Skip to main content

qmmm (c38b2)

Combined Quantum and Molecular Mechanical Hamiltonian


A combined quantum (QM) and molecular (MM) mechanical potential
allows for the study of condensed phase chemical reactions, reactive
intermediates, and excited state isomerizations. This is necessary
since standard MM force fields are parameterized with experimental
data on the potential energy surface which may be far removed from
the region of interest, or have the wrong analytical
form. A full decription of the theory and application is given in
J. Computational Chemistry (1990) 6, 700.

The effective Hamiltonian, Heff, describes the energy and forces on
each atom. It is treated as a sum of four terms, Hqm, Hmm, Hqm/mm,
and Hbrdy.

Hqm Describes the quantum mechanical particles. The semi-
empirical methods available are AM1, PM3 and MNDO. All treat
hydrogen, first row elements plus silicon, phosphorus,
sulfur, and the halogens. MNDO has additional parameters
for aluminium, phosphorus, chromium, germanium, tin, mercury,
and lead. Full details concerning these theoretical methods
can be found in Dewar's original papers, JACS (1985) 107,
3902, JACS (1977) 99, 4899, Theoret. Chim. Acta. (1977) 46, 89.

Hmm The molecular mechanical Hamiltonian is independent of the
coordinates of the electrons and nuclei of the QM atoms.
CHARMM22 is used to treat atoms in this region.

Hqm/mm The combined Hamiltonian describes how QM and MM atoms
interact. This is composed of two electrostatic and one
van der Waals terms. Each MM atom interacts with both the
electrons and nuclei of the QM atoms (therefore two terms).
The van der Waals term is necessary since some MM atoms
possess no charge and would consequently be invisible to
the QM atoms, and in other cases often provide the only
difference in the interaction (Cl vs. Br).

Hbrdy The usual periodic or stochastic boundary conditions are
implemented.


The quantum mechanical package MOPAC 4.0 was interfaced with
Academy. There are several limitations with the current program
implemntation. The current plan is to update the quantum mechanical
procedures to MOPAC 6.0, to include vibrational analysis using
analytical functions, with the possibility of using free energy
perturbation.


1) QUANTUM module
* Syntax | Syntax of QM/MM Commands
* Description | Brief Description of Quantum Commands
* GLINK | GHO Method and GLNK Command
* DECO | DECO Command
* DAMP | DAMP Command
* PERT | PERT Command
* pBOUNd | Simple Periodic Boundary Condition
* GROUp | GROUp keyword
* CHDYn | CHDYn keyword
* NEWD | NEWDS Command
* EXTE | EXTErnal File Command
* LEPS | LEPS Command
* SVB | SVB Command
* 2DSVB | 2D Usage of SVB

2) SQUANTM module
* SQUANTUM | SQUANTUM Module
* SQM_Syntax | Syntax of the SQUANTM commands
* SQM_Install | Installation of SQUANTM in CHARMM environment


Top
Syntax for QUANTUM commands

QUANtum [atom-selection] [GLNK atom-selection] [LEPS int1 int2 int3]
[DECO]
[PERT REF0 lambda0 PER1 lambda1 PER2 lambda2 TEMP finalT]
[NGUEss int]
[NEWD int] ewald-spec

[IFIL int] [ITRMax int]
[CAMP] [KING] [PULAy]
[SHIFt real] [SCFCriteria real]
[UHF] [C.I.] [EXCIted] [NMOS int] [MICR int]
[TRIPLET|QUARTET|QUINTET|SEXTET]

[AM1|PM3|MNDO] [CHARge int] [NOCUtoff]
[ALPM real] [RHO0 real] [SFT1 real]
[EXCIted] [BIRADical] [C.I.]

[ITER] [EIG2] [ENERGY] [ PL ]
[DEBUG [1ELEC] [DENSITY] [FOCK] [VECTor]]

[LEPS LEPA int LEPB int LEPC int
D1AB real D1BC real D1AC real
R1AB real R1BC real R1AC real
B1AB real B1BC real B1AC real
S1AB real S1BC real S1AC real
D2AB real D2BC real D2AC real
R2AB real R2BC real R2AC real
B2AB real B2BC real B2AC real
S2AB real S2BC real S2AC real]

ewald-spec::= { [ KMAX integer ] } KSQMAX integer
{ KMXX integer KMXY integer KMXZ integer }

MULLiken

ADDLinkatom link-atom-name atom-spec atom-spec

RELLinkatom link-atom-name atom-spec atom-spec

link-atom-name ::= a four character descriptor starting with QQ.

atom-spec::= {residue-number atom-name}
{ segid resid atom-name }
{ BYNUm atom-number }


Top
Description of QUANtum Commands


Most keywords preceed an equal sign followed by an appropriate
value. A description of each is given below.

1ELECtron The final one-electron matrix is printed out. This matrix
is composed of atomic orbitals; the array element between
orbitals i and j on different atoms is given by,

H(i,j) = 0.5 (beta(i) + beta(j)(overlap(i,j))

The matrix elements between orbitals i and j on the same
atom are calculated from the electron-nuclear attraction
energy, and also from the U(i) value if i=j.

The one-electron matrix is unaffected by (a) the charge and
(b) the electron density. It is only a function of the
geometry. Abbreviation: 1ELEC.


0SCF The data can be read in and output, but no actual
calculation is performed when this keyword is used.
This is useful as a check on the input data.

All obvious errors are trapped, and warning messages printed.
A second use is to convert from one format to another.
The input geometry is printed in various formats at the
end of a 0SCF calculation. If NOINTER is absent, cartesian
coordinates are printed.
Unconditionally, MOPAC Z-matrix internal coordinates are
printed, and if AIGOUT is present, Gaussian Z-matrix
internal coordinates are printed. 0SCF should now be used
in place of DDUM.


1SCF When users want to examine the results of a single SCF
calculation of a geometry, 1SCF should be used. 1SCF can
be used in conjunction with RESTART, in which case a single
SCF calculation will be done, and the results printed.

When 1SCF is used on its own (that is, RESTART is not also
used) then derivatives will only be calculated if GRAD is
also specified. 1SCF is helpful in a learning situation.
MOPAC normally performs many SCF calculations, and in
order to minimize output when following the working of the
SCF calculation, 1SCF is very useful.

AM1 The AM1 method is to be used. By default MNDO is run.

PM3 The PM3 method is to be used. By default MNDO is run.

NOCUtoff QM/MM cutoffs are disabled, such that the QM region
interacts with all MM charges. By default the QM region
only interacts with those charges that are within the
standard CHARMM nonbond cutoffs.

ANALYTical By default, finite difference derivatives of energy with
respect to geometry are used. If ANALYT is specified,
then analytical derivatives are used instead. Since the
analytical derivatives are over Gaussian functions --
a STO-6G basis set is used -- the overlaps are also over
Gaussian functions. This will result in a very small
(less than 0.1 Kcal/mole) change in heat of formation.
Use analytical derivatives (a) when the mantissa used is
less than about 51-53 bits, or (b) when comparison with
finite difference is desired. Finite difference
derivatives are still used when non-variationally optimized
wavefunctions are present.

