became a chemical problem. However, due to the complexity of the molecular system force
eld methods have been developed. Initially such molecular
mechanics methods were based on spectroscopic data and later
further development utilized optimized potential energy functions.
Yet the reliability of the analytic potentials used today may still be
an open question. On the one hand very extensive expansion may
be accurate but impractical to use and truncated expansion may be
practical but not sufciently accurate. Thus some compromise is
needed. The present Letter aims to analyse what functions are
practical to use yet may give accurate enough results for such
studies.
1.2. Historic background
For several decades, during the second half of the 20th century,
it was an unexplained phenomenon that the barrier to internal
rotation (torsion) about a CAC single bond in ethane (CH3ACH3)
Corresponding author. Fax: +45 8619 6199.
E-mail address: kemskj@chem.au.dk (S.J. Knak Jensen).
http://dx.doi.org/10.1016/j.cplett.2014.03.029
0009-2614/ 2014 Elsevier B.V. All rights reserved.

could be successfully calculated by just about any method,

whereas the lower anti barrier along the OAO single bond in
hydrogen peroxide (HOAOH) was very difcult to compute. The
apparent discrepancy was claried in 1978 when Cremer [1]
studied hydrogen peroxide with a large polarized basis set which
produced remarkably accurate results. Moreover, Peterson and
Csizmadia [2] demonstrated that all critical points of the
Potential Energy Hypersurface (PEHS) of three independence
variables of n-butane (H3CACH2ACH2ACH3) could be reproduced
at a modest level of theory. Further studies at that time [35]
indicated that the presence of lone pairs made the description
more difcult than the modest distortion of CAH bonding
electron pair.
The mathematical representation of the rotation potential using
analytic functions has its own story. The forecast of the internal
rotation and its energy changes is essential because it reect the
extent of intramolecular interactions [6]. The potential energy
curves and surfaces can be represented by suitably chosen analytical functions. Such functions can describe the atomic motion within a molecule, or within complexes formed by the interaction of the
molecules [7]. If this function is in a suitable form then it can be
used as an approximate analytic solution of the Schrdinger equation. Such equations can also be used for estimating rates of internal rotation [8,9].
Several functional groups have a unique internal rotational degree of freedom around a symmetric axis [10]. In such cases xed

170

A. Rgyanszki et al. / Chemical Physics Letters 599 (2014) 169174

bond angles may be assumed for the internal torsional movements,

thus only a single bond is engaged in the internal rotation [11] and
the potential energy depends on the dihedral angle only [12,13].
In the 1970s the Fourier-series were investigated by Radom
and Pople to determine the internal rotation around a single bond
[9,14]. In most cases Fourier-series, with relative few terms were
used for the tting of the curves.
In the late 1980s Chung developed a method for calculating
torsional energies with n-term Fourier-series. The full term expansion Fourier-series is recommended for higher accuracy in contrast
to minimum number of terms [13]. Nevertheless the practical
question still remains; how many terms are needed to obtain reliable results for r-bonds.

As a third question we wish to examine the rotation around a

single bond (CAO or OAO) in compounds containing more than
one r-bonds resulting in variable chemical environments. In these
case studies we would examine the effect

1.3. Scope
We are considering the dihedral torsion or internal rotation of
typical organic functional groups, like a methyl group (H3C).
The simplistic view-point is that the rotation is always the same
no matter wherever it is located in any molecule. A more sophisticated point of view is that the characteristic of the methyl rotation
is inuenced somewhat by its environment and most certainly by
the group it is attached to. The rst question we wish to investigate
is how large such a nearest neighbour interaction is in compounds
H3C-X, where X may be chosen horizontally or vertically along the
periodic table:

of the presence of a methyl group in an OAO torsion(III) and in a

H3CAO torsion(I) and the introduction of an OH group in a
H3CAO torsion(II).
In this Letter we explore rotations around bonds in small molecules. However, the perspective is to extend the study to sidechain rotations in amino acid residues. As a modest step in that
direction we have included the simplest chiral amino acid residue
(alanine) among the considered species.
2. Methods
Let DE be the energy change associated with rotation of a functional group relative to the lowest energy. DE can be expanded in a
Fourier-series with the general form:

DE DE 0
The second question we will explore is if it is possible to quantify
the changes in energy for the rotation about homonuclear diatomic
units such as (XX) as the number of lone pairs of X increases, as in
the series:


n 
X
j2pk/
j2pk/
aj cos
bj sin
360
360
j1

where DE0 is a constant, n is the number of terms in the expansion

and k is a constant indicating the periodicity of the rotating group
(i.e., k = 3 for a CH3 group).

