Gaussian 03 Online Manual
Last update: 2 October 2006

Obsolete Keywords

The following options from early revisions of Gaussian 03 have been removed in revision D.1:

  • Int=NoSchwartz (default behavior changed; replaced by Schwartz option)
  • Int=NoVFMM
  • Int=JigGrid

The following table lists obsolete keywords used by previous versions of Gaussian. While all of them are still supported by Gaussian 03, we strongly recommend converting to the up-to-date equivalents given in the table.

Obsolete Keyword Replacement Keyword & Option
Alter Guess=Alter
BeckeHalfandHalf BHandH
Camp-King SCF=Camp-King
MP2=Stingy and VeryStingynone (options are a no-op)
Savenone (Save is a no-op)

Obsolete Utility

The chkmove utility, which converted checkpoint files to and from binary and text formats for transfer between different computer architectures, is no longer provided. Its functionality is now handled by formchk and unfchk.

Specifies a coupled cluster calculation using double substitutions and evaluation of the contribution of single and triple excitations through fourth order using the CCD wavefunction. It is superseded by CCSD(T). ST4CCD is a synonym for CCD+STCCD.

Invert the A-matrix directly. The default is the iterative solution, which is always preferable.

This properties keyword can be used to evaluate molecular orbitals, the electrostatic potential, the electron density, density gradient, the norm of the density gradient, and Laplacian of the density over a 3 dimensional grid (cube) of points. Its use is deprecated in favor of the cubegen utility.

Requests that a formatted version of the checkpoint file be written at the end of a successful run. This keyword is deprecated in favor of the formchk utility. The formatted checkpoint file always has the name Test.FChk (note the mixed case), and it is placed into the default directory from which the job is run. This keyword cannot store transition densities or natural orbitals in the formatted checkpoint file.


All: Write everything to the formatted checkpoint file.
: Write forces in internal coordinates.
ForceCart: Write forces in Cartesian coordinates.
: Write the electric field properties (in Cartesian coordinates).
: Write the intermediate structures from an optimization in internal coords.
: Write the intermediate structures from an optimization in Cartesian coords.
: Write the basis set data (exponents, coefficients, etc.).
: Write the Molecular orbitals.
: Write separate α and β components (default=total density).
: If densities are requested, use the natural orbital representation (the default is the density lower triangle).
: Write the SCF density.
: Write the generalized density for the current method.
: Write all available densities.
: Write the transition density between the ground and current state.
: Write the transition densities between the ground and all excited states.
: Write all trans. densities involving either ground or current state.
: Write all transition densities.
: Write the CI-Singles 1PDM for the current state.
: Write all CI-Singles 1PDMs.

Indicates that the geometry specification is in Cartesian coordinates. Cartesian coordinates can be included in molecule specifications without any special options being necessary.

LST and LSTCyc
Requests that an initial guess for a transition structure be generated using Linear Synchronous Transit [575]. The LST procedure locates a maximum along a path connecting two structures and thus provides a guess for the transition structure connecting them. LST is not valid with AM1.

Note that an LST calculation does not actually locate a proper transition state. However, the structure resulting from an LST calculation may be suitable as input for a subsequent Opt=TS. In general, however, the LST method has been superseded by Opt=QST2.

The Massage keyword requests that the molecule specification and basis set data be modified after it is generated. This keyword is deprecated in favor of ExtraBasis, Charge, Counterpoise and other keywords. See below for its full description.

Requests an optimization using a pseudo-Newton-Raphson method with a fixed Hessian and numerical differentiation of energies to produce gradients. This option requires that the Hessian be read in via ReadFC or RCFC. It can be used to locate transition structures and higher saddle points.

Requests the Fletcher-Powell optimization algorithm [144], which does not require analytic gradients.

Requests a gradient optimization, using the default method unless another option is specified. This is the default whenever analytic gradients are available and is invalid otherwise.

Requests that the MNDO (or AM1, if possible) force constants be computed and used to start the (presumably ab initio) optimization.

Specifies the Murtaugh-Sargent optimization algorithm [145]. The Murtaugh-Sargent optimization method is an obsolete alternative, and is retained in Gaussian 03 only for backwards compatibility.

Requests that a unit matrix be used instead of the usual valence force field guess for the Hessian.