BIRADical NOTE: BIRADICAL is a redundant keyword, and represents a
particular configuration interaction calculation.
Experienced users of MECI (q.v.) can duplicate the effect
of the keyword BIRADICAL by using the MECI keywords
OPEN(2,2) and SINGLET. For molecules which are believed to
have biradicaloid character the option exists to optimize
the lowest singlet energy state which results from the
mixing of three states. These states are, in order, (1) the
(micro)state arising from a one electron excitation from
the HOMO to the LUMO, which is combined with the microstate
resulting from the time-reversal operator acting on the
parent microstate, the result being a full singlet state;
(2) the state resulting from de-excitation from the formal
LUMO to the HOMO; and (3) the state resulting from the single
electron in the formal HOMO being excited into the LUMO.

Microstate 1 Microstate 2 Microstate 3
Alpha Beta Alpha Beta Alpha Beta Alpha Beta


LUMO * * * *
--- --- --- --- --- --- --- ---

+

HOMO * * * *
--- --- --- --- --- --- --- ---

A configuration interaction calculation is involved here.
A biradical calculation done without C.I. at the RHF
level would be meaningless. Either rotational invariance
would be lost, as in the D2d form of ethylene, or very
artificial barriers to rotations would be found, such as in
a methane molecule "orbiting" a D2d ethylene. In both
cases the inclusion of limited configuration interaction
corrects the error. BIRADICAL should not be used if
either the HOMO or LUMO is degenerate; in
this case, the full manifold of HOMO x LUMO should be
included in the C.I., using MECI options. The user should
be aware of this situation. When the biradical
calculation is performed correctly, the result is normally
a net stabilization. However, if the first singlet
excited state is much higher in energy than the
closed-shell ground state, BIRADICAL can lead to a
destabilization. Abbreviation: BIRAD. See also MECI,
C.I., OPEN, SINGLET.

CAMKINg

CHARge When the system being studied is an ion, the charge, n, on
the ion must be supplied by CHARGE=n. For cations n can be
1 or 2 or 3, etc, for anions -1 or -2 or -3, etc.

EXAMPLES

ION KEYWORD ION KEYWORD

NH4(+) CHARGE=1 CH3COO(-) CHARGE=-1
C2H5(+) CHARGE=1 (COO)(=) CHARGE=-2
SO4(=) CHARGE=-2 PO4(3-) CHARGE=-3
HSO4(-) CHARGE=-1 H2PO4(-) CHARGE=-1

C.I. Normally configuration interaction is invoked if any of the
keywords which imply a C.I. calculation are used, such as
BIRADICAL, TRIPLET or QUARTET. Note that ROOT= does not
imply a C.I. calculation: ROOT= is only used when a C.I.
calculation is done. However, as these implied C.I.'s
involve the minimum number of configurations practical,
the user may want to define a larger than minimum C.I., in
which case the keyword C.I.=n can be used. When C.I.=n is
specified, the n M.O.'s which "bracket" the occupied-
virtual energy levels will be used. Thus, C.I.=2 will
include both the HOMO and the LUMO, while C.I.=1 (implied
for odd-electron systems) will only include the HOMO
(This will do nothing for a closed-shell system, and leads
to Dewar's half-electron correction for odd-electron
systems). Users should be aware of the rapid increase
in the size of the C.I. with increasing numbers of M.O.'s
being used. Numbers of microstates implied by the use of
the keyword C.I.=n on its own are as follows:

Keyword Even-electron systems Odd-electron systems
No. of electrons, configs No. of electrons, configs
Alpha Beta Alpha Beta

C.I.=1 1 1 1 1 0 1
C.I.=2 1 1 4 1 0 2
C.I.=3 2 2 9 2 1 9
C.I.=4 2 2 36 2 1 24
C.I.=5 3 3 100 3 2 100
C.I.=6 3 3 400 3 2 300
C.I.=7 4 4 1225 4 3 1225
C.I.=8 (Do not use unless other keywords also used, see below)

If a change of spin is defined, then larger numbers of
M.O.'s can be used up to a maximum of 10. The C.I. matrix
is of size 100 x 100. For calculations involving up to 100
configurations, the spin-states are exact eigenstates of
the spin operators. For systems with more than 100
configurations, the 100 configurations of lowest energy are
used. See also MICROS and the keywords defining spin-states.

Note that for any system, use of C.I.=5 or higher normally
implies the diagonalization of a 100 by 100 matrix. As a
geometry optimization using a C.I. requires the derivatives
to be calculated using derivatives of the C.I. matrix,
geometry optimization with large C.I.'s will require more
time than smaller C.I.'s.

Associated keywords: MECI, ROOT=, MICROS, SINGLET, DOUBLET, etc.


C.I.=(n,m) In addition to specifying the number of M.O.'s in the
active space, the number of electrons can also be defined.
In C.I.=(n,m), n is the number of M.O.s in the active
space, and m is the number of doubly filled levels to be used.

EXAMPLES
Keywords Number of M.O.s No. Electrons

C.I.=2 2 2 (1)
C.I.=(2,1) 2 2 (3)
C.I.=(3,1) 3 2 (3)
C.I.=(3,2) 3 4 (5)
C.I.=(3,0) OPEN(2,3) 3 2 (N/A)
C.I.=(3,1) OPEN(2,2) 3 4 (N/A)
C.I.=(3,1) OPEN(1,2) 3 N/A (3)

Odd electron systems given in parentheses.


DEBUG Certain keywords have specific output control meanings, such as
FOCK, VECTORS and DENSITY. If they are used, only the final
arrays of the relevant type are printed. If DEBUG is supplied,
then all arrays are printed. This is useful in debugging
ITER. DEBUG can also increase the amount of output produced
when certain output keywords are used, e.g. COMPFG.

DCART The cartesian derivatives which are calculated in DCART for
variationally optimized systems are printed if the keyword
DCART is present. The derivatives are in units of
kcals/Angstrom, and the coordinates are displacements
in x, y, and z.

DENSITY At the end of a job, when the results are being printed,
the density matrix is also printed. For RHF the normal
density matrix is printed. For UHF the sum of the alpha
and beta density matrices is printed.
If density is not requested, then the diagonal of the
density matrix, i.e., the electron density on the atomic
orbitals, will be printed.

DOUBLet When a configuration interaction calculation is done,
all spin states are calculated simultaneously, either
for component of spin = 0 or 1/2. When only doublet states
are of interest, then DOUBLET can be specified, and all
other spin states, while calculated, are ignored in the
choice of root to be used.

Note that while almost every odd-electron system will have
a doublet ground state, DOUBLET should still be specified
if the desired state must be a doublet.

DOUBLET has no meaning in a UHF calculation.

EIGS
EIG2
ENERGY

ESR The unpaired spin density arising from an odd-electron
system can be calculated both RHF and UHF. In a UHF
calculation the alpha and beta M.O.'s have different
spatial forms, so unpaired spin density can naturally
be present on in-plane hydrogen atoms such as in the phenoxy
radical.

In the RHF formalism a MECI calculation is performed. If the
keywords OPEN and C.I.= are both absent then only a single
state is calculated. The unpaired spin density is then
calculated from the state function. In order to have
unpaired spin density on the hydrogens in, for example,
the phenoxy radical, several states should be mixed.

EXCIted The state to be calculated is the first excited open-shell
singlet state. If the ground state is a singlet, then the
state calculated will be S(1); if the ground state is a
triplet, then S(2). This state would normally be the
state resulting from a one-electron excitation from the
HOMO to the LUMO. Exceptions would be if the lowest
singlet state were a biradical, in which case the EXCITED
state could be a closed shell.

