Gyanendra J.Opt 2015
Gyanendra J.Opt 2015
Gyanendra J.Opt 2015
E-mail: [email protected]
Abstract
Resonance characteristics, chaotic dynamics and bistable response of the erbium-doped fibre ring
laser (EDFRL) are studied in an experiment under cavity-loss modulation at frequencies in the
range of 1–35 kHz. With an increase in pumping ratio and modulation index, the resonance
frequencies of the ring cavity laser change in agreement with the expectations from the model. A
comparison with the calculated results of the pump-modulated laser helps to bring out the
differences in the two modulation schemes. The loss-modulated laser is found to be more
sensitive in the sub-harmonic range to its dynamical state while the pump-modulated laser is
more sensitive in the super-harmonic range. To achieve chaos, pump modulation requires high
values of modulation index (∼28%) while it is seen even at very low values (<1%) in loss
modulation. Even though both the schemes follow the same period-doubling route to chaos,
features such as the presence of bifurcations or chaos at the resonance frequencies are quite
different. Loss modulation is more suited to study the chaotic features in EDFRL due to the
range of available control parameters. By comparing the strange attractors and Lyapunov
exponents of both cases, it is found that the quality of chaos is better in loss modulation.
Keywords: erbium-doped fibre laser, ring cavity, laser dynamics, loss modulation, chaos
2
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
dN lsFN N alternate model for EDFL dynamics that includes the effect of
=- + D ( N0 - N ) - , (2 )
dt tr ts amplified spontaneous emission (ASE), excitation state
absorption and multimode operation, but these factors are not
where S is the rate of spontaneous emission, F is the photon required in our case, since the loss variation is very less so
density, N and N0 are the population inversions in the that the center wavelength of lasing remains unchanged and
presence and absence of lasing respectively, σ is the cross- the excited state absorption at the lasing wavelength in our
section for stimulated emission, R is the reflection coefficient work (1570 nm) is insignificant because of the low pumping
of the output mirror, tr = l /u is the round trip transit time of ratio (1.35). The simple and well-established model used here
the light in the cavity, u is the speed of light in fibre, ts is shows good agreement with the experimental results as dis-
lifetime of the upper lasing level, l is the cavity length, W is cussed in the next section.
the cavity loss at the lasing wavelength and D is the relative
volumetric pumping rate [27]. Equations (1) and (2) can be
written with scaled parameters as
4. Results and discussion on loss modulation
dx
= s + gx (y - a) , (3 )
dq 4.1. Nonlinear regime of oscillation
dy
= a - y (x + 1 + d ) , (4 ) In an earlier work, we have reported the evolution of the
dq
dynamics of erbium-doped fibre ring laser (EDFRL) from
where x = usts F, y = 2slN , g = ts /tr , s = ust 2s S, n = fts,
linear to nonlinear regime by varying the loss in the ring
1
()
q = t /ts, a = ln R + lW , a = 2lsts DN0 and d = Dts. cavity with the change of modulation index and modulation
The parameters s, x, y and a are indicative of spontaneous frequency [31]. In the present work, we analyse the nonlinear
emission rate which contributes to the laser intensity, laser regime, mainly for its resonance characteristics and chaotic
intensity, population inversion and cavity loss respectively, dynamics, both experimentally as well as theoretically by
while a and d are constants dependent on pumping rate and g using the appropriate model [14]. For a low pumping ratio of
is a constant. The symbols t and f represent time and 1.35, the laser output at four different RF signal peak-to-peak
frequency respectively. The cavity loss includes the insertion amplitudes (Vpp) of 10 mV (m=0.0006), 20 mV
losses of all components used in the ring and the pump-to- (m=0.0012), 30 mV (m=0.0018) and 40 mV
signal overlap factor. (m=0.0024) are shown in figure 2(a) while varying the
To observe the nonlinear dynamical effect of chaos, a modulation frequency. Apart from the fundamental, under-
third variable (other than x and y) is essential [28]. This can be tones and overtones of the driving frequency are also
provided with an external modulating signal. In the case of observed.
