Enabling pulse compression and proton acceleration in a modular ICF driver for nuclear and particle physics applications
ABSTRACT The existence of efficient ion acceleration regimes in collective laser-plasma interactions opens up the possibility to develop high-energy physics facilities in conjunction with projects for inertial confinement nuclear fusion (ICF) and neutron spallation sources. In this paper, we show that the pulse compression requests to make operative these acceleration mechanisms do not fall in contradiction with current technologies for high repetition rate ICF drivers. In particular, we discuss explicitly a solution that exploits optical parametric chirped pulse amplification and the intrinsic modularity of the lasers aimed at ICF.
arXiv:physics/0509096v2 [physics.acc-ph] 7 Dec 2005
Enabling pulse compression and proton acceleration
in a modular ICF driver for nuclear and particle
F. Terranovaa,, S.V. Bulanovb,c, J.L. Collierd, H. Kiriyamab, F. Pegoraroe
aI.N.F.N., Laboratori Nazionali di Frascati, Frascati (Rome), Italy
bAdvanced Photon Research Centre, JAERI, Kizu-cho, Kyoto-fu, Japan
cA. M. Prokhorov General Physics Institute of RAS, Moscow, Russia
dCentral Laser Facility, Rutherford Appleton Laboratory, Didcot, UK
eDip. di Fisica, Univ. di Pisa and CNISM, Pisa, Italy
The existence of efficient ion acceleration regimes in collective laser-plasma inter-
actions opens up the possibility to develop high-energy physics facilities in con-
junction with projects for inertial confinement nuclear fusion (ICF) and neutron
spallation sources. In this paper, we show that the pulse compression requests to
make operative these acceleration mechanisms do not fall in contradiction with
current technologies for high repetition rate ICF drivers. In particular, we discuss
explicitly a solution that exploits optical parametric chirped pulse amplification
and the intrinsic modularity of the lasers aimed at ICF.
PACS: 41.75.Jv, 42.65.Yj, 52.38.Kd
Particle acceleration through collective short wavelength electromagnetic effects has
been pursued for decades . The advent of laser has made it possible to use the
interaction of the laser light with the charged particles  and the plasma  and this
technique is still considered a viable alternative to traditional RF-based boosters. Such
a confidence has been strengthened after the advent of wideband oscillators based on
Ti:sapphire and the revolution in power ultrafast lasers due to the development of
Chirped Pulse Amplification  (CPA). Production of energetic ions and electrons has
been reported by several experimental groups and new breakthroughs are expected
after the commissioning of the next generation of multi-petawatt lasers. Nonetheless,
there is large consensus on the fact that a well-understood and stable regime of ac-
celeration hasn’t been achieved, yet. The lack of a reference mechanism is the root
of the best known drawbacks of laser acceleration: strong dependence on the initial
conditions, large energy spread, poor light-to-particle energy conversion efficiency and
significant shot to shot variations. These limitations, however, are unlikely to be in-
trinsic features of laser acceleration. This is particularly clear for ion acceleration. It
has long been understood that fast ion generation is related to the presence of hot
electrons . A variety of effects occur when the main source of acceleration is charge
displacement due to the electron motion in the plasma or inductive electric fields from
self generated magnetic fields . However, for higher laser intensities (1023W/cm2)
acceleration results from the laser pressure exerted to the comoving electron-ion sys-
tem; the latter acts as a progressively more opaque screen for the laser light and, in
this case, charge displacement only plays the role of rectifier for the transversal laser
field and as a medium for electron-ion energy transfer during light absorption. This
mechanism is described in details in [7, 8, 9]. It provides energy transfer efficiency com-
parable to RF-based synchrotrons or even cyclotrons and, more importantly, decouples
the final ion energy from the accelerated ion current (see Sec.4). If this mechanism
were confirmed experimentally, it could represent the first serious alternative to syn-
chrotrons suited for high energy physics (HEP) applications. As noted in , a proof of
principle of this radiation-pressure dominated (RPD) acceleration mechanism is at the
borderline of current technology, but the possibility of using this technique to overcome
the limitations of traditional proton accelerators faces many additional difficulties. In
particular all present high power lasers operate at very low repetition rate. This is a
classical problem e.g. in inertial confinement fusion (ICF), where the basic principle