The EXCITED state will be calculated from a BIRADICAL
calculation in which the second root of the C.I. matrix is
selected. Note that the eigenvector of the C.I. matrix is
not used in the current formalism.
Abbreviation: EXCI.

NOTE: EXCITED is a redundant keyword, and represents a
particular configuration interaction calculation.
Experienced users of MECI can duplicate the effect of the
keyword EXCITED by using the MECI keywords OPEN(2,2),
SINGLET, and ROOT=2.

FOCK

FORCE A force-calculation is to be run. The Hessian, that is
the matrix (in millidynes per Angstrom) of second
derivatives of the energy with respect to displacements of
all pairs of atoms in x, y, and z directions, is
calculated. On diagonalization this gives the force
constants for the molecule. The force matrix, weighted
for isotopic masses, is then used for calculating the
vibrational frequencies. The system can be characterized
as a ground state or a transition state by the presence of
five (for a linear system) or six eigenvalues which are
very small (less than about 30 reciprocal centimeters). A
transition state is further characterized by one, and
exactly one, negative force constant.

A FORCE calculation is a prerequisite for a THERMO calculation.
Before a FORCE calculation is started, a check is made to
ensure that a stationary point is being used. This check
involves calculating the gradient norm (GNORM) and if it is
significant, the GNORM will be reduced using BFGS.

All internal coordinates are optimized, and any symmetry
constraints are ignored at this point. An implication of
this is that if the specification of the geometry relies on
any angles being exactly 180 or zero degrees, the
calculation may fail.

The geometric definition supplied to FORCE should not rely
on angles or dihedrals assuming exact values. (The test
of exact linearity is sufficiently slack that most
molecules that are linear, such as acetylene and but-2-yne,
should not be stopped.) See also THERMO, LET, TRANS,
ISOTOPE.

In a FORCE calculation, PRECISE will eliminate quartic
contamination (part of the anharmonicity). This is
normally not important, therefore PRECISE should not
routinely be used.

In a FORCE calculation, the SCF criterion is automatically
made more stringent; this is the main cause of the SCF
failing in a FORCE calculation.

ITER The default maximum number of SCF iterations is 200.
When this limit presents difficulty, ITRY=nn can be used
to re-define it. For example, if ITRY=400 is used, the
maximum number of iterations will be set to 400. ITRY
should normally not be changed until all other means of
obtaining a SCF have been exhausted, e.g. PULAY CAMP-KING etc.

INTERP
LPULAY

LARGE Most of the time the output invoked by keywords is
sufficient. LARGE will cause less-commonly wanted, but
still useful, output to be printed.

1. To save space, DRC and IRC outputs will, by default,
only print the line with the percent sign. Other output
can be obtained by use of the keyword LARGE, according to
the following rules:

Keyword Effect
LARGE Print all internal and cartesian coordinates
and cartesian velocities.
LARGE=1 Print all internal coordinates.
LARGE=-1 Print all internal and cartesian coordinates
and cartesian velocities.
LARGE=n Print every n'th set of internal coordinates.
LARGE=-n Print every n'th set of internal and cartesian
coordinates and cartesian velocities.

If LARGE=1 is used, the output will be the same as that of
Version 5.0, when LARGE was not used. If LARGE is used, the
output will be the same as that of Version 5.0, when LARGE
was used. To save disk space, do not use LARGE.

MECI At the end of the calculation details of the Multi Electron
Configuration Interaction calculation are printed if MECI
is specified. The state vectors can be printed by
specifying VECTORS. The MECI calculation is either
invoked automatically, or explicitly invoked by the use of
the C.I.=n keyword.

MICRos The microstates used by MECI are normally generated by use
of a permutation operator. When individually defined
microstates are desired, then MICROS=n can be used, where
n defines the number of microstates to be read in.

Format for Microstates

After the geometry data plus any symmetry data are read in,
data defining each microstate is read in, using format
20I1, one microstate per line. The microstate data is
preceded by the word "MICROS" on a line by itself. There
is at present no mechanism for using MICROS with a reaction path.

For a system with n M.O.'s in the C.I. (use OPEN=(n1,n) or
C.I.=n to do this), the populations of the n alpha M.O.'s
are defined, followed by the n beta M.O.'s. Allowed
occupancies are zero and one. For n=6 the closed-shell
ground state would be defined as 111000111000, meaning one
electron in each of the first three alpha M.O.'s, and one
electron in each of the first three beta M.O.'s.

Users are warned that they are responsible for completing
any spin manifolds. Thus while the state 111100110000
is a triplet state with component of spin = 1, the state
111000110100, while having a component of spin = 0 is
neither a singlet nor a triplet. In order to complete the
spin manifold the microstate 110100111000 must also be included.

If a manifold of spin states is not complete, then the
eigenstates of the spin operator will not be quantized.
When and only when 100 or fewer microstates are supplied,
can spin quantization be conserved.

There are two other limitations on possible microstates.
First, the number of electrons in every microstate should
be the same. If they differ, a warning message will be
printed, and the calculation continued (but the results
will almost certainly be nonsense). Second, the component
of spin for every microstate must be the same, except for
teaching purposes. Two microstates of different
components of spin will have a zero matrix element
connecting them. No warning will be given as this is a
reasonable operation in a teaching situation. For example, if
all states arising from two electrons in two levels are to
be calculated say for teaching Russel-Saunders coupling,
then the following microstates would be used:

Microstate No. of alpha, beta electrons Ms State

1100 2 0 1 Triplet
1010 1 1 0 Singlet
1001 1 1 0 Mixed
0110 1 1 0 Mixed
0101 1 1 0 Singlet
0011 0 2 -1 Triplet

Constraints on the space manifold are just as rigorous, but
much easier to satisfy. If the energy levels are
degenerate, then all components of a manifold of degenerate
M.O.'s should be either included or excluded. If only
some, but not all, components are used, the required
degeneracy of the states will be missing.

As an example, for the tetrahedral methane cation, if the user
supplies the microstates corresponding to a component of
spin = 3/2, neglecting Jahn-Teller distortion, the minimum
number of states that can be supplied is 90 = (6!/(1!*5!))*
(6!/(4!*2!)).

While the total number of electrons should be the same for all
microstates, this number does not need to be the same as
the number of electrons supplied to the C.I.; thus in the
example above, a cationic state could be 110000111000.

The format is defined as 20I1 so that spaces can be used
for empty M.O.'s.

MNDO The default Hamiltonian within MOPAC is MNDO, with the
alternatives of AM1 and MINDO/3. To use the MINDO/3
Hamiltonian the keyword MINDO/3 should be used. Acceptable
alternatives to the keyword MINDO/3 are MINDO and MINDO3.

NGUEss The number of steps to regenerate initial guess during molecular
dynamics. The default is 100 step. If NGUEss <= 0, then it will
be used previous density all over the dynamics. Only applied in
the molecular dynamics.

PRECISE The criteria for terminating all optimizations, electronic and
geometric, are to be increased by a factor, normally, 100.
This can be used where more precise results are wanted. If
the results are going to be used in a FORCE calculation,
where the geometry needs to be known quite precisely, then
PRECISE is recommended; for small systems the extra cost in
CPU time is minimal.

PRECISE is not recommended for experienced users, instead
GNORM=n.nn and SCFCRT=n.nn are suggested. PRECISE should
only very rarely be necessary in a FORCE calculation: all
it does is remove quartic contamination, which only affects
the trivial modes significantly, and is very expensive
in CPU time.

