HAL Id: jpa-00210303
https://hal.archives-ouvertes.fr/jpa-00210303
Submitted on 1 Jan 1986
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Semi-empirical evaluation of Sr-Ar difference potentials
Le Quang Rang, D. Voslamber
To cite this version:
Le Quang Rang, D. Voslamber. Semi-empirical evaluation of Sr-Ar difference potentials. Journal de
Physique, 1986, 47 (7), pp.1149-1154. �10.1051/jphys:019860047070114900�. �jpa-00210303�
Semi-empirical evaluation of Sr-Ar difference potentials
Le Quang Rang
Département de Recherches Physiques, Tour 22, Université Paris VI, 75230 Paris Cedex 05, France and D. Voslamber
Association EURATOM-CEA, Département de Recherches
surla Fusion Contrôlée,
B.P. N° 6, 92265 Fontenay-aux-Roses, France (Reçu le 27 dgcembre 1985, accepté le 6
mars1986)
Résumé. 2014 En essayant de faire coïncider
unprofil d’absorption théorique
avec unprofil expérimental,
nousavons
tenté, pour le
casde la raie de résonance de strontium perturbé par argon, d’évaluer les deux premières
différences de potentiels prises
sousla forme de Lennard-Jones. Le profil expérimental utilisé est celui obtenu
dans
uneexpérience de Carlsten et al. Pour le profil théorique
nous avonsemployé
uneexpression
«d’aile unifiée » basée
surl’approximation des trajectoires rectilignes pour les noyaux et
surl’approximation adiabatique pour les états électroniques.
Abstract.
2014By fitting
atheoretical to
anexperimental absorption profile of the strontium
resonanceline per- turbed by argon,
wehave attempted
anevaluation in the Lennard-Jones form of the first two Sr-Ar difference potentials. The experimental absorption profile used in the procedure is taken from Carlsten et al’s redistribution
experiment. For the theoretical profile
aunified wing formula is used which is based
onthe classical straight path approximation for the nuclei and
onthe adiabatic approximation for the electron states.
Classification
Physique Abstracts
32.70J - 34.15
1. Introduction.
The amount and accuracy of information about inter- atomic interactions and collision dynamics have been
increased substantially by recent light-scattering expe- riments using tunable dye lasers [1-4]. In particular,
the redistribution of radiation close to
anatomic
resonance has proved to be an effective tool for stu- dying such collision complexes in which an optically
active atom in a selected state interacts with an
optically passive perturber. To the extent that the
collision occurs adiabatically, the scattering process is significantly determined by the corresponding
interatomic potentials and dipole strengths. These
are
therefore essential for the characteristics of the
spectroscopic data (such as e.g. line shapes and frequency-dependent polarization degrees) and may, in turn, be determined from these data
onthe basis of suitable theories.
In practice, the determination of potential and dipole curves from measured line shapes or polariza-
tion degrees, is difficult because the functional depen-
dence between the observed spectral data and the
quantities to be determined is rather involved. Com-
plications may arise from the dynamics of the combined collisional and radiative process, from the multiplicity
of the transitions associated with a given frequency interval, and from the statistics of the problem.
Therefore, many of the previous investigations have
not strictly proceeded to determine the interatomic parameters from the observed spectral data. Instead they were confined to comparing these data with
corresponding theoretical results which were based
on
interatomic parameters obtained by independent
methods. The extent of agreement between experiment
and theory then provided some information about the quality of the interatomic parameters used in the
theory. On the other hand, in those papers where interatomic parameters
wereextracted directly from
observed data, the complexity of the problem had to
be reduced by simplifying assumptions such
asthe impact theory for the line body, the quasistatic theory for the far line wings and, in nearly all cases,
the existence of only one transition (i.e. one lower
and one upper potential) contributing to the line.
Thus if
aquasistatic line wing is due to a single tran-
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019860047070114900
1150
sition and if one disregards the complications arising
from multiple Condon points, it can be directly
converted into the corresponding potential
curves[5-7]. To extract
adifference potential from the impact-broadened line body,
oneusually employs
the method introduced by Behmenhurg [8] : the potential is assumed to be of the Lennard-Jones form whose two constants
canthen be determined from the width and the shift of the line. Either
orboth of these methods have been used in nearly all papers
dealing with e.g. alcaline-earth lines perturbed by
noble gases. For the Sr-Ar case, which will be studied in the present paper, see e.g. references [1] and [9-13].
