Women in Science – New Frontiers of Research

Multiscale modelling: hybrid quantum

mechanics/molecular mechanics as an example
and some recent developments
Debashree Ghosh*
Physical and Materials Chemistry Division, CSIR-National Chemical Laboratory, Pune 411 008, India

looks at the average pattern of flow it is very similar to

Most of the physical phenomena are multiscale in
nature and therefore, to depict it properly one requires fluid dynamics through a constrained space.
multiscale modelling techniques, i.e. physical models Traditionally models have been developed that
that are accurate over multiple length and time scales. describe the physics associated with a single time and
The seminal work by Warshel and Levitt marks the length scale. However, since the real-world phenomena
beginning of hybrid quantum mechanics/molecular requires one to span over more than one scale, simulation
mechanics (QM/MM) method as a successful strategy or modelling techniques have evolved such that they can
towards the understanding of chemistry and physics handle multiple scales in a seamless fashion 1.
in condensed phases and especially in biological sys-
tems. Recently, these methods have been extended to
problems such as light–matter interaction, where the Analytical multiscale methods
QM sub-system is excited from the ground to the
excited states. The MM environment provides a field Analytical multiscale approaches have been developed
that changes the potential energy landscape of both for quite a few decades now. In this approach, one
the ground and excited states in a distinctly different derives the rules (or phenomenological models) of a
way. In this review, we discuss the general strategy of macroscale system from that of a microscale model.
multiscale modelling with emphasis on hybrid Renormalization group (RG) is a classic example of such
QM/MM and the recent developments in excited state
analytical multiscale method2. RG has been used to study
QM/MM methods.
phase transitions and critical phenomena. The need for
renormalization arises when there are two or more differ-
Keywords: Biological systems, hybrid quantum ent scales (time or length) and there is no clear demarca-
mechanics/molecular mechanics, multiscale. tion between them. RG is a technique to reduce the
degrees of freedom by integrating out less important
Multiscale modelling degrees of freedom in a consistent fashion.
Numerical RG technique was developed by Wilson and
MANY real-world phenomena are multiscale in nature, i.e. co-workers to solve the Kondo problem 3. The idea is eas-
they show complex behaviour that spans over a large ily understood in a 1D or 2D spin block model. Let us say
range of length and time scales. Length scales can span we have a 2D system of atoms (Figure 1) arranged in a
from few Å in case of bond lengths to a few microns in lattice and let us assume that only the nearest neighbours
case of living cell. On the other hand, time scales can interact with each other with a strength J (coupling
range from femtosecond (10–15 s) for bond vibrations to constant). Now, the Hamiltonian of the system can be
milliseconds (10–3 s) for protein folding. In most cases,
the different scales (microscopic and macroscopic when it
is the length scales) might be a continuum of varying
scales. The microscopic behaviour in most cases is spe-
cific and complex, while the macroscopic behaviour can
be defined by more general rules. For example, the flow
of traffic or pedestrians in a crowded street. If one tracks
the trajectory of individual pedestrians, they have a spe-
cific destination and, therefore, they move towards that
destination while avoiding the nearest neighbour interac-
tions or collisions. However, when one zooms out and
Figure 1. Renormalization group provides the tools to coarse grain
*e-mail: [email protected] interactions and have been successfully used in 2D lattice systems.

CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017 1455

 Special Section:
described as a function of J. However, we can divide the provides the basis for electrostatics. The dispersion–
system into super-blocks containing 2  2 atoms. Instead repulsion and cavitation energies are derived from
of defining the interactions in terms of atoms, we can the overlapping van der Waals radii of the solute
proceed to define it in terms of these super-blocks of at- atoms. The various types of PCM models are dielectric
oms where we have a renormalized coupling constant J. PCM (D-PCM)8, conductor like PCM (C-PCM or
Similar ideas have also been used to solve the many- COSMO)9 and other models such as SMx models10. The
body quantum mechanical wave function, where continuum models are computationally very fast but lack
renormalization is performed using information from the specific interactions since they do not have atomistic
density matrices. The density matrices describe the im- details.
portance of various electronic configurations and, there-
fore, can be used as a measure to eliminate less important
degrees of freedom. This is called density matrix renor- Molecular mechanics
malization group (DMRG) and has been developed for
strongly correlated electronic Hamiltonians4–6. Force field or molecular mechanics are empirical poten-
tials for interactions between atoms in the molecules of
the system. This empirical potential, once fitted with
Hierarchy of physical models
respect to either experimental data or more accurate
(quantum mechanical) calculations, can be used to study
Physical models have been built to deal with numerical
the dynamics of the system. This is the basic principle of
solutions in different length and time scales. As men-
classical molecular dynamics (MD), i.e. the force on the
tioned before, traditional modelling approaches focus on
system can be written as
one or other time and length scale (Figure 2).
 
F  V (r ), (1)
Continuum

where the potential V (r ) is the empirically fitted poten-
Materials are made up of atoms and molecules, but from tial. The motion of the atoms or molecules are derived
a length scales significantly larger than inter-atomic from the forces using Newton’s laws of motion
distances, it can be considered as continuous, and the
dynamics can be defined by fundamental physical laws  
F  ma , (2)
and differential equations derived from them. Continuum
models disregard the atomistic or discrete nature of mate- 
where m is the mass of the molecule or atom and a is the
rials.
acceleration.
Polarizable continuum model (PCM) is a popular con-
The force field or the fitted potential can be divided
tinuum model that is used to describe solvation effects7.
into bonded and non-bonded interaction terms
Tomasi and co-workers defined the free energy of solva-
tion as composed of electrostatics, dispersion-repulsion
VMM  Vbonded  Vnon-bonded . (3)
and cavity effects. The solvents do not have any atomistic
detail but the overall dielectric constant of the medium
Classical MD can be of two broad types depending on
the level at which the force fields are computed –
atomistic and coarse grain.

Coarse grain MD: The system size and time scales

required for understanding some processes precludes the
use of full atomistic details, but on the other hand might
require more degrees of freedom than a complete contin-
uum description. In such situations, force fields can be
developed which do not depend on the position of indi-
vidual atoms11,12, but rather depend on groups of atoms.
For example, in MARTINI (a popular coarse grain force
field)13, each residue on the backbone of an amino acid is
described as a single bead at the centre of mass of the
residue. This reduces the number of degrees of freedom
significantly from the full atomistic description. These
Figure 2. Different physical models that can be used for various time residues or beads interact with each other with bonded
and length scales. and non-bonded interaction energy terms.

1456 CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017

Women in Science – New Frontiers of Research
Atomistic MD: For a more detailed (accurate) descrip- Hˆ   E , (6)
tion of the system, one requires a full atomistic descrip-
tion, where the force field depends on the positions of all is solved to calculate the wavefunction and energy of the
the atoms in the system. The force field is again divided system.
into bonded and non-bonded interaction terms.
The bonded interactions can be classified into Semi-empirical: Semi-empirical methods are the most
affordable among the quantum mechanical methods.
Vbonded  Vbond  Vangle  Vdihedral Semi-empirical methods are based on mean field descrip-
tions, such as Hartree-Fock, with added approximations
and empirical data. Huckel’s theory and extended Huckel
  kb (r  r0 )2   k a (  0 )2
methods are successful examples of semi-empirical
bonds angles
methods. Such methods can deal with conjugated systems
1 and  electronic excited states. The  orbitals are not
  Vn [1  cos(n   )], (4) described explicitly in these methods and the  orbitals
torsions 2 are described by empirical parameters, i.e. the Hamilto-
nian Ĥ is a reduced empirical Hamiltonian. CNDO,
where the terms Vbond, Vangle and Vdihedral represent the MNDO, INDO, etc. are other examples of semi-empirical
interaction energies between covalently bonded atoms, quantum mechanical methods.
between orbitals in adjacent bonds, and that due to the
twisting of a bond, i.e. bond order. The bond and angle Density functional theory: Density functional theory
terms are most commonly parametrized as harmonic (DFT) is one of the most successful approaches in quan-
potentials and the dihedral energy term as a periodic tum mechanics used for a wide variety of systems. In
function such as cosine. The non-bonded interactions can order to reduce the degrees of freedom, instead of the
be classified as electrostatics and van-der Waals’ inter- complex n particle wavefunction ( (x1, y1, z1,…, xn, yn,
actions zn)), one solves the electron density ( ) which is a func-
tion of space (x, y, z) and time15. The Hohenberg-Kohn
Vnon-bonded  Velectrostatics  Vvdw
theorem states that the electron density (  (x, y, z)) deter-
mines all the properties of the ground state of the
N N  qi q j  C12,ij C6,ij  system 16,17.
     12  6  , (5)
i 1 j i 1  rij  rij rij  Thus, by focussing on the calculation of electron den-
  