loss modulation, the parameter a is varied sinusoidally as For Vpp=10 mV, three resonances are seen near the
a = a0 (1 + m sin (2pnq )). Here α0 is the total cavity loss of frequencies of 6.9 kHz, 11.6 kHz and 26 kHz. The ROF
the ring. In the case of pump modulation, one modulates measured independently for the laser is 12.3 kHz [31]. The
parameters a and d for sinusoidal modulation as mathematical expression of ROF can be derived by lineariz-
a = a 0 (1 + m sin (2pnq )) and d = d 0 (1 + m sin (2pnq )). ing the equations (3) and (4) [32]. The value of ROF depends
Here a0 and d0 are the pump dependent constants, when there on the length of the cavity, type of active material, the input
is no external modulation. Moreover, m, q and ν are the pump power and the total cavity loss [32]. When this natural
modulation index, scaled time and scaled modulation fre- frequency of the laser is comparable to the sinusoidal mod-
quency respectively. Thus the modulated laser is modeled ulation frequency, the fundamental resonance at 11.6 kHz is
through differential equations of a driven system. As expec- obtained. When the natural frequency is nearly twice the
ted, the driven system displays a behaviour characteristic of modulation frequency, the first super-harmonic resonance at
the driving amplitude, frequency and the initial conditions. 6.9 kHz is obtained. When the natural frequency is a fraction
But, the two cases of pump and loss modulation do not give of the modulation frequency, a sub-harmonic resonance is
identical results. The similarities and differences between obtained at 26 kHz. For the pumping ratio used here, it is
these two modulation schemes are brought out in the next two noticed that the fundamental resonance frequency (11.6 kHz)
sections. for Vpp=10 mV is slightly lower than the ROF (12.3 kHz)
The values of the various parameters relevant to the [16]. On increase in Vpp to 20 mV, the fundamental, sub- and
discussion here are g = 1.2 ´ 10 5, a0=19.44, s = 0.3, super-harmonic resonances are observed near 10.6 kHz,
α0=11 and d0=0.245 [29]. The frequency range of mod- 25.1 kHz and 6.8 kHz respectively, with all the three reso-
ulation is chosen from 1 to 35 kHz here in order to restrict the nance frequencies being shifted towards the lower fre-
discussion to the range of the ROF of the laser and one sub- quency side.
and super-harmonic of it. The length of EDF is 8 m and the On further increase in Vpp to 30 mV and 40 mV, the
output wavelength of the laser is 1570 nm. The entire analysis resonance frequencies get lowered further. Here all the curves
assumes unidirectional ring cavity for the laser. The pumping are slightly asymmetric in shape which is elaborated in [33]. It
ratio defined as the ratio of operating pump power to the is noticed that the change in fundamental resonance frequency
threshold pump power for lasing has the value of 1.35 unless with increase in Vpp is lesser than the change in sub-harmonic
specified otherwise. Stewart et al [30] have reported an frequency. It implies that the system is more sensitive in the
3
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
Figure 2. Nonlinear responses of EDFRL under the cavity loss modulation for different modulation indices (m) taking modulation frequency
as the control parameter. (a) Experiment, (b) calculation.
sub-harmonic resonance regime. This is discussed further in bifurcation diagram for Vpp=60 mV (m=0.0036) in the
this section. experiment is shown in figure 3(a). For low frequencies up to
Trends similar to the experimental results are also noticed 4.2 kHz, the laser gently follows the modulating signal. On
in the calculated results, especially in the range of resonance further increase in modulation frequency, the laser dynamics
frequencies, which are shown in figure 2(b). For these cal- has higher periods such as period-2, period-4 and period-8
culations, equations (3) and (4) are solved by using the fourth eventually leading to chaos in the modulation frequency span
order Runge–Kutta method with a small step size of 10−6 of 18.5–24 kHz under the period doubling route [31].
under the initial conditions of x=0.001 and y=0.001. The On further increase in frequency, the laser dynamics gets
solution provides the output response of the laser since the x back to periodic states. This implies that for a fixed pumping
and y parameters represent the laser intensity and population ratio, the laser output has a chaotic nature for certain com-
inversion respectively. Here, xmax represents the maximum binations of the modulation index and modulation frequency
value of the laser intensity (x) and is extracted through a peak- periodically [34].
search for a certain time span at each value of modulation This experimental result is compared with the calculation
frequency. If the maximum of laser intensity has an identical from the model and is shown in figure 3(b). Trends are quite
value for every pulse, then there is only one value of xmax for similar and a few minor deviations would be due to non-ideal
that particular frequency. If the intensity peaks differ in their experimental conditions. It is worth noting that the frequency
amplitude at different times, then more than one xmax value range of chaos is in the sub-harmonic range for the case of
will be present leading to bifurcation. loss modulation and at very low values of modulation index.