could be demonstrated, for instance, at the National Ignition Facility (NIF) in US 
on a single-shot basis. However, the ultimate use of ICF to produce electric power
will require repetition rates of the order of tens of Hz, an increase of several order of
magnitude compared to the shot rate achievable with state-of-the-art fusion laser tech-
nology. This rate cannot be achieved with flashlamp-pumped neodymium-doped glass
lasers that requires a significant interpulse cooling time. There is, currently, a very
large effort to find alternative solutions and promising setups based on excimer lasers
or diode-pumped solid state lasers have been identified, offering, in principle, the repe-
tition rates needed for ICF. In , we noted that the solution of the problem of thermal
stability and the exploitation of the RPD mechanism would open up the fascinating
possibility of a multipurpose driver for ICF and HEP applications. However, we did
not address explicitly the question whether the power requirements for ICF could fall
in contrast with the pulse compression requests needed for the RPD acceleration to be
fully operative. More precisely, the strong constraints on the choice of the amplifying
medium (and therefore on the gain bandwidth) could make impossible an appropriate
pulse compression, so that the intensity needed to operate in the RPD regime would
not be reached by a multi-pulse device. In this paper we show that there exists at least
one particular configuration that is able to fulfill simultaneously the two requirements;
i.e. it offers a high repetition rate device that can be implemented to design a multi-
GeV proton booster with a technology suited for ICF energy production. This solution
is based on the exploitation of optical parametric chirped pulse amplification [11, 12]
(OPCPA) and the intrinsic modularity of ICF drivers1(see Sec.2); it is described in
details in Sec. 3, while its potentiality as a new generation proton driver is discussed
in Sec. 4
2 Drivers for Inertial Confinement Fusion
Large scale flashlamp-pumped solid state lasers built for ICF are inherently single shot
devices, requiring several hours to recover from thermal distortions. They are aimed
at a proof-of-principle for ICF but their scaling to a cost effective fusion reactor is
highly non trivial. In the last decade, three main research lines have been investigated
and much more efforts will be put in the forecoming years if the proof-of-principle
programs for ICF at NIF or at Laser Megajoule (LMJ) in France is successful. The
first one exploits traditional RF-based technologies for ion acceleration to transfer
energy to the target, trigger ignition and sustain burning. This approach profits of
the enormous experience gained in particle accelerators since the 50’s and the large
efficiency (∼ 30−35%) obtained at HEP facilities  but, as a matter of fact, the mean
intensities and the required uniformity of target illumination are well beyond current
technology. The second and third ones exploit lasers to ignite and sustain fusion and
are aimed at developing systems with much higher thermal yield than Nd:glass. They
are based on diode-pumped solid state lasers (DPSSL) or high power excimer lasers.
1These drivers are constituted by an ensemble of independent beamlines. Synchronous operation
of the beams is requested for fusion applications since the power must be delivered in a single shot
(maximum peak power) and illuminate uniformly the fuel target. However, in most of the applications
related to particle and nuclear physics, the performance of the facility mainly depends on the achieved
mean power and asynchronous operation modes can be envisaged. This simplifies substantially the
technological challenge of the feedback system for light synchronization.
2.1 Diode-pumped solid state lasers
The possibility of building ICF lasers with high repetition rates and efficiency us-
ing solid state materials mainly relies on the substitution of flashlamps with low-cost
laser diode arrays and the development of crystals for greater energy storage and ther-
mal conductivity than Nd:glass . Yb:crystals cooled by near sonic helium jets are
presently favorite candidates. The main advantage of this approach is that it retains
most of the features of Nd:glass systems, first of all the possibility (still to be demon-
strated) of < 1% smooth irradiation on-target for direct drive in a timescale of fractions
of ns. A DPSSL-based fusion reactor would be - like NIF - highly modular. One possi-
ble vision is based on 4 kJ DPSSL composed of 1 kJ beamlets operating at a repetition
rate of the order of 10 Hz and assembled to reach the overall MJ power per shot. To our
knowledge, the most advanced R&D project is the Mercury/Venus laser system .