sity (and not the wavefunction), an effective one-electron
Schrödinger equation can be derived and solved. The
where qi and rij denote the charge on atom i and the
problem is, thus, reduced to finding the kinetic and
distance between ith and jth atoms respectively. The C6
exchange correlation energies. There are a variety of
and C12 terms depend on the van der Waals radii of atoms
functional forms of the kinetic and exchange correlation
and the strength of van der Waals interaction energy be-
part of the Hamiltonian that gives rise to various DFT
tween the non-bonded atoms. Thus, electrostatics forms
functionals.
the long-range interactions while the van der Waals inter-
actions are more short range in nature. Force fields such Many-body theory: Many-body theory requires consid-
as AMBER, CHARMM, GROMOS, OPLS, etc. follow eration of the full electronic Hamiltonian and its solution.
this general form of parametrization or small changes the- It is the most rigorous form of quantum mechanical
reof. They form the general class of non-polarizable force calculations and is computationally expensive. One starts
fields14. with the mean field approximation to get a reference
wavefunction, i.e. Hartree-Fock theory. This reference
Quantum mechanics or electronic structure wavefunction is refined most often in a perturbative
approach giving rise to a variety of post Hartree Fock
While molecular dynamics in both atomistic and coarse methods, such as MP2, CCSD, etc.
grain scale are powerful tools for understanding large
scale dynamics of systems, it is still incapable of describ- Multi-physics models
ing chemical reactions. Chemical reactions deal with
breaking and forming chemical bonds and, therefore, the As mentioned before, there are many phenomena in
motion of electrons. For understanding such phenomena nature that requires consideration of a multitude of time
one requires not only atomistic but also electronic and length scales and, therefore, a single physical model
degrees of freedom, and therefore, quantum mechanical is not adequate to describe it. Therefore, multi-physics or
or electronic structure methods, i.e. Schrödinger equa- multiscale models have been developed. Multiscaling can
tion be achieved in sequential and concurrent fashion.
CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017 1457
 Special Section:
 Sequential multiscale methods: In sequential multis- HQM/MM(sub) = HMM (full system)
caling, one has a macroscale model, the parameters
for which are pre-computed using microscale models. + [HQM (active site) – HMM (active site)], (7)
This is essentially parameter-passing and is conceptu-
ally simple since it does not require consideration of where the subscript QM and MM denotes the levels of
two physical models in the same computation. theory used in the calculation. Since QM calculation can-
 Concurrent multiscale methods: In concurrent multis- not be performed on the full system, that is tackled at the
caling, the macro- and microscale are treated together MM level of theory. To look at the phenomena in the ac-
in a hybrid framework and the interaction between tive site that requires electronic degrees of freedom, QM
them is computed on the fly. Most the multiscale level of theory is used. Finally to remove the double
methods such as hybrid quantum mechanics/molecular counting, MM level calculation on the active site is per-
mechanics (QM/MM) discussed in this article are formed and subtracted. Thus, one needs to perform three
concurrent multiscale methods. decoupled calculations – (i) full system at the MM level
of theory; (ii) active site at the QM level of theory, and
(iii) active site at the MM level of theory. This sort of
Domain decomposition methods
embedding, while reasonable for decomposition of
energy, is not a good way of calculating the properties of
The idea of domain decomposition methods is often used
the system since the QM wavefunction is not modified by
in multiscale methods. The system of interest is divided
the field due to the MM environment. An extension of the
into domains or sub-systems. The properties of sub-
subtractive embedding, where more than 2 levels of
systems are then calculated separately and at different
theory are used is called ONIOM20.
levels of theory. Domains can either be overlapping or
non-overlapping. Hybrid QM/MM belongs to the general
class of domain decomposition methods1,18,19 .

