You are on page 1of 102

HIGH INTENSITY LASER PLASMA INTERACTION

J.L. BOBIN Universit Pierre et Marie Curie (Paris 6), Laboratoire de Physique et Opaque Corpusculaires, Tour 12 E.5 4, Place Jussieu, 75230 Paris Cedex 05, France

NORTH-HOLLAND-AMSTERDAM

PHYSICS REPORTS (Review Section of Physics 1_etters) 122, No. 4 (1985) 173274. North-Holland, Amsterdam

HIGH INTENSITY LASER PLASMA INTERACTION


J.L. BOBIN
Unirersit Pierre ci Marie (uric (Paris 6), Laboratoire de Physique ci Optique (orpuscu/alres, Tour 12 ES. 4. Place Jussieu, 75230 Paris Cedex ((5~Frnace

Received November 984

Contents:

I. Introduction 2. Light absorption and plasma motion 1. The Linear Regime 3. Equations of the linear regime 4. Dimensional analysis for collisional absorption 5. Dimensional analysis for plasma flow dynamics 6. Collisional absorption as a function of the incident intensity 7. Structure of the conduction zone: classical transport 8. Long pulses, short pulses 9. Laser driven detonation I)). Limits of the linear regime II. Effects of the Ponderomotive Force II. 12 13 14. 15. 16. 17. 18. 19. 2)). Ponderomotive force and pressure Resonant propagation and absorption Electron bunchingCavities Steady structures, profile steepening Isothermal regimes Anomalous transport Possible physical origins of flux limitation Flux limitation form transport theory Filamentation and self focusing Self generated magnetic fields

175 75 77 177 179 18)) 182 183 188 189 191 93 193 195 199 201 206 21)7 210 213 2)7 22)

III. Wave Couplings

226

21. 22. 23. 24. 25. 26.

Purely growing modes and 3-wave instabilities Instability dynamics Influence of the pump tuning and intensity Pump depletion Coherence and incoherence Scattering instabilities in homogeneous and inhomogeneous plasmas 27. Filamentation revisited 28. Laser interaction with an inhomogeneous plasma flow 29. Wave breaking IV. Harmonic Generation 30. v X B non-linearities 31. Oscillating mirrors 32. 33. 34. 35. Raman upeonversion. Couplings and cascades Growth rates and thresholds Spectrum of harmonic lines Intensities of harmonic lines

226 229 235 237 241) 242 247 251 263 255 255 257 259 263
264

268

36 Conclusion References

27)) 271

Single orders for this issue PHYSICS REPORTS (Review Section of Physics Letters) 122. No. 4 (1985) 173-274. Copies of this issue may he obtained at the price given below. All orders should he sent directly to the Publisher. Orders must he accompanied by check. Single issue price Dfl. 66.00, postage included.

0 370-l573/85/$35.70

Elsevier Science Publishers By. (North-Holland Physics Publishing Division)

J.L. Bobin, High intensity laser plasma interaction

175

Abstract: Basic mechanisms, linear and non-linear, are reviewed. Light absorption may take place linearly through inverse bremsstrahlung. Together with usual heat transport by electrons it leads to a linear gas dynamical regime whose main formulas are given. When absorption occurs non-linearly, by resonance in a density gradient or through wavewave couplings, several non-linear regimes may show up. They are investigated in connection with the ponderomotive force, soliton formation, wavebreaking, , , , The processes responsible for harmonic generation are given special developments.

1. Introduction When a high intensity laser beam is focused onto a solid surface, matter is strongly heated. The temperature is so large that a dense plasma is formed and set into motion. As pointed out long ago [13], the process is of interest for controlled thermonuclear fusion. Subsequent developments led to the concept of laser driven implosions [4]. Obviously the physics of laserplasma interaction are of paramount importance: the feasability and the efficiency of the proposed fusion devices depend on the way energy is transferred from the light to the plasma. One is then induced to ask the following questions: How does the plasma absorb radiation? What is its thermodynamical state? Which forces act on it? What is its motion? Answers to these questions are not independent from one another. Many unknowns still remain after some twenty years of extensive research. This is due to the complexity of the basic mechanisms which furthermore are strongly intermingled. However, a separation naturally occurs between linear and non-linear regimes. The main features of the former were determined some 12 years ago [58]. The latter are still under investigation. They belong to the more general field of plasma non-linearities which is rapidly growing and changing. In the case of laser interaction three main parts may be distinguished: effects of the ponderomotive force, wavewave couplings and harmonic generation. They can all be related to specific non-linear terms in the coupled equations describing the fields and the plasma. The present review is organized accordingly. The viewpoint is rather theoretical: fields obey Maxwells equation whilst the plasma is described fluid dynamically. The equations are presented first. Then, four chapters deal with the linear regime and the three classes of non-linear phenomena introduced above. Most of the paper deals with a qualitative picture with computational and experimental checks. Recently, due both to theoretical considerations and to improvements in frequency conversion, a special emphasis was put on the wavelength dependence of laser interaction in the visible and near U.V. parts of the electromagnetic spectrum. The Nd-glass versus CO2 radiation is an old debate, not completely settled. and It is accordingly generally admitted lower frequencies favour non-linearities (the well2 scaling) lower thethat absorption. The gas dynamical behaviour becomes less known Lk controlable, a major drawback for those interested in inertially confined thermonuclear devices. However the details of the wavelength dependence are still poorly known, except in the linear regime.