Table 1
Fits of the Fourier series (1) to the rigid- and relaxed torsional potential energy curves (PEC), DE (kJ/mol), obtained by DFT [B3LYP/6-31G(d)] calculations for fully hydrogenated CX and X-Y systems, (X,Y = O, N, S). The data are arranged according to the number of terms, n, in the Fourier series.

DE 0
C-X family (n = 1)
CAC
Rigid
Relax
CAN
Rigid
Relax
CAO
Rigid
Relax
CAS
Rigid
Relax

6.68
5.86
4.05
4.97
1.96
2.97
1.34
2.72

a1

b1

a2

b2

a3

b3

R2

6.63
5.99
4.04
4.83
1.97
2.99
1.35
2.75

1.12  102
1.07  101
6.31  103
6.42  101
1.04  104
2.50  101
3.79  103
4.30  105

1
0.9936
1
0.9963
1
0.9996
0.9999
0.9968

X-Y family (n = 2)
OAO
Rigid
Relax
SAS
Rigid
Relax
SAO
Rigid
Relax

14.0
12.3
11.5
15.2
17.0
13.5

10.1
17.1
3.16
5.65
1.82
7.02

7.34  101
1.62
2.28  101
5.80  101
3.92  102
6.789  101

13.4
8.12
11.3
14.2
17.1
12.7

1.95
1.55
1.64
2.95
7.42  101
2.48

0.9997
0.9994
0.9997
0.9972
0.9997
0.9987

X-Y family (n = 3)
NAO
Rigid
Relax
NAS
Rigid
Relax
NAN
Rigid
Relax

27.0
28.7
24.8
19.6
13.9
12.1

2.07
3.33
2.17  101
4.31
18.2
12.2

1.03
1.37
3.46  101
5.44  101
1.34
2.52  101

14.9
11.5
8.81  101
17.0
7.39
11.7

4.24
2.06
1.82
8.16  101
1.09
4.85  101

22.6
1.97
3.41
4.42

2.54
3.47  101
7.66  101
2.74  101

0.9992
0.9994
0.9995
0.9994
0.9977
0.9904

A. Rgyanszki et al. / Chemical Physics Letters 599 (2014) 169174

DE was calculated in two ways, (i): by quantum mechanics

using the B3LYP/6-31G(d) implementation of the density functional theory [15] in the GAUSSIAN09 software package [16] a
method extensively used for small and medium sized molecules
and (ii) by force eld methods, Amber/GAFF and Amber99SB
[1719] tools routinely used for big molecules like proteins.
The side chain methyl rotation in N-acetyl Alanine-methylamide
was also studied at the B3LYP/6-31G(d) level of theory and by
the force eld method, Amber99SB. Using force-eld methods for
the molecules of the sizes studied here may seem like overkill
but is it useful for estimation of the associated energy changes
when proteins are studied. DE was calculated for a range of dihedral angles, /, in interval [0,2p].

171

and a1 for n = 1 inspired us to adopt a simple even tting function

for DE as

DE


n 
DEmax X
DEmax
j2pku

cos
n 1 j1
360
2j

where DEmax is determined from the t. For example, the H3C-X

family can be represented by an expansion with one term as

DE

DEmax DEmax

cos0:052u
2
2

For the X-Y family -excluding the NAN and NAS members- the
function is:

DE

DEmax DEmax
DEmax

cos0:0175  2u
cos0:0175u
3
2
22

In the case of NAN and NAS we obtain the function:

The calculation of DE was done as a Potential Energy Curve (PEC) of

rigid rotation, DErigid, as well as a PEC of relaxed rotation, DErelax.
The former is considered the less useful because in rigid rotation
the functional groups are rotated like mechanically rigid wheels
without geometrical adjustment. In contrast, in relaxed rotation
the groups involved in the internal rotation would have the opportunity to adjust their geometries during rotation. The PEC curve is
lower for the relaxed rotation than for the rigid rotation. However,
the energy lowering needs not be the same for the transition states
as for the potential minima. DErigid and DErelax were both calculated
at the B3LYP/6-31G(d) level. DErelax was also calculated using force
eld methods to assess the quality of the analysis of data obtained
from the two levels of description.
3. Results
A Fourier series (1) was tted to the DErigid and DErelax data for
the two families of compounds (H3C-X and X-Y where X, Y = O, N,
S). The results are shown in Table 1. It appears that the t is very
simple in the case of the H3C-X family as a single term (n = 1) will
present the data very well. We note that the b1 coefcient is much
smaller than the a1 coefcient, suggesting the PEC is close to being
an even function. It is also worth noting that a1 is very close to DE0.
The variation of DE with / is shown in Figure 1. The quality of the
t is estimated by calculating the deviation between the tted
function and the raw data as a function of the dihedral angle. We
nd the maximum deviation to be smaller than 0.1 kJ/mol.
For the second family (X-Y) two or three terms are required.
Again we note from Table 1 that the b-coefcients are much smaller than the a-coefcients. The DE plots for the X-Y family are
shown in Figure 2.
Inuences of substituents on DE were studied using the compounds I, II and III. The variations of DE in these cases are shown
in Figure 3. This gure shows, that H3CAO rotations (left hand side
of Figure 3) can be represented fairly accurately by single term
(n = 1) while the rotation about a OAO bond requires a two term
(n = 2) expansion in the Fourier-series.
It appears from Figure 2 that hydrazine has a very shallow minimum at the anti conformation [20], which increases the demand
for more terms in the Fourier expansion. Figure 4 illustrates how
the quality of the t increases for hydrazine as the number of terms
increases. The case of n = 3 does a perfect job indeed.
The observation that the b-coefcients in (1) are much smaller
than the a-coefcients in most cases along with the closeness of E0