Hybrid QM/MM

Proteins are typically made up of 1000 s of atoms and the

system is even larger when one considers the surrounding
water molecules. All these atoms and molecules contribute
to the properties of the protein, in both ground and ex-
cited states. Therefore, for treatment of proteins and other
biological systems, classical force fields are required.
However, chemists have always been fascinated by re-
actions in nature, i.e. in proteins and for that one needs to
consider the electronic structure of the bio-system. There-
fore, for treatment of chemical reactions in condensed
phases, hybrid treatment is required. The important parts
of the system, i.e. the active site, requires QM treatment.
In the active sites the electronic degrees of freedom are
important for phenomena such as bond breaking or for-
mation, or electron transfer processes. The rest of the sys-
tem is viewed in a more spectator-like fashion, i.e. it
provides an effective field to the active site. Figure 3. Hybrid QM/MM scheme shown in green fluorescent pro-
tein (GFP). The GFP chromophore lies at the centre of barrel formed by
Thus, in such active site-driven processes in proteins, a -sheets. The chromophore (which absorbs visible and fluoresces) is
small part of the system is treated at a higher level of the- treated quantum mechanically, while the rest of the system is treated at
ory, typically a quantum-mechanical method, while the a molecular mechanical level.
rest of the system is treated at a lower level of theory,
typically a molecular mechanical method (Figure 3 shows
such a division of a enzymatic reaction where only the
active site is treated quantum mechanically). Hybrid
QM/MM embedding can be achieved in two ways –
subtractive and additive.
1. Subtractive embedding: In subtractive embedding
(shown in Figure 4), the total QM/MM Hamiltonian of a
system can be written as Figure 4. Subtractive scheme of QM/MM embedding.

1458 CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017

Women in Science – New Frontiers of Research
2. Additive embedding: In the additive scheme of Electronic embedding
embedding, one writes the total Hamiltonian of the sys-
tem as To alleviate the deficiency of mechanical embedding, in
electronic embedding, the effect of charges in the MM
HQM/MM(sub) = HQM (active site) region on the QM charge density is accounted for. In
other words, the effective field due to MM region on the
+ HMM (environment) + HQM/MM, (8) QM region is included as a perturbation to the Hamilto-
nian of the QM part. Therefore, in the electronic embed-
where the HQM/MM denotes the interactions between ding, although the QM/MM interaction energy takes a
the QM and MM sub-systems, i.e. the active site and the similar form as eq. (9), the charge density  (r) is cor-
environment. Thus, it is the proper implementation of rected due to the effect of MM environment.
the QM and MM subsystem interactions that is the most Furthermore, since the Hamiltonian is changed pertur-
crucial and complicated part. Depending on the different batively due to the presence of the MM region
methods of handling the interaction terms there are three
different approaches – mechanical embedding, electronic Hˆ   Hˆ QM  VˆQM/MM . (10)
embedding and polarization embedding.
The charge density  (r) obtained from solving this Ĥ 
Additive scheme of hybrid QM/MM gives access to all the perturbatively corrected properties
of the system.
Mechanical embedding
Polarization embedding
In case of mechanical embedding the QM and MM sub-
systems, i.e. HQM and HMM are solved separately at dif- While electronic embedding accounts for the effect of
ferent levels of theory. The interaction between the QM MM region on the QM region, the QM region can also
and MM sub-systems is added as energy corrections at affect the MM region which might self-consistently
the end of the calculation, as Coulomb interaction energy change the QM Hamiltonian. This self-consistent effect
between the charge density in the QM region and the is called polarization embedding. There are various
point charges in the MM region approaches of polarization embedding and this continues
to be an active field of reach. Drude oscillator model,
_  (r )qi fluctuating charge model, induced dipoles, etc are some
E QM/MM =  dr , (9) of the polarization embedding methods. Effective frag-
ri
ment potential (EFP)21–23 and polarizable embedding
(PE)24 are two sophisticated polarizable force fields,
where  (r) denotes the charge density of the QM region
which have been used for both ground and excited state
at point r, qi denotes the charge of the ith point in MM
properties in the hybrid QM/MM framework.
region and ri denotes the distance from the ith MM point
In all the embedding schemes, a crucial component for
and position r in the QM region. The charge density (r)
the proper functioning of the hybrid QM/MM formalism
of the QM region and qi are calculated separately and do
is the accurate treatment of boundary atoms, i.e. the at-
not affect each other. This is the major drawback of this
oms that are very near to the QM and MM boundaries.
scheme.
Since, in the case of MM atoms, the force field is
Since this is a correction to the energy and not the
depicted as charges and higher multipoles (permanent or
Hamiltonian, this approach suffers from the same defi-
induced), it can cause unphysical electric fields on the
ciencies of the subtractive scheme. Since the Hamiltonian
nearest neighbouring QM atoms and, therefore, charge
of the QM region is not changed due to the QM/MM in-
damping schemes are often used. This situation is further
teractions, the wavefunction does not change due to the
complicated in a biological system, such as protein,
interactions and, therefore, properties apart from energy
where the QM boundary atom and MM boundary atom
cannot be correctly predicted with this approach.
are bonded to each other. This requires satisfying the
valency of the boundary QM atom. Methods such as link
atom (LA)25, and molecular fractionation with conjugate
capping (MFCC)26 have been developed for this purpose.