2. Light absorption and plasma motion


A laser created plasma is electrically neutral but for local fluctuations. An intense oscillating electromagnetic field acts primarily and directly on the far less massive electrons. Action onto the ions mainly takes place through locally induced spacecharge fields.

176

J.L. Bobin, High intensity laser plasma interaction

Let EL and BL be the electric and magnetic fields, respectively, of the laser wave. Let E5 be the space charge electrostatic field. The electrons have a velocity v, a particle density n, exert a pressure Pc and undergo collisions with an effective frequency ii. The dynamics of the electrons are described by an equation of motion:
(9 v VPe m[-_+(v.V)] v = _e(Es+EL+XBL)-- miv

(2.1)

coupled to a wave propagation equation deduced from Maxwells equations


~2 I?
UL..L

22

at
E

c V EL = 4ire

at nv.

5 is related to charge density fluctuations through the Poisson equation V~E~41T(~n)e. Three sources of non-linearities are clearly visible: (v V) v, v

(2.3)

x BL,

nv.

(2.4)

The first one is at the origin of a non-linear force [9, 10]. The second results in the generation of radiation whose frequencies are the harmonics of the incoming one, see e.g. [11]. The third term represents wavewave couplings [12, 13]. The non-linear terms may be neglected through the usual linearization procedure. Assume that v and n, both have two components: a slowly varying one, u, n0 plus a high frequency perturbation , . Furthermore, terms of order v/c are neglected. The high frequency part of (2.1) reduces to

ai~/at+vii

eE/m.

(2.5)

When coupled to (2.2) linearized i.e. 2EL/(9t2 + C2V2EL = 4iren (9 0 (9ii/(9t,


(2.6)
Wp

it allows the straightforward calculation of an energy absorption coefficient Kr,,: for w greater than 2(1 K~ = cw and for w close to w, K~, = (2w~v)t2/c where v~, is the plasma frequency
~2/~2)112

(2.7)

(2.8)

J.L. Bobin, High intensity laser plasma interaction

177

(2.9) 4zrn0e2/m. The fluid dynamical derivation of K~ is actually incomplete. In order to deal with the whole of the relevant physics, it is necessary to start from a kinetic equation. This was done in the case of the VlasovMaxwell system of equations [14, 15]. The absorption is calculated in presence of discrete fixed ions. The result is a high frequency conductivity which includes a factor:
=

exp{ik (r,

ri)})

(2.10)

i.e. the spatial correlations of the ions. Here r, denotes the location of ion i. The absorption is different for randomly distributed ions or for ions whose spatial distribution is consistent with correlations at equilibrium. This factor is usually overlooked when computing laser driven implosions. However when the target is made of high Z material the ion plasma parameter (T being the temperature in energy units) 2e2(n 113/T (2.11) 1=Z 1) can be of order 1, even at electron temperatures of several hundreds keV, indicating strong ion correlations which could contribute to an enhanced absorption. Since in the following, absorption will be considered as an energetic boundary condition for the flow, the fluid approach is to be taken up. Now, the low frequency part of (2.1) is F (9u/3t+(uV)u= eEs/m Vp~/nom~0 (2.12)

since the slowly varying velocity u, i.e. the flow velocity, is much smaller than the electron thermal velocities, it can be neglected together with its derivatives. Then both sides of (2.12) are approximately zero, thus yielding a Boltzmannian distribution for the electrons. On the other hand, the equation of motion for the ions with mass M and charge Z is

0u/(9t + (u V) u = ZeEs/M ZVp


.

1/noM.

(2.13)

Eliminating E~ between (2.12) and (2.13), one gets an equation of motion for the neutral plasma as a whole 3u/(9t + (u V) u = ~ZV(pe
+

p1)/noM = Vp/p

(2.14)

where p is the fluid mass density and p the total pressure, ions plus electrons.

