Sie sind auf Seite 1von 3

March 1, 1993 / Vol. 18, No.

5 / OPTICS LETTERS 367

42-fs pulse generation from a mode-locked fiber laser started


with a moving mirror

M. H. Ober, M. Hofer, and M. E. Fermann*


A-1040 Vienna, Austria
Technische Universitat Wien, Abteilung Quantenelektronik und Lasertechnik, Gusshausstrasse 27/359/9,

Received September 25, 1992


for what is
Passive mode locking initiated with a moving mirror is demonstrated in a neodymium fiber laser
pulses with a width of 42 fs and energies as high as
to our knowledge the first time. Near-bandwidth-limited
1 nJ are generated.

The recent development of the femtosecond the phase coherence between the cavity modes can
Ti:sapphire oscillator' has driven a renaissance in be increased further by translating one of the cavity
ultrafast optical technology. The rapid progress has end mirrors, which not only effectively increases T,
been possible since the gain medium is solid state but also provides a larger seed signal than produced
and offers a wide bandwidth for a large tuning range. from mode beating alone. However, the exact origin
Pulse formation can easily be initiated passively by of this effect still remains unclear.
using an additional slow saturable absorber
2' 3 or sim- Our experimental efforts thus centered on opti-
ply by moving one of the cavity mirrors.
4 6 Single- mizing T, and evaluating the moving-mirror tech-
mode rare-earth-doped fibers7 offer some unique nique for self-starting of waveguide lasers. The fiber
properties for simplifying femtosecond oscillators laser cavity is shown in Fig. 1 and is essentially
even further. Fiber lasers are equally solid state the same as that described in Ref. 10. It consists of
and offer a wide tunability range. In addition, their a Fabry-Perot resonator that contains a polarizing
waveguide structure allows one to reduce some of dispersive delay line and an active neodymium-doped
the requirements on the pump source and should fiber butted to one of the cavity mirrors. The wedged
potentially lead to more compact femtosecond pulse output coupler had a reflectivity of 54%. A weakly
sources. A major limitation of previous attempts to linearly birefringent, 20-cm-long piece of silica fiber
(numerical aperture 0.17, core diameter 5 ptm, and
produce sub-100-fs pulses from fiber lasers has been
that an active modulator was required to initiate Nd3 + doping level of 1700 parts in 106)was employed
the pulses,8 which is unacceptable in the context of in the experiments. The polarization state in the
simplicity. fiber was adjusted with a single twisted fiber loop.
Here we demonstrate for what is to our knowledge A Kr+-ion laser operating simultaneously on the 752-
the first time femtosecond pulse generation in a Nd and 799-nm lines was used as the optical pump
fiber laser without resorting to any intracavity start- source. With an absorbed pump power of 320 mW,
ing mechanism. We employ a moving mirror for the the maximum cw and mode-locked output powers
start-up of mode locking, since for the generation of were 120 and 73 mW, respectively. Passive mode
the shortest possible pulses this technique may be locking was initiated by manually translating the
preferred. It does not require any additional cavity output coupler and sustained 8 by nonlinear polar-
components, and no additional bandwidth modula- ization evolution in the fiber. Once pulses were
tion is incorporated. The method yields pulses as initiated the mirror movement was stopped and re-
short as 42 fs with pulse energies as high as 1 nJ. liable mode locking without dropout for periods of
A minimum self-starting threshold may be calcu- hours was achieved.
lated from the mutual coherence9 time r, = 1/17Av When we maximized the cw lasing signal by ad-
of the axial modes in the cavity, where Av is the justing the polarization state inside the fiber, the
half-width of the first beat note of the free-running half-width of the first beat note of the free-running
laser. In order to obtain self-starting mode locking, laser was measured to be 30 kHz, which implies