This requests output of an integral file in one variant of the format originated for the PolyAtom integrals program. The format produced by default is that used by the Caltech MQM programs, but the code in Link 9999 is easily modified to produce other variations on the same theme.

Write an MO coefficient file in Caltech (Tran2P5) format. This is only of interest to users of the Caltech programs.

The PCM model present in Gaussian 94 may be accessed using this option to SCRF. It requires the dielectric constant of the solvent and the number of points per sphere as input. The radii of the spheres may optionally be specified for each atom type by including the ReadRadii option. Alternate radii for each atom for use in fitting potentials may be input via the ReadAtRadii option.

Uses the polarizable dielectric model [285,286,287], which corresponds to the Gaussian 98 SCRF=PCM option except for some minor implementation details [302]. This model is no longer recommended for general use. The default SCRF method is IEF-PCM.

Force numerical SCRF rather than analytic. This keyword is required for multiple orders beyond Dipole. This option implies the use of spherical cavities, which are not recommended. No gradients are available for this option.

The options Dipole, Quadrupole, Octopole,nd Hexadecapole specify the order of multipole to use in the SCRF calculation. All but Dipole require that the Numer option be specified as well.

Begin the SCRF=Numer calculation with a previously computed reaction field read from the input stream, immediately after the line specifying the dielectric constant and radius (three free-format reals).

Used to specify the location of the .SCR scratch file.

Retain symmetry restrictions. NoSymm relaxes symmetry restrictions and is the default.

Description of Cube

The Cube properties keyword can be used to evaluate molecular orbitals, the electrostatic potential, the electron density, density gradient, the norm of the density gradient, and Laplacian of the density over a 3 dimensional grid (cube) of points. Its use is deprecated in favor of the cubegen utility.

By default, Cube evaluates the electron density (corresponding to the Density option). Which density is used is controlled by the Density keyword; use Density=Current to evaluate the cube over the density from a correlated or CI-Singles wavefunction instead of the default Hartree-Fock density.

Note that only one of the available quantities can be evaluated within any one job step. Save the checkpoint file (using %Chk), and include Guess=(Read,Only) Density=Checkpoint in the route section of a subsequent job (or job step) in order to evaluate a different quantity without repeating any of the other steps of the calculation.

Gaussian provides reasonable defaults for grids, so Cube does not require that the cube be specified by the user. However, the output filename must always be provided (see below).

Alternatively, Cube may be given a parameter specifying the number of points to use per "side" (the default is 80). For example, Cube=100 specifies a grid of 1,000,000 points (1003), evenly distributed over the rectangular grid generated by the program (which is not necessarily a cube). In addition, the input format used by earlier versions of Gaussian is still supported; Cube=Cards indicates that a grid will be input. It may be used to specify a grid of arbitrary size and shape.

The options Coarse, Medium and Fine may also be specified as the parameter to Cube. They correspond to densities of 3, 6 and 12 points/Bohr, respectively. These options are designed to facilitate uniform quality in grid sampling across the range of molecular sizes.

The files created by Cube can be manipulated using the cubman utility.

Note that Pop=None will inhibit cube file creation.


When the user elects to provide it, the grid information is read from the input stream. The first line-required for all Cube jobs-gives a file name for the cube file. Subsequent lines, which are included only with Cube=Cards, must conform to format (I5,3F12.6), according to the following syntax:

Output-file-name                            Required in all Cube jobs. 
IFlag, X0, Y0, Z0                            Output unit number and initial point. 
N1, X1, Y1, Z1                               Number of points and step-size in the X-direction. 
N2, X2, Y2, Z2                               Number of points and step-size in the Y-direction. 
N3, X3, Y3, Z3                               Number of points and step-size in the Z-direction. 

IFlag is the output unit number. If IFlag is less than 0, then a formatted file will be produced; otherwise, an unformatted file will be written.

If N1<0 the input cube coordinates are assumed to be in Bohr, otherwise, they are interpreted as Angstroms (this keyword is not affected by the setting of the Units keyword). |N1| is used as the number of X-direction points in any case. Note that the three axes are used exactly as specified; they are not orthogonalized, so the grid need not be rectangular.

If the Orbitals option is selected, the cube filename (or cube filename and cube specification input) is immediately followed by a list of the orbitals to evaluate, in free-format, terminated by a blank line. In addition to numbers for the orbitals (with β orbitals numbered starting at N+1), the following abbreviations can appear in the list:

The highest occupied molecular orbital

The lowest unoccupied molecular orbital

All occupied (α) orbitals

All β occupied orbitals for UHF

All orbitals

All occupied non-core orbitals

All virtual orbitals

See the examples section for sample input files.