PULAy The default converger in the SCF calculation is to be
replaced by Pulay's procedure as soon as the density matrix
is sufficiently stable. A considerable improvement in
speed can be achieved by the use of PULAY. If a large
number of SCF calculations are envisaged, a sample calculation
using 1SCF and PULAY should be compared with using 1SCF on
its own, and if a saving in time results, then PULAY should
be used in the full calculation. PULAY should be used with
care in that its use will prevent the combined package of
convergers (SHIFT, PULAY and the CAMP-KING convergers)
from automatically being used in the event that the system
fails to go SCF in (ITRY-10) iterations.

The combined set of convergers very seldom fails.

QUARTet RHF interpretation: The desired spin-state is a quartet,
i.e., the state with component of spin = 1/2 and spin =
3/2. When a configuration interaction calculation is done,
all spin states of spin equal to, or greater than 1/2 are
calculated simultaneously, for component of spin = 1/2.
From these states the quartet states are selected when
QUARTET is specified, and all other spin states, while
calculated, are ignored in the choice of root to be used.
If QUARTET is used on its own, then a single state,
corresponding to an alpha electron in each of three M.O.'s
is calculated.

UHF interpretation: The system will have three more alpha
electrons than beta electrons.

QUINTet RHF interpretation: The desired spin-state is a quintet,
that is, the state with component of spin = 0 and spin = 2.
When a configuration interaction calculation is done, all
spin states of spin equal to, or greater than 0 are
calculated simultaneously, for component of spin = 0.
From these states the quintet states are selected when
QUINTET is specified, and the septet states, while
calculated, will be ignored in the choice of root to be
used. If QUINTET is used on its own, then a single state,
corresponding to an alpha electron in each of four M.O.'s
is calculated.

UHF interpretation: The system will have three more alpha
electrons than beta electrons.

ROOT The n'th root of a C.I. calculation is to be used in the
calculation. If a keyword specifying the spin-state is
also present, e.g. SINGLET or TRIPLET, then the n'th root
of that state will be selected. Thus ROOT=3 and SINGLET
will select the third singlet root. If ROOT=3 is used on
its own, then the third root will be used, which may be a
triplet, the third singlet, or the second singlet (the
second root might be a triplet). In normal use, this
keyword would not be used. It is retained for educational
and research purposes. Unusual care should be exercised
when ROOT= is specified.

SCFCrt The default SCF criterion is to be replaced by that defined
by SCFCRT=. The SCF criterion is the change in energy in
kcal/mol on two successive iterations. Other minor
criteria may make the requirements for an SCF slightly more
stringent. The SCF criterion can be varied from about
0.001 to 1.D-25, although numbers in the range 0.0001 to
1.D-9 will suffice for most applications.

An overly tight criterion can lead to failure to achieve a
SCF, and consequent failure of the run.


SEXTet RHF interpretation: The desired spin-state is a sextet:
the state with component of spin = 1/2 and spin = 5/2.
The sextet states are the highest spin states normally
calculable using MOPAC in its unmodified form. If SEXTET
is used on its own, then a single state, corresponding to
one alpha electron in each of five M.O.'s, is calculated.
If several sextets are to be calculated, say the second
or third, then OPEN(n1,n2) should be used.

UHF interpretation: The system will have five more alpha
electrons than beta electrons.

SHIFt In an attempt to obtain an SCF by damping oscillations
which slow down the convergence or prevent an SCF being
achieved, the virtual M.O. energy levels are shifted up or
down in energy by a shift technique. The principle is that
if the virtual M.O.'s are changed in energy relative to
the occupied set, then the polarizability of the occupied
M.O.'s will change pro rata. Normally, oscillations are
due to autoregenerative charge fluctuations.

The SHIFT method has been re-written so that the value of
SHIFT changes automatically to give a critically-damped
system. This can result in a positive or negative shift
of the virtual M.O. energy levels. If a non-zero SHIFT
is specified, it will be used to start the SHIFT technique,
rather than the default 15eV. If SHIFT=0 is specified,
the SHIFT technique will not be used unless normal
convergence techniques fail and the automatic "ALL
CONVERGERS..." message is produced.

SINGLet When a configuration interaction calculation is done, all spin
states are calculated simultaneously, either for component
of spin = 0 or 1/2. When only singlet states are of
interest, then SINGLET can be specified, and all other spin
states, while calculated, are ignored in the choice of root
to be used.

Note that while almost every even-electron system will
have a singlet ground state, SINGLET should still be
specified if the desired state must be a singlet.

SINGLET has no meaning in a UHF calculation, but see also TRIPLET.

TRIPLet The triplet state is defined. If the system has an odd
number of electrons, an error message will be printed.

UHF interpretation. The number of alpha electrons exceeds
that of the beta electrons by 2. If TRIPLET is not
specified, then the numbers of alpha and beta electrons are
set equal. This does not necessarily correspond to a singlet.

RHF interpretation.

An RHF MECI calculation is performed to calculate the
triplet state. If no other C.I. keywords are used, then
only one state is calculated by default. The occupancy of
the M.O.'s in the SCF calculation is defined as
(...2,1,1,0,..), that is, one electron is put in each of
the two highest occupied M.O.'s.

See keywords C.I.=n and OPEN(n1,n2).

UHF The unrestricted Hartree-Fock Hamiltonian is to be used.

VECTors The eigenvectors are to be printed. In UHF calculations
both alpha and beta eigenvectors are printed; in all cases
the full set, occupied and virtual, are output. The
eigenvectors are normalized to unity, that is the sum of
the squares of the coefficients is exactly one. If DEBUG
is specified, then ALL eigenvectors on every iteration of
every SCF calculation will be printed. This is useful in
a learning context, but would normally be very undesirable.


Top
Description of the Generalized Hybrid Orbital (GHO) method
and the GLNK Command

[GLNK atom-selection]

atom-selection: contains a list of atoms that are boundary atoms.

Restrictions: The current implementation of the method requires that
ALL boundary atoms are placed at the end of the QM residue, or at
the end of the QM atom list. It is also strongly advised to treat
the entire QM fragment as a single residue, without any GROUPping
of atoms. This is because the delocalized nature of molecular
orbitals does not allow for arbitrarily excluding a particular
fragment or orbitals from interacting with other parts of the system.


Description: In addition to the link atom approach, a generalized
hybrid orbital (GHO) approach for the treatment of the division across
a covalent bond between the QM and MM region. The method recognizes
a frontier atom, typically carbon which is the only atom that has
its parameters optimized at this time, both as a QM atom and an MM
atom. Thus, standard basis orbitals are assigned to this atom.
These atomic orbitals on the frontier atoms are transformed into a
set of equivalent hybrid orbitals (typically the frontier atom is
of sp3 hybridization type). One of the four hybrid orbitals, which
points directly to the direction of the neighboring QM atom, is
included in QM-SCF orbital optimizations, and is an active orbital.
The other three hybrid orbitals are not optimized. Thus, they are the
auxillary orbitals. Since hybridization (contributions from s and
p orbitals to the hybrid orbitals) is dependent on the local geometry,
change of bond angles will lead to bond polarization in the active
orbital. Also, since the active orbital is being optimized in the
SCF procedure, charge transfer between the frontier atom and the
QM fragment is allowed. Consequently, the GHO method provides a
convenient way for smooth transition of charge distribution from the
QM region into the MM region.