In our present investigation we have attempted to
go beyond the procedures mentioned above by using
a
dynamical (as opposed to quasistatic) theory for the semi-empirical evaluation of a pair of difference potentials corresponding to a E - E and a.E - 17 transi- tion. We assume that both potentials can be approxi-
mated in the Lennard-Jones form, so that the task is to determine the four parameters AC6-, ðC;2’
AC6’, OC 2. We have chosen the resonance transition of strontium perturbed by argon,
asfor this case
detailed experimental data are available (see references quoted above and in particular Carlsten et al. [1],
henceforth referred to as CSR). In section 2 we will
present
abrief survey of the line broadening theory employed in
ourinvestigation. Section 3 will summa-
rize the results for the Sr-Ar difference potentials
which we have obtained from
abest-fit with the redistribution experiment of CSR. Some concluding
comments on these results will be given in section 4.
2. Theory.
It has been emphasized in a previous publication [14]
that results obtained from light scattering experiments
are closely related to those obtained in emission or
absorption spectroscopy. In particular, the total integrated fluorescence signal as a function of the
incident detuning from the atomic resonance is to a
very good approximation proportional to the absorp-
tion profile of the resonance transition considered. We therefore have chosen to take as point of departure a previously established formalism for the wings of
aspectral line broadened by heavy perturbers (Le Quang Rang and Voslamber [15], henceforth referred to as LV). In the spectral range where the broadening
occurs adiabatically, this formalism provides a
«uni-
fied » wing expression in the sense that it describes the transition from the close (dynamic) to the far (possibly quasistatic) parts of the line wings. The
formalism is based on the following three assumptions : (a) the one-perturber approximation,
(b) the classical straight-path approximation for the
two nuclei involved in the collision complex,
(c) the adiabatic approximation for the electronic states.
As opposed to the unified Franck-Condon theory
of Szudy and Baylis [16]
ourtreatment, though less sophisticated with regard to the dynamics of the nuclei,
has the advantage of including the adiabatic rotation of the dipole associated with the radiative transition.
It might be argued that
acollision is
neveradiabatic in all its stages; in particular, after radiative excitation of
acollision complex at small internuclear separation,
the two particles fly apart to spend a long time (of the
order of the natural life time) at large distances before
reemitting the fluorescence photon. During this time, the dipole decorrelates from the nuclear motion, and
the present treatment might seem unsatisfactory. The
answer to such an objection is that the absorption profile is only sensitive to what happens during the
radiative excitation process, i.e. during the relatively
short instant when the interatomic potential is near-
resonant with the incident radiation. During this
instant the rotation of the dipole is important. Whether
later
onthe dipole continues to rotate or not is irrele-
vant to the total integrated fluorescence which is
uniquely determined by the strength of the transition
dipole, not by its direction.
While assumptions (a)4c) suffice for deriving the general line-wing expression, equation (1) presented below,
wehave made three further approximations
for numerical convenience :
(d) the use of
arepresentative «thermal velocity »
instead of
anaverage over velocities,
(e) the use of the dipole strengths of the unperturbed optically active atom for all atom-perturber distances, (f) the neglect of the Boltzmann factor (pa =1
in Eqs. (2), (3), (7) below).
As for the quality of approximations (a)-(f) in the
case we will be considering, we first note that there
should be
noproblem with approximation (a) because
the frequency domain investigated lies well beyond
the halfwidth of the spectral line. The validity of approximations (b) and (c), however, depends on the potentials to be determined and must be confirmed
a
posteriori. Since the order of magnitude of the potentials was known in advance, the approximate
domain of adiabaticity could be delimited from the outset; this led us to exclude the experimental point at d
= -5 cm-1 (see Sect 3) from our best-fit procedure. The final results then confirmed that the adiabatic approximation breaks down for I.J (
7 cm - 1 but holds well in the remaining part of the wing. The overall validity of the straight-path approxi-
mation was also confirmed, except perhaps in the region of the E -II potential well (mainly made up
by the excited II potential) whose depth is about 30 %
of the thermal energy of a particle. Since on the line profile this corresponds to the far-wing satellite at
=- 278 cm-1, the shape of this satellite may suffer
some influence from approximation (b). The bearing
of approximation (d) can hardly be judged without
performing the average over velocities; however,
there is an indication from the work of Caby-Eyraud
et al. [17] on Stark broadening by electrons that the
error connected with using a thermal velocity should
be less than 10 %. (It is in fact of the order of 5 % in the
case considered by these authors.) Approximation (e) is currently used in line-shape calculations because
as
amatter of experience the dipole strength of an
allowed line usually depends little on the atom- perturber separation. Approximation (f) is of a similar
nature as approximation (b) but refers uniquely to
the groundstate potential which is not determined in the present paper. The dependence of this potential
on the atom-perturber distance is expected to be
weaker than that of the excited potentials so that approximation (f) should be at least as good as approximation (b). An ultimate judgement of this approximation, however, can only be given once
the groundstate potential has been determined, too.
