HAL Id: jpa-00209202
https://hal.archives-ouvertes.fr/jpa-00209202
Submitted on 1 Jan 1979
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.
Relaxation in the 3T1u state of F centres in CaO
J. Cibert, P. Edel, Y. Merle d’Aubigné, R. Romestain
To cite this version:
J. Cibert, P. Edel, Y. Merle d’Aubigné, R. Romestain. Relaxation in the 3T1u state of F centres in CaO. Journal de Physique, 1979, 40 (12), pp.1149-1160. �10.1051/jphys:0197900400120114900�.
�jpa-00209202�
Relaxation in the 3T1u state of F centres in CaO
J. Cibert, P. Edel, Y. Merle d’Aubigné and R. Romestain
Laboratoire de Spectrométrie Physique (*), Université Scientifique et Médicale de Grenoble,
B.P. 53X, 38041 Grenoble Cedex, France
(Reçu le 1 er juin 1979, accepté le 29 août 1979)
Résumé.
2014Nous avons détecté optiquement la relaxation dans l’état excité 3T1u du centre F dans CaO en utili-
sant des hyperfréquences en bande X (9 GHz). A 1,7 K, on observe deux composantes : une composante lente égale à la durée de vie radiative (3,7 ms), due à la variation de la population totale du niveau 3T1u ; une compo-
sante rapide, due à des transitions d’un puits Jahn-Teller à l’autre. On montre que les composantes tétragonales
et les composantes trigonales des déformations induites par les phonons doivent être prises en compte. Ces tran- sitions dépendent beaucoup de la valeur des contraintes internes statiques. En utilisant la répartition de contraintes, déterminée dans un travail précédent, on peut reconstruire par un calcul numérique les signaux expérimentaux.
Abstract.
2014We have studied by optical detection the relaxation in the 3T1u excited state of F centres in CaO
using microwaves in the X band. At 1.7 K both a slow and a fast recovery are observed. The slow component, equal to the radiative lifetime (3.7 ms) is due to a variation of the total population of the 3T1u level. The fast component is due to transitions from one Jahn-Teller well to another. Evidence is given that both the tetragonal
and the trigonal components of the phonon induced strain field have to be considered. All these relaxations depend
very much on the static internal strains. Numerical calculations using the distribution of strains determined in previous work fit the experimental data.
Classification Physics Abstracts
76.30M - 78.55
76.30M201378.55
Introduction.
-The spectroscopic behaviour of the F centre in CaO is now well understood : optical
spectra have been studied by Henderson et al. [1, 2], optically detected magnetic resonance (ODMR) by
Edel et al. [3]. Uniaxial stress experiments [4, 5] show
the effect of strain on either the zero phonon line or
the ODMR lines, confirm the inhomogeneous nature
of these lines, and give some information concerning
the distribution of internal strains [6]. References to
previous work are given in several papers [7, 8].
The excited state is both an orbital and a spin triplet 3Tlu. It is strongly coupled to Eg tetragonal modes
of vibration and undergoes a static Jahn-Teller effect
(JTE) [9]. The static nature of the JTE can be seen by
ODMR [3] : three equivalent tetragonal spectra are observed, each of these spectra corresponding to one
of the Jahn-Teller wells. The ground vibronic level is still a Tlu triplet [9]. Inside this triplet, matrix elements of off-diagonal electronic operators (when expressed
in the real basis 1 X >, 1 Y), 1 Z » such as spin-orbit interaction, dipole-dipole interaction and coupling to
the trigonal T2g strains are strongly reduced [9]. This
is not the case for the effect of the tetragonal Eg strains
eo, ee which is represented in the real basis by operators
(*) Laboratory associated with the Centre National de la Recher- che Scientifique.
having only diagonal matrix elements. As a result of the coupling to internal strains of Eg symmetry, the
orbital degeneracy of the 3Tlu ground vibronic level is lifted as shown in figure 1. In the relaxation effects
we will see that both the tetragonal and the trigonal
components of the time-dependent-strain induced by
the phonon field have to be taken into account.
Fig. 1.
-Level scheme of the 3T 1 u multiplet. The magnetic field
is aligned along the z cubic axis [001]. The six ODMR lines are
refered to by the six numbers 1 to 6. The numbers in parenthesis give the relative deexcitation rates.