the critical buildup time Tcrit = TR/[KPav ln(m)] of a
pulse in the cavity has to be smaller than 'r, where POLARIZATION
TR is the round-trip time, Pavis the average cw intra- CONTROL
HIGH
cavity power, m is the number of initially oscillating OUTPUT REFLECTOR
modes, and K is the passive amplitude-modulation 4
COUPLER LENS
coefficient. 10 As proposed by Haus and Ippen, rc
~~~~~~PUMP
DDL
is mainly limited by intracavity reflections, which BEAM
cause uneven frequency shifts of the cavity modes
and thus lead to a broadening of the beat notes.
Hence a good measure of a clean cavity is a r, as Fig. 1. Cavity design for a passively mode-locked Nd
long as possible. Recently Liu et al. suggested that
6 fiber laser oscillator. DDL, dispersive delay line.
0146-9592/93/050367-03$5.00/0 © 1993 Optical Society of America
368 OPTICS LETTERS / Vol. 18, No. 5 / March 1, 1993

beat-note spectrum by approximately an order of


3.0 l magnitude.
Depending on the polarization state inside the
I 2.5 - fiber, either pulses with a cw background or clean
pulses could be initiated. Clean pulse formation
relied on maximizing the amount of passive ampli-
'i2.0
tude modulation, which also produces the largest
C mode-locked powers. Clearly the linear cavity loss
1. 5 under these conditions was higher than that present
for non-background-free pulses. In general, clean
pulses were the most difficult to initiate and required
the fastest movement of the mirror. But it was also
1 .0 . .. . . .
-400 -200
possible to first initiate pulses with a background
0 200 400 and then subsequently to optimize the polarization
Time Delay [fs] state to obtain clean pulses. For a polarization state
at which background-free pulses could be initiated
8 we deduced the following start-up conditions for
7 an absorbed pump power of -320 mW: r, = 3 /s,
6 ln(m) : 10, Pa,- 30 mW, and K 8 X 10-5 W'1,
which gives a Tcrit of 600 As. Thus when neglecting
5 passive phase-modulation effects,13 we are a factor of
.C_1
4 200 away from self-starting. In contrast, to start up
3
non-background-free pulses, we measured T0 = 3.8
(n As, ln(m) - 10, Pav 52 mW, and K 1.6 x 10-4
:C 2 W-', which gives a Tcritof 170 Us. Hence the moving
1 mirror allows one to lower the self-starting threshold
O
by some 2 orders of magnitude. The minimum
I . . . absorbed pump power to start up the mode-locking
-200 -1 00 0 100 200 process was 140 mW at an output coupling of 20%,
Time Delay [fs] which is within the range of current diode pump
Fig. 2. Collinear intensity (top) and fringe resolved (bot- sources.
tom) autocorrelation traces of a 53-fs pulse. The over- When initiating pulse formation with a moving
layed sech2 envelope confirms that the pulses are nearly mirror, we obtained clean 53-fs FWHM pulses with
transform limited. SH, second harmonic. a time-bandwidth product of 0.39 (assuming a sech2
shape) and energies as high as 1 nJ. Autocorrelation
a r, = 10 ,s. A reduction in the output coupling traces of the pulse trains are shown in Fig. 2, and
narrowed Aiv by up to a factor of 2, whereas a slight the corresponding spectrum is displayed in Fig. 3.
misalignment of the cavity produced a beat note as At this point the contrast ratio was measured to
narrow as 1.5 kHz (measurement limited). A possi- be larger than 2000:1 (measurement limited) 0.5 ps
ble explanation for this reduction may be that mis- away from the pulse center, which indicates that the
alignment produces an optimum phase relationship pulses were of high quality and did not have any
between the cavity reflections,4 where the frequency small shoulders. In contrast to all-fiber passively
shifts between the cavity modes are minimized. The mode-locked Er fiber lasers,'4 the pulse trains were
obtained beat-note bandwidths are as narrow as in stable and had a fixed repetition rate of 70 MHz.
bulk Nd:glass lasers," which clearly demonstrates The average mode-locked power of as much as 70 mW
that fiber lasers can be virtually as reflection free as
their bulk counterparts.