Let us now turn to the line-shape formalism derived by LV. We define 4ro) to be the pressure-broadened absorption profile (without Doppler broadening)
normalized to the total sum of dipole strengths contributing to the spectral line under consideration.
As compared to equations (1) and (9) of LV, we omit
the factor (colcoo)P (p =1) because the absorption profile occurring in the redistribution function is not defined to be the shape of the absorption coefficient
K(ro) (absorbed intensity per unit length and unit frequency) but is the shape of K(ro)/C1iro (absorbed
number of photons per unit length and unit frequency).
As shown in LV, the above assumptions (a)-(c) then
lead to the following expression for the wings of the profile L(w) :
Here, n denotes the number density of perturbers and fM(v) is the Maxwell distribution of relative atom-
perturber velocities
v(involving the reduced atom-
perturber mass). The subscripts
aand fl number the
lower (a) and upper (P) molecular electronic energy
eigenstates corresponding to given positions of the
nuclei in the collision complex. In order to specify
the quantities A’ , a# B.’ .8 (k = 1, 2, 3), we have to dis-
tinguish between two types of radiative transitions
a H
p, namely between those involving no change of
the quantum number A associated with the electronic
angular momentum about the internuclear axis
(AA 0,
«parallel transitions ») and those for which AA
=+ 1 (o perpendicular transitions »).
2.1 PARALLEL TRANSITIONS.
-For transitions a +-+ p
with AA
=0, the A’ are given by (see Eqs. (16) of LV)
These expressions involve the Boltzmann factor
and the phase integral
where ga is the multiplicity of the lower states a, and
Ey(r) is the energy eigenvalue of state y (for inter-
nuclear distance r) plus the repulsion energy of the
two nuclei in the collision complex. Further, the quantity dflfJV bZ + r) is defined from the relation
where Dczp(b, cp, z) is the transition dipole for a relative (perturber to atom) position characterized by the cylindrical coordinates b, (p,
z(the z-axis pointing parallel to the relative velocity v), and eb and ez are unit vectors in the collision plane pointing respectively perpendicular and parallel to the z-axis. As can be
seen from the unit vector (beb + zez)IJb2 + r, the
transition dipole Dczp points parallel to the internuclear axis.
2.2 PERPENDICULAR TRANSITIONS.
-For transitions
a
- fl with AA = ± 1, the formal expressions for
A 1 , B,,’P; A2and B 2 differ from those given in equa- tions (2) and (3) only by a factor of 1/J2. If, as is
usually done, the transitions A --+ A’(A’ = A ± 1)
and - A - - A’ (which yield the same contribution
to the line intensity) are considered as one transition
(I A I -+ I A’ I) the factor of 1/,/2- does not occur and the A;p (k = 1., 2) are formally the same as in equations (2)
and (3).
1152
The quantities A2* and 82*
aregiven by
Again the factor 1/g/2 in this expression is to be dropped if
nodistinction is made between the tran- sitions A -+ A’ and - A - - A’. The functions pa and 4JafJ occurring in A’ (k
=1, 2, 3) are formally
the same as in the case of parallel transitions (see Eqs. (4) and (5)), while the function dafJ is now defined
from the relation
where etp
=e,,
xeb. As can be seen from this equation, D,,,.8 has only components perpendicular to the inter-
nuclear axis.
3. Semi-empirical determination of Lennard-Jones constants.
We have applied the formalism of section 2 to the
resonance line 51 So - 5 1 Plat 460.73
nmof Sr per- turbed by Ar,
asmeasured in the light scattering experiment of CSR. This line is characterized by two
transitions occurring from the E: ground potential to
the excited E and H potentials of the Sr-Ar collision
complex. We assume that the corresponding difference potentials can be approximated in the Lennard-
Jones form, i.e. (wither being the resonance frequency)
where
adenotes the lower (E) potential and P either
of the upper potentials E and H. For convenience,
we also introduce the dimensionless constants C6, Cf2 defined from
Inserting the rhs of equation (8) into equation (5) yields
with
Using this result and choosing the physical para- meters in accord with the experiment of CSR,
wehave
evaluated the wing formula (1) numerically
onthe
basis of approximations (d)-(f), and have converted it to the
crosssection for collision-induced fluores-
cence.
