September 2, 2009: G AUSSIAN 09 Revision A.

02

Vibrationally-resolved electronic spectra in G AUSSIAN 09

V. Barone, J. Bloino, M. Biczysko
Scuola Normale Superiore, Piazza dei Cavalieri, 7 I-56126 Pisa, Italy

General overview
In a first approximation level, electronic spectra (UV-vis, photoelectron, X-ray, etc.) can be simulated by computing
vertical excitation energies at geometrical structures optimized for the ground electronic state, with the resulting stick
spectra subsequently convoluted by Gaussian or Lorentzian functions. However, such a treatment neglects completely
dynamic effects, which influence significantly the spectrum line-shape in many cases. To overcome this limitation,
a new procedure to compute vibrationally-resolved electronic spectra has been included in G AUSSIAN. It relies on
the Franck-Condon principle [1, 2] to evaluate the intensities of transitions between two vibronic states. This is done
by considering that the nuclear positions are mostly unaltered by the electron jump which takes place during the
electronic transition.
To deal with medium-to-large systems, an efficient pre-
screening scheme [3, 4] is used to choose the most relevant
transitions. It is designed to be fast and simple enough to have
little incidence on the computational cost of the spectrum gen-
eration, while providing an accurate choice of the transitions to
deal with in order to produce accurate spectra. One of its main
features is that it relies on the number of integrals to compute
and not on the dimensionality of the system. As a consequence,
G AUSSIAN can be tailored to perform fast calculations while
retaining the general shape of the spectrum at first, as shown on
figure 1. It is noteworthy that this prescreening scheme works
as a black-box procedure and performs a full dimensional treat-
ment, considering all vibrational modes of the studied system. Figure 1: Convergence of the phosphoresence spec-
Additionally, G AUSSIAN accepts a wide range of settings to trum for chlorophyll c2 with different maximum
adjust the calculations or to modify the layout of the spectrum number of integrals to compute (N ) for each class.
I
when needed. Because calculations to generate the theoretical
spectrum are independent on the lower and upper bound of the latter, the energy range on which the simulation is
performed has a very little impact on the computational costs1

1
However, a high resolution, given by a small interval between two measuring points of the discretized spectrum, associated to a broad band
can noticeably slow down the calculations.

1
Contents
1 Introduction 2

2 Theoretical background 2

3 Quick start 10

4 List of options 13

5 Some practical examples 17

6 Description of the errors reported by G AUSSIAN 18

1 Introduction
In this new release of G AUSSIAN, it is possible to compute vibrationally-resolved absorption and emission spectra.
This white paper aims at introducing the general theory underlying such calculations and presenting the possibilities
offered to generate electronic spectra inside G AUSSIAN.
The document is divided as follows. After a brief introduction of the basic theory used in G AUSSIAN to generate
vibronic spectra, a quick presentation of its basic usage is given. The main features of the input and output will then
be described. The second part of the document is devoted to more advanced cases and to the tools to control and
fine-tune the calculations to meet specific needs. The available options and their purpose will be listed, followed by
examples of applications showing the usage of the most common ones. The last part summarizes the most common
errors reported by G AUSSIAN and possible workarounds. This section is not aimed at being exhaustive and alternative
solutions can work as well or even be more adapted to some particular cases.
The following typographic conventions will be used thereafter:
• ′ : placed next to a letter/symbol, refers to the initial state
• ′′ : placed next to a letter/symbol, refers to the final state
• NA : represents the number of atoms in the molecule or system of interest
• N : represents the number of normal modes in the molecule or system of interest
• M: designates a matrix or vector
• Mi or M(i): represents the i-th element of vector M
• Mij or M(i, j): represents the i, j element of matrix M
Note: For convenience, the notation adopted in this document is slightly different from the spectroscopic conventions
since the relative energies of the initial and final states are not taken into account when using ′ and ′′ . This allows to
treat indifferently the similar cases of absorption and emission without duplicating the formulae.

2 Theoretical background
The theoretical description in this section is not intended to be exhaustive but rather to give a sufficient background in
order to use G AUSSIAN for the simulation of vibrationally resolved electronic spectra. It is divided in three subsec-
tions. The first one presents a general theoretical background and the approximations used to compute the transition
dipole moment and the transition intensities. This will help to understand the possibilities given by the Franck-Condon
principle as well as the limit of such a method. The second part deals with the prescreening method used by G AUS -
SIAN to choose a priori the most intense transitions. The parameters on which it relies as well as their influence are
described there. Finally, a discussion on the guessing of spectrum convergence is carried out.

2
Calculation of the transition dipole moment
Line intensities in one-photon absorption and emission spectra depend on the square of the transition dipole moment
h Ψ′ | µ | Ψ′′ i as well as on the radiation frequency (ω)2 :

4π 2 ω X ′ X ′ 2
σ(ω)abs = ρ h Ψ | µ | Ψ′′ i δ(E ′′ − E ′ − h̄ω) (1)
3c ′ ′′

4ω 3 X ′ X ′ 2
σ(ω)emi = 3
ρ h Ψ | µ | Ψ′′ i δ(E ′′ − E ′ + h̄ω) (2)
3h̄c ′ ′′

where ρ′ is the Boltzmann population of each initial state and σ(ω) is the rate of photon absorption (or emission) per
molecule and per unit of radiant energy. Ψ′ and Ψ′′ represent the molecular wave functions of the initial and final
state, respectively.
In the frame of the Born-Oppenheimer approximation, it is possible to write the molecular wave function Ψ as a
product of an electronic wave function ψe and a nuclear wave function ψn . The transition dipole moment can then be
written,
h Ψ′ | µ | Ψ′′ i = h ψe′ ψn′ | µe | ψe′′ ψn′′ i + h ψe′ ψn′ | µn | ψe′′ ψn′′ i (3)
Since the electronic wave functions are orthogonal to each other, the second term of the rhs in the previous equation
vanishes.
We further assume that the Eckart conditions [5] are met, so that the nuclear wave function can be, with a good ap-
proximation, separated into a rotation and a vibrational contribution, the translational part being completely separated
from the other two and discarded here. In the treatment presented here, rotational energies are considered sufficiently
small and will not be taken into account.
Using both approximations, the transition dipole moment can finally be written,

h Ψ′ | µ | Ψ′′ i ≈ h ψv′ | µif | ψv′′ i (4)

where µif = h ψe′ | µ | ψe′′ i is the electronic transition dipole moment.