In particular, the Mercury R&D is aimed at a 100 J, 10 Hz laser based on gas cooled
Yb:S-FAP crystals grown up to 20 cm diameter . The laser operates at 1047 nm
(1ω) with a 2 ns pulsewidth, a 5x diffraction limited beam quality and an efficiency
greater than 5%. Fusion drivers are better operated at higher frequencies to increase
the rocket efficiency and reduce laser-plasma instabilities . Hence, DPSSL are op-
erated at 3ω (349 nm) with a conversion efficiency greater than 80%. Gain bandwidth
is of the order of 5 nm for Yb:S-FAP, significantly lower than for Nd:glass (28 nm) so
that the time duration of a DPSSL Fourier-limited (chirp-free) TEM00output beam
would be bandwidth limited to ∼ 0.3 ps pulses.
Excimer power lasers have been developed both for laser fusion and defense uses. The
current main candidate for ICF is krypton-fluoride. Electron beam pumped KrF sys-
tems offer superior beam spatial uniformity, short wavelength and high laser efficiency
(∼ 10%). As for DPSSL, an excimer-based fusion reactor is highly modular and sin-
gle beamlines could provide up to 50 kJ of laser light . Again, the thermal yield
and the efficiencies requested for a viable commercial power plant  represent major
technological challenges. The laser operates at 248 nm but a certain degree of tun-
ability is offered by the fact that the same system design can be re-used for other gas
mixtures  (e.g. ArXe lasing at 1733 nm or XeF at 351 nm). In particular, XeF
has been the leading candidate for defense applications and large aperture lasers with
energy yield per pulse in the 5 kJ range has been built since the late 80’s . XeF
has also been considered for laser fusion but it is less effective than KrF due to its
lower efficiency and because it behaves spectrally inhomogeneous, precluding efficient
narrow-band operation .
3 OPCPA pulse compression
As a by-product of its peculiar design (see Sec.2), a multi-shot ICF driver offers a
large number of beamlines operating, probably, in the near-UV region with a rather
limited spectral bandwidth and an energy per pulse ranging from 1 to 50 kJ. We do not
expect a Ti:sapphire CPA system being able to use efficiently neither this pump source
nor its outstanding average power regime. On the other hand, Optical Parametric
Chirped Pulse Amplification offers, in principle, a higher degree of tunability and
could be successfully adapted to exploit a narrow band, energetic pump pulse 
and its average power . Pulse compression should be enough to trigger the RPD
acceleration mechanism and exploit the high repetition rate to increase the average ion
current2. Optical parametric amplification is a nonlinear process that involves a signal
wave, a pump and an idler wave . In a suitable nonlinear crystal, the high intensity
and high frequency pump beam (ωp) amplifies a lower frequency (ωs), lower intensity
signal. Energy conservation is fulfilled through the generation of a third beam (“idler”)
whose frequency is constrained by
ωp= ωs+ ωi. (1)
Parametric gain is achieved over the coherence length, defined as the length over which
the phase relationship among the three waves departs from the ideal condition (“phase
matching”). Phase matching corresponds to momentum conservation and can be ex-
?kp,?ks,?kibeing the wave vectors of pump, signal and idler, respectively. Clearly, en-
ergy and momentum conservation cannot be fulfilled simultaneously in a linear crystal
but birefringence offers a way out. In spite of the variety of nonlinear crystals devel-
oped so far for frequency multiplication, only a few can be grown to large size (tens
of cm) to handle the pump energy available and offer an adequate fluence limit for
high power applications. Here, we mainly concentrate on Potassium Dihydrogen Phos-
phate (KDP), a negative uniaxial crystal commonly used for frequency multiplication
of Nd:YAG lasers3. In this case, phase matching can be achieved for parallel beams
2We do not intend this technique as a route for Fast Ignition since the spectrum of the ions is too
energetic once the RPD regime is operational.
3In the rest of the paper we assume for KDP the following Sellmeier’s equations:
λ2− 0.012942625+ 13.00522
for the ordinary index and
λ2− 0.012281043+ 3.2279924
for the principal extraordinary index (λ is the wavelength in µm). The nonlinear coefficients are
d36≃ d14= 0.44 pm/V.