We have performed this calculation for various
sets of constants C6, C 2, C6 , C 2 so
asto get optimal agreement with the measurements
aspresented in figure 11 of CSR. This
«best-fit procedure » had to be
restricted to the red wing of the line. The
reasonwhy
any attempts to also include the blue wing were
without success, is probably that this part of the profile
is determined,
onthe one hand, by very small inter- nuclear distances where the Lennard-Jones approxi-
mation becomes poor, and on the other hand, by
transitions from the virtual level (laser detuning above resonance) to the real potential curves below reso-
nance ; it appears that such transitions are not satis-
factorily accounted for by the adiabatic formalism of section 2. All calculations we carried out using
realistic Lennard-Jones parameters led to blue wing
intensities which were about five times smaller than the experimental ones.
The best-fit procedure on the red side was facilitated
by the fact that the far red line wing (detunings
I L1 > 60 cm-1) is anti-static for the E - E transition and is thus uniquely determined by the E - 17 tran-
sition, i.e. by the constants C6 and C 2. In this
spectral range the procedure could therefore be
limited to a two-parameter fit which led to the result
C6
=4.7, C 2
=0.10. For these values the agreement with the three experimental points existing in that
range (without regarding their error bars)
wasbetter
than 3 %. The corresponding profile shows a satellite
near A - 250 cm-1 (situated between the last two
experimental points of the red wing). This satellite is much closer to line centre than that conjectured by
Lewis et al. [18] and is also contrary to the way in which CSR have drawn the curve through their experimental points. Inclusion of the error bars would open the possibility of drawing a somewhat patholo- gical curve which, if touching the extreme bar limits, could have negative curvature in the frequency
domain considered. Such a curve, which seems rather
unlikely, might be compatible with a satellite far out in the wing as considered by Lewis et al.
In contrast to the far line wing, the close part of the wing is not determined by a single transition, but is
made up by contributions from both the E - E and the Z - 17 transition with comparable order of magnitude.
We therefore subtracted from the experimental profile
the contribution of the Z - H transition determined above and took the remaining intensity contribution
as a
basis for a best-fit procedure with calculated f - E curves. Small detunings (I L1 ( 7 cm-1)
wereexcluded from the procedure in order to avoid falsi- fication arising from non-adiabatic effects. The two-
parameter fit in the remaininlspectral domain then
led to the result C6
=7.3, C;2
=2.46. For these
values, the agreement with the experimental points
was better than 8 %.
The degree to which the theoretical intensities can
be brought to coincide with the experimental ones (3 % for the Z - H transition, 8 % for the L - L tran-
sition) is not necessarily
ameasure for the precision
of the Lennard-Jones constants obtained or for the
precision with which Lennard-Jones potentials can approximate the real potentials. Indeed, in the best-fit
procedure described above, any errors of the theore- tical and experimental intensities have been ignored.
While it seems hard to determine the theoretical uncertainties connected with approximations (a)-(f)
in section 2, the experimental uncertainties are known
as they
aremarked by error bars in figure 11 of CSR.
On the far wing, which has served to determine the
L - H transition, the experimental errors are about 15 %. As can be
seenfrom the various trial calculations
we
have performed, this corresponds to
anuncertainty
of about 25 % for the Cf-constant and of about 50 %
for the C 2-constant Note, however, that the uncer-
tainties of the two constants are not independent of
each other since the ratio (C6 )2/C 2 was found
empirically to be invariant to better than 5 % (which
can be interpreted as an invariance of the potential
well depth). Thus, multiplying C6 =4.7 by 1 + e(l e 1;$
0.25) brings the theoretical intensity out of the experi-
mental 15 % error bars if Cf2 is not multiplied by (1 + S)2 ± 0.05.
In the close wing region which has served to deter- mine the E - E transition, the experimental errors
attain about 20 % as referred to the total intensity.