The charge density on the auxilary orbitals are determined by equally
distributing the MM partial charge on the frontier atom. Thus,
P(mu mu) = 1 - q(mm)/3. The neutral group convention adopted by
the CHARMM force field makes it possible not to alter, to add, or
to delete any MM charges. Furthermore, no extra degrees of freedom
is introduced in the GHO approach.

The GHO method based on Unrestricted HF theory (GHO-UHF) is implemented
at semiempirical level (AM1, PM3) in the quantum module. With this
extension, GHO boundary treatment can be used for open shell QM fragments
in combined QM/MM calculations.

For a GHO-UHF wavefunction, we have two sets of auxiliary hybrid
orbitals for alpha spin and beta spin electrons respectively.
The charge density assigned to each of these auxiliary hybrid orbitals
is 0.5(1.0-q(mm)/3.0), while q(mm) denotes the MM partial charge of the
GHO boundary atom. Similar to GHO-RHF, the hybridization basis
transformation is carried out between the density matrix and Fock matrix,
both for the alpha and the beta sets.

Analytical gradients and Mulliken population analysis are also implemented
for GHO-UHF.

Limitations: The present implementation allows up to 5 QM-boundary
atoms, which uses psuedo-atomic numbers 91-95. Thus, elements 91
through 95 can not be used in QM calculations.

Reference: Reference made to the following paper, which contains
a more thorough description and discussion of test cases, is appreciated.

Jiali Gao, Patricia Amara, Cristobal Alhambra, and Martin J. Field,
J. Phys. Chem. 102, 4714-4721 (1998). "A Generalized Hybrid Orbital
(GHO) Approach for the Treatment of Link-Atoms using Combined
QM/MM Potentials."


Top
Description of the DECO Command

[DECO]

The lone command DECO initiates an qm/mm interaction energy
decomposition calculation on the fly during a molecular dynamics simulation
using the QUANtum command. It is currently implemented only for
semiempirical Hamiltonians. The analysis is based on the method reported
in J. Gao and X. Xia, Science, 258, 631 (1992). It decomposes the total
QM/MM electrostatic interaction energy into a vertical interaction energy
Evert, and a polarization term Epol. The latter is further separated into
electrostatic stabilization Estab, and charge distortion Edist. These
terms are defined as follows (Y is the wave function of the qm system
in the presence of mm charges, and Yo is the wave function of the qm
system in the absence of mm charges, i.e., in the gas phase):

Eqm/mm = <Y|Hqm+Hqmmm(elec)|Y>
= Evert + Epol
Evert = <Yo|Hqmmm(elec)|Yo>
Epol = Eqm/mm - Evert

Epol = Estab + Edist
Estab = <Y|Hqmmm(elec)|Y> - <Yo|Hqmmm(elec)|Yo>
Edsit = <Y|Hqm|Y> - <Yo|Hqm|Yo>

where Hqm is the Hamiltonian of the qm system, and Hqmmm(elec) is
the electrostatic part of the QM/MM interaction Hamiltonian. Note that
the van der Waals term is kept track of separately within CHARMM's
general energy terms. In addition, the decomposition also averages
the average "gas-phase" energy <Egas> of the QM system during the QM/MM
simulation. Egas, of course, is NOT the true average gas-phase energy,
but it is one that is restrained by the presence of the MM field.
It is, however, interesting to note that <Egas> - <Yo|Hqm|Yo> gives
the "strain energy" due to geometrical strain in the condensed
phase/protein environment. JG 12/00


Top
[ DAMP real ]

A simple density damping option is added to the SCF driver for the quantum
module. The motivation of adding this option is to provide a possibility to
overcome SCF convergence difficulties. Currently, this damping accelerator is
only used to limit oscillation behavior in GHO-UHF type calculations.

For an SCF iteration with density damping turned on, the actual density matrix
used for next iteration is computed by a linear combination of the current
density with the previous one:

P = a x P + (1-a) x P
i i-1 i

The damping factor "a" is a user defined floating point number between 0 and 1.
One can specify this damping factor as "DAMP a" in the QUANtum command line.
The default of this damping factor is 0.0, i.e., no damping at all. Any
damping factor being less than 0 or greater than 1 will incur a level -5
warning.

In the current implementation, several "damped" steps (with a user defined
damping factor "a" ) are carried out until the alpha and beta density matrices
are partially converged (density changes are smaller than 100 times the density
convergence criterion), then "undamped" steps (a=0.0) follow until the final
convergence is reached.


Top
Description of the PERT Command

[PERT REF0 lambda0 PER1 lambda1 PER2 lambda2 TEMP finalT]

REF0 lambda0 the reference Lambda value in a FEP calculation
PER1 lambda1 the forward perturbation Lambda value in a FEP calculation
PER2 lambda2 the reverse perturbation Lambda value in a FEP calculation
TEMP finalT the target or final temperature of the MD simulation
NOTE: this is required. Otherwise an error will occur.

The PERT command performs electrostatic free energy decoupling
calculation for QM/MM interactions on the fly of a molecular dynamics
simulation. The algorithm is based on a method described in J. Gao,
J. Phys. Chem. 96, 537 (1992). Through a series of simulations, the
electrostatic component of the free energy of solvation can be determined.
See, other free energy simulation documents.

Delta G(L0->L1) = -RT < exp(-[E{H(L1)}-E{H(L0)}]/RT > _E{H(L0)}

where

E{H(Li)} = <Y|Hqm + Li * Hqmmm(elec) + Hqmmm(vdW)|Y>

JG 12/00


Top
Simple Periodic Boundary Conditions for QM/MM calculations

This code is an extension of the algorithm already implemented in
duplication of coordinates to save memory in QM/MM calculations. It takes
advantage of the minimum image convention for a periodic cubic (or rectangular
or any other shapes) box such that crystallographic images are not required to
be generated in the psf (» images ).

[Syntax]

BOUNd {CUBOUNdary } {BOXL <real> } CUTNB <real>

CUBOUN = CUbicBOUNd
BOXL = length of the box edge
CUTNB = cutoff for generating "virtual" images

Note: QM/MM PBOUND is INCOMPATIBLE with atom based non-bonded list.

RESTRICTIONS:

1. Information about the periodic boundary must be given
to the program through the command READ IMAGE (» image )

2. The system must be centered using commands:
a) IMAGE (» image ) when solute and solvent
are small molecules (3-5 atoms)
b) CENT keyword in DYNA command line (» dynamc ) when solute is
a protein or large organic molecule.

3. The mm and qm/mm nonbonded lists (electrostatic and Van der Waals
interactions)
must be generated by groups, i.e:

update group fswitch vdw vswitched vgroup ...

4. Compile with PBOUND in pref.dat

Example:
...