(“collinear geometry”) when the pump beam is at an angle θmwith respect to the KDP
nep(θm)ωp = nosωs+ noiωi. (5)
Note that in the present configuration the pump beam is polarized along the extraor-
dinary direction, while both the signal and the idler beam have ordinary polarization
(“Type I” phase matching). Recalling
ep(θm) = sin2(θm)n−2
nepand nopbeing the principal extraordinary and ordinary refractive indexes at pump
wavelength, we get:
It is worth mentioning that θmshows a less pronounced dependence on the wavelength
for Type I phase matching than for type II, i.e. the case when only the idler or the
signal has ordinary polarization . This is an additional advantage when broad
amplification bandwidth is sought for.
Fig.1 shows the FWHM amplification bandwidth for a KDP-based Type I amplifier
operated in collinear geometry. The bandwidth has been computed assuming a pump
wavelength of 349 nm (see Sec.2) and a pump intensity of 2 GW/cm2. The latter
is determined by the fluence F at which the crystal is operated and the pump pulse
duration τ. Following , we assumed here F=1.0 J/cm2for KDP and τ=0.5 ns4.
The gain bandwidth has been computed assuming no pump depletion so that the gain
G can be approximated as [12, 26]
G = 1 + (ΓL)2
where B ≡ [(ΓL)2− (∆kL/2)2]1/2; Γ represents the gain coefficient
Γ ≡ 4πdeff
2ǫ0nep(θm) nosnoic λsλi;(9)
the quantity L is the length of the crystal and ∆k ≡ kp−ks−kiis the phase mismatch
among signal, idler and pump. Note that in collinear geometry, this quantity is scalar
since the wave vectors lay along the same axis. deffis the effective nonlinear coefficient:
for Type I phase matching in KDP
deff= −d14sinθsin2φ (10)
4In fact, KDP can be operated at higher fluencies since its optical damage threshold is greater than
5 GW/cm2but no long-term reliability studies are available for these extreme values. Note also that
competing nonlinear processes like self-focusing or self-phase modulations have been neglected.
450 500550 600650700 750800
Figure 1: FWHM bandwidth expressed in wavenumbers (1/λ) versus signal wavelength
for λp= 349 nm. The KDP-based amplifier (G=1000 at central wavelength) is operated
in Type I collinear mode (see text for details).
where θ is the angle between the propagation vector and the optic axis and φ is the
azimuthal angle between the propagation vector and the xz crystalline plane5. Hence,
θ = θmand φ can be chosen to maximize deff(φ = π/4). In Fig.1, as well as in Ref. ,
L has been equalized in order to attain G = 1000. In particular, for λs= 700 nm, such
a gain is reached at L = 2 cm.
Fig.1 points toward the existence of a window for full exploitation of the original
pump power. More precisely, a faint, chirped, wideband seed signal could be amplified
by a chain of Type I amplifiers  and finally enter the power KDP-based amplifier
depleting the intense pump wave6. However, a significant improvement in bandwidth
can be achieved operating the system in non-collinear mode. In this case the pump and
signal wave vectors are no more parallel but form an angle α between them (Fig.2). The
angle is independent of the signal wavelength. Again, the idler frequency is fixed by
energy conservation but the emission angle Ω varies with λs. Therefore, the matching
∆k|| = kpcosα − ks− kicosΩ = 0
∆k⊥ = kpsinα − kisinΩ = 0
The additional degree of freedom coming from the introduction of α can be exploited
to improve the gain bandwidth. In particular, it helps achieving phase matching at
first order for small deviations from the central signal wavelength. It corresponds to
5For the axis notation see .
6A complete numerical analysis of the signal evolution in pump depletion mode is beyond the scope
of this paper. A full calculation for Nd:glass pumps has been carried out in .
Figure 2: Phase matching triangle for noncollinear OPCPA.
together with the energy conservation constraint (i.e. a finite increase ∆ω of the signal
frequency corresponds to a finite decrease −∆ω of the idler).