This amounts to more than 40 % after subtraction of the L - 17 transition. Note that only small additional uncertainties arise from the errors of the subtracted L - n transition. Indeed, for the close line wing,
this transition occurs in the van der Waals region
where the intensity is proportional to (C6 ) 1l2 (in the quasistatic approximation). The 25 % error in C6
stated above would thus entail a 12 % error in the
Z - H intensity, and the relative error of 0.40 mentioned above would be increased to only (0.42 + 0.122)1/2
=0.42. Unfortunately, a 40 % error
in the intensity of the E - Z transition implies a much larger error in the determination of the constants C6 and C12’ Calculations for C6
=4, C12 = 0.56 and Cl = 14, Ci2
=6.9 showed that these values are
still compatible with the outermost points of the experimental error bars. A pessimistic error estimation
therefore yields the result C6 = 7.3 only up to
afactor of 2, and the result C 2
=2.46 only up to a factor of 3 or 4. Again, the two errors are not inde-
pendent of each other but linked by the approximate
invariance of the ratio (Cl)2/Cf2 which in the case of the Z - Z transition holds to about 24 %.
It is worth noting that the error limits stated above appear much more tolerable if they
aremapped into
those of the depth Eo
=(åcg)2/4 åcf2 and distance rfl
=(2 ACO fl /6 of the potential well. These
constants may be taken as Lennard-Jones parameters instead of ACf and Acf2- As mentioned above, for
the Z - II transition the error of Eo is only 5 %.
Since ra
=(ACf 12 Eff)1/6, the 25 % error for AC6
reduces to an error of about 4 % for ro’. In the case of
the Z - Z transition, the error of E0 was stated to be
24 %. The error of r§
=(AC6/2 Et)1/6 is composed of
a 12 % error arising from the factor of 2 uncertainty in ACI and a 4 % error arising from the 24 % uncertainty
of E5. The total relative error of r5 may therefore be estimated as (0.122 +0.04 2)1/2 0.13.
In terms of the depths and distances of the potential wells, our results may finally be expressed
asfollows :
In figure 1 we have plotted the red wing of the cross
section
6for collision-induced fluorescence, calculated
with the above values for Eo, r-, Eo and ro (solid line). The experimental results (circles with error bars)
of CSR from which these values have been determined
by
abest-fit procedure, are shown for comparison.
In order to display the effects of nuclear motion,
we
have confronted
ourdynamical profile with the quasistatic one, based on the same Lennard-Jones parameters (dotted line). Also shown (dashed line)
is the result of a dynamical calculation using Eo = 1 000 cm-’, r’ = 3.4 A, Eo = 20 cm -1,
Fig. 1.
-Red wing of the
crosssection
afor collision- induced fluorescence. Circles with
errorbars : experiment
of CSR. Solid
curve :theoretical result obtained from
abest-fit with CSR (corresponding to the Lennard-Jones
constants given in Eqs. (9)). Dashed
curve :analogous
theoretical result based
onthe Lennard-Jones constants
conjectured in reference [18]. Dotted
curve :quasistatic
result based
onthe Lennard-Jones constants given in equa-
tions (9).
1154
r’
=6.6 A. These values had been conjectured by
Lewis et al. [18] (by also considering the experiment
of CSR) for a calculation of the polarization of redis-
tributed light. While their constants for the E - f transition are remarkably close to ours, there is
asubstantial difference for the E - H transition, espe-
cially with regard to the depth of the potential well
which determines the position of the second satellite
on the far wing.
For I A I 7 cm-1 the dynamical curves deviate significantly from experiment. As mentioned already,
this domain has been excluded from the best-fit
procedure because it is dominated by non-adiabatic effects which are not included in our formalism.
The main non-adiabatic effect is that for large atom- perturber distances the radiating dipole does no longer follow the rotation of the internuclear axis.
Therefore, one of the main causes for dynamical broadening (amplitude modulation due to the rotation of the dipole) disappears when approaching the line
centre. This is the reason why the experimental near wing intensity in figure 1 drops below the calculated
adiabatic intensity.
4. Concluding remarks.
A best-fit procedure based on Carlsten et al’s experi-
ment [1] ] and on an adiabatic line wing formalism developed previously (Le Quang Rang and Voslamber [15]) has led us to a new determination of Lennard- Jones constants for the first two Sr-Ar difference
potentials. Agreement with the measurements on the far red wing could only be obtained by assuming a
satellite located between the last two experimental points. We suggest that this satellite be confirmed
(or refuted) by measurements at intermediate detun-
ings. In case it cannot be confirmed experimentally,
we