In the method implemented in G AUSSIAN, the harmonic approximation is also used. As a consequence, it is
possible to write the multidimensional vibrational wave function ψv as a product of monodimensional wave functions
ψvi (Qi ), where Qi is one of the N normal coordinates, describing the i-th vibration of the molecule.
However, since the analytic form of the electronic transition dipole moment is unknown, these approximations are
not yet sufficient to compute the transition dipole moment. An additional approximation is done by using the Franck-
Condon principle, proposed by Franck [1] and mathematically formalized by Condon [2]. The underlying theory is
that, during a transition, the electron jump in a molecule takes place in such a short time that the relative positions
and velocities of the nuclei are nearly unaltered by the molecular vibrations. Following this principle, schematically
shown in figure 2, the most intense transition will be from the chosen initial state to a final state being at a minimum
of the lowest potential surface vertically upward. Since this approximation is very restrictive by assuming that the
electronic transition dipole moment remains constant during the transition, a more flexible approach is adopted here.
Indeed, since the shifts of the nuclear positions in the molecule are often fairly small during the electronic transition,
it is possible, with a good accuracy, to develop the electronic transition dipole moment in a Taylor series about the
equilibrium geometry of one state of reference, which is the final state in our case:
N  N XN
!
∂µ

∂ 2µ
X if 1 X if
µif (Q′′ ) ≃ µif (Q′′0 ) + Q′′ + Q′′k Q′′l + . . . (5)
∂Q′′k 0 k 2 ∂Q′′k ∂Q′′l
k=1 k=1 l=1 0

where Q′′0 represents the equilibrium geometry of the final state.

2
With respect to the formulae presented here, G AUSSIAN does not take into account the constant part to calculate the intensity of the
spectrum lines

3
 The zeroth order term corresponds to a static electronic transition dipole. This
is a direct application of the Franck-Condon principle mentioned above. As a con-
sequence, it is referred as the Franck-Condon (FC) approximation. When dealing
with fully-allowed transitions (|µif (Q′′0 )| ≫ 0), this approximation generally
gives very good results.
However, in the case of weakly-allowed (|µif (Q′′0 )| ∼ 0) or dipole-forbidden
(|µif (Q′′0 )| = 0) electronic transitions, this approximation is not describing
correctly the overall spectrum, missing the most intense vibronic transitions. It is
necessary to consider a variation of the dipole moment during the transition. A
first approximation is to account for a linear variation of the dipole moment with
the normal coordinates. This corresponds to the first order of the Taylor expansion
and is referred as the Herzberg-Teller [6] (HT) approximation3 . Hereafter, the FC
approximation considers only the first term in the rhs of eq. 5, HT the second one,
and FCHT both terms, as approximation of µif . Figure 2: The Franck-Condon prin-
Using eq. 5, the transition dipole moment can be written: ciple is shown by the vertical dotted
line. Following this principle, the high-
′
hΨ |µ|Ψ i= ′′
µif (Q′′0 ) h ψv′ | ψv′′ i est probability of transition corresponds
to h 0′ | 2′′ i where the overlap of the
N  wave functions is the highest.
∂µif
X 
+ h ψv′ | Q′′k | ψv′′ i
∂Q′′k 0 (6)
k=1
N
N X
!
1 X ∂ 2 µif
+ h ψv′ | Q′′k Q′′l | ψv′′ i
2 ∂Q′′k ∂Q′′l
k=1 l=1 0

The overlap integral h ψv′ | ψv′′ i is also referred to as the Franck-Condon inte-
gral. The integrals in the second and third terms of equation 5 depend on the normal coordinates and therefore cannot
be, in this form, generalized to any problem. To simplify the calculation of the integrals given in equation 6, it is
necessary to transform them in more versatile and usable Franck-Condon integrals. This is done by using the creation
and annihilation operators from the second quantization, respectively a†i and ai , defined as
r  
† ωi d
ai = Qi − (7)
2h̄ dQi
r  
ωi d
ai = Qi + (8)
2h̄ dQi
Hence,
 1/2 h
h̄ i
Qi = ai + a†i (9)
2ωi
In the current implementation, the Taylor expansion of the electronic transition dipole moment is supported up to
the second term in the rhs in eq. 5. As a consequence, the approximation of µif will be limited to the Franck-Condon
and Herzberg-Teller terms. In addition, the vibrational wave function wil be replaced from now on by a vector “v ”
containing the quantum numbers vi representing the same vibrational state described by ψv :
N
Y
| v i = | ψv i = | ψvi i
i=1

Using eq. 9, the second term integral in the rhs of eq. 6 can be rewritten,
s
′ ′′ ′′ h̄ hq ′′ q i
h v | Qk | v i = ′′ vk h v | v − 1k i + vk′′ + 1 h v ′ | v ′′ + 1′′k i
′ ′′ ′′
(10)
2ωk
3
In their original work, Herzberg and Teller considered the initial state as the state of reference for the linear variation of the electronic
transition dipole moment. We use here a more general interpretation of their work by applying to any state of reference.

4
where | v ′′ + 1′′i i represents, from the energy perspective, a vibrational state which differs from the state represented
by | v ′′ i by the eigenvalue of the monodimensional Schrödinger equation Evi′′ +1 = h̄ωi′′ (vi′′ + 1/2) = Evi′′ + h̄ωi′′ .
Inserting eq. 10 in eq. 6, it is possible to write the transition dipole moment integral as a combination of overlap
integrals:

h Ψ′ | µ | Ψ′′ i =µif (Q′′0 ) h v ′ | v ′′ i

N   s
X ∂µif h̄ hq ′′
+ vk h v ′ | v ′′ − 1′′k i (11)
∂Q′′k 0 2ωk′′
k=1
q i
+ vk′′ + 1 h v ′ | v ′′ + 1′′k i

The overlap integrals are given by the equation,

N
!−1/2
1/4
Y ′ ′′
h v ′ | v ′′ i = π −N/2 det Γ′ Γ′′ det(J)1/2 2vi 2vi (vi′ !)(vi′′ !)
 

i=1
N
!
Z +∞ Y 1/2 1/2 (12)
× dQ′′1 . . . dQ′′N Hvi′ (Γ′i Q′i )Hvi′′ (Γ′′i Q′′i )
−∞ i=1
h 1 i
× exp − (Q′ Γ′ Q′ + Q′′ Γ′′ Q′′ )
T T

2
where Γ′ and Γ′′ are the diagonal matrices of the reduced frequencies from the initial and final states, respectively,
and Hvi (qi ) is a Hermite polynomial of degree vi defined by:

2 dvi −qi2
Hvi (qi ) = (−1)vi eqi e (13)
dqivi

A first issue to calculate the integral in eq. 12 arises from the fact that each vibrational wave function is expressed
in a different set of normal coordinates. To overcome this, the linear transformation proposed by Duschinsky [7] to
express the normal coordinates of one state with respect to the other’s is used,

Q′ = JQ′′ + K (14)

where J is called the Duschinsky matrix and represents the mixing of the normal modes during the transition, and K
is the shift vector due to the changes in geometry between the initial and final states. This transformation is generally
a good approximation, when the molecule does not undergo a noticeable distortion during the transition.
The rotation or Duschinsky matrix and shift vector are given by:

−1
J = L′ L′′ (15)

−1 1/2
K = L′ M ∆X (16)

where L′ and L′′ are the transformation matrices from mass-weighted Cartesian coordinates to normal coordinates,
M is the diagonal matrix of atomic masses and ∆X is a vector representing the shift of nuclear Cartesian coordinates
between the initial and final states.
J and K can be printed in G AUSSIAN with the keyword PrtMat in the ReadFCHT section (with the respective
values 1 and 2 as shown in the keywords list).
In addition, while the overlap integral between the vibrational ground states h 0′ | 0′′ i can be straightforwardly
calculated, this is generally not the case for h v ′ | v ′′ i where v ′ and/or v ′′ are not null. Among the several compu-
tational approaches proposed till now, we have selected the Sharp and Rosenstock’s one [8]. However, their original
work was based on analytic formulae for each possible combination of v ′ and v ′′ and it is not adapted when dealing
with a large number of normal modes. A more versatile method based on recursive formulae [9], which follows a
similar formalism is described here.