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0197900400120114900
We first present experimental results on relaxation
in the triplet 3T 1 u level (sections 1 and 2). Then we
show that from the results of the static studies it is
possible to derive a dynamical spin Hamiltonian (3)
which explains well the dynamical behaviour in high magnetic field and at low temperature (4).
1. Experimental methods.
-Contrary to a conven-
tional Tl relaxation time measurement, we do not monitor the population difference between the two sublevels which are perturbed by the microwaves.
The dynamic behaviour of the populations or, at least, of some linear combinations of them, is obtained by monitoring the intensity of the emitted light of a given polarization after having switched on or off the reso-
nant microwaves [10]. Very little change has to be
made to the ordinary ODMR apparatus to observe relaxation. We have worked at X band (about 9 GHz)
using a Varian electromagnet (typically 0.3 T) ; the microwaves are switched by a PIN diode. The sample
is cooled down to pumped liquid helium temperature
(typically 1.7 K). The F centre luminescence is excited
by a mercury lamp. The light emitted in the whole fluorescence band is collected through a polarizer on
a photomultiplier (EMI 9558 or RCA 6217). The output current is recorded, via a preamplifier with adjustable bandwidth and a multichannel analyzer having a time resolution of 0.2 gs. Signal averaging
is performed on typically 104 microwave pulses.
2. Experimental results.
-At fixed temperature the signal is expected to depend on :
-
the pair of levels which is perturbed by the microwaves, i.e. which of the six ODMR lines has been selected,
-
the direction of the magnetic field : we have
worked with B parallel to the [001] axis,
-
the polarization with which the emitted light is
observed. In our case, we monitored signals in a and
in n polarizations, i.e. linear polarizations respectively perpendicular and parallel to the magnetic field.
The results at 1.6 K are reported in figures 2 and 3, schematically for all cases and with complete details
for some of them. When the microwaves are switched
on or off, one observes a signal which contains both a fast and a slow component (Fig. 2a). The general
appearance of this signal is the same for all the low field lines (lines 1, 2, 3 in figure 2a) but depends strongly
on the polarization of the monitoring light. For the high field lines, the signal has the opposite sign. We
will describe the signal observed when switching off
the microwaves.
2.1 FAST COMPONENT.
-Depending on the line being saturated and the polarization of the monitoring light, this fast component can or cannot be fitted to a single exponential (Fig. 3). The signal to noise ratio
was not good enough to try to extract more than one
exponential from the data. When the signal can
Fig. 2a.
-Schematic diagram of the relaxation signal.
Fig. 2b.
-Experimental results on ODMR line 6 (see figure 1 for
the identification of the line).
reasonably be fitted to an exponential law (e.g. Fig. 3a
in n polarization and the end of the signal in 6 pola- rization), the time constant is found to be about 20 gs at 1.7 K. This rapid variation is opposite in sign on the
two polarizations of the fluorescence and it involves
a change in the distribution of the populations among the different sublevels emitting light of different
polarizations.
2.2 SLOW COMPONENT.
-It is exponential with a
time constant equal to the whole 3Tlu level lifetime
(i.e. 3.7 ms) as measured by simply switching off the
excitation light. The sign is the same on both polari-
zations and here it involves a change in the total
population of the 3Tlu level.
At 4.2 K the rapid components of all signals (n or
Q whatever the line considered) show a quasi expo- nential behaviour with a time constant of about 4.5 JlS, the signs of the signals being the same as at 1.6 K.
3. Theory.
-3.1 RATE EQUATIONS.
-To com-
pute the relaxation signal we have to consider the
different mechanisms allowing transitions between the various levels. We are here interested in the evolution of the populations of the nine sublevels of the 3Tlu
level and we will have to consider four types of tran-
sitions, as illustrated in figure 4 :
Fig. 3.
-Detail of the rapid component of the relaxation signal.
3a) Signal for the centre parallel to the magnetic field (line 6).
3b) and 3c) Signal for the centres perpendicular to the magnetic
field. The magnetic field was slightly misaligned from the [001]
direction in order to distinguish the two lines that would be other- wise superimposed.