Figure 1. Fits of (1) with n = 1 to the torsional potential energy curves (PEC), DE (kJ/
mol), for fully hydrogenated C-X families, (X = C, N, O, S). Open (blue) and lled
(red) symbols indicate data for rigid- and relaxed rotation, respectively. Data
obtained at the B3LYP/6-31G(d) level of theory. (For interpretation of the references
to colour in this gure legend, the reader is referred to the web version of this
article.)

172

A. Rgyanszki et al. / Chemical Physics Letters 599 (2014) 169174

Figure 2. Fits of (1) with n = 2, 3 to the torsional potential energy curves (PEC), DE (kJ/mol), for fully hydrogenated X-Y families, (X, Y = N, O, S). Open (blue) and lled (red)
symbols indicate data for rigid- and relaxed rotation, respectively. Data obtained at the B3LYP/6-31G(d) level of theory. (For interpretation of the references to colour in this
gure legend, the reader is referred to the web version of this article.)

DE

DEmax DEmax
DEmax

cos0:0175  2u
cos0:0175u
4
2
22
DEmax

cos0:0175  3u
23

In Table 2 we compare the results obtained from tting (2) to the

DE data derived from the relaxed dft calculations to the force eld
results.
4. Discussion
It appears from Table 1 that not all r-bonded functional groups
behave in an identical way during internal rotation (torsion). Some
of them like H3C-X exhibit rather high symmetry with three identical transition states [13]. These can be represented with a single
(n = 1) cosine function. Rotation about other r-bonds, where the
three TS are not identical or where only a pair of non-identical
TS is in the potential, require a longer Fourier expansion (n = 2 or
3). All of the potentials however were tted very accurately with
R2 values ranging from 0.9904 to 1 (Table 1).
The TS obtained by molecular mechanics (MM) simulations
agreed reasonably well (Table 2) for the C-X family. In the case
of the X-Y family noticeable deviations were observed in the cases
of the SAO, SAS, NAO and NAN linkages, 33.9 kJ/mol, -18.6 kJ/mol,
13.9 kJ/mol and 26.3 kJ/mol, respectively. This is understandable
because traditional force eld software were written to simulate
protein structure and protein folding. Thus discrepancies may be

expected in cases of bonds that are not common in proteins. Since

MM simulations, in general, aim to reproduce geometries and relative stabilities of minima, the achieved TS energy approximation
are really very good.

5. Conclusion
Fourier series, comprised by sin and cos terms are capable to t
any function. However, the potential energy functions considered
here are close to being even functions which makes a pure cos
expansion satisfactory. Depending on the structural complexity
of a given molecule, a 1-term or 2-term or 3-term expansion
turned out to be adequate, using only cos expansion. For CAC,
CAN, CAO and CAS bonds the 1-term expansion was accurate enough. For OAO, SAS and SAO bonds the 2-term expansion was satisfactory. In contrast, 3-term expansion was necessary to t the
potential energy function associated with the internal rotation
about the NAO, NAS and NAN bonds. The CAC rotation potential,
tted by Molecular Mechanics (MM), agreed very well with the
present Fourier series t for the methyl rotation in ethane as well
as in alanine-diamide.
The ultimate purpose is to build a very accurate multi variable
mathematical function for the Potential Energy Hypersurface of
exible molecules, such as peptides, of several internal bond-rotations. The present study made for a single dihedral rotation suggests that such an ultimate goal is achievable.

A. Rgyanszki et al. / Chemical Physics Letters 599 (2014) 169174

173

Figure 3. Fits of (1) to the potential energy curves for case studies of the environmental effects. Open (blue) and lled (red) symbols indicate data for rigid- and relaxed
rotation, respectively. Data obtained at the B3LYP/6-31G(d) level of theory. (For interpretation of the references to colour in this gure legend, the reader is referred to the
web version of this article.)

