Professional Documents
Culture Documents
1, JANUARY 1994
supported by the following observation. It is well known The ansatz (1.2) introduced into (1.2) introduced into (1. I )
that the spectrum of a soliton-like pulse, perturbed by the gives dependences of pulse width and chirp parameter on
periodicity of the excitations in a resonator, deteriorates if dispersion, with the SPM coefficient as a parameter. This is
the nonlinear phase shift of the pulse approaches 360 per shown in Fig. 4 (taken from [23]). We see that the Kerr effect
pass [22]. This observation suggests that the cleanest spectra calls for negative GVD compensation in order to produce a
are obtained when the nonlinear phase shift per pass is small. chirp free pulse. The fibers in our experiment had negative
But this means that the master equation [23], developed for dispersion and it is this case on which we concentrate.
small nonlinear changes per pass, is legitimately applied to When p = 0, the pulse is soliton like. In fact in the
the analysis, at least qualitatively. The same cautionary note ensuing analysis we shall often invoke the soliton picture.
does not apply to linear processes (linear loss, dispersion) that It should be noted, however, that in our approach we imply
can cause large changes per pass. In these cases the master that the cavity length is much less than the distance within
equation has to be modified appropriately. which a pulse acquires a nonlinear phase shift of 2 x. The
1 ) The APM Operation of Fiber Ring Laser: Figure 2 ex- solitons that are predicted from the master equation are
plains the APM principle for a nonlinear polarization rotation average solitons similar to those that occur in repeaterless
inferometer constructed of a polarizer, wave plate, an isotropic transoceanic transmission [21]. These can experience large
Kerr medium (fiber), and an analyzer. All these components attenuation and amplification within the distance between am-
can be realized with fibers. The wave plate transforms linear plifiers (in our case one cavity roundtrip). The nonlinear phase
polarization into elliptic polarization. The elliptic polarization shift within this distance is small, however. Our entire analysis
:p6=o
202 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 30, NO. I, JANUARY 1994
y=i If the pulse is soliton like, then the area of the pulse is
proportional to A07 and is fixed. Equation (1.3) is maximized
at constant area when:
P \.,.,
-2
4
\ ,..-.. 6.1
8
% 6-4 5
-
--.__
->.
This fixes the peak power of the soliton. Its width is given by
4 . 2 0 2 4 4 2 0 2 4
D Dthe area theorem. Thus maximization of APM action quantizes
the soliton. This accounts for the observation of many solitons
Fig. 4. Chirp parameter and pulsewidth as a function of normalized dis-
oersion. with SPM coefficient as oarameter. D.... is the chim r
narameter.
~ ~ ~ ~c~
in long fiber
j ....,
;he normalized GVD parameter, ;
~~~~~~~~
the normalized equivalent fast saturable A potential limit to the pulse duration has been revealed
absorber, and /i the normalized parameter in recent studies on the periodic perturbation of solitons [22],
[26]. A soliton becomes highly unstable when the period of
perturbation approaches the length over which the nonlinear
phase shift of the soliton is 27r. Because a mode-locked laser
periodically perturbs the pulse at the cavity roundtrip length
L, the shortest soliton that can be stably supported must
accumulate less than 360 phase shift.
The occurrence of many solitons is, in general, undesirable,
because their relative timing cannot be easily controlled. A
design of a modelocked laser should strive for stability with
a single soliton per roundtrip (unless there are other ways in
which the timing and distribution of multiple solitons could
be controlled). This means that the loop should be made as
short as possible and the gain should be adjusted to be low
enough so that the energy of one single quantized soliton fully
depletes the gain. This consideration underlies the design of
the fiber ring laser.
2) Analysis of Fiber Ring Laser: In the following, we de-
velop a simple model of APM action in a fiber ring. The
Fig. 5. Multiple solitons in one roundtrip. fiber used in our experiment, and in other experiments, even
though called nonbirefringent has spurious birefringences
is predicated on this picture. In the ring laser experiments with beat length of the order of 2 m. Hence, it is expected
described here it was strictly valid. When not valid in figure- that in experiments with a fiber of 4.8 m, as reported in [12],
eight laser configurations, our analysis cannot lay claim to the model of simple elliptic polarization rotation cannot hold
fully explaining phenomena in such lasers. precisely. However, correction of the spurious polarization
Before we proceed with the discussion of the fiber ring transformation effects in the fiber by the second polarization
laser, we want to look at a phenomenon that has been widely transformer preceding the polarizer removes these effects to
observed in fiber laser systems, namely the occurrence of many lowest order. Hence it is expected that the analysis presented
pulses within one round trip time, the pulses all being of here will give the correct order of magnitude of the self-
standard energy [24]. Sometimes, the detector shows pulses amplitude modulation coefficient.