5
 First, two dummy vectors, T and U, are defined to be near-null so that,
′
!
′ 1/2 ′ ∂v  2 T ′ 1/2 ′

Hv ′ (Γ Q)= exp −T + 2T Γ Q (17)
∂Tv ′
T=0

and !
′′
′′ 1/2 ∂v 1/2
Q′′ ) = exp −U2 + 2UT Γ′′ Q′′
 
Hv ′′ (Γ (18)
∂Uv ′′
U=0
where ′′ ′′ ′′ ′′ ′ ′ ′ ′
∂v ∂ v1 ∂ v2 ∂ vN ∂v ∂ v1 ∂ v2 ∂ vN
′′ = ′′ ′′ . . . v ′′
and ′ = ′ ′ . . . v′
(19)
∂Uv v
∂U11 ∂U22
v
∂UNN ∂Tv v
∂T11 ∂T22
v
∂TNN
Replacing the Hermite polynomials in eq. 12 by their definitions given in eqs. 17 and 18, and using the Duschinsky
transformation, the following formula is obtained:
1/4  ′ ′′ −1/2
h v ′ | v ′′ i = π −N/2 det Γ′ Γ′′ det(J)1/2 2v 2v v ′ !v ′′ !


′ ′′
∂v ∂v h1
Z
′′
2 2
Q′′ JT Γ′ JQ′′
T
 
× exp −T − U dQ exp
∂Tv ′ ∂Uv ′′ 2
(20)
+ Q′′ JT Γ′ K + KT Γ′ JQ′′ + KT Γ′ K + Q′′ Γ′′ Q′′
T T


!
i
1/2 1/2 1/2
+ 2UT Γ′′ Q′′ + 2TT Γ′ JQ′′ + 2TT Γ′ K
T=U=0

After developing and integrating the previous equation, we obtain:

−1/2 ∂ v ′ ∂ v ′′  1/2
′ ′′

v ′ v ′′ ′ ′′N/2
 ′ ′′ 1/4 det(J)
h v | v i = 2 2 v !v ! 2 det Γ Γ
∂Tv ′ ∂Uv ′′ det JT Γ′ J + Γ′′


h 1 1 i
× exp − KT Γ′ K + KT Γ′ J(JT Γ′ J + Γ′′ )−1 JT Γ′ K
h 2 2 i
1/2 1/2
× exp UT (2Γ′′ (JT Γ′ J + Γ′′ )−1 Γ′′ − I)U
(21)
h i
1/2
× exp −2KT Γ′ J(JT Γ′ J + Γ′′ )−1 Γ′′ U
h i
1/2 1/2
× exp TT (2Γ′ J(JT Γ′ J + Γ′′ )−1 JT Γ′ − I)T
h i
1/2
× exp −2KT (Γ′ J(JT Γ′ J + Γ′′ )−1 JT − I)Γ′ T
h i
1/2 1/2
× exp UT 4Γ′′ (JT Γ′ J + Γ′′ )−1 JT Γ′ T

To simplify the notation, 5 matrices referred to hereafter as Sharp and Rosenstock matrices are defined:
1/2 1/2
A = 2Γ′ J(JT Γ′ J + Γ′′ )−1 JT Γ′ −I
1/2
B = −2Γ′ (J(JT Γ′ J + Γ′′ )−1 JT Γ′ − I)K
1/2 1/2
C = 2Γ′′ (JT Γ′ J + Γ′′ )−1 Γ′′ −I
′′ 1/2
D = −2Γ (JT Γ′ J + Γ′′ )−1 JT Γ′ K
1/2 1/2
E = 4Γ′′ (JT Γ′ J + Γ′′ )−1 JT Γ′

These matrices can also be printed in G AUSSIAN with the keyword PrtMat in the ReadFCHT section (with the
respective values 3, 4, 5, 6 and 7 as shown in the keywords list).

6
 In addition, setting v ′ = v ′′ = 0, we obtain from eq. 21 the overlap integral between the vibrational ground
states:
 1/2
′ ′′ N/2
 ′ ′′ 1/4 det(J)
h0 |0 i=2 det Γ Γ
det JT Γ′ J + Γ′′


(22)
h 1 1 T ′ i
T ′ T ′ ′′ −1 T ′
× exp − K Γ K + K Γ J(J Γ J + Γ ) J Γ K
2 2
Hence, eq. 21 can be written,
′ ′′
′ ′′

v ′ v ′′ ′ ′′
−1/2 ∂v ∂v
h v | v i = 2 2 v !v ! h 0′ | 0′′ i
∂Tv ′ ∂Uv ′′
! (23)
h i
× exp UT CU + DT U + TT AT + BT T + UT ET
T=U=0

This equation is the starting point to generate two recursive formulae depending on the initial differentiation, with
respect to Ti or Ui . The procedure, using the Leibniz product rule, is to differentiate once by Ti or Ui and the result by
each Tj and Uj once, with j running from 1 to N . The procedure is straightforward so only the results are presented
here:
1 h
q
h v ′ | v ′′ i = p ′ Bi h v ′ − 1′i | v ′′ i + 2(vi′ − 1) Aii h v ′ − 2′i | v ′′ i
2vi
s
N q N (24)
X X vj′ i
+ 2vj′ Aij h v ′ − 1′i − 1′j | v ′′ i + Eji h v ′ − 1′i | v ′′ − 1′′j i
2
j=1 j=1
j6=i

and
1 h
q
h v ′ | v ′′ i = p Di h v ′ | v ′′ − 1′′i i + 2(vi′′ − 1) Cii h v ′ | v ′′ − 2′′i i
2vi′′
s
N q N (25)
X
′ ′′ ′′ ′′
X vj′ i
+ ′′
2vj Cij h v | v − 1i − 1j i + Eij h v ′ − 1′j | v ′′ − 1′′i i
2
j=1 j=1
j6=i

with the first equation corresponding to an initial differentiation by Ti and the second one to Ui .
In the current version in G AUSSIAN, all evaluated transitions start from the vibrational ground state of the initial
electronic state (T=0). As a consequence, only the last recursive equation is used without the last term in the right-hand
side:
1 h
q
h 0′ | v ′′ i = p ′′ Di h 0′ | v ′′ − 1′′i i + 2(vi′′ − 1) Cii h 0′ | v ′′ − 2′′i i
2vi
N q
X i (26)
+ 2vj′′ Cij h 0′ | v ′′ − 1′′i − 1′′j i
j=1
j6=i