While the features of frequency dependence match well This will be contrasted with the result in pump modulation in
between the experimental and calculated results, we observe a section 5.
discrepancy between the values of the modulation index used Reconstructed attractor is a way to visualize the phase
in the experiment and calculations. The experiment uses space of the experimental data [35, 36]. Here it is obtained by
modulation indices of 6×10−4, 1.2×10−3, 1.8×10−3 and using the method of delay and the delay τ is decided by the
2.4×10−3. To observe similar features in the modeling, we autocorrelation function [36]. The dense reconstructed strange
require modulation indices that are five times lower. This attractor from the experimental data of Vpp=60 mV at a
could be due to two reasons. Firstly, the laser system in the modulation frequency of 22.5 kHz and modulation index of
modeling is idealized while the experimental system will have 0.0036 is shown in figure 3(c) for τ of 6.25 μs. The strange
several hidden losses. Secondly, the values of absorption/ attractor calculated from the model for a modulation index of
emission cross-section and population inversion used in the 0.000 72 and frequency of 22.5 kHz is shown in figure 3(d) for
calculations are mere estimates from the experiment and are τ of 6.5 μs. The total time span for both figures 3(c) and (d) is
not necessarily very precise. However, the increment in the same at 50 ms and the colour variation evolves as black-red-
modulation index is kept in a similar proportion in both green-blue-violet when the time varies as 0–10 ms, 10–20 ms,
experiment and modeling. The modulation index (m) used for 20–30 ms, 30–40 ms, and 40–50 ms. The strange attractors have
the experimental data takes into account the total passive the clear indication of chaos. The axes are named as V(t) and x
cavity loss (11 dB). Total cavity loss includes the loss induced (t) in the two figures to indicate that one of them is from
by numerous splices and the insertion loss of different com- experimental data and the other is from calculation.
ponents used in the cavity such as WDM, couplers, PC, In all the above results, we analysed the behaviour of the
isolator and MZM. loss-modulated laser at discrete values of modulation index
while continuously varying the modulation frequency, in
order to understand the frequency dependence of its nonlinear
4.2. Chaotic regime
dynamics. It is equally important to study the laser response
The laser shows more complex behaviour at higher modula- with a continuously varying modulation index at certain
tion amplitudes as the extent of nonlinearity increases. The discrete values of modulation frequency. We choose to
4
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
Figure 3. Bifurcation diagrams of EDFRL dynamics under the cavity loss modulation with modulation frequency as the control parameter, as
obtained in the experiment (a) and calculation (b). In (c) and (d), the strange attractors for experimental laser output at modulation index and
modulation frequency of m=0.0036 and f=22.5 kHz is shown along with the strange attractor calculated for a modulation index of
0.000 72 and frequency of 22.5 kHz.
analyse this near the natural frequency and its first undertone Vpp=30 mV, the super-harmonic, fundamental and sub-har-
and overtone frequencies. monic resonance peaks are observed at 6.4 kHz, 10.3 kHz and
At the modulation frequency of 6.15 kHz (corresponding 23.2 kHz while increasing the modulation frequency from 1 to
to the first super-harmonic of the measured ROF), the bifur- 35 kHz. This is shown in figure 5(a) in blue colour. On the
cation starts at m=0.0024, while period-2 (P2), period-4 other hand, the resonance peaks are observed at 17.9 kHz,
(P4), and period-8 (P8) are observed at modulation indices of 8.7 kHz and 6.3 kHz (shown in red) with relatively higher peak
0.0027, 0.0041 and 0.0059 respectively, as indicated by intensities while decreasing the modulation frequency from 35
vertical lines in figure 4(a), eventually leading to chaos when to 1 kHz. Thus, two hysteresis curves are seen here, indicated
the modulation index is in the range of 0.0062–0.0090. On by ‘A’ and ‘B’ in figure 5(a). It is noticed that the hysteresis for
further increase in the modulation index, the laser dynamics the sub-harmonic resonance regime is larger. It is due to higher
returns back to the periodic states. Likewise, at a modulation sensitivity of the laser dynamics in the sub-harmonic regime, as
frequency of 12.3 kHz (measured ROF), equal to the natural already seen in figures 3 and 4(c), (f). The bistable behaviour
frequency, higher periods and chaos are observed but the calculated under loss modulation is shown in figure 5(b). The
modulation index span for chaos is different and wider than modulation index is chosen here to match with the appropriate
the previous one, as shown in figure 4(b). At a modulation frequency range seen in the experiment and to get the correct
frequency of 24.6 kHz (first sub-harmonic of the measured order of magnitude variation between the low and high
ROF), wider ranges of periodic chaotic regimes are noticed, amplitudes of the hysteresis curve.