! Set-up image information for cubic periodic boundaries
! cubig.img file in the /test/data/ directory
set 6 58.93044
set 7 58.93044
set 8 58.93044
open unit 1 read form name cubic.img
read image card unit 1
close unit 1


IMAGe byseg xcen 0.0 ycen 0.0 zcen 0.0 select segid prot end
IMAGe byres xcen 0.0 ycen 0.0 zcen 0.0 select sol end


BOUNd CUBOUND BOXL 58.93044 CUTNB 12.0


UPDAte group fswitch noextend cdie vdw vswitched eps 1.0 -
cutnb 12.0 ctofnb 11.5 ctonnb 10.5 vgroup WMIN 1.2 -
inbf 25 imgfrq 1000 cutim 12.0

QUANtum group sele qms end glnk sele bynu 68:69 end am1 charge 1 -
scfc 0.000001

DYNAmics vverlet rest nstep 15000 timestp 0.001 -
ilbfrq 0 iseed 324239 firstt 239.0 finalt 298.15 -
teminc 5.0 ihtfrq 5.0 iasor 0 iasvel 1 iscvel 0 -
ichecw 1 ieqfrq 200 nprint 100 nsavc 00 -
nose tref 298.15 qref 50.0 isvfrq 100 -
tstruc 298.15 -
twindh 5 twindl -5 iprfrq 2000 wmin 0.9 -
iunrea 9 iunwri 10 iuncrd 11 iunvel -1 kunit -12 -
CENT ncres 162

....



Top
Description of the GROUp keyword

[ GROUp]

The QM/MM module that was initially implemented into CHARMM allows
for separate QM group and MM group interactions, where a "QM" molecule
can be divided into several groups. The GROUp option allows the QM
molecule to be partitioned into separate groups for generating non-bonded
list, but keeps the interactions between the ENTIRE QM molecule and any
MM group that is whithin the cutoff of any one qm group, avoiding the
possibility that some MM group only interact with part of the QM molecule.
This is necessary because the QM molecule is not divisible as the wave
function is delocalized over the entire molecule. (June, 2001)

See also description of the GLNK keyword.


Top
Description of the CHDYn keyword

[CHDYn]

CHDYn allows the computation of average Mulliken population charges
on quantum atoms during a molecular dynamics simulation. It prints the
averaged atomic charges at every IPRFRQ steps.
When CHDYn is used along with DECO it will result in the calculation
of average atomic charges for the same trajectory in the presence of
the MM bath (condensed phase) and absence of the MM charges (gas phase).

RESTRICTIONS: CHDYn is only implemented for molecular dynamics calculations
with the Leapfrog Verlet and Velocity Verlet integrators.

TESTCASE : qmfep.inp


Top
Description of the NEWDS Command

[ NEWD int ] ewald-spec

ewald-spec::= { [ KMAX integer ] } KSQMAX integer
{ KMXX integer KMXY integer KMXZ integer }

A simple Ewald sum method is implemented into the QM/MM potential. A full
description of theory is described in J. Chem. Theory. Comput. (2005) 1, 2.
This is based on regular Ewald sum method and share similar keywords
ewald ).

The defaults for the QM/MM-Ewald calculations are set internally
and are currently set to NEWD 1, KMAX=5, KSQMax=27, where the
KMAX keyword is the number of kvectors (or images of the
primary unit cell) that will be summed in any direction. It is the
radius of the Ewald summation. For orthorombic cells, the value of
kmax may be independently specified in the x, y, and z directions with
the keywords KMXX, KMXY, and KMXZ. But, different from regular Ewald in
in MM part.

The KSQMax key word should be chosen between KMAX squared and 3 times
KMAX squared, and KAPPA value share the exact same number you use in Nonbond
options.

Specific for SQUANTM: implementation of QM/MM-PME method.
PME version fo QM/MM-Ewald summation is the default for SQUANTM module. (You
can use regualr QM/MM-Ewald method by adding a keyword NOPMEwald.) Whenever
MM-MM long-range electrostatic interactions are computed by using PME method,
QM-MM interactions in the QM/MM-Ewald are also computed by PME method, where
the potential at the QM atom sites are computed by PME algorithm. The
gradients component by QM-MM interactions at the QM and MM atom sites are
also computed by PME algorithm, which is separatedly calculated from the
MM-MM PME contributions. Still, the long-range QM-QM interactions are
computed by a regular QM/MM-Ewald method, this requires users to specify
the ewald-spec options, otherwise defauls values will be invoked. The
QM/MM-PME method is much faster and requires less memory.


Top
[EXTErnal PUNIt int]

EXTE Allows reading the semi-empirical parameters from an external file.
The format is as follows (free format):

PARNAME1 ATOMTYPE1 PARVALUE1
PARNAME2 ATOMTYPE2 PARVALUE2
...
END

Empty lines are ignored
Acceptable parameters names are:
ALFA(ALP),BETAS,BETAP,BETAD,ZS,ZP,ZD,USS,UPP,UDD,GSS,GPP,GSP,GP2,HSP,
VS,VP,K1(FN11),L1(FN21),M1(FN31),K2(FN12),L2(FN22),M2(FN32),K3(FN13),
L3(FN23),M3(FN33),K4(FN14),L4(FN24),M4(FN34)

The following derived parameters are computed automatically:
EISOL,DD,QQ,AM,AD,AQ

Example:
USS H -11.3336021333
BETAS H -6.1735981344
END


Top
Description of the LEPS Command

[LEPS LEPA int LEPB int LEPC int -
D1AB real D1BC real D1AC real -
R1AB real R1BC real R1AC reat -
B1AB real B1BC real B1AC real -
S1AB real S1BC real S1AC real -
D2AB real D2BC real D2AC real -
R2AB real R2BC real R2AC real -
B2AB real B2BC real B2AC real -
S2AB real S2BC real S2AC real]

Description: The motivation behind the semiempirical valence bond term (SEVB)
is to improve the quality of the potential energy surface (PES) when using
semiempirical hamiltonians (AM1 or PM3) to model the reactive event in
enzyme active sites. NDDO based hamiltonias represent a cheap alternative
to describe reactions in enzyme active sites. They allow for a quantum
mechanical description of the active site together with an extensive sampling
of the protein configurational space when combined qmm/mm techniques are used.
However the savings in computer time come with sacrifices in the quality of
the PES due to the NDDO approximation. The SEVB term is introduced in the
hamiltonian of the system to palliate this problem. It contains two extended
London-Eiring-Polany-Sato (LEPS) equations for the three body subsystem
{A,B,C}. This reduced subsystem mimics the transfer of the particle B
between centers A and C in the active site. In most of the applications
B is a light atom like hydrogen and A and C correspond to the donor and
acceptor sites,

A-B + C ---> A + B-C

Each of the two extended LEPS functions have different parameters and depend
on the distances r(A-B), r(B-C), and r(A-C). One LEPS potential (V(ref))is
fitted to reproduce high ab initio or experimental data for a model reaction
whilst the second one (V(NDDO)) is fitted to the NDDO hamiltonian in use.
Finally the SEVB correction is introduced in the hamiltonian of the system
as the difference V(ref)-V(NDDO).

Syntaxis: The keyword LEPS in the command line QUANtum turns on the
routine that evaluates the SEVB correction. The three atoms needed to
evaluate the distances r(A-B), r(B-C), and r(A-C) are indicated by,

LEPA - donor center.
LEPB - transferred atom.
LEPC - acceptor center.

int - corresponds to the psf number of the respective atom.