A priori evaluation of the transition intensities

Using the recursive eqs. 24 and 25, it is possible to express any overlap integral h v ′ | v ′′ i with respect to h 0′ | 0′′ i,
which is calculated exactly. Although, there is in principle an infinite number of combinations to take into account,
most of them correspond to transtions of negligible intensity.
In G AUSSIAN, an a priori scheme [3, 4] selects the transitions which must be taken into account. This method
relies on a categorization of the latter with respect to the number of simultaneously excited modes in the final state,

7
called classes. For instance, class 1 (C1 ) represents all transitions to final vibrational states with a single excited mode
i, h v ′ | 0′′ + vi′′ i and class 0 contains the overlap integral between the vibrational ground states, h 0′ | 0′′ i. Based on
this division, the overlap integrals in classes 1 and 2 are used as reference data to evaluate those to compute in each
“higher” class, each class being calculated one after the other, increasing the number of excited modes in the final
state. The advantages of using the overlap integrals of these classes are twofold. The first one is that these integrals
are computationally cheap and are generated quickly even in the case of large molecules. The second interest lies in
the information provided by the reference data they provide.
Depending on the approximation of the electronic transition dipole moment, FC or FCHT/HT, 2 or 3 sets of
reference data are used for the prescreening. The first set FC1 is defined during the calculation of the transtions from
class 1 and is used for FC, FCHT and HT calculations. Its elements are defined as:

 1 h q i2
′′ ′ ′′ ′′ ′′ ′′ ′ ′′ ′′ ′′
FC1 (i, vi ) =  p ′′ Di h 0 | 0 + vi − 1i i + 2(vi − 1) Cii h 0 | 0 + vi − 2i i 
 (27)
2vi

where the he factors Cii and Di give respectively information on the effect of the shifts in equilibrium positions and
the frequencies on the overlap integrals of overtone, and more precisely on the vibrational progression of mode i.
In case of HT or FCHT calculations, a second set of data is extracted from class 1, HC1 , which stores the square
of the pure Herzberg-Teller contribution for a given mode i and the corresponding transition h 0′ | 0′′ + vi′′ i:
 s i2
 ∂µif

h̄ h
q q
′′ ′′ ′ ′′ ′′ ′′ ′′ ′ ′′ ′′
HC1 (i, vi ) = 
 vi h 0 | 0 + vi − 1i i + vi + 1 h 0 | 0 + vi + 1i i  (28)
∂Q′′ i 02ω ′′ i

Finally, a last set FC2 , used for all calculations, is obtained in class 2, and gives information about the Duschinsky
mixing of the normal modes. It contains all combinations of modes i and j but only considering the cases of an equal
number of quanta for both modes (vi′′ = vj′′ ):

2 FC (i, v ′′ ) × FC1 (j, vi′′ )

FC2 (i, j, vi′′ = vj′′ ) = h 0′ | 0′′ + vi′′ + vj′′ i − 1  i

(29)
h 0′ | 0′′ i2


From the previous description, a problem arises to define the dimensions of the arrays FC1 , FC2 and HC1 , and
more precisely the maximum quantum numbers vi′′max for classes C1 and C2 . Practically, these are set manually. Since
calculations of the Franck-Condon integrals in these classes are computationally cheap, a relatively high limit, chosen
to work in most common cases, is considered; C1max = 20 for class 1 and C2max = 13 for class 2. These values can be
set manually in the ReadFCHT section with the keywords MaxC1 (or MaxOvr) for C1max , and the keywords MaxC2
(or MaxCmb) for C2max .
The procedure to generate a vibronic spectrum is as follows. First, calculations are carried out for class C1 and
then class C2 . For class C3 and above, these are done in two steps. The evaluation method is initially called to set
the maximum number of quanta vi′′max to process for each mode i. To perform this prescreening, an additional limit,
NImax , which represents the maximum number of integrals to compute in a given class is set. This number controls
roughly the computational cost of the overall simulation, but at the risk of a loss in accuracy of the generated spectrum
if it is set too low. A good compromise between resources usage and precision of the spectrum is set by default
(NImax = 108 ). This parameter can be modified in G AUSSIAN with the keyword MaxInt given in the ReadFCHT
section (see the keywords list for an explanation of the syntax).
Two thresholds, ǫ1 and ǫ2 , are defined for the sets of data FC1 and FC2 , respectively. In case of FCHT or HT
calculations, a third threshold ǫH 1 for HC1 is also set. For each mode i, two maximum quantum numbers, respectively
C1 C2
vimax and vimax , are obtained by decreasing the number of quanta vi′′ until the conditions FC1 (i, vi′′ ) ≥ ǫ1 for the first
set and FC2 (i, j, vi′′ ) ≥ ǫ2 with j 6= i for the second set, are met. For FCHT/HT calculations, a similar procedure is
performed with HC1 and the condition HC1 (k, vk′′ ) ≥ ǫH H
1 to obtain the maximum quantum number vimax . The highest
of the maximum quantum numbers defined previously is chosen as vi′′max . Once the set v ′′max has been defined, the
corresponding number of integrals to calculate is roughly estimated, for a given class Cn , as NI = N Cn × hv ′′max in ,
where N Cn represents the number of combinations of the n excited oscillators and hv ′′max i is the arithmetic mean of
the N maximum quantum numbers. If the number of integrals to compute, NI is higher than the allowed limit NImax ,

8
the thresholds ǫ1 and ǫ2 (and ǫH 1 for FCHT and HT calculations) are increased and the set of maximum quantum
numbers v ′′max is re-estimated. When the condition NI ≤ NImax is fulfilled, all transition dipole moment integrals are
computed up to the chosen maximum number of quanta vi′′max for each mode.

Evaluating the reliability of the prescreening with the sum rules

To check the reliability of the prescreening method, the total intensity calculated after each class is checked with the
total intensity expected using analytic sum rules. The sum of all transition intensities is given by,

h Ψ′ | µ | Ψ′′ i2 ≈ h v ′ | µif | v ′′ i2
X X X X
I tot = ρ′ ρ′
 
(30)
′ ′′ v′ v ′′

where the first sum symbol represents the summation over all possible vibrational initial states and the second one the
ensemble of all final states. ρ′ is the Boltzmann population of each initial state.
Expanding the square in the previous equation and accounting for the orthonormality of the vibrational wave
functions, the intensity I tot can be written:
X
I tot = ρ′ h v ′ | µif · µif | v ′ i (31)
v′

In case of FC calculations, where µif = µif (Q′′0 ), eq. 31 can be written:

2 X ′ ′ 2
I tot = µif (Q′′0 ) ρ h v | v ′ i = µif (Q′′0 )
 
(32)
v′

In the framework of the Herzberg-Teller approximation of the electronic transition dipole moment (FCHT and HT
calculations), the computation of I tot is more involved. Considering the Taylor expansion of the electronic transition
dipole moment given in eq. 5, and eq. 31, it is straightforward to see that annihilation and creation operators will be
needed to obtain the sum rules. However, the development of µif in the Taylor series about the equilibrium geometry
of the final state is not immediate because the annihilation and creation operators obtained this way act on the quantum
states of the final state and not on the initial state, with respect to our previous definition in eq. 5. A simple solution
would be to develop µif in a Taylor series about the equilibrium geometry of the initial state. However, while the
equality µif = µif (Q′ ) = µif (Q′′ ) is true, this is generally not the case when considering the Taylor series around
each equilibrium geometry. As a consequence, two analytic limits are defined depending on the state of reference,
tot for the initial state and I tot for the final state.
IQ ′ Q′′
tot is rather straightforward to evaluate:
The first analytic limit, IQ ′