All values in the cube file are in atomic units, regardless of the input units.

Using the default input to Cube produces an unformatted output file (you can use the cubman utility to convert it to a formatted version if you so desire). When the Cards option is specified, then the IFlag parameter's sign determines the output file type. If IFlag>0, the output is unformatted. If IFlag<0, the output is formatted. All values in the cube file are in atomic units, regardless of the input units.

For density and potential grids, unformatted files have one row per record (i.e., N1*N2 records each of length N3). For formatted output, each row is written out in format (6E13.5). In this case, if N3 is not a multiple of six, then there may be blank space in some lines.

The norm of the density gradient and the Laplacian are also scalar (i.e., one value per point), and are written out in the same manner. Density+gradient grids are similar, but with two writes for each row (of lengths N3 and 3*N3). Density + gradient + Laplacian grids have 3 writes per row (of lengths N3, 3*N3, and N3)

For example, for a density cube, the output file looks like this:

NAtoms, X-Origin, Y-Origin, Z-Origin 
N1, X1, Y1, Z1                       # of increments in the slowest running direction 
N2, X2, Y2, Z2 
N3, X3, Y3, Z3                       # of increments in the fastest running direction 
IA1, Chg1, X1, Y1, Z1                Atomic number, charge, and coordinates of the first atom
IAn, Chgn, Xn, Yn, Zn                Atomic number, charge, and coordinates of the last atom
(N1*N2) records, each of length N3    Values of the density at each point in the grid

Note that a separate write is used for each record.

For molecular orbital output, NAtoms will be less than zero, and an additional record follows the data for the final atom (in format 10I5 if the file is formatted):

NMO, (MO(I),I=1,NMO)                  Number of MOs and their numbers

If NMO orbitals were evaluated, then each record is NMO*N3 long and has the values for all orbitals at each point together.


If one wishes to read the values of the density, Laplacian, or potential back into an array dimensioned X(N3,N2,N1) code like the following Fortran loop may be used:

  Do 10 I1 = 1, N1 
  Do 10 I2 = 1, N2       
     Read(n,'(6E13.5)') (X(I3,I2,I1),I3=1,N3)    
  10 Continue

where n is the unit number corresponding to the cube file.

If the origin is (X0,Y0,Z0), and the increment is (X1,Y1,Z1), then point (I1,I2,I3) has the coordinates:

X-coordinate: X0+(I1-1)*X1+(I2-1)*X2+(I3-1)*X3

Y-coordinate: Y0+(I1-1)*Y1+(I2-1)*Y2+(I3-1)*Y3

Z-coordinate: Z0+(I1-1)*Z1+(I2-1)*Z2+(I3-1)*Z3

The output is similar if the gradient or gradient and Laplacian of the charge density are also requested, except that in these cases there are two or three records, respectively, written for each pair of I1, I2 values. Thus, if the density and gradient are to be read into arrays D(N3,N2,N1), G(3,N3,N2,N1), RL(N3,N2,N1), a correct set of Fortran loops would be:

  Do 10 I1 = 1, N1   
  Do 10 I2 = 1, N2       
     Read(n,'(6F13.5)') (D(I3,I2,I1),I3=1,N3) 
     Read(n,'(6F13.5)') ((G(IXYZ,I3,I2,I1),IXYZ=1,3), I3=1,N3) 
10 Continue

where again n is the unit number corresponding to the cube file.


Number of points to use per "side" (the default is 80). For example, Cube=100 specifies a grid of 1,000,000 points (1003), evenly distributed over the rectangular grid generated by the program (which is not necessarily a cube).

3 points/Bohr.

6 points/Bohr.

12 points/Bohr.


Compute just the density values. Cannot be combined with the Volume keyword or the Cube=Orbitals option.

Compute the electrostatic potential at each point.

Compute the density and gradient.

Compute the Laplacian of the density ∇2ρ). Divergence is a synonym for Laplacian.

Compute the norm of the density gradient at each point.

Compute the values of one or more molecular orbitals at each point. MO is a synonym for Orbitals. Cannot be combined with the Volume keyword or the Cube=Density option.