-
non radiative transitions between two sublevels of the 3Tlu manifold with a rate R,j for the transitions from the sublevel j to the sublevel i. The mechanisms
of these internal transitions will be detailed later in this section,
-
transitions from one of these sublevels to the
ground state 1 Aig. We will assume that they are purely
radiative with a rate A; i and that there is no other deexcitation mechanism for the 3Tlu level. Their
Fig. 4.
-Definition of the different transition rates.
relative values are recalled on figure 1 by the numbers in parenthesis,
-
transitions from the IT 1 u level to the sublevel i
of the 3Tlu manifold. These feeding transitions have
an unknown rate Pi. We will see however, that since
the internal transitions inside the 3Tlu manifold are
faster than the deexcitation rates A i, the values of Pi
have only a slight effect on the final populations and knowledge of them is not necessary for interpretation
of the signals,
-
microwave induced transitions with a rate
Wik
=Wki which is obviously different from zero
only when i and k are the two sublevels between which microwave transitions are induced.
We assume that the conditions of weak optical pumping are satisfied. The ground level population
is then practically left unchanged and the populations
ni of the sublevels of the 3Tlu manifold are controlled
by the nine following linear differential equations :
The time dependence of n; is then given by the sum
of nine exponential functions :
The nine time constants r. depend only on the tran-
sition rates Rjj, A ;, Wik and not on the feeding rates Pi as long as the approximation of weak pumping
remains valid whereas the amplitudes B. and Bm depend also on the initial conditions.
3.2 INTERNAL TRANSITION RATES.
-These tran- sitions between the different sublevels of the 3T 1 u
manifold are due to the coupling of the F centre to
the phonon field. Here we give only an outline of the calculation. More details are given in the appendix.
Considering the low temperatures at which the experiments have been performed, we will assume
that we have only direct transitions between the sub- levels : either direct transitions between spin sublevels
of a given Jahn-Teller well or transitions between sublevels of different Jahn-Teller wells.
We will make use of two simplifying features of the
problem : the phonon bath can be considered to a
very good approximation as being isotropic and the phonon wavelengths are large.
(i) The isotropy of the phonon bath is demonstrated
by the results of the measurements on sound velocity by Son and Bartels [llJ : vt
=5 x 103 m/s and
vj 1 = 8.3 x 103 m/s for transverse and longitudinal polarizations respectively, whatever the direction of
propagation.
(ii) The long wavelength approximation (Van Vleck [12]) is usual when dealing with phonons of low energy,
i.e. with a wavelength much larger than the interatomic
separation (here a typical value of the random splitting
between the three orbital states is 3 cm -1, leading to
a phonon wavelength of 500 Â). This approximation
states that the time dependent strain may be described
by the usual strain tensor. The coupling to the phonon
strain field is therefore described by the usual orbit lattice coupling Hamiltonian [4, 5] :
The exact definition of the different operators and of the strain components are given in the appendix,
ea, ee and ee’ e,, el and e, are linear combinations of the strain components that transform respectively
like the Ai (totally symmetric), E (tetragonal) and T2 (trigonal) representations of the cubic group.
From stress experiments [4, 5], we know that : (i) Vi is small and anyway we will see that the
term V, ea la is ineffective in spin lattice relaxation.
(ii) The value of V2 is 4.5 x 104 cm-1.
(iii) V3 is not measurable : it is strongly reduced by Jahn-Teller effect ; we can assign 5 x 102 cm-1
as an upper limit of the reduced term V3 R, where R is the Ham reduction factor [9].
We can write the Hamiltonian JCoL in a condensed
form :
where F is the index of the representation (Al’ E or T2) and y the component of F (y = 0 or e, if F = E).
It is shown in the appendix that the transition rate
Rji from the sublevel i to the sublevel j is given by :
a is a numerical factor involving the sound velocity
and the density of CaO. Its value is 0.15 ± 0.04 when all energies are expressed in cm -1.
Using expression (1) we can now determine the
various relaxation paths inside the 3Tlu level.
3 . 2.1 Transition inside a given well.
-We can use
the values ( 1 1 0 Ty 1 j) that have been measured by
Le Si Dang et al. [5] when 1 i > and 1 j > belong to the
same well : these matrix elements are all smaller than 3 cm-1 and lead to transition rates smaller than 0.2 s -1 : these rates are much smaller than the deexci- tation rate and we will neglect them as they would
lead to relaxation times of the order of seconds, much longer than the measured times.