Enzyme catalysis with hybrid QM/MM

Hybrid QM/MM as a viable multiscale method was first

Figure 5. Additive scheme of QM/MM embedding. formulated in the seminal work by Warshel and Levitt1.
CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017 1459
 Special Section:
In their work, they studied the cleavage of glycosidic The Coulomb potential of the EFs includes charges,
bond by lysozyme and the stability of the carbonium ion dipoles, quadrupoles and octupoles, and fitted using dis-
that is formed in the reaction. The main idea of the work tributed multipole analysis. Apart from static multipoles,
is that the enzyme provides an environment that gives an polarizability is an important factor especially for excited
effective electrostatic field, thereby preferentially stabi- state properties. Polarizability is the self-consistent
lizing the transition state and increasing the speed of the change in charge density of one fragment due to the pre-
reaction. The main reaction site is treated quantum- sence of other fragments. In EFP, the polarizability is
mechanically since it involves a bond breaking and re- estimated as the self-consistently calculated induced
quires electronic degrees of freedom. It was shown in this dipole moments
study that the effective field provided by the environment  
need not be described at the quantum mechanical level    F , (12)
for understanding the mechanism of catalysis. 
Since then numerous studies have been carried out on where  denotes the polarizability tensor and F is the
elucidating the mechanisms of various enzyme catalytic field due to the rest of the multipoles. The dispersion
reactions in this hybrid framework. interaction can be expanded in the London dispersion
series, where only the first term is retained in EFP, i.e. it
takes the C6/R 6 form. Exchange repulsion, which has its
New developments origin in the Pauli exclusion principle, is an inherently
quantum mechanical part of the interaction. It is also the
Excited states with QM/MM most expensive part of the EFP calculation and depends
on the overlap integrals of localized molecular orbitals
Understanding the relative changes between ground and (LMOs).
excited states is crucial to the exploration of light–matter In the hybrid QM/MM framework with EFP as the MM
interactions. Light–matter interactions are important for method of choice, the Hamiltonian can be written as
processes such as photo-synthesis, light harvesting and so
on. Excited state processes are also important for under- Hˆ QM/EFP   |Hˆ QM  VCoul  VPol |    EEFP EFP , (13)
standing biological processes, such as chemiluminis-
cence. Given the vast number of biological processes that where the Hamiltonian of the QM part is perturbed by the
involve both ground and excited states of the system, it is Coulomb and polarization terms of the EFs in the envi-
instrumental to develop computational methods that are ronment. Since polarization is an inherently many-body
capable of dealing with both ground and excited states of property, it is solved self-consistently. The solution of
large systems. this equation would give a wavefunction that is corrected
While hybrid QM/MM methods have been developed for the presence of the environment. This is the ground
quite extensively over the last three decades for chemical state QM/EFP wavefunction.
reactions in ground state and especially for enzyme The charge density of the excited state wavefunction is
catalysis, its use for the understanding of excited state
different from the ground state and, therefore, the many-
processes and spectroscopic measurements in condensed body effect that it exerts on environment is different.
phases is a more recent phenomena. It requires the use of Therefore, the polarization due to excited state can be
not only state-of-the-art quantum mechanical methods for drastically different from the ground state. This is espe-
excited states, but also its seamless interface with MM cially true in case of charge transfer excited states. To
methods. Since the charge densities of ground and excited account for that, an added perturbative polarization cor-
states can be quite different, especially for charge transfer rection is included for the excited state that depends on
excited states, the field due to MM region can also be
the induced dipole moments on the environment (MM
significantly altered in the excited state, thus, requiring a
region) at the excited state.
proper description of the polarizability of MM region.
Using this formalism, we have been able to accurately
Effective fragment potential (EFP) is one such polariz-
predict the ionization energies and excitation energies of
able and sophisticated MM method22,23. It is derived from
biological systems such as DNA bases, GFP chromopho-
an a priori ab initio calculation on a fragment (typically a
res etc in solvation27–29.
solvent molecule) and is therefore, non-empirical. EFP was
originally developed and implemented in GAMESS to
study water. It was later extended to any general molecule. Redox potential with QM/MM
The potential due to the effective fragments is divided
into four terms – electrostatics (Coulomb), polarization, While excitation and ionization energies are important
dispersion and exchange repulsion properties, experimentally it is easier to observe quanti-
ties such as redox potential. Therefore, it is important to
EEFP – EFP = E Coul + EPol + E Disp + EEx – rep. (11) estimate and predict redox potentials. Traditionally redox