I. The Linear Regime 3. Equations for the linear regime When the oscillatory velocity is much smaller than the thermal velocity of the electrons, the electronion collision frequency i- is a function of the electron temperature T~ (in energy units). It reads after Scheuer [16]
4fle 2

3 m~(21TTe)3~ In A

ZC

(3.1)

I 75

J.L. Bohin. High intensity laser plasma interaction

where the Coulomb logarithm deals with the ratio .1


=

3( T)312 2Ze3(~n~)2

(3.2)

of the Dehye length to the impact parameter for a 900 collision. Substituting (3.1) into (2.5) yields a field amplitude independent absorption coefficienl K,,,. The linear collisional absorption of light is readily included in the energy equation in the form ~ (p ~+ pU) u(p pU + p) + I + F) 0
(3.3)

~.

~+

in which U is a specific internal energy, F a heat flow, and I the absorbed intensity such that denoting by I its value outside the plasma and by ~ the radiative intensity:
I=I~4

with

V.~=K,,,.

(3.4)

The set of equations (2.14). (3.3), (3.4) should he supplemented by the continuity equation
Ilp/9t + V~(pu) =
~.

(3.5)

One thus gets a complete description of the fluid dynamical behaviour of a single component fluid (plasma), irradiated by a laser beam and linearly absorbing. Such a system of equations cannot be analytically solved in general and requires the use of numerical methods. Computer codes were devised and widely used mainly for calculating laser driven implosions: Medusa [17], Lasnex [18]. However, analytical solutions can be found in some special instances: self similar flows or even simpler, steady one-dimensional situations. Indeed, in the latter case, one gets the following first integrals in a reference frame moving with the structure under investigation Pu
=

J.
=

JV,

(V= lip)

(3.6)
(3.7)

J2V+p

J2V+p y
yl

/J2V2

J2V~ y
y~l

(3.8)

Here an ideal fluid is assumed. J is a constant mass flux. The state labelled 0 is a uniform subsonic radiation-free reference state. Its density is of course greater than the critical value p. for which w = m Al 4n~eZ (39)

Pci~

Equations (3.6)(3.8) can be solved for any interesting quantity. However if one is not interested in the

J.L. Bobin, High intensity laser plasma interaction

179

details of the structure, much of the relevant physics can be unraveled from dimensional considerations, a common feature in fluid dynamics.

4. Dimensional analysis for collisional absorption Assume a laser beam is impinging onto a very gentle plasma density gradient: the characteristic length L is much longer than any absorption length 1. The underdense absorbing layer has an almost constant density p <Pc. The geometry is plane one-dimensional (fig. 4.1). The absorption coefficient is written in compact form K~= ap2/T312 (4.1)

in which induced emission has been ignored. An implicit assumption is thus /1w ~ T, a convenient approximation for laser interaction in which one typically has /1w 1 eV, T 100 eV. The constant in (4.1) has the dimension
~ -~

[a] = [L]8[t]3 [A4]t12


.

(4.2)

On the other hand one also has [I] = [M][t]31 [K~]=[L} (43)

J
(4.4)

The second equation (4.3) induces one to consider a self regulating situation, i.e. Kj
=

const.

Accordingly, using (4.1) one gets the proportionality T (p2al)23 (4.5)

which shows that for a given 1, T is an increasing function of p. Furthermore, combining dimensional equations for a, I, p and p, yields for a constant intensity the following relationships: p p a118115116t7116 ~
a3/8113h16t2hh16J

46
.

Fig. 4.1. Plane 1-D geometry for collisional absorption.

180

fL. Bobin, High intensity laser plasma interaction

and independently of a and t p I213p3. (4.7)

The second equation (4.6) indicates that p increases with time, i.e. the absorption zone in the inhomogeneous profile, progressively moves towards higher densities, as time elapses. Moreover, the length 1 of the zone also increases according to 1 a11~I~6t9~6
.

(4.8)

The same behaviour occurs for the temperature since from (4.5) one also has T
ah/4 19/8tts/s .

(49)

Thus, the existence of the self regulating regime is limited in time. Since monochromatic radiation cannot propagate beyond Pc, as soon as the absorbing zone reaches the vicinity of Pc, another regime necessarily takes place. Now (4.7) tells us that for given I and Pc, p has a fixed value. Due to the equation of state (of the ideal gas type e.g.) Z+ 1

M p~T

(4.10)

where Z is the average charge of ions with an average mass M T no longer varies either. Combining then (4.7) and (4.10), one gets a relationship independent from a and 1, i.e. from any assumption about the absorption mechanisms. It suffices to state that the absorption takes place at the cut-off density, so that 213. (4.11) T (I/pc) Now, one may use (3.9) to link p and T to radiative quantities only: intensity and wavelength A 2i~c/w.One thus gets T (IA2)213 p (I/A)~3~I
~.