of twice or three times the height, but this can be generally The fiber ring laser contains an isolator forcing unidirec-
attributed to the response time of the detector, which registers tional propagation, and a polarizer. After the polarizer, the
two pulses spaced by less than its response time as a single mode is linearly polarized. This is followed by a polarization
pulse of twice the height. Figure 5 shows such an occurrence transformer, which generates elliptic polarization from the
as seen by us. We claim that this phenomenon is due to a linear polarization. The elliptic polarization rotates by virtue
quantization of soliton pulses in an APM system that has of the Kerr effect, the rotation being proportional to intensity.
excessive gain. The rotated ellipse passes again through the polarizer, now
The argument we present is the following: Because of the acting as an analyzer and thus an intensity-dependent loss
soliton area theorem, a mechanism limiting either the peak is generated. Another point of view is to consider the two
power or pulse width determines the energy of the soliton. polarizations to be the two arms of a nonlinear Mach-Zehnder
The APM action represented by 7 ( n I 2in (1.1) is produced interferometer, the analyzer providing one output port of the
interferometrically, and is represented more accurately by sin interferometer, the absorbed polarization representing the other
-ylaI2, indicating saturation of APM action. The energy gain port. This point of view interprets the intensity dependent
of a pulse passing through such an interferometer is given by: loss as intensity-dependent interference at the Mach-Zehnder
output, i.e., as additive-pulse modelocking. Clearly, gain must
d t 1o,I2sin(yl(i.12) (1.3) be provided to compensate for the loss. We shall neglect the
effect of the gain on the APM action.
HAUS et al.: ADDITIVE PULSE MODELOCKING 203
We start with the column matrix representation of a polar- where X is the free space wavelength, n2 is the nonlinear
ized mode amplitude:
a= [z]
The amplitudes are so normalized that their squares given the
index coefficient defined for linear polarization, L K is ~the ~
length of the Kerr medium and d,ffis the effective mode
cross-sectional area.
It is useful to note that when K is transformed to the
linear polarization basis, one is left with a rotation matrix
~
power. Passage through a polarization transformer produces by the angle A@& = (a+ - @-)/2 (multiplied by a com-
an output b from an input a through multiplication by the mon nonlinear phase factor, the self-phase modulation). The
polarization transformation matrix T: polarization ellipse experiences intensity dependent rotation,
but the ellipticity is unchanged. This clearly illustrates the
b=Ta . (2.2) equivalence between the nonlinear polarization rotation and
the APM point of view.
The most general polarization transformation matrix of a
The polarization suppresses the y-component so that the
lossless structure is:
input into the polarization transformer is:
four real parameters equation (2.2) has three such parameters; The circular polarization excitation after the polarization trans-
the fourth is omitted since it simply represents the common former is UTa. From the components of this column matrix
phase shift of the two polarizations through the transformer. we may evaluate the phase shifts @*caused by E+.
Next, the field passes through the fiber Kerr medium. In
general, a standard fiber has a weak linear birefringence
due to imperfections, such as ellipticity of the core. An
exact description of the nonlinear polarization evolution in The output of the polarizer is
the presence of birefringence requires the use of Jacobian
elliptic functions [25]. However, if the net amount of linear
birefringence is small (< 1 beat-length), most of the key
bl =[1 O]Ut ["
-j*+
e-ja-
O 1 UTa
physics of operation may be discerned within the approxi-
mation of an isotropic medium. In this limit, the rotation of
elliptic polarization rotation occurs. This may be a realistic
regime because the presence of many linear birefringent beat- (2.11)
lengths creates bandwidth modulation and allows for local (in
frequency) loss minima for CW radiation which may hinder
where the nonlinear phase shift matrix transforms the circu-
mode-locking.
larly polarized components, and multiplication by Ut trans-
In an isotropic medium, it is convenient to handle the Kerr
forms back into linear polarization. The throughput power
effect in the basis of the circular polarization modes E+
is
and E-. These related to the linear modes by the unitary
transformation, IAI2 [I - (1 - 2t2)cos(@+ - @-)
(b# = -
2
(2.4)
- 2tdC-F sin (a+ - a-) sin ( 4 + 0)](2.12)
from which one may evaluate the loss parameter e and the
where
U = - [l l j ] SAM parameter of the master equation.