which is shown in figure 4(c), implying that the system is
more sensitive in the sub-harmonic frequency range. Similar
trends are observed in the calculated results shown in
figures 4(d)–(f) for these three frequencies in the range of 5. Calculated results on pump modulation and
modulation indices studied here. Minor deviations seen comparison with loss modulation
between the experiment and calculation are attributed to the
slight variations between the real conditions of the experiment Pump modulation is a well-established scheme to generate
and the ideal conditions of the calculations. chaos in fibre lasers. Several reports exist on the nonlinear
dynamics of EDFRL studied experimentally under pump
modulation [7, 10–13]. We are discussing it here only to bring
4.3. Bistable response
out the similarities and differences with loss modulation
Hysteresis is seen in the response of the laser dynamics to a discussed in the previous section, and therefore we will pre-
change in modulation frequency. At a fixed value of sent only the calculated results. In this scheme, input pump
5
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
Figure 4. Bifurcation diagrams from the experiment and calculation in the left and right column respectively, for loss modulated EDFRL,
with the variation of modulation index at some selected modulation frequencies. (a), (d) f=6.15 kHz, (b), (e) f=12.3 kHz, (c), (f)
f=24.6 kHz, pumping ratio=1.35.
Figure 5. Experimental (a) as well as calculated (b) bistable responses of the EDFRL at a fixed value of modulation index while increasing
and decreasing the modulation frequency.
power is changed as a sinusoidal function of time. Pump fundamental and super-harmonic frequency are comparable in
modulation also shows the super-harmonic, fundamental and pump modulation. In the case of loss modulation studied
sub-harmonic resonances but at much higher modulation earlier in figure 2, the maximum amplitudes were comparable
indices in comparison to loss modulation. The extent of this at the fundamental and sub-harmonic frequencies. This dif-
increase is nearly two orders of magnitude. Calculated laser ference also manifests itself when we compare figure 3 with
output responses at four different modulation indices of figure 6(b). The frequency range of chaos was reached in the
m=0.05, 0.1, 0.15 and 0.20 are shown in figure 6(a). At a sub-harmonic range with loss modulation, while it is reached
higher modulation index of m=0.28, the laser shows the in the super-harmonic range for pump modulated laser.
chaotic response seen in figure 6(b) at frequencies lying in the The extent of nonlinearity in the system affects its natural
super-harmonic resonance regime. resonance frequency. With stronger modulation of the laser, it
The distinct difference between figures 2 and 6(a) is is driven further into the nonlinear regime resulting in the
worth the emphasis. The maximum amplitudes at the generation of sub- and super-harmonics, ultimately resulting
6
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
Figure 6. Nonlinear responses of EDFRL under the pump modulation for different modulation indices taking modulation frequency as the
control parameter. (a) m=0.05, 0.1, 0.15, and 0.2, (b) m=0.28. Pumping ratio is 1.35.
Figure 7. Variation of super-, fundamental and sub-resonance frequencies with pumping ratio for (a)–(c) experimental data and (d)–(f)
calculation under the loss modulation, while (g)–(i) shows the calculation for pump modulation. ROF is the relaxation oscillation frequency.
in strong instabilities and chaos. This is an extremely of achieving population inversion. But, all resonance fre-
important feature while using the modulated laser for appli- quencies decrease significantly when the modulation index is
cations near its ROF. It is possible to drive the laser system increased. An increase in modulation index implies a larger
into a dynamically nonlinear regime by changing the mod- driving force. Interestingly, at a fixed value of pumping ratio,
ulation index and/or by changing the extent of pumping. The the change in the fundamental resonance frequency with
fundamental, sub- and super-harmonic frequencies change modulation index is more than the change in sub- and super-
with the modulation index and pumping ratio as shown in harmonic resonance frequency. By comparing figures 7(d)–(f)
figure 7. In the case of loss modulation, the measured results with figures 7(g)–(i), one notices that the super-harmonic
seen in figures 7(a)–(c) match well with those calculated by resonance frequency in loss modulation does not change
the model shown in figures 7(d)–(f). The calculated results for much with modulation index while the sub-harmonic fre-
pump modulation are shown in figures 7(g)–(i). In both the quency remains unchanged by a change in modulation index
cases of loss and pump modulation, the fundamental as well in pump modulation.