(4 12)

when collecting experimental results from the literature (e.g. [19])it turns out that for IA2 greater than 1012, T is proportional to (IA2)213. Fitting the formulas (4.12) to measurements, and putting T in keV, I in W/cm2, A in ~imand p in megabars, yields T= 109(1A2)213
p
= 1.4

(4.13)
(4.14)

x i0~(1+ 1/Z)(I/A)213.

5. Dimensional analysis for plasma flow dynamics No motion was considered in the preceding section. In order to introduce a flow velocity into the dimensional analysis, one starts from the dimensions of p and I:

J.L. Bobin, High intensity laser plasma interaction

181

[p][L] [t]_l

[I]

(5.1)

i.e., there exists a characteristic velocity linking the pressure obtained in the plasma to the absorbed intensity. Now when a plasma flow is driven by a laser beam impinging onto a solid surface (or less commonly onto a high density gas jet [20])part of the flow is unsteady: a raref action wave sweeps out a low density plasma. A stable transition to a stationary zone is possible provided the particle velocity is the local sound velocity: c~ = [(Z+ 1)T/M]H2 = (p/p)U2.

(5.2)

Therefore when absorption takes place near a sonic point, one gets from (5.1) and (5.2) the proportionality p312p~112cc I (5.3)

already obtained in (4.7) for a non-moving plasma. When the sonic point is at the critical density, eqs. (4.12) are recovered. Still no assumptions are made concerning absorption mechanisms. One may then interpret the empirical law T (1A2)213 (5.4)

as denoting absorption at critical density which is also a sonic point for the flow. This is valid for any absorption process linear or non-linear. The validity domain of (5.4) is fairly well known from experiments in which I and Te can be measured accurately enough. In the plane ln T, ln(1A2) of fig. 5.1 experimental points lie within a factor of 2 around a skeleton which includes a segment with slope 2/3. Everyone agrees with: (5.4) holds for the range
1012< IA2 <

(23) X i0~ Wp.2/cm2.

For intensities higher than 3 x 103/A2 which correspond to a threshold for the onset of parametric instabilities, two electron temperatures are found [21].Only the higher one seems to follow (5.4) up to Te(keV)

IJp/crnt

10~2 1O~
Fig. 5.1. Electron temperature versus IA2. Experimental points lie in the shaded area.

182

iL. Bobin, High intensity laser plasma interaction

1015/A2 although different power laws were put forward specially at Livermore [22]
Te

(IA2)~5.

(5.5)

The cold component exhibits a much slower variation which is poorly known. When (5.4) is not satisfied, this may be due to two reasons: either the collisional absorption does not take place at a well-defined density, an unsteady situation for the whole duration of the laser pulse, or the sonic point is not where absorption is. Both effects are likely to occur simultaneously for the lower intensities (1A2< 1012 Wp.2/cm2). Then the plasma flow is driven by a volume absorption in the rarefaction at densities well below critical. For the highest fluxes (IA2> 1014 Wti.2/cm2) the situation is not that clear. Indeed a competition sets up between instabilities due to the non-linear coupling of electromagnetic waves with plasma modes, and ponderomotive effects which tend to steepen the density profile.

6. Collisional absorption as a function of the incident intensity In the previous sections, scaling laws were set up. They all involve the absorbed intensity I. This is only part of the incoming intensity I~. The connection between the plasma parameters such as the temperature and I~ required the quantitative knowledge of the absorption processes and the subsequent flow dynamics. In the linear regime, one may assume that the absorbing underdense zone is an exponential rarefaction
p=p~e~

(6.1)

where L is a time dependent characteristic length L=c.,t.


(6.2)

The plasma flow is always slower than electromagnetic wave propagation. Accordingly, one may solve the transfer equation (3.4) at any instant. Now the incoming light propagates in the underdense plasma up to Pc. At that point it undergoes a mirror reflection and then propagates backwards. Along most of the profile the absorption coefficient is given by eq. (2.5). The exception is the vicinity of the critical density. However it turns out that one does not make an important error by using (2.7) for the whole density profile. Then the ordinary differential equation d4 dx
=

AA2p~ e_sx/L

(6.3)

T312(l

e~~)2

is separable. Its solution with the boundary condition: ~ is I~ at infinite x. is In ~ = 8AA2p


0L 312

(6.4)

~ 3 T Now, one may rewrite (4.13) as

J.L. Bobin, High intensity laser plasma interaction

183

T312/A2= BI. Furthermore Pc is proportional to A2. One thus gets ~ Jo IA4

(6.5)

(6.6)

In (6.4), (6.5), (6.6) the quantities A, B and C are constants. The result (6.6) is better expressed in terms of an absorption efficiency
=

I/Ia

(6.7)

as CL/1 4 0A i.e., the absorption scales as I 4. Figure 6.1 displays 0A (A = 1 ~m) [23].
flA

ln(1

7)A)

(6.8)
i~A versus I

as it results from numerical computations

1013

10~1~ \,.J/cm2

Fig. 6.1. Absorption efficiency versus I for a CH plasma: (a) L = 50 ~sm, (b) L = 100 tim, (c) L = 500 tim.