e=1-t (2.13)
J z 1 -j
The coherence term vanishes in the circular basis, and the Kerr
effect can be represented by the diagonal matrix,
The SPM parameter is from (2.9)
3 ) The Figure Eight and the Sagnac Rejector Laser: FIGURE EIGHT
Faraday nonreciprocal phase shifters give simpler design rules takes care of the reversal of the y-axis. Thus, we find from
but are in general less desirable because of their cost and (4.1)--(4.4) for the scattering matrix TRof the controller turned
limited bandwidth. around (reversed):
4 ) The Linear Loss-Free Resonators and the Scattering Ma-
trices ofthe Polarisation Controllers: We shall now determine TR = PTtP (4.5)
what transformers lead to loss-free modes of the figure-eight where the subscript t indicates a transpose.
and Sagnac reflector laser. The loop with a perfectly balanced Let us now consider the transmission of the loop, as used
(50/50)coupler can act either as a perfect reflector, a perfect in the figure-eight configuration. We want to make it perfectly
transmitter, or somewhere in between, depending upon the transmissive in the absence of gain and nonlinearity. Denoting
adjustment of the polarization controller [27]. When it is by b the output as shown in Fig. 8 we have
close to a perfect transmitter, it is one of the loops in the
b = i[T - P T R P ] ~ . (4.6)
figure-eight laser. When it acts more like a reflector it is the
reflecting loop in the Sagnac loop reflector laser. The analysis The - sign in the above expressions come from double
calls for a more sophisticated treatment of the scattering passage through the coupler with scattering matrix S
matrix describing the transformation of polarization by the
polarization controllers in Fig. 6. The loop is traversed both
s = -1[ 1 j] (4.7)
forwards and backwards, and this requires careful analysis of J z j 1
time reversed propagation. We note that the transmission will be zero, when T = PTRP
We use linear polarization as the basis as in (2.1). Passage and can be made total, for both polarization when
through the controller multiplies a by T as in (2.2). If the T = -PTRP. How does one achieve the second condition?
IiAUS et U/.: ADDITIVE PULSE MODELOCKING 205
(4.9)
L J
and the other polarization experiences
The polarization has been transformed from I[: to y and from
y to -I[:, respectively. Let us proceed further with the figure- a- = ~.elA(~
206 IEEE JOURNAL OF QUANTUM ELECTRONICS. VOL. 30. NO. 1. JANUARY 1994
y=-
G-1
4
ne sin (5) Fig. 10. The shortest limit on the stability was slow drift in the
alignment of the 980 nm Ti:sapphire pump system. Even so, it
ran stably for hours. With diode pumping, the stability should
only be limited by environmental effects on the fibers, which
e=i-cos (5) over days required some minor adjustments of the polarization
controller settings.
An autocorrelation of the pulse at 192 pW signal output
where y and 6 have been referred to the amplitude A incident power is shown in Fig. 11. The FWHM of the trace is 700 fs,
upon the loop. The Sagnac ring reflector can be made to corresponding to a 452-fs pulse if a sech profile is assumed.
operate very similarly if one uses the configuration with A long-range scan of over 80 ps was also performed in
R given by (4.11), a configuration that also rotates the order to verify the presence of only a single pulse within
polarizations traveling through the ring. Then the loss bias the 80 ps resolution limit of our sampling oscilloscope. The
and the nonlinear phase shift are produced in the same way corresponding peak power is estimated to be 101 W inside
as in the figure-eight ring laser, and the SAM, SPM and loss the cavity.
coefficients are also given by (5.6)-(5.8). A typical spectrum is shown in Fig. 12 and reveals sharp
6. Experimental Results on Fiber Ring Laser: We now spikes along with the broad spectrum of the modelocked
discuss experimental results obtained with the laser depicted in signal. These spikes are found to occur at specific frequencies,
Fig. 1 [ 121. A fiber pigtailed polarization sensitive isolator was independent of the adjustment of the polarization controllers.
used to force unidirectional operation of the ring. The pigtails Only their amplitudes are affected by such adjustments. We
were constructed of SMF28 fiber. The output was taken from attribute them to spurious CW oscillations in the ring. Proper
a 90/10 output coupler constructed of dispersion-shifted fiber. filtering could remove them, but it is difficult to introduce
A 980/1550 nm wavelength-division multiplexing (WDM) filters into an all-fiber ring.
coupler constructed of Coming Flexcor 1060 fiber was used The average dispersion of the fiber was estimated to be
to launch pump light from a Tisapphire laser tuned to 980 nm. approximately -19 ps.ps/km. For a soliton of 452 fs FWHM,
HAUS et al.: ADDITIVE PULSE MODELOCKING 207
11. DISCUSSION
The ring laser shows the way for the construction of self-
starting APM all-fiber lasers with negative GVD producing
pulses with low background (clean spectra). The rules for the
design of such lasers are the following:
I. One pulse, with energy quantized by the saturating APM
action should be able to saturate the gain. This calls for
relatively short, high-repetition-rate resonators.