The value of the parameters to build the functions V(NDDO) and V(ref) are,

. for the NDDO LEPS functions,

D1AB - dissociation energy for the diatomic A-B
D1BC - dissociation energy for the diatomic B-C
D1AC - dissociation energy for the diatomic A-C
R1AB - equilibrium distance for the diatomic A-B
R1BC - equilibrium distance for the diatomic B-C
R1AC - equilibrium distance for the diatomic A-C
B1AB - beta exponent for the diatomic A-B
B1BC - beta exponent for the diatomic B-C
B1AC - beta exponent for the diatomic A-C
S1AB - Sato parameter for the diatomic A-B
S1BC - Sato parameter for the diatomic B-C
S1AC - Sato parameter for the diatomic A-C

. for the reference LEPS functions,

D2AB - dissociation energy for the diatomic A-B
D2BC - dissociation energy for the diatomic B-C
D2AC - dissociation energy for the diatomic A-C
R2AB - equilibrium distance for the diatomic A-B
R2BC - equilibrium distance for the diatomic B-C
R2AC - equilibrium distance for the diatomic A-C
B2AB - beta exponent for the diatomic A-B
B2BC - beta exponent for the diatomic B-C
B2AC - beta exponent for the diatomic A-C
S2AB - Sato parameter for the diatomic A-B
S2BC - Sato parameter for the diatomic B-C
S2AC - Sato parameter for the diatomic A-C

A real value is expected after each one of them.

Limitations: The current implementation is only intended for a single SEVB
correcting term.

Reference: A detailed description of the LEPS potential energy functionals
as well as the application to an enzymatic hydride transfer can be found in,

C. Alhambra, J. Corchado, M. L. Sanchez, M. Garcia-Viloca, J. Gao &
D. G. Truhlar, Journal of Physical Chemistry B, 2001, 105, 11326-11340.
"Canonical Variational Theory for Enzyme Kinetics with the Protein Mean
Force and Multidimensional Quantum Mechanical Tunneling Dynamics. Theory
and Application to Liver Alcohol Dehydrogenase."


Top
Description of SVB command

[ LEPS SVB LEPA int LEPB int LEPC int -
D1AB real D1BC real D1AC real -
R1AB real R1BC real R1AC reat -
B1AB real B1BC real B1AC real ]

Description: A simple analytical function is included in combined QM/MM
potential energy functions using semiempirical Hamiltonian for enzyme
reactions to obtain more accurate energetic results. The motivation
behind the simple valence bond (SVB) term is to introduce small energy
corrections at critical points (reactants, transition state, and products)
on the QM potential energy surface. The underlying assumption is that the
general shape of the QM potential energy surface at the semiempirical level
is in reasonable accord with high-level ab initio result. The SVB term
is a simplified version of the semiempirical valence bond term (SEVB)
invoked by the command LEPS.

The SVB term is a combination of two Morse potentials, which depend on
the bond distances of the breaking and making bonds, respectively, and
a coupling term that is typically (but not exclusively) a function of
the donor-acceptor distance.

Specifically, for the reaction A-B + C ---> A + B-C

with r1 = distance A-B
r2 = distance B-C
r3 = distance A-C

the SVB correction along a given reaction coordinate that depends on r1 and r2
is:

VSVB = 1/2 [ M1(r1)+M2(r2) - [(M2(r2)-M1(r1))**2+4V12**2)]**1/2]

where M1(r1) and M2(r2) are Morse potentials:

M1(r1) = D1AB [ exp(-2*B1AB*(r1-R1AB))-2*exp(-B1AB*(r1-R1AB)) ]

M2(r2) = D1BC [ exp(-2*B1BC*(r2-R1BC))-2*exp(-B1BC*(r1-R1BC)) ]

and the coupling term has the form:

V12 = D1AC * exp(-B1AC*(r3-R1AC))

where,

D1AB = difference in dissociation energy between the reference calculation
or experimental value and the dissociation energy given by
semiempirical method (for the AB bond)

D1BC = difference in dissociation energy between reference calculation
or experimental value and the dissociation energy given by
the semiempirical method (for the BC bond)

B1AB and B1BC = related to the bond force constants (kij)
and to the bond dissociation energies (D1ij) by
B1ij = sqrt (kij/2*D1ij). These values can be obtained from
experimentally determined frequencies or from high level
calculations.

R1AB, R1BC and R1AC = equilibrium bond length for bonds AB, BC, and AC,
respectively.

D1AC,R1AC = adjustable parameters to obtain the desired barrier height.

Note: D1AB and D1BC may be also adjusted to obtain the desired reaction energy.
The difference D1AB-D1BC is the relative correction of the product state energy
respect to the reactant state energy. It is recommended to avoid negative values
for these variables.

Reference: A detailed description of the SVB method
as well as the application to the nucleophilic addition reaction catalized by
haloalkane dehalogenase is found in:

Devi-Kesavan, L.S.; Garcia-Viloca, M.; Gao, J. Theor.Chem.Acc. 2002, in press.


Example:

...

QUANtum group sele qms end glnk sele bynu 68:69 end am1 charge 1 -
scfc 0.000001 -
LEPS SVB LEPA 57 LEPB 58 LEPC 13 - ! atoms involved
D1AB 36.0 D1BC 15.0 D1AC 15.0 - ! energies
R1AB 1.101 R1BC 1.1011 R1AC 2.707 - ! equil. bond lenghts
B1AB 1.393 B1BC 1.409 B1AC 1.0 - ! exponents

...


File: qmmm, Node: 2DSVB, Up: Top, Previous: SVB, Next: SQUANTUM

The following options allow one to construct a two-dimensional simple
VB-like function to correct the (presumably) semiempirical QM/MM
potential energy surface. This extends beyong the one-dimensional
correction described above.

[LEPS
[LEPA int LEPB int LEPC int
D1AB real D1BC real D1AC real
R1AB real R1BC real R1AC real
B1AB real B1BC real B1AC real
S1AB real S1BC real S1AC real
D2AB real D2BC real D2AC real
R2AB real R2BC real R2AC real
B2AB real B2BC real B2AC real
S2AB real S2BC real S2AC real]

[SVB [SURF] [TEST] [GCOU|G1CO|G2CO]
LEPA int LEPB int [LEPC int]
D1AB real D1BC real [D1AC real]
R1AB real R1BC real [R1AC real]
B1AB real B1BC real [B1AC real]
[LEPD int LEPE int [LEPF int]
D1DE real D1EF real [D1DF real]
R1DE real R1EF real [R1DF real]
B1DE real B1EF real [B1DF real]
[VC12]]]

SVB SVB turns on the Simple Valence Bond correction function.
It requires correction values as difference between experimental
or high-level QM level data and the semi-empirical level used.
The following keywords work only with the SVB option.
SURFace A two-dimensional correction term is employed
TEST Additional output regarding SVB term at each energy step. Useful
when testing the extent of correction needed.
GCOUpling Use a gaussian coupling term which is a function of the reaction
coordinate. If SURF is also switched on, this keyword will switch
on gaussian terms for both coordinates.
G1COuple Use gaussian coupling term for first coordinate.
G2COuple Use gaussian coupling term for second coordinate.
VC12 Add coupling term between two coordinates. This term is currently
only a constant number.

If only two atoms are specified (LEPA and LEPB or LEPD and LEPE), a simple
exponential term between the two atoms will be added. If a gaussian term
is switched on via the GCOU,G1CO, or G2CO keywords, then a gaussian term
will be employed instead.


Top
**********************************************************************
****** SQUANTUM MODULE ******
Combined Quantum Mechanical and Molecular Mechanics Method
Based on SQUANTM in CHARMM

The F90 semiempirical code written by Ross Walker (TSRI, AMBER) and
Mike Crowley (TSRI, AMBER and CHARMM), the interface to CHARMM has
been implemented by Kwangho Nam (UMN, nam@chem.umn.edu) including
GHO and Swithing function implementation. The QM/MM-Ewald summation
implementation is done as a joint project between TSRI and
UMN (University of Minnesota).

