Professional Documents
Culture Documents
High-Energy Pulse Compression by Use of Negative Phase Shifts in Cascaded Second Order Non Linearity
High-Energy Pulse Compression by Use of Negative Phase Shifts in Cascaded Second Order Non Linearity
Pulse compression is an established technique for gen- was demonstrated, with energy efficiency of close to
erating optical pulses shorter than those produced by 50%. Dubietis and co-workers have also demonstrated
lasers or amplif iers. Most commonly, additional band- the phase-matched generation of second-harmonic
width is generated by self-phase modulation in an pulses shorter than the input fundamental pulse
optical fiber. The group-velocity dispersion (GVD) re- through pulse tilting.6
quired for compression of the pulse is typically provided Here we show that negative phase shifts produced in
by gratings or prisms. Compressors based on single- phase-mismatched type I second-harmonic generation
mode fibers are limited to nanojoule pulse energies by can be exploited for effective pulse compression. Our
higher-order nonlinear effects and ultimately by dam- research is conceptually similar to that employed in
age to the fiber. Thus, new approaches are needed traditional compressors: In a first stage the pulse
for compression of the high-energy pulses that are now accumulates a nonlinear phase shift, and the pulse is
available from chirped-pulse amplif iers,1 for example. then compressed by dispersive propagation in a second
Bulk materials can be used for pulse compression stage [Fig. 1(a), inset]. The use of a self-defocusing
based on third-order nonlinearity.2 Continuum gen- nonlinearity reduces or eliminates the problems that
eration arising from self-focusing can severely distort can arise from Kerr self-focusing. Positive GVD is
the beam and impede effective compression. These needed for compression, and this can be provided
effects limit the utility of compression in bulk third- by a suitably chosen piece of transparent material.
order media. 120-fs pulses are compressed by a factor of ⬃4, and
One solution to this problem was reported by Nisoli higher compression ratios should be possible. The
et al.3 They achieved large spectral broadening by compressor is eff icient, with the compressed pulse
propagating pulses through a high-pressure noble gas amounting to at least 85% of the input-pulse energy.
conf ined in a hollow-core waveguide of fused silica. It is well know that the cascading of x 共2兲 共v; 2v, 2v兲
Excellent results were obtained, including compres- and x 共2兲 共2v; v, v兲 processes leads to a nonlinear phase
sion from 140 to 10 fs. Pulse energies as high as shift DF NL in a pulse that traverses a quadratic
240 mJ were produced with 660-mJ input pulses. Al- medium under phase-mismatched conditions for
though the compressed pulse energy is a substantial second-harmonic generation (SHG). The phase shifts
improvement on that which is achievable with ordi- can be either positive or negative, depending on the
nary fibers, these results do point out a limitation of sign of the phase mismatch DkL 共Dk 苷 k2v 2 2kv 兲.
this approach: Because the pulse does not propagate Bakker and co-workers performed a theoretical study
as a guided mode, the waveguide is lossy. Additional of the phase shifts generated by three-wave interac-
drawbacks include the susceptibility of the waveguide tions,7 and cascade phase shifts were later measured
to optical damage, the complexity associated with han-
dling the high-pressure gas, and a lack of commercially
available components.
Recent research has shown that second-order non-
linearities can be exploited for pulse compression.
Following research by Wang and Luther-Davies,4
Dubietis et al. demonstrated that pulses can be com-
pressed in phase-matched type II second-harmonic
generation.5 This approach relies on group-velocity
mismatch (GVM) among the three interacting waves
and requires division of the input pulse into o and e
waves as well as on an appropriate predelay of one of Fig. 1. (a) Calculated spectrum and ( b) intensity prof ile of
the input pulses. Compression from 1.3 ps to 280 fs the compressed pulse. Dashed curves, the input pulse.