Absorption increases strongly with L. At large L, the incoming intensity does not even reach the cut-off density: volume absorption. At high radiation fluxes and small L, ~A goes to zero but the linear model no longer holds. Indeed non-linear mechanisms take place and tend to modify the fractional absorption. Although simplistic, and not too realistic, the model used in this section allows one to draw some general conclusions: A significant scaling is 1 4. The transition between 0A a situation with a strong collisional absorption and the one with a small absorption takes place over about two decades of the significant parameter 1 4. 0A part. Actually, it turns Due to the presence of L in (6.8), plasma dynamics might play an important out that its main influence deals with the scaling. One finds that, in general, ~A scales as 1 0As where S (>4) is an exponent depending upon the plasma dynamical regime. For instance in a self similar situation, S is 5 as shown by Mora [24].

7. Structure of the conduction zone: classical transport We now are interested in the overdense region (p > Pc) in which laser radiation does not enter. Its structure is dominated by the thermal transport term F which appears in (3.2). The so-called classical or normal transport term was derived long ago by e.g. Spitzer Harm [25] and Braginskii [26]. This is

184

IL. Bobin, High intensity laser plasma interaction

the one usually given in plasma textbooks. It reads F=~T512VT (7.1)

where x is a constant. It refers to steady state situations with moderate gradients and energy fluxes. These conditions are not always fulfilled in laser plasma interaction, a point to be discussed later (see section 10). In a stationary regime, the energy equation in one-dimensional geometry is iu2 y \ 1 pu(_+_c~)+FJ=0. ar~ \2 yl For a fully ionized plasma y is 5/3 and one immediately gets a first integral u T A 2+5IKT312~1) 2 ar c. p[~D? where V~= u/c. is the local Mach number, while referring to (7.1) and (5.2) K=~M/(Z+1).
A 0 is the integration constant. It represents the energy exchange between the conduction zone and the external world: laser radiation flux converted into thermal flux at cut-off. Now Ao/c~is of the form JV0/ V (<<J). Thus it is negligible in most of the structure. When the flow is fairly subsonic (~j~2 ~g 5) the factor (~2 + 5) in (7.5) is approximately constant. In the plane case
pu=J,

a r

(7.2)

(7.3)

(7.4)

(7.5)

A0/c~=~0,

(7.6)

the equation is readily integrated. The solution with suitable boundary conditions is 215 T = Tc(1 x/xo) where Tc is the temperature at cut-off and x 0 is the thickness of the conduction zone

(7.7)

2 T~ 4K 25 T~ ft
Pc

(7.8)

separating the critical density from an ablation front where T0, aT/ax~x. (7.9)

IL. Bobin, High intensity laser plasma interaction

185

In the spherical case J = 4irr~pu and one finds T=T~( ~


T~ To

(7.10)

2/5

2/5

(i_~)
T

(7.11)

in which r~ and r0 ~ r~,are the radii of the critical layer and the ablation front respectively. One has 25J 2 16~rKT~ (7.12)

and r~ has a rather complicated expression we may omit. The well-known temperature profiles for the plane and the spherical case are displayed on fig. 7.1.

x~ x
a)

~
b)

Fig. 7.1. Temperature profiles: (a) plane case, (b) spherical case.

In order to complete the profile determination in plane geometry, one uses the first integral of the equation of motion (3.6) which is rewritten P/p+p=Ju~+p~. (7.13)

The pressure is then a linear function of the specific volume V = i/p. Since the point at cut-off is sonic, Ju~=p~, and accordingly
=

(7.14)

(~1R~ + 1)

if one has ~V?~ ~ 1,

(7.15)

where ~V?0 is the Mach number for the ablation front. It is generally negligible so that the ablation pressure is twice the pressure which prevails at cut-off. This pressure is much greater (usually megabars)

186

fL. Bobin, High intensity laser plasma interaction

T
~ Sonic
__~-=<~

[_~

>~
~c

~ Shock Initial

Absorption Raretaction

Conduction

Fig. 7.2. Temperature and density profiles in the linear regime.