1460 CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017

Women in Science – New Frontiers of Research
potentials are calculated by constructing the Hess’s cycle, fields have been developed over the last few decades.
calculating the ionization energies in gas phase (which With the increase in computational power these are
requires a purely quantum mechanical level of theory) increasingly being used to solve chemical and physical
and the solvation energies using continuum models. real world problems. Hybrid QM/MM is one such highly
However, continuum models lack the description of successful technique for solving chemical problems rang-
specific interactions such as H-bonds which are crucial ing from enzyme catalytic reactions to excited state phe-
for the accurate estimation of solvation energies. There- nomena in more recent times.
fore, to achieve the quantitative accuracy in the redox
potential calculations, one requires a hybrid QM/MM 1. Warshel, A. and Levitt, M., Theoretical studies of enzymic reac-
formalism with discrete atoms in the MM region and pre- tions: dielectric, electrostatic and steric stabilization of the
ferably polarization. Thus, we have used hybrid QM/EFP carbonium ion in the reaction of lysozyme. J. Mol. Biol., 1976,
103, 227–249.
as described in the previous section. Furthermore, the
2. Kadanoff, L. P., Scaling laws for ising models near high tc. Phys-
total free energy change of the electron transfer process ics, 1966, 2, 263–272.
can be written as 3. Wilson, K. G., The renormalization group: critical phenomena and
the kondo problem. Rev. Mod. Phys., 1975, 47, 773–840.
1 E 4. White, S. R. and Noack, R. M., Real space quantum renormaliza-
G   ln exp , (14) tion group. Phys. Rev. Lett., 1992, 68, 3487–3490.
k BT k BT i 5. Chan, G. K.-L. and Head-Gordon, M., Highly correlated calcula-
tions with a polynomial cost algorithm: a study of density matrix
where i denotes ensemble average over state i. Using renormalization group. J. Chem. Phys., 2002, 116, 4462–4476.
linear response analysis (LRA), it can be reduced to 6. Ghosh, D., Hachmann, J., Yanai, T. and Chan, G. K.-L., Orbital
optimization in density matrix renormalization group, with appli-
1 cations to polyenes and -carotene. J. Chem. Phys., 2008, 128,
G  ( VIE Red  VEA Ox ), (15) 144117(1)–14417(14).
2
7. Tomasi, J., Mennucci, B. and Cances, E., The IEF version of the
where VIE Red denotes the ionization energy ensemble PCM solvation method: an overview of a new method addressed
to study molecular solutes at the QM ab initio level. J. Mol.
averaged at the reduced state and VEAOx denotes the
Struct. Theochem., 1999, 464, 211–226.
electron affinity ensemble averaged at the oxidized state. 8. Cossi, M., Barone, V., Mennucci, B. and Tomasi, J., Ab initio
Both these quantities can be calculated in the hybrid study of ionic solutions by a polarizable continuum dielectric