as sub- and super-harmonic resonance frequencies increase It is amply clear from the previous discussion that much
with the pumping ratio. Higher pumping ratio implies an ease lower modulation indices are required for loss modulated
7
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya
Figure 8. Pump modulation: (a) strange attractor of the calculated laser output for the modulation index and modulation frequency of
m=0.28 and f=5. 6 kHz respectively. (b) Bistable response of the laser for the modulation index of m=0.091.
laser to enter the nonlinear regime in comparison to pump the loss modulation case. As opposed to loss modulation, the
modulation. This is because the pump modulation involves dominance of peaks in the super-harmonic resonance range
the upper energy state of erbium whose lifetime of 1 μs is two signifies more sensitivity in that frequency regime in pump
orders of magnitude shorter than the lifetime of 10 ms of the modulation.
metastable state perturbed during loss modulation. At kHz
frequencies, the pump modulation scheme requires a higher
perturbation in pump power to introduce a significant non- 6. Conclusion
linearity, as opposed to the modulation of the loss in the latter
scheme, since the pump modulation involves a low lifetime We have studied the resonance characteristics, chaotic
energy state. The fact that the modulation indices required for dynamics and bistable response of an EDFRL by modulating
inducing similar effects in pump and loss modulation will the cavity loss sinusoidally with the help of an intensity
differ by two orders of magnitude for class B lasers is stated modulator in the ring for the limited modulation frequency
in an earlier work [19] and our results support this span of 1–35 kHz. Resonance peaks are observed near the
quantitatively. natural frequency of the laser and its harmonics. With an
Figure 8(a) shows the strange attractor for the laser increase in the modulation index, the resonance peak shifts to
output calculated for pump modulation at the modulation lower frequency. By varying the modulation parameters, the
index and modulation frequency of m=0.28 and chaotic regime is reached through the period-doubling route.
f=5.6 kHz respectively, with a τ value of 7.5 μs. The total At a fixed value of modulation index, sweeping the mod-
time span is the same (50 ms) as in loss modulation discussed ulation frequency from lower to higher gives an output
earlier in figure 3. The colour scheme is also the same as that response that is non-identical with frequency variation from
used for figures 3(c) and (d). These values are complementary higher to lower value due to hysteresis and bistable nature of
to those used in loss modulation. The largest Lyapunov the laser system. Calculation using the appropriate theoretical
exponent is a measure of the divergence of two nearby tra- model gives a good agreement with the experimental results.
jectories with time, and hence a higher value of positive Pump modulated EDFRL is also modeled to compare it with
exponent implies a better quality of chaos [35]. Our calcu- the loss modulated system. Several differences are noted
lation of largest Lyapunov exponent followed the method in between the two modulation schemes. Chaos can be achieved
[37] at those m values and modulation frequencies where in loss modulation at much lower modulation indices in
chaos appears. For the case of loss modulation, the values of comparison to pump modulation. At a fixed value of pumping
largest Lyapunov exponent for the measured (m=0.0036) ratio and modulation index, it is easier to obtain the chaotic
and calculated (m=0.000 72) time series at the modulation nature in the sub- harmonic resonance regime in loss mod-
frequency of f=22.5 kHz are 0.32 and 0.30 respectively, ulation while it is easier in super-harmonic resonance regime
while for the case of pump modulation at f=5.6 kHz and in pump modulation. The quality of chaos is found to be
m=0.28, the largest Lyapunov exponent is 0.25. This is a better in loss modulation as established by calculating the
confirmation of the better quality of chaos in loss modulation largest Lyapunov exponent, and the process of loss modula-
at kHz frequencies. tion gives a finer control on the laser dynamics.
As in the case of loss modulation, the laser shows bis-
table behaviour under the pump modulation which is shown
in figure 8(b). This is also reported in earlier works [10, 13]. Acknowledgments
We have calculated it here only to discuss it with the loss
modulation results for comparable parameters. The modula- The authors acknowledge the Department of Physics of IIT
tion index used here is two orders of magnitude larger than Kanpur for the support to carry out the work.
8
J. Opt. 17 (2015) 125402 G Kumar and R Vijaya