than the pressure in an unperturbed solid. Therefore there should be an intermediate state between the ablation front and the solid material. Due to the pressure ratio, a shock front propagates ahead of the ablation front so that finally one has the well-known scheme of fig. 7.2. Not all the regions in the flow are stationary in the plane case. Steady structures are found for the absorption zone, the conduction zone and the shock front. Both fronts, ablation and shock, propagate with respect to the unperturbed material with different velocities. This is due to thermodynamical conditions (entropy should not decrease across a front) whose details can be found in Frasers article [271. Front velocities are easily calculated through straightforward algebra [71. One finds for the shock velocity

(~ )
Th
PsPo+Ps
=

~ c.

~ 5Ps

/2

(7.16)

assuming Po

4p~, the

usual shock conditions. By the same token, the ablation velocity is:

DA=

[Pc~

(2PcPo~Ps)~2]
Ps Pa

~ (R~~+ ~!~)
c
4Ps

2Ps

Pa
In both formulas, c

5.

(7.17)

5 is the sound speed at cut-off. The knowledge of these velocities allows one to calculate the ratio shock and the one ejected in the ablation:
=

~m

between the mass driven in the

(c, (c. DA)Pc D~p. DA)Pc


2 (Pc/4Ps) + (3pc/4ps)2 (8ps/3pc)

/8p ~

/2

(7.18)

if one has PsIPc~ 1. The ratio ~1m only depends on (ps/Pc). Now, the critical density Pc is proportional to the laser frequency squared, so that eventually ~j is proportional to A. A hydrodynamical efficiency ~ (ratio of internal energy fluxes) is obtained by multiplying ~1mby TI/TC. Since
ToPoPcPoPcPc Tc PcPo
4PcPs 2p 5

(719)

one finds in the reference frame moving with the ablation front

IL. Bobin, High intensity laser plasma interaction


1/2 0

187

T~

\~s/

(7.20)

For the 1.06 p~m radiation of the Nd glass laser, this efficiency is about 0.1 (Fauquignon and Floux [6]). For the spherical geometry, the profiles can be determined after the differential equation of motion, which in its simplest form, reads pu du
= dp. (7.21)

Using an ideal gas equation of state and the definition (7.4) of the Mach number, one may transform the above equation either for u: dlnu= 1 2dln~ 1-~P? or for ~P?: dln~V~+~dln T= 1 2dln~. 1-~P? From (7.22) it is readily deduced that for small ~V?, u has a local maximum at r = (3/2)1/2 ro and in this vicinity 2uccl/p. (7.24) Tx r From (7.23) one gets the behaviour of ~ which can be entirely calculated, at least numerically, after the knowledge of T, either in the overdense region (7.10) or assuming an isothermal plasma in the underdense zone. At large ~JJ~(T= T~) one gets
= 2 ln(r2/T~)cc (7.22)

(7.23)

u2.

(7.25)

Thus, the whole structure, including the underdense rarefaction has a stationary structure according to the scheme of fig. 7.3. In all the above calculations, ions and electrons have one and the same temperature T Actually, laser light heats the electrons which in turn transfer their energy to the ions through the usual relaxation term

r~J~t
Fig. 7.3. Steady flow structure in 1-D spherical geometry.

188

IL. Bobin, High intensity laser plasma interaction

dTe

TT,
Teq

dt

Teq

mM /T fleZe lnA m

~+)

T\312 M

(7.26)

The temperature decoupling can be accounted for [71 with the result that in steady state, the electron temperature is only slightly higher than the ion temperature. At cut-off, one finds
Te~1.5Ti.

(7.27)

8. Long pulses, short pulses Equation (7.8) gives the thickness of the conduction zone in plane one-dimensional geometry. Using the mass number A of the plasma ions it may be rewritten x 0= where x 0 and A are in micrometers and Tc in keV. The numerical coefficient was chosen in order to fit numerical simulations [28]. Obviously, time is needed to build up a zone with such a thickness. The order of magnitude for the minimum required time is 9T312A2 =
tc
=

1/2

2A2 300
(---)

(8.1)

TC

4.5 X 10231A4

(8.2)

1.3

X 10

where

is in seconds, T in keV, I in W cm2 and A in micrometers. The sound velocity is


=

~( A

1/2 )

10~ T~2 cm/s.