Eqs.13 and 14 are equivalent to:
dωicosΩ − kisinΩdΩ
dωisinΩ + kicosΩdΩ
and are simultaneously fulfilled if
= 0. (17)
The derivatives are related to the Sellmeier’s equations for KDP since
so that the signal/idler angle Ω can be explicitly computed. These derivatives cor-
respond to the group index for signal (ngs = cdks/dωs) and idler (ngs = cdki/dωi).
Hence, Eq.17 can be interpreted as the request for signal group velocity to equal the
projection of idler group velocity along the signal direction. Note that it is impossible
to fulfill (17) if the group velocity of the idler is smaller than that of the signal. For
the case under consideration (λp= 349 nm), this generalized matching condition can
be achieved in the signal region between 400 and 700 nm. Fig.3 shows the signal group
velocity as a function of λs= 2πc/ωs(continuous line). The dashed line corresponds
to the idler velocity at ωi= ωp−ωsversus the signal wavelength. Finally, it is possible
to compute the (signal wavelength independent) angle α between the pump and the
signal, which turns out to be
The FWHM bandwidth of the amplified signal versus λsfor a KDP-based Type I ampli-
fier operated in non-collinear geometry is shown in Fig. 4. As for Fig.1, the bandwidth
has been computed assuming a pump wavelength of 349 nm and a pump intensity of
2 GW/cm2. Again, the crystal length L for G=1000 is about 2 cm. Fig.5 shows the
variation of the gain as a function of the wavenumber difference with respect to the
central wavenumber (1/λs), which the angle α has been tuned for. The continuous
line refers to λs= 550 nm (maximum bandwidth). The dotted and dashed lines refer
to λs= 450 nm and λs= 650 nm, respectively. Figs.4 and 5 represent a key result:
in non-collinear geometry the broadband window for the exploitation of the DPSSL
or XeF drivers corresponds to a signal wavelength of about 550 nm. This region is
accessible by a modelocked Ti:sapphire oscillators (signal generator) after self-phase
modulation  and frequency doubling. It provides the seed signal (at the nJ level)
that is amplified by the chain of low power amplifiers7. The pump signal for the low
power amplifiers (<5 J) can be either derived by the main pump or, if necessary, by
a dedicated low energy pump at a more appropriate wavelength. Finally, the signal is
sent to the power amplifier operating in pump depletion mode. The studies performed
in 2002 by I. Ross and coauthors  indicate that extraction efficiencies of the order
of 40% can be obtained. In particular, a 4 kJ pump pulse would provide a broadband
amplified signal of about 1.6 kJ. The actual light on target depends on the quality of
the amplified signal and the compression optics and it is discussed in the next section.
4 Proton production and acceleration
The pulse duration that can be achieved after the amplification process is dominated, to
first order, by the bandwidth of the seed signal and the gain bandwidth of the OPCPA
even if additional effects connected to the beam quality entering the compressor and the
compressor itself should be taken into account. In particular, the spectral phase [26, 27]
generated in an optical parametric amplification when the seed signal is chirped plays
a role in setting up the recompression system. For the case under study, the phase Φ
of the amplified signal is given  by
Φ = atan
B sin(∆k/2L) coshB − (∆k/2L)cosB sinhB
B cos(∆k/2L) coshB + (∆k/2L)sin(∆k/2L) sinhB
7See e.g. Sec.3.1 of .
400450500 550600650 700750 800
No matching conditions
No matching conditions
Figure 3: (Continuous line) Signal group velocity as a function of λs= 2πc/ωs. (Dashed
line) Idler velocity at ωi= ωp− ωsversus signal wavelength.
400 450 500550 600650 700750
No matching conditions
Figure 4: FWHM bandwidth expressed in wavenumbers (1/λ) versus signal wavelength
for λp= 349 nm. The KDP-based amplifier (G=1000 at central wavelength) is operated
in Type I non-collinear mode (see text for details).
-3000-2000 -10000 10002000 3000
Figure 5: Gain versus the wavenumber difference with respect to the central wavenum-
ber (1/λs), which the angle α has been tuned for.
λs= 550 nm (maximum bandwidth). The dotted and dashed lines refer to λs= 450 nm
and λs= 650 nm, respectively. The KDP-based amplifier (G=1000 at central wave-
length) is operated in Type I non-collinear mode (see text for details) at λp= 349 nm.