3.2.2 Transition between diffèrent wells.
-We
have only to consider the tetragonal and the trigonal
components of the strain field : the electronic operator involved in the symmetric term VI ea la being propor- tional to the unity matrix, this term cannot induce
any transition.
3.2.2. x Relaxation induced by the tetragonal
time-dependent-strain.
-We look at the transitions that can be induced by the term v2(ee le + eE lE). If
the vibronic eigenstates are the states 1 X > >, 1 Y», 1 Z )>), we see that a tetragonal time-dependent-strain,
being diagonal in this basis, cannot induce a transition
between these states. If such a transition exists, it
can only be due to a mixing [13] of the orbital states.
Such a mixing can be induced by any off-diagonal
operator. The first case we can think of is a mixing by the spin-orbit coupling Âl. S. Let us look for instance if a transition is possible between the states Z, +1 )) and ) Y, 0 )) (+ 1 and 0 stand for the MS spin quantum number) ; we have to calculate :
where are the perturbed states :
Combining eqs. (2) and (3) one gets :
where R is the Ham reduction factor [9].
We can see that, though tetragonal strains are not
affected by the JTE, we need an off-diagonal operator
to provide a mixing of the states and the effect of this
mixing operator is quenched by the JTE, so that, as
expected, the probability of tunnelling between two
wells has a factor proportional to the square of the
overlap between the wavefunctions of the two wells.
The various transitions made possible by spin-orbit mixing are given in figure 5. On the same figure we
recall the polarization selection rules : we can see
that this spin-orbit driven process (that we will call
now the Eg process) connects levels which emit light
of the same polarization. So we can group the sub- levels into two sets :
-
a group of levels emitting Q light, i.e. light polarized in the plane perpendicular to the magnetic field,
-
a group of levels emitting either n light or nothing.
A similar grouping of levels was made by Bill and
Silsbee in analyzing reorientation of 0- in CaF2 [14].
Other interactions can also provide the mixing :
the most important are the dipole-dipole coupling
which preserves the selection rules of the spin-orbit coupling, and the coupling to the random, static, trigonal strains.
3.2.2. Relaxation induced by the trigonal time- dependent-strain.
-We now look at the transitions that can be induced by the term
Fig. 5u.
-Relaxation rates induced by the spin orbit coupling
and the tetragonal component of the phonon strain field (Eg pro-
cess). E has the value
(see text).
Fig. 5b.
-Relaxation rates induced by the trigonal component of the phonon strain field (T2g process). T has the value
(see text). For both figures, the random splitting between the three orbital states is not represented.
We just replace the general operator Ory of eq. ( 1 ) >
by V3 lç, V3 111, V3 1, and we get transitions without
spin flip (Fig. 5), with a transition rate :
if 1 ) and 1 j ) have the same spin state.
Here no mixing is necessary, due to the fact that the trigonal strains are off-diagonal operators. Their effect is also reduced by JTE so that we see that both
processes, the Eg process and this process (which we
will call T2g process) are reduced by R, the Ham
reduction factor, through either the mixing operator
or the strain operator.
There are two very important différences between the two processes :
-
the Eg process groups the sublevels into two sets characterized by their polarization (7c or u) selec- tion rules whereas the T2g process groups the sub- levels into three sets characterized by their spin quan- tum number,
-
the energy dependence of the Eg process is given
Fig. 6.
-Dependence of the relaxation rates on the phonon energy.
6a) T2g process (1.5 K). 6b) T2g process (4.2 K : the scale has been divided by a factor 30). 6c) Eg process (1.5 K). 6d) Eg process
(4.2 K : the scale has been divided by a factor 3). The functions
are AE’[exp(AEIkT) - 1J-1 for the T2g process and
for the Eg process.
These two dependences are shown in figure 6. We
see that for the Eg process, at very low phonon ener- gies the transition rate does not fall to zero : the increase in the mixing of the orbital states which
varies as (Ei - Ej)’ balances the
dependence given by the coupling to the phonon
bath. We will come back to this important point in
the final discussion. We have also to note that in our case we cannot make use of the high température approximation since kT (- 1.2 cm-’) is smaller than but comparable to the phonon energy (- 3 cm-’).