Remove the SCF core density. This is the default for the density, and is not allowed for the potential. FC is a synonym for FrozenCore.

Evaluate the density including all electrons.

Use the total density. This is the default

Use only the alpha spin density.

Use only the beta spin density.

Use the spin density (difference between alpha and beta densities).

Read grid specification from the input stream (as described above).

Read in a list of arbitrary points.

Density, cubegen

The following job will create a cube file named orbitals.cube containing the HOMO and LUMO.

#n rhf/6-31g* 5d scf=tight cube=(orbitals) test 

HOMO and LUMO in default cube 



The following cube file illustrates the method for defining your own cube via Cube=Cards:

# rhf/6-31g* 5d scf=tight cube=(density,cards) test 

Density cube with user-defined cube 


  -51      -2.0      -2.0      -1.0 
   40       0.1       0.0       0.0 
   40       0.0       0.1       0.0 
   20       0.0       0.0       0.1 

Description of Massage

The Massage keyword requests that the molecule specification and basis set data be modified after it is generated. This keyword is deprecated in favor of ExtraBasis, Charge, Counterpoise and other keywords.

The Massage keyword thus makes it possible to add additional uncontracted basis functions to a standard basis set. Common polarization or diffuse functions can be added in this way to standard basis sets for which these functions are not internally defined. For example, diffuse functions could be added to the 3-21G basis set to form 3-21+G. Similarly, polarization functions might be added to 6-311G to form a 6-311G(5d3f) basis, which is larger than the largest internally stored 6-311G-based basis set, 6-311G(3d1f).

The standard basis functions are assigned to atoms before Massage alterations take place, while the number of electrons is computed from the atomic numbers after the modifications.

Calculations with massaged basis set data cannot generate archive entries, and do not take advantage of molecular symmetry. Some of this functionality of Massage has been superceded by the ExtraBasis keyword. Point charges may also be specified for single point energy calculations using Charge.

Massage may also be used for counterpoise calculations and BSSE (see the examples).


Massage requires one or more lines of input in the following format:

center, func, exp, [cX, cY, cZ ]

where center is the center number (numbering follows the ordering of the molecule specification section), func is a code indicating the type of modification (see below), exp is the exponent of Gaussian or new nuclear charge (a value of 0 says to add a ghost atom), and cX,cY,cZ are the coordinates of the point charge in Angstroms when func is -1 (see below). A blank line terminates this input section.

func can take on these values:

0 or Nuc
Change the nuclear charge.

1 or SP
Add an SP shell.

2 or D
Add a D shell.

3 or P
Add a P shell.

4 or S
Add an S shell.

5 or F
Add an F shell.

-1 or Ch
Add a point charge.

Note that this keyword is not affected by the setting of the Units keyword, and its input is always interpreted as Angstroms.

Charge, ExtraBasis, Gen, Counterpoise

Adding Point Charges. The following input file adds point charges to a calculation on water using the Massage keyword. Note: This is usually done with the Charge keyword and input.

# RHF/6-31G(d) Massage Test
Water with point charges       

0 1 
O -0.464  0.177  0.0 
H -0.464  1.137  0.0 
H  0.441 -0.143  0.0       

0 ch 2.0 1.0  1.0 1.0 
0 ch 2.5 1.0 -1.0 1.0 

Adding Basis Functions. The following input adds functions to the D95 basis set (in order to reproduce a calculation from the literature that used a non-standard basis set). Note: This is usually done with the ExtraBasis keyword and input.

# RQCISD(Full)/D95 Freq=Numer Massage Test 
H2O Frequencies at QCISD(Full)/DZP       

0 1 
H 1 R 
H 1 R 2 A       

1 D 0.85 
2 P 1.0 
3 P 1.0 

Computing Counterpoise Corrections Manually. The following input file performs a counterpoise calculation. Note the the Massage keyword is not used. The atoms to be removed are simply designated with the ghost atom suffix (Bq). Note: The Counterpoise keyword is now used to perform this type of calculation.

# b3lyp/3-21G** nosymm scf=tight test
HBr + H2O manual counterpoise calculation, H2O removed 

 0  1 
H      0.685176   -0.004924   -0.026973 
Br    -0.771917    0.000050    0.001967 
O-Bq   2.536864   -0.000136   -0.051401 
H-Bq   3.015128    0.789231    0.184042 
H-Bq   3.021888   -0.784986    0.185282