(8.3)

The build-up time for the conduction zone is thus independent of the target material. For a given absorbed intensity it increases as the fourth power of the wavelength. It is impossible then to reach a steady state albeit localized in the flow for a shorter pulse. Putting numbers in the formulas, for an absorbed intensity 1013 W/cm2, tc is 4.5 X 10_lOs for radiation with wavelength 1.06 ~m (Nd glass laser), 2.8 x 10- ~ s for radiation with wavelength 0.53 ~im(frequency doubling of Nd laser light), 4.5 x 10_6 s for radiation with wavelength 10 ~m (molecular CO 2 laser). A 100 picosecond pulse is long for radiation at 0.53 p.m, short for the two other wavelengths. Experiments confirm 4 of this statement as shown on fig. 8.1, durations on which are flA those is plotted function of scaling parameter IA is section 6. The wavelength and pulse usedas byaAmiranoff et the al. [29].The steady regime obtained either for 80 ps or for 2.5 ns at 0.53 p.m (same L according to fig. 6.1). On the contrary at 1.06 p.m, L is much smaller for iOops than for 2.5 ns, the steady state being reached in the latter case only. One may also notice that by comparison with the model dealt with in section 6, the maximum fractional absorption is smaller for shorter pulses. The difference comes from the choice of (2.7) for the absorption coefficient. Had we taken up (2.8) together with the same exponential density profile, the transfer equation

IL. Bobin, High intensity laser plasma interaction

189

1
--

=4T2ns

0
I

1 4 0X ~~4a/cm2

1011
Fig. 8.1. Scaling of experimental absorption efficiencies.

d4~/dx = ap~2e~~2~/ would have led to the solution


l =

(8.4)

1 exp( ~aLp~4)=

exp(bL/A312)

(8.5)

where a and b are constants. The correction exp(bL/A32) is larger for smaller ratios L/A3~2.Given A, it will be larger for short pulses since L has no time to reach its steady state value. Note that the way (8.5) was obtained is valid for short pulses only: steep gradient, absorption at cut-off, negligible in the underdense plasma. With low intensities associated with long pulses, the non-total absorption comes from partial reflections along the density gradient (Fresnels formulas, or B.K.W. analysis) none of the previous calculations accounted for. As far as absorption takes place mainly at cut-off, a sonic point, the scaling laws (4.12) are not affected by the steady or unsteady behaviour of the flow in critical and overdense regions. Since, on the other hand for short wavelength long pulse intensities smaller that the instability threshold, absorption is comparatively high and weakly depends upon the intensity, it is not surprising that an experimental scaling law as (1 2)213 be fairly well satisfied for the temperature. Astonishingly enough it also holds for 10 p.m radiation0A in short pulses (low absorption). This could be due to effects represented by (8.5) with the consequence that the curve lA(J0A4) is very flat around 50%, as shown on fig. 8.1 for 100 ps pulses of 1 p.m light.

9. Laser driven detonation So far, it was assumed that the density profile encompasses overdense regions. This corresponds to laser radiation in the visible part of the electromagnetic spectrum impinging on to a solid surface. In other instances such as gas breakdown, no overdense medium is present. This is the main difference with the cases dealt with in the previous sections. Since laser light absorption still creates a high pressure plasma, a shock wave is a necessary transition to the unperturbed material. Now, the restrictions of thermodynamical origin which prevent the shock and the absorbing layer to propagate at

l9))

iL. Bobin. High intensity laser plasma interaction

one and the same speed, no longer hold. Then a stationary structure with the same mass flux i through both fronts may build up. The situation is thus close to the one which prevails for chemical detonation. Assuming that light absorption produces a state in which the particle velocity equals the local sound velocity (ChapmanJouguet condition) one gets the optical analogue of the Zeldovich [30], Von Neuman [311,Dring [32] scheme shown on fig. 9.1. The corresponding temperature and density profiles are shown on fig. 9.2. The light may come from either side of the structure.

\Shock adiaba~(Hugoni&)

~
\
\

Isen~rope
\ //

9 V V0 Vcj V9 Absorphon
Fig. 9.2. Temperature and density profile in the optical detonation.

Fig. 9.1. Pressure/specific volume diagram for the optical detonation: initial conditions po V0, shock conditions p, V~, final state p~ j V~J. Absorption takes place between V0 and Vcj.

In laser induced gas breakdown (for a review see e.g. Dc Michelis [33], Grey Morgan [34]. Ostrovskaya and Zeidel Raizer [361),one usually observes such a detonation front propagating towards the focusing lens. The absorption is then almost total. When this is not the case a second front is also seen which propagates away from the lens. For instance, in an experiment by Gravel et al. [37] using a CO2 laser, the breakdown spark is driven by a first high power pulse of about 100 ns F.W.H.M. (Full Width Half Maximum). The spark is a diverging lens for the radiation in the subsequent microsecond low power tail of the pulse. This radiation is focused beyond the geometrical focus, interacts with the real boundary of the spark, and finally drives a fast detonation propagating away from the lens. The propagation velocity of the laser supported detonation (LSD) is readily calculated, provided the ChapmanJouguet condition holds, i.e. in the final state
[351,

U~j=

YPcj Vcj.