One may also think of relaxation processes induced
by mixing with excited vibronic states. In fact, contrary
to what happens in the static case where second order terms can be more important than first order ones,
higher order terms in relaxation processes between different wells are always affected by the Ham reduc-
tion factor and are therefore negligible compared to
the first order terms.
4. Relaxation model.
-4. 1 GENERAL CONSIDE- RATIONS.
-Having found the different theoretical
expressions for the transition rates, one would in
principle just have to introduce the numerical values into the nine evolution equations and solve them. In fact we know only the order of magnitude of the spin
orbit coupling constant and an upper limit for V3.
We will therefore first examine whether we can make
use of some simplifying features.
The first simplification one can think of would be
to consider that only one of the two Eg and T2g
processes is rapid with respect to the lifetime. A
qualitative discussion of the experimental results (Figs. 2, 3) will show that such a simplification is not possible. If only one of the two processes is rapid,
we can group the levels into two or three sets inside which relaxation is rapid : three sets corresponding
to the three spin states for a T2g process or two n
and J sets for an Eg process, as shown in figure 5.
When switching on or off the microwaves, a quasi-
Boltzmann equilibrium is reached within each set.
Quasi-Boltzmann equilibrium means that a Boltz-
mann equilibrium is not reached (but it does not
mean a Boltzmann equilibrium at a temperature different from the sample temperature). The departure
from Boltzmann equilibrium is of the order of the
ratio of the deexcitation rates to the relaxation rates, and these quasi-Boltzmann equilibria, with or without microwaves, are only slightly different. So, when the microwaves are switched off, the fast component of the relaxation, which under this hypothesis corres- ponds to relaxation inside a set, has a very small
amplitude. One expects to detect only a slow compo- nent which corresponds to a rearrangement of the
populations between the different sets and has a time
constant of the order of the lifetime. This is obviously
not the case : one can see in figure 2 that the fast and the slow components of the relaxation signal
have comparable amplitudes. In other words, in order
to observe a non negligible fast component in the relaxation signal, one has to assume the existence of
a fast relaxation path between the two sublevels involved in the microwave transition. Figure 5 shows
that this can only be the case if both Eg and T2g
processes are rapid compared to the lifetime.
The shape of the rapid signal (independently of its
time scaling since we can consider as constant the
variation due to the slow component) is then governed by the ratio of the Eg and T2g processes, i.e. the ratio
V2 ÂIV3 DE, where AE represents an average energy
separation between levels which is due to intemal strains. We will now consider the two extreme cases
V2 ÀIV3 4E > 1 and V2 ÀIV3 DE 1.
4.2 AN OVERSIMPLIFIED MODEL.
-We will first
assume V2 ÂIV3 AE » 1, which means that the Eg
process is much more efficient than the T2g process,
both being rapid compared to the lifetime. We will
therefore group the different sublevels in the two n
and 6 sets defined before. The Eg process induces very rapid transitions between the sublevels inside each set. Transitions between the two sets are induced
by microwaves and by the T2g process (Fig. 7). The
Fig. 7a.
-Level grouping in the oversimplified model (Eg process
very much efficient than the T2g process, see text).
Fig. 7b.
-Expected ODMR signal in this model.
population distribution at equilibrium within the 3Tlu
manifold when microwaves are applied is then the following :
i) Due to the Eg process quasi-Boltzmann equili-
brium is achieved inside each of the two groups and this whatever the feeding rates since the Eg process
is much more efficient than the radiative transition
rates.
ii) The ratio of the populations of the two levels corresponding to the resonance is intermediate bet-
ween unity (case of complete saturation of the line)
and quasi-Boltzmann equilibrium (no microwave
power), depending on the efficiency of the microwave compared to the average effect of the T2g process.
iii) The total population of the 3Tlu manifold is
given by the ratio between the feeding rate (which
is given by the power of the exciting light) and the
effective deexcitation probability
relative populations of the sublevels in each n or 6
set remain in quasi thermal equilibrium. The T2g pro-
cesses give rise to a transfer of population from one
set to the other, restoring the quasi-Boltzmann equi-
librium among the nine sublevels. The observed rate of recovery is an average of the rates of transfer bet-
ween the various sublevels, and may be easily comput- ed by introducing into the equation of evolution the fact that quasi thermal equilibrium is always achieved
within each n or J set. The signal has opposite sign
when detected on Q or n polarization. The variation
in the populations of the sublevels results in a variation of !eff’
1the average lifetime. As the populating rate
is left unchanged, there is a readjustment of the total population of the 3Tlu level ; this readjustment is
seen by a variation of the total fluorescence, with a
time constant equal to the effective lifetime.