Figure 4. Fits of (1) to torsional potential energy curves (PEC), DE (kJ/mol), for hydrazine with n = 1, 2, and 3 terms. Open and lled symbols indicate data for rigid- and
relaxed rotation, respectively. Data obtained at the B3LYP/6-31G(d) level of theory.

174

A. Rgyanszki et al. / Chemical Physics Letters 599 (2014) 169174

Table 2
Comparison of the maxima (or transition states) obtained by tting (2) to the relaxed
DE data (kJ/mol) from DFT [B3LYP/6-31G(d)] calculations to the force eld data for
fully hydrogenated C-X and X-Y families, (X,Y = O, N, S). Data for N-acetyl Alaninemethylamide are also presented.
A: Higher TS
Method

B: Lower TS
DFT

Force eld

DFT

Force eld

C-X family (n = 1)
CAC
11.78
CAN
10.11
CAO
6.01
CAS
5.51
H3C-Ala
15.62

12.05
9.35
4.82
5.17
14.91

X-Y family (n = 2)
SAO
47.86
NAO
43.29
SAS
34.42
OAO
37.12

13.96
41.86
53.02
35.7

18.77
42.93
23.08
2.92

11.42
37.67
47.86
4.41

X-Y family (n = 3)
NAS
36.66
NAN
40.56

22.74
14.28

36.66
7.64

21.81
5.30

Acknowledgement
The authors would like to thank Balzs Jjrt for helpful contributions and Mikls Krsz for helpful discussions.
The authors acknowledge the nancial support within TMOP4.2.2.A-11/1/KONV-2012-0047 New functional material and their
biological and environmental answers, TMOP-4.2.2.C-11/1/

KONV-2012-0010 Supercomputer the national virtual laboratory, HUSRB/1002/214/193 Bile Acid Nanosystems as Molecule
Carriers in Pharmaceutical Applications and TMOP 4.2.4. A/211-1-2012-0001, National Excellence Program Elaborating and
operating an inland student and researcher personal support system, subsidized by the European Union and co-nanced by the
European Social Fund.
References
[1] D. Cremer, J. Chem. Phys. 69 (1978) 4440.
[2] M.R. Peterson, I.G. Csizmadia, J. Am. Chem. Soc. 100 (1978) 100.
[3] B. Viskolcz, R. Izsak, S.N. Fejer, I.G. Csizmadia, M. Szori, Int. J. Quantum Chem.
107 (2007) 1826.
[4] B. Viskolcz, M. Szori, I.G. Csizmadia, S.N. Fejer, Mol. Phys. 104 (2006) 795.
[5] B. Viskolcz, S.N. Fejer, I.G. Csizmadia, J. Phys. Chem. A 110 (2006) 3808.
[6] M. Head-Gordon, J.A. Pople, J. Phys. Chem. 97 (1993) 1147.
[7] G.H. Penner, J. Mol. Stuct. 137 (1986) 121.
[8] L. Radom, W.J. Hehre, J.A. Pople, W.A. Lathan, J. Am. Chem. Soc. 95 (1973)
693.
[9] L. Radom, J.A. Pople, W.J. Hehre, J. Am. Chem. Soc. 94 (1972) 2371.
[10] J. Schroderus, I. Ozier, N. Moazzen-Ahmadi, J. Chem. Phys. 115 (2001) 1392.
[11] R.H. Hunt, C.W. Peters, K.T. Hecht, R.A. Leacock, J. Chem. Phys. 42 (1965)
1931.
[12] A. Chung-Phillips, J.W.H. Kao, J. Am. Chem. Soc. 101 (1979) 1087.
[13] A. Chung-Philips, J. Chem. Phys. 88 (1988) 1764.
[14] L. Radom, J.A. Pople, J. Am. Chem. Soc. 94 (1970) 4786.
[15] A.D. Becke, J. Chem. Phys. 98 (1993) 5648.
[16] M.J. Frisch et al. GAUSSIAN 09, Revision A.02. Gaussian Inc, Wallingford, CT, USA,
(2009).
[17] J.M. Wang, R.M. Wolf, J.W. Caldwell, J. Comput. Chem. 25 (2004) 1157.
[18] D.A. Case et al., AMBER 12, University of California, San Francisco, 2012.
[19] A. Jakalian, D.B. Jack, C.I. Bayly, J. Comput. Chem. 23 (2002) 1623.
[20] L. Song, M. Liu, W. Wu, Q. Zhang, Y. Mo, J. Chem. Theory Comput. 1 (2005) 394.