The continuous line refers to
Figure 6: Phase of the amplified signal (light line) and gain (dark line) for the amplifier
parameters of Fig.5 and λs= 550 nm.
and it is shown in Fig.6 (light line) for the amplifier parameters of Fig.5 and λs =
550 nm. For sake of clarity, the region where gain is > 1 (dark line) is also shown.
For this class of spectral chirping , nearly ideal recompression can be achieved as
far as cubic phase terms can be compensated. Neglecting the throughput efficiency of
the compressor and the losses due to the spectral clipping on the gratings, the output
power P for a Gaussian profile is:
P ≃ 1.6 kJ∆νFWHM(Hz)
= 240 PW (21)
if the bandwidth is dominated by the OPCPA gain bandwidth at λ=550 nm (2200
cm−1). The actual maximum intensity on target depends on the quality of the optics
and the available compressor gratings. Note, however, that operating near diffraction
limit is not requested in the present case. In has been shown [7, 8] that RPD accel-
eration mechanism is fully operative for I = 1 × 1023W/cm2, although the transition
region between the low intensity regimes and the RPD one is presently unexplored both
from the experimental and from the numerical point of view. In the RPD case, the
energy of the accelerated ions depends on the intensity and pulse duration according
to Eq. (17) in Ref. 
2w + 1
while the number of accelerated ions depends solely on the illuminated area S :
Np = n0Sl0. Here w is proportional to the laser pulse energy, Etot. It is given by
w = Etot/n0l0mpcS. The efficiency of the laser energy transformation into the fast
proton energy is
2w + 1. (23)
In the ultrarelativistic limit, w ≫ 1,the efficiency tends to unity, and it is small in
the nonrelativistic case when w ≪ 1 .
The studies performed in  made use of 27 fs (FWHM) gaussian pulses of I =
1.4 × 1023W/cm2and, at these intensities, protons are accelerated following an t1/3
asymptotic law up to kinetic energies of ∼ 30 GeV. These numerical studies are ex-
tremely challenging even for large parallel computer facilities. In order to reduce com-
plexity, the study has been carried out with laser pulse of relatively small focal spot.
In addition, the dynamical evolution has been followed up until the t1/3asymptotic
behaviour is reached (i.e. before the complete laser-plasma decoupling). The overall
laser energy to proton kinetic energy conversion efficiency at that time is 40%. Ex-
trapolation up to the time of decoupling indicates that an energy conversion efficiency
of 57% can be reached and the maximum kinetic energy for the above parameters ex-
ceeds 30 GeV. In the case under consideration, the illuminated area corresponding to
an intensity on target of I = 1.4 × 1023W/cm2is S = 1.5 × 10−10m2, i.e. a circular
spot of 7 µm radius. Since the kinetic energy reached by the protons is proportional
to the product of intensity and duration, a single shot (7 fs FWHM) would accelerate
particles up to about 8 GeV. The number of accelerated protons Npcorresponds to the
energy on target Etotcorrected for the laser-to-ion energy transfer efficiency (η) divided
by the proton kinetic energy (Ep,kin).
≃ 7 × 1011protons/pulse (24)
There is, however, an important caveat to be stressed. The RPD mechanism is fully
operative if the illuminated area is sufficiently large so that border effects can be
neglected . Due to limited computational resource, a systematic study of the size
and scaling of the border effect is not available at present. Results from  indicates
that, at the intensities mentioned above, the border corona has a depth of the order of
5 µm, This implies that either the beamlet arrangement should be able to illuminate
uniformly a relatively large area, as it is done e.g. for the fuel target when the driver
works in ICF mode, or that a significant higher energy should be used for the pump
in a proper chain of KDP amplifiers. This energy is not used to increase the pump or
the signal intensity but only the surface S, therefore it does not challenge the damage
threshold of the amplifier/compressor components.
From Eq.24, it follows that a single ICF beamline operating at 10 Hz would be
equivalent to a 10 kW proton driver. Clearly, a full ICF facility (2 MJ pump energy
at a repetition rate of 10 Hz) would allow the construction of an extremely ambitious
proton driver in the multi-MW intensity range (see Fig.7). Differently from traditional
proton drivers, a laser driven device is highly modular, while for RF-based accelera-
tors particles have to be stacked in a single lattice; in the RPD regime the stability
conditions during acceleration are much less stringent than for traditional drivers since
acceleration occurs nearly instantaneously and not in long periodical structures. Any-
way, the possibility to develop laser fusion in strict connection with particle acceleration
represents per se a fascinating research line.