This qualitative description is in rather good agreement with the experimental results. Quantitati- vely one has just to introduce into the rate equations,
the persistence of Boltzmann equilibrium within each set : this leads to two differential equations which
can be easily solved in order to obtain an analytical expression for the expected relaxation signal.
The rates given in (4) and (5) however depend strongly on the energy DE of the phonons involved,
and this frequency is related to the local (internal)
strains. When averaging over the distribution of internal strains one gets a relaxation signal which is
very highly non exponential whatever the polariza-
tion of the monitoring light and whatever the ODMR line which is saturated. These conclusions are not
experimentally verified : see e.g. the quasi exponential signal obtained in n polarization in figure 3a.
Note that the opposite assumption, V2 ÀIV3 AE « 1 (i.e. T2g process much more rapid than the Eg process)
would lead to an inverted signal on 7r polarization
and has therefore to be rejected.
4.3 MORE REALISTIC MODEL.
-We are therefore in the intermediate case : V2 ÀIV3 AE - 1. Though
the Eg and the T2g processes are still more efficient
than the lifetime, both have the same order of magni-
tude. In this case only a numerical computation is possible. We have to consider a statistical distribution of the static tetragonal strain components e. and e,.
Their effect is determined by the following results :
i) From the study of the lineshape of both the zero phonon line and the ODMR lines, Le Si Dang et al. [6 ]
were able to determine the str’ain distribution : they
find the distribution for V2 e. and V2 e, is best repre- sented by a Voigt profile but can be approximated by a Gaussian distribution with a standard deviation u = 3.1 cm-’.
ii) The same studies [6] have shown that when the
magnetic field is along the z axis, the shape of the
ODMR line is practically governed by the parameter
V2 ee alone. Our calculations have therefore been
performed by giving to this parameter the value
corresponding to the centre of the ODMR line, where the relaxation measurements have been done
(V2 e(J
=2.2 cm-’) letting V2 e, run over all its dis- tribution. The calculation is then performed in the following way for a given value of V2 ÂIV3 : the
relaxation signal is computed when the microwaves
are switched off, the initial conditions being deter-
mined by solving the relaxation equations at equili-
brium in presence of microwaves. We then compute
the effective signal by summing over V2 e, and using
the strain distribution mentioned before. The same
procedure is followed for different values of the ratio
V2 ÂIV3 to find the best value in order to fit the shape
of the experimental signal. Once the fit is found at 1.6 K, the signal is computed at 4.2 K and compared
to the experimental signal, which is quasi exponential
in both n and J polarizations with a time constant of
about 4.5 ps. A first set of parameters
gave a good fit at 1.6 K but predicted a time constant
of 1.5 gs at 4.2 K, in poor agreement with the experi-
mental value. This suggests that the 1.6 K temperature
was underestimated because of heating of the crystal (about 50 mW of light). In fact one can measure the temperature fairly accurately by looking at the magnetic circular polarization, which has been shown to vary as th (g,uB B12 kT) [3]. We have checked this law with different powers of excitation and have found that at the levels of light at which the micro-
waves experiments were performed the effective tem-
perature fitting the th (gllB B12 kTeff) law was of the
order of 2 K when the temperature of the helium bath measured by its vapor pressure was 1.6 K. A new
fit at 2 K then gave
V2 ÂR
=700 cm - 2 and V3 R = 200 cm-l,
these values leading to a time constant of 3.4 gs at 4.2 K. From the stress experiment [4] one knows the
value V2
=4.5 x 104 cm-’ - so we can deduce a
value of the reduced spin-orbit coupling constant :
ÀR
=1.5 x 10-2 cm-1.
Figure 8 shows the comparison of the experimental
and computed signals for the line 6
1