The new semiempirical code, SQUANTM, is envisioned to eventually
replace the current semiempirical QM code, which was originally
incorporated into CHARMM by Martin Field and Paul Bash based on
Stewart's MOPAC version 5 program. The SQUANTM was written in
Fortran90, and the result is a substantial improvement in
computational speed. However, the two packages are not in conflict
as long as they are not compiled together. Therefore, all the
original MOPAC-based QM/MM options and commands are kept.
Should a user choose to use the SQUANTM or the MOPAC-based QM/MM
algorithm, one simply follow the compiling steps highlighted below.


Top
Syntax for SQUANTM commands

QUANtum [atom-selection] [GLNK atom-selection]
[REMOve] [SWITched]

[AM1|PM3|MNDO|PDP3|PDMN] [CHARge int] [NOGAussian] [NDIS]
[SCFCriteria real]
[DOUBLET|TRIPLET]

[NEWD int] ewald-spec NOPMEwald

[DUALqm] { [PERTurbation pert-spec] }
{ [MLAYered mlay-spec] }

ewald-spec::= { [KMAX int] } [KSQMAX int]
{ [KMXX int] [KMXY int] [KMXZ int] }

pert-spec ::= [atom-selection] [KHARge int] [GLN2 atom-selection]
[REF0 real] [PEF1 real] [PEF2 real]
[TEMP real]
[PKAP] { [NOMM] }
{ [NOAN] [NODI] [NOIM] }

mlay-spec ::= [atom-selection] [KHARge int] [LINK atom-selection]
[RCUT real]

(**mlay-spec: refer the comments below.)


GLNK: GHO method implementation (refer qmmm.doc).

REMOve: Classical energies within QM atoms are removed.

SWITched: Use switching function from CTONNB to CTOFNB values
based on GROUp method (refer nbonds.doc). It is
incompatible with NEWD options

AM1|PM3|MNDO|PDP3|PDMN: AM1 model, PM3 model, MNDO model,
PDDG/PM3 (PDP3) model, and PDDG/MNDO (PDMN) model.
Note: currently, GHO method only support AM1 and
PM3 model.
For PDDG/PM3 and PDDG/MNDO model, the reference
is Repasky et al. J. Comput. Chem. (2002), 23, 1601.

NOGAussian: This will turn off Gaussian core-core interaction
between QM and MM pairs in the AM1 and PM3 model.
As a default, the Gaussian core-core interactions
will be computed.

NDIS : Turn off DIIS converger. (default is .false.)

NEWD and ewald-spec: refer the description for QUANTUM module.

NOPMEwald: Use regualr QM/MM-Ewald summation methods. (Default is false.)
This is added for backward compatibility. Note that only for
SQUANTM supports QM/MM-PME (refer details for NEWD description.)

DUALqm : Use dual qm/mm calculations. (Multiple qm/mm region
or FEP calculation.)

PERTurbation and pert-spec : Do FEP pKa calculations.
The reactant state (lambda=0) should be
the protonated state, and the product state
should be the deprotonated state (lambda=1).
The perturbation is only applied to the
electrostatic interactions between QM and
MM pairs.

H_eff = (1-lambda)*H_reactant + (lambda)*H_product

atom-selection: atom selection for the product state (lambda=1).
Mostly, this atom selection lacks the H atom that is
annihilated in the FEP calculations.
KHARge : charge for 2nd qm region (product state).
GLN2 : use GHO method in the 2nd qm region.
REF0/PEF1/PEF2/TEMP : refer section for "Description of the PERT Command"
(above).
PKAP : turn on mm valency terms for pKa calculation, in particular,
for dummy H-atom.
In the product state, the H atom (for the protonated state)
is changed into mm dummy atom that is not interact with other
atoms (mm and qm atoms) electrostatically, while the bond,
angle, dihedral, and improper terms are maintained to avoid
the end point problem during the FEP calculations. (Refer the
approach of Riccardi et al. J. Phys. Chem. B (2005), 105,
17715.)
NOMM : turn off the angle/dihedral/improper terms for dummy H-atom.
NOAN / NODI / NOIM : specifically turn off either angle, dihedral, or
improper terms.
("NOMM" is same as "NOAN NODI NOIM.")
**Note that the bond terms are maintained.**

MLAYered and mlay-spec: Do multi-layered qm/mm calculations.
This is similar to the electronically embedded ONIOM
type potential. The 1st qm/mm layer has a larger qm
selection, and the 2nd qm/mm layer has a smaller qm
selection but is a subset of the 1st qm selection.
The total energy is extrapolated from the ab initio
and semiempirical qm/mm calculations. Therefore, it
assumes the potential at the semiempirical level for
the larger qm region is corrected at the ab initio
level for the smaller qm region, and the 2nd qm region
should be chosen with care.

E_tot = {E_semi-qm/mm (1st qm/mm region)}
+ {E_ai-qm/mm (2nd qm/mm region) - E_semi-qm/mm (2nd qm/mm region)}

**Note that only GAMESS-UK is tested with SQUANTM, and check for future
update for GAMESS-US. To compile for this, use both U and SQ in the
install.com.

atom-selection: atom selection for the 2nd qm region. (This should be a subset
of 1st qm selection.)
KHARge : charge for the 2nd qm region.
LINK atom-sele: selection for (2nd) qm atoms to which the link H-atom will be
connected. (For the 2nd qm region, qm/mm boundary will be
handled only by the H-link atom approach. However, the user
does not need to invoke ADDLink for this. The position will be
internally determined and the gradients are projected out.)
RCUT : cutoff distance for mm atoms that are included in the 2nd
qm/mm calcualtion.
**This is one nice way to include long-range electrostatics for the ab initio
qm/mm calculations, in which the long-range electrostatics are included at
the semiempirical level. For example, users can select the 1st qm region
to be same as the 2nd qm region, and specify RCUT for the 2nd qm region,
while the 1st qm region can be calculated with either QM/MM-Ewald sum or
no-cutoff.


Note: Currently, the SQUANTM module does not support UHF calculations.
Thus, if you want to run UHF calculations, use QUANTUM or MNDO97
module available for CHARMM QM/MM calculations.

For QM-MM interaction, the interaction doesn't include the
Gaussian core-core interactions in the PDDG/PM3 model
(refer future implementation and changes). However, in the AM1
and PM3 model, it is computed as default unless turned off
by using NOGAussian keyword. Depending on your keyword selection,
the results could be different from the original implementation in
QUANTUM or MNDO97 based on J. Comput. Chem. (1990) 6, 700.


Top
Installation of SQUANTM module in CHARMM

To compile SQUANTM with CHARMM, one uses:

install.com [machine] [size] SQ [other Sw]

The "SQ" specifies to compile SQUANTM with CHARMM by replacing QUANTUM
keyword in pref.dat file. Currently, the platform should support
compilers that can compile F77 and F90 code simultaneously. The
platform and compilers tested include ALTIX and GNU using intel fortran
compilers (ifort and efc), and IBMAIX and IBMAIXMP platform using xlf90
and related compiler. For other platforms, install.com and Makefile
need to be modified accordingly.