N 
 ∂µif 2 ′ 2
 
X 2 X
tot
ρ′ h v ′ | µif (Q′0 ) +  Q | v′ i

IQ ′ =

 ∂Q′  k (33)
v′ k=1 k 0

Using the annihilation and creation operators, the analytic limit is given by the formula,
N   2
tot
 ′
2 X ′ X  ∂µif (τ )  h̄ ′
IQ ′ = µif (Q0 ) + ρ 
 2ω ′ (2vk + 1)
 (34)
 ∂Q′ k 0 k
v′ k=1

tot , additional steps are needed to evaluate the analytic limit:

In the case of IQ ′′

N  2
 ∂µif

X
′ ′ ′′
2 X  ′′ 2
tot  Q | v′ i

IQ ′ = ρ h v | µif (Q0 ) + 
 ∂Q′′  k (35)
v′ k=1 k 0
 N  N
2
∂µif
X  X  X 
′ ′  ′′ ′
| v′ i
T

T


= ρ h v | µif (Q0 ) + J Q + J K (36)

∂Q ′′ kl l kl l 
′v
 k 0 k=1 l=1

9
where the inverse of the Duschinsky transformation given in eq. 14 has been used to express Q′′ with respect to Q′ .
tot is given by the relation:
After using the operators of the second quantization, IQ ′′

"
2 X X
I tot′′ = µif (Q′′0 ) + ρ′

Q
τ =x,y,z v ′
N 
X  N N   
∂µif (τ ) X
X T
∂µif (τ ) T

h̄ ′
J kl J ml ′ (2vl + 1)
∂Q′′k 0 l=1
∂Q′′m 0 2ωl
k=1 m=1
(37)
 N   N 
′′
X ∂µif (τ ) X T

−2µif (Q0 , τ ) J kl Kl
∂Q′′k 0 l=1
k=1
 X N   N N   N #
∂µif (τ ) X T
X ∂µif (τ ) X T

+ J kl Kl J mn Kn
∂Q′′k 0
∂Q′′m 0
k=1 l=1 m=1 n=1

where the summation over the Cartesian coordinates (x, y, z) is explicitly given.
The choice of IQ tot or I tot in G AUSSIAN depends on the available data. In the case of absorption, the excited
′ Q′′
state is Q , so µif (Q′′0 ) and ∂µif /∂Q′′ are known. Hence, the analytic sum chosen as a reference to evaluate the
′′

convergence of the spectrum is IQ tot . In case of emission, the reference is I tot .

′′ Q′
a
The chosen analytic limit, I (I tot for FC calculations, IQ tot or I tot for FCHT/HT calculations) is used to calculate
′ Q ′′

the spectrum convergence (called in the G AUSSIAN output the spectrum progression) with the following relation:
′′ i2
P ′ P  ′ 
′ ρ ′′ h v | µif | v
spectrum progression = v v
(38)
Ia
In the current implementation of G AUSSIAN, the temperature is not yet taken into account, so the relation used in
practice is:
h 0 | µif | v ′′ i2
 ′ 
spectrum progression = (39)
Ia

3 Quick start
In its most concise form, the generation of a vibrationally-resolved electronic spectrum in G AUSSIAN requires a single
keyword representing the terms of the Taylor series given in eq. 5 that must be taken into account.

• FC: zeroth-order term of the Taylor series µif = µif (Q′′0 ) , suitable for fully-allowed transitions
 

h i
• FCHT: zeroth- and first-order terms of the Taylor series µif = µif (Q′′0 ) + N ′′ )Q′′ , suitable
P
k=1 (∂µ if /∂Qk k
for most cases spanning from fully-allowed to dipole-forbidden transitions
h i
• HT: first-order term of the Taylor series µif = N ′′ )Q′′ , suitable for dipole-forbidden transi-
P
k=1 (∂µ if /∂Qk k
tions or to show the contribution of the Herzberg-Teller approximation to the spectrum.

Additionally, G AUSSIAN will need the checkpoint file containing the data of the final state. This file must be given
after the standard input (in general, below the molecule specification), the data for the initial state being taken from
the current calculation.
Two methods can be distinguished to generate the vibrationally-resolved electronic spectra. The first one is to
simulate the latter as soon as the frequencies of the initial state have been calculated. In this case, two G AUSSIAN jobs
are required, assuming the structures of the initial and final states are already optimized:

1. Calculation of the frequencies for the final state

2. Calculation of the frequencies for the initial state followed by the generation of the spectrum.

10
 In the input file for the second step, the keywords FC, FCHT, HT is given in Frequency as follows (the example
below is limited to the case of the Franck-Condon approximation):

Link 0 Commands

Route section (# lines) with Frequency=(FC,SaveNM) NoSymm

Title section

Molecule specification

Name of the checkpoint file generated in step 1

In the example above, two additional keywords are used. While not mandatory, SaveNM will make G AUSSIAN
save the frequencies and normal modes in the checkpoint file. This option is necessary to deal with QM/MM calcula-
tions or when dealing with frozen atoms to extract correctly the normal modes and frequencies. As a consequence, it
is advisable to use it every time when computing the frequencies for both states and avoid possible issues caused by
the impossibility of G AUSSIAN to recompute correctly the eigenvalues and eigenvectors of the Hessian matrix.
Since this method requires the computation of frequencies each time a spectrum must be generated, it is not well
suited when diverse parameters must be tried for the generation of the spectrum. A second way of using G AUSSIAN
is through a 3-steps procedure where the last step of the previous method is broken down in two separate steps, first
the calculation of the frequencies and then the generation of the spectrum. Schematically, this gives the following
procedure:

1. Calculation of the frequencies for the initial state

2. Calculation of the frequencies for the final state

3. Generation of the electronic spectra

The order in which the first two steps are performed is not important and they can be done simultaneously. We
will however consider this sequence is respected in the following description
While this method seems more complex, it is also more flexible as G AUSSIAN works as a stand-alone program
dedicated to the simulation of the electronic spectra. The input files for the first two steps are classical ones to compute
frequencies with the keyword SaveNM used in Frequency. The input file for the last step is as follows:

Link 0 Commands
%Chk=Name of the checkpoint file generated in step 1

#P Geometry=AllCheck Frequency=(ReadFC,FC,SaveNM) NoSymm

Name of the checkpoint file generated in step 2

The options Geometry=AllCheck and Frequency=ReadFC are described in the G AUSSIAN manual. Ba-
sically, they let us extract all necessary data from the checkpoint file given in the line starting by %Chk= without the
need to perform long calculations before the procedure to generate the spectrum is called.
As mentioned before, this approach is particularly interesting when investigating the best settings to generate an
accurate electronic spectrum, using the parameters described in the section below.