QM/EFP framework described above. Using QM/EFP model. Chem. Phys. Lett., 1998, 286, 253–260.
along with LRA we have been able to predict the redox 9. Cossi, M., Rega, N., Scalmani, G. and Barone, V., Energies, struc-
tures, and electronic properties of molecules in solution with the
potentials of small molecules very accurately30. The
C-PCM solvation model. J. Comput. Chem., 2003, 24, 669–681.
details of the excited state QM/EFP developments are 10. Cramer, C. J. and Truhlar, D. G., Implicit solvation models: equi-
explained by Ghosh31. libria, structure, spectra, and dynamics. Chem. Rev., 1999, 99,
2161.
11. Rudd, R. E. and Broughton, J. Q., Coarse-grained molecular
Complex biological systems dynamics and the atomic limit of finite elements. Phys. Rev. B,
1998, 58, 5893–5896.
Much of the on-going research is towards developing hy- 12. Ensing, B., Nielsen, S. O., Moore, P. B. and Parrinello, M.,
Energy conservation in adaptive hybrid atomistic/coarse-grain
brid QM/MM formalisms for complex biological systems. molecular dynamics. J. Chem. Theory Comput., 2007, 3, 1100–
Molecular tailoring approach (MTA)32, molecular frac- 1105.
tionation by conjugate capping (MFCC)26, fragment 13. Marrink, S. J., Risselada, H. J., Yefimov, S., Tieleman, D. P. and
molecular orbitals (FMO)33, and molecule in molecule34 de Vries, A. H., The martini force field: coarse grained model for
are some examples of the divide and conquer approaches biomolecular simulations. J. Phys. Chem. B, 2007, 111, 7812–
7814.
that have been extremely successful. MFCC-like ap- 14. Jorgensen, W. L., Maxwell, D. S. and Tirado-Rives, J., Develop-
proaches have also been developed for hybrid QM/EFP ment and testing of the OPLS all-atom force field on conforma-
very recently35. Other fields of development are towards tional energetics and properties of organic liquids. J. Am. Chem.
faster and more accurate polarizable force fields. The Soc., 1996, 118, 11225–11236.
proper treatment of boundary between QM and MM 15. Parr, R. G., Density functional theory of atoms and molecules. In
Horizons of Quantum Chemistry, Springer, pp. 5–15.
regions, where they are covalently bonded, is another 16. Hohenberg, P. and Kohn, W., Inhomogeneous electron gas. Phys.
developing field. Methods such as link atom scheme25 Rev. Lett., 1964, 136, 864–871.
and frozen molecular orbitals (FMO)33 are noteworthy 17. Kohn, W. and Sham, L. J., Self-consistent equations including
among these. exchange and correlation effects. Phys. Rev. Lett., 1965, 140,
1133–1138.
18. Lin, H. and Truhlar, D. G., QM/MM: what have we learned, where
Conclusion are we, and where do we go from here? Theor. Chem. Acc., 2007,
117, 185–199.
Since the processes in nature are multiscale in nature, 19. Senn, H. M. and Thiel, W., QM/MM methods for biological sys-
multi-physics methods both in analytical and numerical tems. Topics Curr. Chem., 2007, 268, 173–290.

CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017 1461

 Special Section:
20. Dapprich, S., Komaromi, I., Byun, K. S., Morokuma, K. and 29. Bose, S., and Ghosh, D., Effect of solvation on electron detach-
Frisch, M. J., A new oniom implementation in Gaussian. Part I. ment and excitation energies of a green fluorescent protein chro-
the calculation of energies, gradients, vibrational frequencies and mophore variant. J. Phys. Chem. B, 2016, 120, 4410–4420.
electric field derivatives. J. Mol. Struct. Theochem., 1999, 461, 30. Ghosh, D., Roy, A., Seidel, R., Winter, B., Bradforth, S. E. and
1–21. Krylov, A. I., A first-principle protocol for calculating ionization
21. Gordon, M. S., Freitag, M. A., Bandyopadhyay, P., Jensen, J. H., energies and redox potentials of solvated molecules and ions:
Kairys, V. and Stevens, W., The effective fragment potential Theory and application to aqueous phenol and phenolate. J. Phys.
method: a QM based MM approach to modelling environmental Chem. B, 2012, 116, 7269–7280.
effects in chemistry. J. Phys. Chem. A, 2001, 105, 293–307. 31. Ghosh, D., Hybrid equation-of-motion coupled-cluster/effective
22. Ghosh, D. et al., Non-covalent interactions in extended systems fragment potential method: a route toward understanding photo-
described by the effective fragment potential method: theory and processes in the condensed phase. J. Phys. Chem. A., 2017,
application to nucleobase oligomers. J. Phys. Chem. A, 2010, 114, 121(4), 741–752.
12739–12754. 32. Sahu, N. and Gadre, S. R., Molecular tailoring approach: a route
23. Ghosh, D. et al., Effective fragment potential method in Q-Chem: for ab initio treatment of large clusters. Acc. Chem. Res., 2014, 47,
A guide for users and developers. J. Comput. Chem., 2013, 34, 2739–2747.
1060–1070. 33. Fedorov, D. G. and Kitaura, K., Extending the power of quantum
24. Sneskov, K., Schwabe, T., Kongsted, J. and Christiansen, O., The chemistry to large systems with the fragment molecular orbital
polarizable embedding coupled cluster method. J. Chem. Phys., method. J. Phys. Chem. A, 2007, 111, 6904–6914.
2011, 134, 104108(1)–104108(18). 34. Mayhall, N. J. and Raghavachari, K., Molecules-in-molecules: an
25. Ferre, N. and Olivucci, M., The amide bond: pitfalls and draw- extrapolated fragment-based approach for accurate calculations on
backs of the link atom scheme. J. Mol. Struc. Theochem., 2003, large molecules and materials. J. Chem. Theory Comput., 2011, 7,
632, 71–82. 1336–1343.
26. Zhang, D. W. and Zhang, J. Z. H., Molecular fractionation with 35. Gurunathan, P. et al., The extension of the effective fragment
conjugate caps for full quantum mechanical calculation of protein potential method to macromolecules. J. Phys. Chem. B, 2016, 120,
molecule interaction energy. J. Chem. Phys., 2003, 119, 3599– 6562–6574.
3605.
27. Ghosh, D., Isayev, O., Slipchenko, L. V. and Krylov, A. I., The
effect of solvation on vertical ionization energy of thymine: from ACKNOWLEDGEMENTS. The author would like to thank CSIR XII
microhydration to bulk. J. Phys. Chem. A, 2011, 115, 6028–6038. Five-year Plan Multi-scale Modelling, DAE-BRNS and DST-SERB for
28. Ghosh, D., Perturbative approximation to hybrid equation of funding.
motion coupled cluster/effective fragment potential method.
J. Chem. Phys., 2014, 140, 094101(1)–094101(9). doi: 10.18520/cs/v112/i07/1455-1462

1462 CURRENT SCIENCE, VOL. 112, NO. 7, 10 APRIL 2017