Unfortunately, a polarization setting that provides
maximum output power minimizes the amplitude
modulation. Therefore passive mode locking could
not be initiated in spite of the narrow beat-note
linewidth. A polarization state in the fiber at which
efficient passive mode locking is possible induces a
large linear loss in the cavity (up to 90%), since 0)
C:
in these highly nonlinear lasers a large amount of
passive amplitude modulation is required to keep C

the pulses stable.' 0 The large linear loss'2 causes


a broadening of Av of as much as 400 kHz. This
was verified by adjusting the polarization state to
obtain maximum cw lasing power and introducing 1020 1040 1060 1080 1100
a loss with an intracavity variable aperture. A Wavelength [nm]
reduction in intracavity power by a factor of 2 Fig. 3. Spectrum of a 53-fs pulse. The time-bandwidth
then equally gave rise to a broadening of the product AvAt = 0.39.
March 1, 1993 / Vol. 18, No. 5 / OPTICS LETTERS 369

8
the Fonds zur F6rderung der Wissenschaftlichen
Forschung in Osterreich, project No. P8024-TEG,
7 and by the Osterreichische Nationalbank project
6 No. P4049. M. E. Fermann acknowledges financial
5
support from the Alexander von Humboldt Stiftung.
*Present address, IMRA America, Inc., 1044 Wood-
I l | - l A
0) 4 bridge Avenue, Ann Arbor, Michigan 48105.
3
C
21 References
(n
1. D. E. Spence, P. N. Kean, and W. Sibbett, Opt. Lett.
0 11 I
16, 42 (1991).
2. N. Sarukura, Y. Ishida, T. Yanagawa, and H. Nakano,
-2 200 - 100 0 100 200 Appl. Phys. Lett. 57, 229 (1990).
[fs]
Time Delay 3. U. Keller, G. W. 'tHooft, W. H. Knox, and J. E. Cun-
ningham, Opt. Lett. 16, 1022 (1991).
Fig. 4. Interferometric autocorrelation trace of a 42-fs 4. H. A. Haus and E. P. Ippen, Opt. Lett. 16, 1331 (1991).
pulse. The structure in the pulse wings arises from 5. N. H. Rizvi, P. M. W. French, and J. R. Taylor, Opt.
residual nonlinear chirp. Lett. 17, 279 (1992).
6. Y. M. Liu, K. W. Sun, P. R. Pruncal, and S. A. Lyon,
makes this sort of oscillator attractive for seeding Opt. Lett. 17, 1219 (1992).
bulk glass amplifiers.'5 Since we did not observe 7. S. B. Poole, D. N. Payne, and M. E. Fermann, Electron.
any pulse breakup and a split spectrum, which arises6 Lett. 21, 737 (1985).
when the pulse straddles the zero-dispersion point,' 8. M. Hofer, M. E. Fermann, F. Haberl, M. H. Ober, and
we estimate that for the 53-fs pulses the group- A. J. Schmidt, Opt. Lett. 16, 502 (1991).
velocity dispersion was negative across the whole 9. F. Krausz, T. Brabec, and Ch. Spielmann, Opt. Lett.
pulse spectrum. 16, 235 (1991).
10. M. Hofer, M. H. Ober, F. Haberl, and M. E. Fermann,
With further optimization of the dispersion and po- IEEE J. Quantum Electron. 28, 720 (1992).
larization, we generated slightly nonlinearly chirped 11. J. Zehetner, Ch. Spielmann, and F. Krausz, Opt. Lett.
pulses with a FWHM of 42 fs and a time-bandwidth 17, 871 (1992).
product of 0.49, as shown in Fig. 4. The shortest pos- 12. F. Krausz, M. E. Fermann, T. Brabec, P. F. Curley,
sible pulses had a width of 32 fs but exhibited clearly M. Hofer, M. H. Ober, Ch. Spielmann, E. Wintner, and
visible shoulders. However, these pulses required A. J. Schmidt, IEEE J. Quantum Electron. 28, 2097
an acousto-optic modulator for start-up. (1992).
In conclusion, a moving mirror was successfully 13. C. Spielmann, F. Krausz, T. Brabec, E. Wintner, and
employed to initiate femtosecond pulse formation in a A. J. Schmidt, IEEE J. Quantum Electron. 27, 1207
passively mode-locked Nd fiber laser for the first time (1991).
The oscillator reliably generates 14. I. N. Duling, Opt. Lett. 16, 539 (1991).
to our knowledge.
15. M. Hofer, M. H. Ober, F. Haberl, M. E. Fermann,
50-fs pulses with an energy content as high as 1 nJ. E. R. Taylor, and K. P. Jedrzejewski, Opt. Lett. 17,
We are indebted to Y. Silberberg for stimulating 807 (1992).
discussions and F. Krausz for a critical reading of 16. A. S. Gouveia-Neto, M. E. Faldon, and J. R. Taylor,
the manuscript. This research was supported by Opt. Lett. 13, 770 (1988).

Das könnte Ihnen auch gefallen