in KTP (Ref. 8) and periodically poled LiNbO3 .9 We where E1 and E2 are in units of the initial value
recently employed cascade nonlinearities for Kerr-lens of the peak fundamental field E0 . The length that
mode locking with DF NL , 0 (Ref. 10) as well as for the characterizes the nonlinear interaction is LNL 苷
production of pulses that are simultaneously solitons nl兾关px 共2兲 E0 兴. Time is measured in units of the input-
in time and space.11 pulse duration t0 , and position is measured in units
We take the simplest approximate approach to ob- of LNL . We solved the wave equations numerically,
tain guidelines for compressor design and performance and the results are compared below with experimental
and then ref ine these through numerical solutions of results.
the appropriate wave equations; a systematic analy- Input pulses of duration 120 fs and energy 600 mJ
sis of the compressor will be presented elsewhere. In at 795 nm were produced by a Ti:sapphire regenera-
general the maximum compression ratio will be propor- tive amplifier. The 7-mm-diameter beam from the
tional to DF NL impressed upon the initial pulse. The amplif ier is compressed by a factor of 2 with a telescope
phase shift produced by the cascade process can be ap- to produce intensities of 20 60 GW兾cm2 on the BBO
proximated as8 crystal. In the second stage of the compressor, a prism
pair or a piece of bulk material provides the GVD
G 2 L2 , to compensate for the phase shift accumulated in the
DF NL 艐 2
DkL quadratic medium.
p With DkL 苷 200p, the spectrum is expected to
with G 艐 vdeff jE0 j兾c n2v nv . The cascade process broaden by a factor of ⬃3 and to develop a multi-
produces a nonlinear frequency chirp in the presence ply peaked structure [Fig. 1(a)]. The GVD required
of GVM. However, with large enough phase mis- for optimum pulse compression is ⬃1500 fs2 , and this
match, the cycles of conversion and backconversion should produce an ⬃30-fs compressed pulse [Fig. 1(b)].
that generate DF NL occur before the pulses move Since we neglect GVD in the SHG crystal, the GVD
away from each other because of GVM. Then DF NL that must be supplied is smaller than the calculated
follows the pulse intensity, so the frequency chirp value. However, the compressed pulse duration is not
is linear near the pulse peak. (Calculations that extremely sensitive to the GVD, so we provided the
illustrate this issue are presented in Ref. 10.) To approximate calculated value. The measured spec-
quantify this point, we introduce the characteristic trum [Fig. 2(a)] and pulse autocorrelation [Fig. 2(b)]
group-mismatch length LGVM 苷 ct0 兾共n1g 2 n2g 兲, produced with a prism pair controlling GVD agree
where n1g 共n2g 兲 is the group index of the fundamental reasonably with calculations. The zero-phase Fourier
(harmonic) frequency and t0 is the initial pulse dura- transform of the experimental spectrum [Fig. 2(c)]
tion. For a crystal of length L 苷 NLGVM we desire produces an autocorrelation [shown in Fig. 2(b)] that
at least 2N conversion –backconversion cycles, which
we arranged by setting DkL 苷 4Np. Crystals with
large values of deff and LGVM are naturally best for
this application. For the compression of 120-fs pulse
at 800 nm we chose to use barium metaborate (BBO),
which has deff 艐 2 pm兾V and LGVM 苷 0.6 mm. With
a 17-mm BBO crystal (provided by Casix, Inc.), this
implies that DkL $ 120p, and we chose DkL 苷 200p.
With intensities of ⬃50 GW兾cm2 , DkL 艐 200p yields a
net DF NL 艐 2p, which is necessary for a compression
ratio of roughly 4. Materials with small Kerr nonlin-
earities are desired for the generation of net negative
phase shifts. The nonlinear index of BBO is small,
n2 艐 5 3 10216 cm2 兾W. A final advantage of BBO is
its lack of two-photon absorption at 400 nm.
In the reference frame of the fundamental pulse, the
equations that govern the interaction between funda-
mental and harmonic fields E1 and E2 propagating in
the z direction, neglecting GVD, are
≠
E1 苷 iE1 ⴱ E2 exp共iDkz兲 1 i2p共n2 I0 兲
≠z