2. Unidirectional operation should be enforced with an
isolator, thus favoring self-initiation of modelocking.
3. The nonlinear phase shift of the pulses should be small
compared with 27r per roundtrip in order to suppress
sideband generation.
We have presented a simple analysis of the figure-eight laser
and the Sagnac loop reflector laser based on the assumption
that the changes per pass are small. We argued that this
Fig. 10. The pulse train. assumption is justified, especially for comparison, because
operation under conditions in which the nonlinear changes
per pass through the nonlinear elements are large may lead
to inferior performance and instability.
There is another observation that suggests soliton behavior
within components of the system is to be avoided, namely
the operation of the Sagnac loop nonlinear reflector. If the
pulses circulating in the Sagnac loop strip themselves of their
continuum radiation after entry into the loop and really become
solitons, their interference at the coupler deviates from the
standard APM action. Solitons have constant phase across their
Time (ps) profile, and hence their shaping cannot be likened to that of
Fig. I 1. The autocorrelation trace. an equivalent saturable absorber [23].
The negative GVD is not essential for the operation of the
laser. If the GVD is positive, chirped solutions are obtained
I I from (1.1). Just as in the case of (perturbed) solitons, it
was found from computer simulations that stable pulses are
obtained if the SPM-induced phase shift of the pulse is less
than 27r. Hence, similar criteria as outlined here for negative
GVD ought to prevail for a laser operating with positive GVD.
In a more general vein, it appears that self-starting could
be promoted in all solid-state APM and KLM systems by
implementation in a ring configuration with Faraday isolation.
Wavelength (microns)
REFERENCES
Fig. 12. The spectrum
F. Ouellette and M. Piche, Pulse shaping and passive modelocking with
the peak soliton power calculated for the given fiber ge- a nonlinear Michelson interferometer, Opt. Comm..vol. 60, p. 99, 1986.
ometry and dispersion is 110 W, which is comparable to
N
K. L. Blow and B. P. Nelson, Improved mode locking of an F-center
laser with a nonlinear nonsoliton extemal cavity, Opt. Lett., vol. 13,
the measured level. The distance within which the nonlinear p. 1026, 1988.
phase shift of the soliton is 27r(8Z0) is on the order of J. Mark, L. Y. Liu, K. L. Hall, H. A. Haus, and E. P. Ippen,
SO m, which is much longer than the loop length. For the Femtosecond pulse generation in a laser with a nonlinear external
resonator. O p f .Lett., vol. 14, p. 48, 1989.
single-pulse regime, the pulsewidth was found to remain D. E. Spence, P. N. Kean. and W. Sibbett, Sub-100 fs pulse gen-
constant as the pump power was increased to the point eration from a self-modelocked titanium-sapphire laser. in Digest of
Conference on Lasers and E1ecfr.o-Optics(Optical Society of America.
of multiple pulsing. The autocorrelation trace revealed the Washington, DC. 1990), paper CPDP10-I.p. 619.
development of a large pedestal extending over 40 ps, which D. K. Negus, L. Spinelli, N. Goldblatt, and F. Feuget, Sub-100
was not observed at lower pump powers. The pedestal may femtosecond pulse generation by Kerr lens modelocking in Ti:AI rO:<,
in Di,qest of Topical Meeting on Advuttced Solid State Lasers (Optical
be attributed to increased coupling of pulse energy to the Society of America, Washington, DC, 1991). postdeadline paper.
continuum. This point of view is reinforced by the observation M. PichYe, Beam reshaping and self-mode-locking in nonlinear laser
that large nonlinear phase shifts per pass cause coupling to the resonators, Opt. Comm.. vol. 86, pp. 156. 1991.
H. A. Haus, J . G . Fujimoto, and E. P. Ippen, Analytic theory of additive
continuum from the unavoidable nonuniformity of gain and pulse modelocking and Kerr lens modelocking, IEEE ./. Qiiunfiim
loss along the fiber. Elwfron., to be published.