11
Description of the output
Whatever the chosen procedure, the output will be similar and structured as follows. The block in the output related
to the generation of the spectrum starts with the following output:

**********************************************************************

Generation of the Franck-Condon spectrum

**********************************************************************

General information is then given on the type of calculation, the approximation of the electronic transition dipole
moment and the type of transition:

Approx. of the electronic transition dipole moment: FC

Type of transition requested: ONE-PHOTON ABSORPTION

The output presented above is an example showing the case of an absorption spectrum with the electronic transition
dipole moment considered as constant during the transition.
G AUSSIAN then describes the procedure of extraction of the data from the sources. In the case of our quick start
procedure. the output will be:

Data for initial state taken from current calculation.

Passed-in normal modes used.

Data for final state taken from checkpoint file "checkpoint file of step 1"
Normal modes recovered from file.

Initial state structure is set in Eckart orientation.

Final state structure is superposed to it.

The last line is followed by the geometry of the system in both states after reorientation. As explained in the de-
scription of the Franck-Condon principle, the calculation of the vibronic spectra requires the minimization of nuclear
displacements during the transition. Although rigorous separation between rotations and vibrations is, of course, not
possible, their coupling for the initial electronic state can be minimized by putting the molecule in the Eckart orienta-
tion (origin in the mass center and Cartesian axes along the principal inertia moments) using the standard G AUSSIAN
orientation for the ordering of Cartesian axes. Two options are next possible for the final electronic state, namely to
use its own Eckart orientation, or to minimize the angular momentum between initial and final structures. The second
one is the default because it minimizes the contribution of rotations to the shift vector and the Duschinski matrix.
Next, the molecule in the final electronic state is firstly rotated using the same rotation matrix of the reference
state and next a new rotation matrix is computed in order to minimize the angular momentum between initial and final
structure
NA
X
Mii X′i ∧ X′′i


(40)
i=1
where M is the diagonal matrix of the atomic masses, X′i and X′′i are the triplets of Cartesian coordinates for the
nucleus i in the initial and final states, respectively.
This two step procedure [10] allows to avoid numerical problems when the geometries and/or starting orientations
of initial and final states are strongly different. There are two options to build the rotation matrix, using Euler angles
[11–13] or quaternions [14,15], respectively. Although the latter approach (which is the default) should be more stable
and direct, there are cases in which it fails, so that, after a predetermined number of iterations, the program shifts to
Euler angles.
The actual calculation is then presented:

12
 --------------------------------------------------
Overlap integrals computation
--------------------------------------------------

Energy of the 0-0 transition: Calculated value cmˆ(-1)

Notes about the overlap integrals description:

- DeltaE is the relative energy of the transition (wrt 0-0)
- TDMIˆ2 is the square of the transition dipole moment integral
- Intensity is the line intensity (absolute value)

A list of the most intense transitions is then printed. They are listed by incrementing class. The general format is:

Initial State: <0|

Final State: |xˆx,xˆx>
DeltaE = x.xxxx | TDMIˆ2 = x.xxxxExxx, Intensity = x.xxxxExxx
........................................

The first line represents the initial state and the second line the final state. The third line contains the energy of the
transition with respect to the energy of the transition between the vibrational ground state, h 0′ | 0′′ i, the value
|h v ′ | µif | v ′′ i|2 , and the intensity of the corresponding spectrum line. The character x given above represents a
digit or a sign.
Once a class has been treated, the convergence of the spectrum is printed:

++ Spectrum progression: xx.xx%

Finally, the spectrum generated by G AUSSIAN is printed:

+------------------+
| Final Spectrum |
+------------------+

Axis X = Energy (in cmˆ-1)

Axis Y = Intensity (in arbitrary units)

------------------------------------------------------------
xxxxx.xxxx x.xxxxxxDxxx

4 List of options
While the default behavior of G AUSSIAN is set to give satisfactory results in most cases, it can be not well suited for
particular cases or poorly adapted to specific requirements from the user. A number of options exist to extensively
parameterize the calculations and provide a satisfactory analysis of the case of interest. These options need to be given
in a specific section referred to as the “ReadFCHT section”. This section lies below the standard G AUSSIAN input
and is activated by adding the keyword “ReadFCHT” in the option Frequency. The resulting G AUSSIAN input is
structured as follows:

13
 Link 0 Commands

Route section (# lines) with Frequency=(ReadFCHT)

Title section

Molecule specification

Optional additional sections used by links called before l716

ReadFCHT section

The description of the different sections above the ReadFCHT section can be found in the G AUSSIAN 09 manual
and is outside the scope of this document. The options available for the simulation of vibronic spectra using the
Franck-Condon principle are listed in the table below, organized by functions. To make the reading of the latter easier,
the lines are differentiated by alternating backgrounds of different colors. Following this scheme, two keywords in the
same “line” are equivalent (ex: ABS and ABSORPTION). Additionally, blocks are defined between two consecutive
horizontal black lines. All keywords given in the same block of lines correspond to the same option and are mutually
exclusive (ex: FC, FCHT and HT). In general, when several keywords related to the same option are given, only the
last one is taken into account and the others are discarded.
Note: Regarding the last group of options dealing with forced calculations, it is highly advisable to avoid their
use by default as they will bypass protections of G AUSSIAN regarding the soundness of the calculations. As a conse-
quence, they should be used carefully.

Keyword Values Description

General parameters of the calculations
FC None (Default) The electronic dipole moment is assumed constant during the
transition. This approximation is generally well suited for fully-allowed
transition.
FCHT None The electronic transition dipole moment is developped in a Taylor series
about the equilibrium geometry of the final state up to the first-order
term included. This approximation is more general than FC and works
on diverse cases, from fully-allowed to dipole-forbidden transitions.
HT None Only the linear variation of the electronic transition dipole moment with
the normal coordinates of the final state is taken into account (first-order
term of the Taylor series given in eq. 5).
ABSORPTION None (Default) The simulated vibronic spectrum corresponds to a one-photon
ABS absorption transition.
EMISSION None The simulated vibronic spectrum corresponds to a one-photon emission
EMI transition.
Sources of input data
CALC1 None (Default) Data for the initial state are taken from the current G AUSSIAN
job.
CHK1 None Data for the initial state are taken from the checkpoint file provided
below the ReadFCHT section.
OUT1 None Data for the initial state are taken from the G AUSSIAN output file pro-
vided below the ReadFCHT section.
CALC2 None Data for the final state are taken from the current G AUSSIAN job.

14
Keyword Values Description
CHK2 None (Default) Data for the final state are taken from the checkpoint file pro-
vided below the ReadFCHT section.
OUT2 None Data for the final state are taken from the G AUSSIAN output file pro-
vided below the ReadFCHT section.
NSTATE 1 Chooses the electronic state from which the data relative to the elec-
tronic transition dipole moments and the derivatives are taken.
User-given input data or modification options
InpDEner 0.0 (= Off) Replaces the calculated difference of energy between both electronic
states involved in the transition by the value given by the user. Such
value is expected in Hartree and is not taken in account if null.
SclVec None A mode-specific scaling algorithm based on the Duschinsky matrix is
used to compute new frequencies for the final state using the differences
between the calculated frequencies for the initial states and user-given
ones listed below the ReadFCHT section. The frequencies are expected
in cm−1 . See ref. [16] for more details on its behavior.
JDusch None (Default) Uses the correct Duschinsky matrix to perform the calcula-
tions.
JIdent None Uses an identity matrix instead of the Duschinsky matrix. In this case
the mode-mixing is ignored.
Parameterization of the G AUSSIAN output
PRTMAT 0 (= Off) Prints the Sharp and Rosenstock matrices or the Duschinsky matrices
and vectors. A digit is assigned to each matrix as follows:

1. Duschinsky matrix J. Please note that, with respect to eq. 14, JT

is printed.