If experimentally confirmed, the radiation-pressure dominated (RPD) acceleration mech-
anism  will offer a stable and efficient operation regime for laser-driven proton boost-
ers. In particular, it has been emphasized  the opportunity to develop these facil-
ities in conjunction with projects for inertial confinement fusion (ICF) and neutron
spallation sources. In this paper, we have shown for the first time that the pulse com-
pression requests to make operative the RPD regime do not fall in contradiction with
the power requests of an ICF driver and we discussed explicitly one solution based on
optical parametric chirped pulse amplification (OPCPA). Compatibility regions have
been identified for OPCPA amplifiers based on Potassium Dihydrogen Phosphate op-
erated in non collinear mode. In this configuration, bandwidths exceeding 2000 cm−1
(FWHM) have been obtained.
The authors would like to thank M. Borghesi and P. Migliozzi for insightful discussions
Figure 7: Average proton current versus proton kinetic energy for various existing
facilities. The two green lines correspond to a driver obtained by a single 4 kJ beamlet
and by the full ICF facility (2 MJ at 10 Hz rap.rate).
 V. I. Veksler, in Proceedings of CERN Symposium on High Energy Acceler-
ators and Pion Physics, Geneva, Vol.1, p.80, 1956; G.I. Budker, Ibid, p.68;
Ya.B. Fainberg, Ibid. p.84.
 K. Shimoda, Appl. Optics 1 (1962) 33.
 T. Tajima, J.M. Dawson, Phys. Rev. Lett. 43 (1979) 267.
 D. Strickland, G. Mourou, Opt. Commun. 56 (1985) 219.
 A.V. Gurevich, L.V. Pariiskaya, L.P. Pitaevskii, Sov. Phys. JETP 22 (1965)
449 [Zh. Eksp. Teor. Fiz. 49 (1965) 647]; A.V. Gurevich, L.V. Pariiskaya,
L.P. Pitaevskii, Sov. Phys. JETP 36 (1973) 274 [Zh. Eksp. Teor. Fiz. 63 (1972)
516]; S.J Gitomer et al., Phys. Fluids 29 (1986) 2679.
 For a recent review and references to the original papers see e.g. A. Maksimchuk
et al., Plasma Phys. Rep. 30 (2004) 473.
 T.Zh. Esirkepov, M. Borghesi, S.V. Bulanov, G. Mourou, T. Tajima, Phys. Rev.
Lett. 92 (2004) 175003.
 S.V. Bulanov, T.Zh. Esirkepov, P. Migliozzi, F. Pegoraro, T. Tajima and F. Ter-
ranova, Nucl. Instrum. Meth. A 540 (2005) 25.
 G. Mourou, T. Tajima, S.V. Bulanov, Rev. Mod. Phys. in press.
 J.A. Paisner, J.D. Boyes, S.A. Kumpan, W.H. Lowdermilk, M. Sorem, Laser
Focus World 30 (1994) 75.
 A. Dubietis, G. Jonusauskas, A. Piskarskas, Opt. Commun. 88 (1992) 437.
 I. Ross et al., Opt. Commun. 144 (1997) 125.
 R. Bieri et al., Fusion Technol. 21, (1992) 1583.
 J.L. Emmett, W.F Krupke, Sov. J. Quant. Electron.13, (1983) 1; W.F Krupke,
Fusion Technol. 15 (1989) 377.
 C. Bibeau et al., “Diode Pumped Solid State Laser Driver for Inertial Fusion
Energy”, Talk at 16thANS Topical Meeting on the Technology for Fusion Energy,
Madison, WI, 14-16 September 2004.
 C. Bibeau et al., Proceedings of the International Forum on Advanced High
Power Lasers and Applications, UCRL-JC-133970, 1999.
 J.D. Sethian et al., Proc. IEEE, 92 (2004) 1403.