208 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 30, NO. 1, JANUARY 1994
[SI I. N. Duling, Sub-picosecond all-fiber erbium laser, Electron. Lett., lasers, Opt. Lett.. to be published.
vol. 27, pp. 544-545, May 1991. [21] L. F. Mollenauer, S. G. Evangelides, Jr., and H. A. Haus, Long-
[9] D. J. Richardson, R. I. Laming, D. N. Payne, M. W. Phillips, and V. J. distance soliton propagation using lumped amplifiers and dispersion
Matsas, 320 fs soliton generation with passively modelocked erbium shifted fiber, J . Lighrwave Tech., vol. 6, pp, 194-197, Feb. 1991.
fibre laser, Electron. Letf., vol. 27, pp. 730-732, 1991. [22] S. Kelly, Characteristic sideband instability of periodically amplified
[IO] E. Yoshida, Y. Kimura, and M. Nakazawa, Laser diode-pumped fem- average soliton, Electron. Lett., vol. 28, p. 806, 1992.
tosecond erbium-doped fiber laser with a sub-ring cavity for repetition [23] H. A. Haus, J. G. Fujimoto, and E. P. Ippen, Structures for additive
rate control, Appl. Phys. Letf.,vol. 60, pp. 932-934, 1992.
pulse mode locking,J. Opt. Soc. Am. B.. vol. 8, pp. 2068-2076, 1991.
[111 M. Fermann and A. Weiner, Poster paper at the Ultrafast Phenomena [24] D. J. Richardson, A. B. Grudinin, and D. N. Payne, Soliton energy
Conference, June 6 1 2 , 1992, Antibes, France.
[I21 K. Tamura, H. A. Haus, and E. P. Ippen, Self-starting additive pulse quantization in the output power of a passive femtosecond soliton laser,
modelocked erbium fiber ring laser, Electron. Lett., vol. 28, pp. Proc. CLE092, paper JWA5, p. 384, Anaheim, CA, 1992.
22262227, 1992. [25] H. G. Winful, Self-induced polarization charges in birefringent optical
[I31 V. J. Matsas, T. P. Newson, D. J. Richardson, and D. N. Payne, fibers, Appl. Phys. Lett., vol. 47, pp. 213-215, 1985.
Selfstarting passively mode-locked fibre ring soliton laser exploiting [26] N. Pandit, D. U. Noske, S. M. J. Kelly, and J. R. Taylor, Characteristic
nonlinear polarisation rotation, Electron. Len., vol. 28, pp. 1391-1393, instability of fibre loop soliton lasers, Electron. Lerr., vol. 28, pp. 455,
1992. 1992.
1141 D. U. Noske, N. Pandit, and J. R. Taylor, Subpicosecond soliton [27] M. E. Fermann, F. Haberl, M. Hofer, and H. Hochreiter, Nonlinear
pulse formation from self-mode-locked erbium fibre laser using intensity amplifying loop mirror, Opt. Lett., vol. 15, p. 752, 1990.
dependent polarisation rotation, Electron. Left., vol. 28, p. 2185, 1992.
[I51 C.-J. Chen, P. K. A. Wai, and C. R. Menyuk, Soliton fiber ring laser,
Opt. Lett.. vol. 17, pp. 417, 1992.
[I61 R. H. Stolen, J. Botineau, and A. Ashkin, Intensity discrimination of
optical pulses with birefringent filters, Opt. Left., vol. 7, p. 512, 1982. H. A. Haus, photograph and biography not available at the time of publication.
1171 M. Hofer, M. E. Fermann, F. Haberl, M. H. Ober, and A. J. Schmidt,
Mode locking with cross-phase and self-phase modulation, Opt. Left.,
vol. 16, pp. 502-504, 1991.
[I81 J. D. Kafka, T. Baer, and D. W. Hall, Modelocked erbium-doped fiber
laser with soliton pulse shaping, Opr. Left., vol. 14, pp. 1269-1271, E. P. Ippen, photograph and biography not available at the time of publication.
1989.
[201 g.,,.
[I91 H. A. Haus and E. P. Ippen, Self-staring of passively mode-locked
Opt. Lett.. vol. 16, pp. 1331-1333, 1991.
Tamun, J. Jacobson, E. P. Ippen, H. A. Haus, and J. G. Fujimoto,
Unidireclional ring resonators for self-starting passively modelocked K. Tamura, photograph and biography not available at the time of publication.