2. shift vector K

3. Sharp and Rosenstock matrix A

4. Sharp and Rosenstock vector B

5. Sharp and Rosenstock matrix C

6. Sharp and Rosenstock vector D

7. Sharp and Rosenstock matrix E

The digits corresponding to the matrices to be printed are placed side

by side to form a number that is given as the value of PRTMAT. Ex:
PRTMAT=12 corresponds to printing the Duschinsky matrix and shift
vector.
ALLSPECTRA None Prints the spectra corresponding to each class in addition to the final,
total spectrum.
PRTINT 0.01 Prints data (assignment, energy, intensity) of the transitions which have
an intensity (|h v ′ | µif | v ′′ i|2 ) above or equal to 1% of the total
analytic limit.
Setting of the spectrum layout

15
 Keyword Values Description
SPECMIN -1000.0 Sets (in cm−1 ) the lower bound of the spectrum. The value is normally
given relative to the energy of the transition between both vibrational
ground states.
SPECMAX +8000.0 Sets (in cm−1 ) the upper bound of the spectrum. The value is normally
given relative to the energy of the transition between both vibrational
ground states.
NORELI00 None Using this keyword, G AUSSIAN expects the lower and upper bound of
the spectrum to be given in absolute values (not relative to the 0-0 tran-
sition). This keyword should be used together with both SPECMIN and
SPECMAX to avoid an unexpected behavior.
SPECRES 8.0 Resolution of the spectrum given by the difference of energy between
two contiguous measuring points (in cm−1 ). It is noteworthy that a
very high resolution (given by a small value of SPECRES) can greatly
hamper the computational speed of the spectrum generation.
SPECHWHM 135.0 Sets the Half-Width at Half-Maximum (in cm−1 ) for the convolution of
the spectrum using a Gaussian distribution. A large value of SPECHWHM
coupled to a small value of SPECRES can cause a steep increase of the
computational times.
Modification of the behavior of the a priori scheme to chose the most relevant transition
MAXOVR 20 Sets the maximum quantum number that has to be considered for each
MAXC1 mode of C1 (C1max ).
MAXCMB 13 Sets the maximum quantum number to consider for each mode involved
MAXC2 in the combination bands in C2 (C2max ).
MAXINT 100 Sets the maximum number of integrals (NImax , expressed in million) to
compute for each class above C2 .
NOINTAN None Deactivates the usage of analytic formulae to compute the overlap inte-
grals in classes C1 and C2 . By default, analytic formulae are used when
available for these classes.
Termination criteria for the spectrum generation
MAXBANDS 7 Sets the maximum class to consider. Calculations can end before if
there are not enough excited modes or DELTASP is also defined and the
condition imposed by the latter is met before.
DELTASP 0.0 Sets a condition on the spectrum convergence. If the spectrum progres-
sion with respect to the total intensity between two consecutive classes
is below DELTASP (in absolute value), then the calculations end.
Forcing calculations4
ForceFCCalc None Forces the calculation of the Franck-Condon integrals even if the overlap
integral between the vibrational ground states of the initial and final
states is very low (|h 0′ | 0′′ i|2 < 10−4 ).
ForcePrtSpectrum None Forces the printing of the final spectrum even if the total progression is
low (below 50%).

The structure of the G AUSSIAN input file starting from the ReadFCHT section is as follows, each optional block
of data (“SclVec” and “Sources”) being terminated by a blank line:
4
Warning: The keywords describe in this section should be used with care as they might lead to results whose reliability is unpredictable

16
 ReadFCHT section
[blank line]
User-given frequencies for the initial state (necessary for SclVec) ‘‘SclVec’’ block
[blank line]
Name of the file containing the data for the initial state (if different from the cur- ‘‘Sources’’ block
rent G AUSSIAN job)
Name of the file containing the data for the final state (if different from the cur-
rent G AUSSIAN job)
[blank line]
By default, only the name of the checkpoint file containing the data of the final state needs to be given after the
ReadFCHT section.

5 Some practical examples

The examples described here are taken from reference [17] and shows the parameters used in the ReadFCHT section
to carry out the calculations. The G AUSSIAN inputs presented here are based on the 3-steps usage of G AUSSIAN,
where the last step is dedicated to the generation of the spectrum.

Photodetachment spectrum of SF−

6 : parameterizing the prescreening method

In the photodetachment spectrum of SF− 6 , a significant elongation of the S-F bond upon electron attachment is found.
As a consequence, the potential curves of the normal mode corresponding to the S-F symmetric stretching are no-
ticeably displaced (see Fig. 2 in ref [18]). When generating the photodetachment spectrum of SF− 6 with the standard
parameter, an insufficient value of MAXC1 is set and only a partial spectrum is obtained. As a consequence, the
following input file is used for the generation of the complete spectrum:

%Chk=SF6_neutral_freq.chk
#P Geom=AllCheck Freq=(ReadFC,FC,SaveNM,ReadFCHT) NoSymm

PRTMAT=12 MAXC1=100 MAXBANDS=4 SPECHWHM=0.5

SF6_anion_freq.chk

C1max is set to a safe, high value since the cost of the prescreening is cheap. When performing the calculation,
G AUSSIAN will find that it does not need to go up to class 7 and will safely stop the calculations. Here, we set
manually the correct limit.

S1 ← S0 absorption spectrum of Anisole: parameterizing the spectrum layout

To compare the theoretical spectrum obtained with G AUSSIAN and the experimental REMPI spectrum, it is necessary
to modify the default values of the parameters corresponding to the spectrum layout. This is done with the following
input

%Chk=anisole_S0.chk
#P Geom=AllCheck Freq=(ReadFC,FC,SaveNM,ReadFCHT) NoSymm

PRTMAT=12 NORELI00 SPECMIN=37900 SPECMAX=42000 SPECHWHM=1. SPECRES=1.

anisole_S1.chk

17
 The absolute values of the spectrum bounds are given here and a very fine spectrum is requested.
Although the spectrum layout obtained this way is very similar to the experimental one, an energy shift was
observed for several peaks, with the frequencies calculated at the harmonic level. To correct this, anharmonic fre-
quencies had to be used. However, they could not be straightforwardly obtained for the final, excited state with the
time-dependent density functional theory (TD-DFT) computations. Hence, the mode-specific scaling scheme [16] im-
plemented in G AUSSIAN and available with the keyword SclVec was used. The anharmonic frequencies calculated
for the ground state [19] were used as input as follows:

%Chk=anisole_S0.chk
#P Geom=AllCheck Freq=(ReadFC,FC,SaveNM,ReadFCHT) NoSymm

PRTMAT=12 NORELI00 SPECMIN=37900 SPECMAX=42000 SPECHWHM=1. SPECRES=1.

SclVec

86.39 200.38 252.59 257.52 417.46 437.57 510.86 554.26 623.74

696.93 757.21 784.32 811.61 883.10 958.95 981.60 994.62 1018.63
1040.32 1082.75 1146.70 1164.42 1177.75 1177.40 1238.75 1305.54 1336.45
1467.80 1454.61 1479.55 1465.73 1494.69 1582.76 1601.40 2869.62 2902.51
2993.57 3006.94 3042.47 3047.04 3068.53 3077.61

anisole_S1.chk

G AUSSIAN then corrects the harmonic frequencies of the excited state with a mode-specific scaling to account for
the anharmonicity. Using the harmonic frequencies present in the checkpoint (ω ′harm ) and the user-given data (ω ′anh ),
a scaling vector α′ is calculated:
αi′ = ωanh
′ ′
/ωharm
Using the square elements of the Duschinsky matrix, the scaling vector for the excited state α′′ is deduced from
α′ with the relation,
N
X
αj′′ = Jij 2 αi′
i=1

The new anharmonic frequencies ω ′′anh are then calculated as:

′′
ωanh = αi′′ ωharm
′′

G AUSSIAN prints in output the scaling vectors α′ and α′′ as well as the frequencies that will be used in each state
for the calculation of the vibonic spectrum.

6 Description of the errors reported by G AUSSIAN

Several controls are performed in G AUSSIAN to check the validity of the calculations. Depending on the origin of
the problem found, calculations can be aborted with an error or a simple warning can be issued. While the former
will result in a termination of the calculations, the latter are there to draw the attention of the user on potential issues.
In this section, we present a brief overview of the most common error and warning messages, their origin as well as
possible workarounds or leads that should be investigated to solve them.

Errors
- Cannot extract data for both states from the current calculations
This error is due to the fact that G AUSSIAN attempts to extract all data from the current job.

18
 It is likely that CALC1 has been given without setting the source of data for the final state. By default, G AUSSIAN
will have set the current calculation for the final state.
- WARNING: Charge (or Multiplicity) is different between initial and final states. Calc. cannot be done out
with the trans. dip. mom. in the data files
This error is raised if a FCHT or HT calculation is requested and G AUSSIAN cannot use the transition electronic
dipole moment and its derivatives that were found in the source files.
For this kind of calculation, the transition is generally fully-allowed and the FC approximation should be sufficient.
- ERROR: The Franck-Condon factor corresponding to the overlap integral between both vibrational ground
states is too small: |<0’|0">|ˆ2 = x
This error indicates that it is very likely that the underlying theory used to generate the spectrum, the Franck-
Condon principle, is unsuitable for the case of interest. This is a typical issue when the equilibrium geometry of the
initial and final states are noticeably different. This can happen when one of the structures is planar in one electronic
state and not planar in the other electronic state.
Another indicator of this problem is the Duschinsky matrix (which can be displayed with the option “PRTMAT=1”
in the ReadFCHT section) that shows the projection of the normal modes of one state on those of the other state. In
this case, the Duschinsky matrix is rather diffuse, that is to say that many normal coordinates of the final state are
expressed by a combination of a rather large number of normal modes of the initial state.
If the calculations must be carried anyway, this error can be bypassed with the keyword ForcedFCCalc. How-
ever, please note that the resulting spectrum will be likely unreliable and the corresponding calculations should be
considered as tests.
- ERROR: The final spectrum progression is too low and the resulting spectrum may be unreliable. To print
the spectrum nevertheless, please use the keyword FORCEPRTSPECTRUM.
This error is often related to the same reasons as the previous one, that is to say that the Franck-Condon principle
can be unsuitable to the case of interest.
If this it is not the case, the problem may be caused by too low number of calculated transitions. Several options
may resolve this problem. The first one is to increase MAXBANDS which correspond to the highest class to be treated.
If it has been set very low, then the spectrum convergence will be limited. Other possibility is that the values of
MAXC1/MAXOVR and/or MAXC2/MAXCBI are set too low. If this is the case, warnings should have been issued during
the calculations:
Warning: insufficient num. of quanta for mode i,
Finally, this error can also be caused by a low number of transitions to compute in each class (MAXINT).

Warnings
- WARNING: Value for PRTINT is too low. Reset to 1.0D-6. Be careful that it can cause a very verbose output
of Gaussian.
Setting a very low value of PRTINT might cause the generation of a very large output file. PRTINT is given as a
fraction of the total analytic sum. As a consequence, setting PRTINT=1.D-3 for example will print the assignment
of all integrals which contribute to at least 0.1% to the total intensity.
- Warning: insufficient num. of quanta for mode i,
G AUSSIAN found that the condition FC1 (i, C1 max ) ≥ ǫ1 was true for the normal mode i. Such a condition means
that the parameter MAXC1 must be set with a higher value that the current one (by default, MAXC1=20).

References
[1] J. Franck, Transactions of the Faraday Society 21, 536 (1926).

[2] E. Condon, Physical Review 32, 858 (1928).

[3] F. Santoro, R. Improta, A. Lami, J. Bloino, and V. Barone, The Journal of Chemical Physics 126, 084509 (2007).

[4] F. Santoro, R. Improta, A. Lami, J. Bloino, and V. Barone, The Journal of Chemical Physics 128, 224311 (2008).

19
 [5] C. Eckart, Physical Review 47, 552 (1935).

[6] G. Herzberg and E. Teller, Zeitschrift für Physikalische Chemie - Abteilung B 21, 410 (1933).

[7] F. Duschinsky, Acta Physicochimica URSS 7, 551 (1937).

[8] T. E. Sharp and H. M. Rosenstock, The Journal of Chemical Physics 41, 3453 (1964).

[9] P. T. Ruhoff, Chemical Physics 186, 355 (1994).

[10] B. Van der Veken, G. Pieters, and M. Herman, Theochem 91, 387 (1983).

[11] S. Ehrenson, Journal of the American Chemical Society 96, 3778 (1974).

[12] W. Miller, B. Ruf, and Y. Chang, The Journal of Chemical Physics 89, 6298 (1988).

[13] C. Zhixing, Theoretical Chemistry Accounts 75, 481 (1989).

[14] G. Kneller, Molecular Simulation 7, 113 (1991).

[15] G. Kneller, The Journal of Chemical Physics 88, 2709 (1991).

[16] J. Bloino, M. Biczysko, O. Crescenzi, and V. Barone, The Journal of Chemical Physics 128, 244105 (2008).

[17] V. Barone, J. Bloino, M. Biczysko, and F. Santoro, Journal of Chemical Theory and Computation 5, 540 (2009).

[18] J. C. Bopp, J. R. Roscioli, M. A. Johnson, T. M. Miller, A. A. Viggiano, S. M. Villano, S. W. Wren, and W. C.

Lineberger, Journal of Physical Chemistry A 111, 1214 (2007).

[19] V. Barone, The Journal of Chemical Physics 122, 014108 (2005).

20