(9.1)

One then finds, Ramsden and Savic [38], 2l)I/pg]US (9.2) D = [2(y where p 5 is the density of the unperturbed gas. Similarly one gets the hot plasma temperature. Champetier [391,

IL. Bobin, High intensity laser plasma interaction

191

1---~~ T~~= 1+a 2 (y+1)2

(9.3)

where a is the degree of ionization. One finds again a j2/3 dependence of the temperature. But as far as the whole structure is underdense, there is no influence of the laser wavelength.

10. Limits of the linear regime When performing experiments with increasing intensities on targets, one finds a transition around 2 to 3 x 1013 Wp.2/cm2, then as shown in fig. 5.1, two electron temperatures show up. The lower is almost constant around 500eV [21], the higher still obeys the scaling law (4.11). Furthermore light absorption turns out to be higher than predicted by the linear theory [40,41]. The actual behaviour is displayed on fig. 10.1.
11A 1

____

_____

Fig. 10.1. Fractional absorption versus intensity: experiments (hatched area) and linear theory.

Obviously beyond (23) x 10~~ Wp.2/cm2, the linear theory, as presented in the previous sections, no longer holds. Now, in section 2, 3 non-linear terms were identified in the equation. They all involve the electron velocity. In a laser irradiated plasma, the electron velocity has two components: an oscillatory part

eE/wm

(10.1)

where E is the electric field of the wave, a thermal part whose distribution and mean values are known provided a temperature does exist

Ve = (V)

(T/m)~2.

(10.2)

In the linear regime, it is assumed that Ve is dominant and, accordingly, 5 is neglected. Non-linear processes are expected to be important wherever the ratio 5/Ve is no longer small. Hence an a priori requirement
52

e2E2

J~A~> 1

(10.3)

v~ w2mT

ircT

where the following relationships were used

192

fL. Bobin. High intensity laser plasma interaction

2/mc2 (classical radius of the electron) 11 = e In (10.3) a scaling law in 1 2 is again found. 0A The linear absorption coefficient is actually K~.,,= ae2/v3
=

cE2/4i~,

(10.4)

(10.5)

where v is the electron velocity. So far it was identified with the thermal velocity, hence the T312 in the denominator. For high irradiances one may rather substitute L~ instead of Ve. One then gets the proportionality K 0~,
j~3/2,

(10.6)

i.e., a strongly decreasing absorption. As shown on fig. 10.1, this is not the case proving that mechanisms, other than collisional, contribute to light absorption. Another feature of the linear regime comes from neglecting radiative effects: on the one hand, the pressure and the energy density associated with the impinging high intensity laser beam; on the other hand, radiative losses ~ due to thermal emission. Now these processes should be accounted for in the more accurate equation of motion and the energy balance equation. In order to discuss the order of magnitudes, it suffices to restrict oneself to steady situations. In plane. l-D geometry the energy equation without radiative terms can be rearranged as follows [71: J /y+l 1P1 I(~~)+F= 2~ Y Pa Y1Po L--)~
~

1 ~

/y+ I I

)po

l\ P

~1
I.

(10.7)

When including the energy density and the pressure due to the laser radiation, one gets in the right-hand side of (10.7) the supplementary term: J 2 1
fJ

1 Po

2EL
Pc

J1 0
Pc

a
C

(10.8)

since both EL and PL are of the order of 10/c. This term will be important if u 10/c is comparable with the left-hand side of (10.7). This is possible if i) either
+

F ~ I~

(10.9)

i.e., low absorption and transport in presence of a high incoming intensity. This situation is also obtained if, even with absorption, most of the energy is reradiated so that a small fraction only: I~ Ia), actually drives the flow. Thermal radiation emission increases with the average charge number (Z) of the plasma ions. Hence the experimental observation of enhanced non-linear effects when shining a given laser intensity on to a higher Z target [42], ii) or I+F-~-I() (10.10)

IL. Bobin, High intensity laser plasma interaction

193

in a relativistic flow. Since no such flow occurs in actual laser interaction experiments, this case will be henceforth discarded.

II. Effects of the Ponderomotive Force 11. Ponderomotive force and pressure They result from averages over a period of an oscillating field. The electron equation of motion should include non-linear terms. Only the friction term in (2.1) is to be neglected. One then obtains a complete equation for the low frequency components (u, n0, E, B) by averaging the high frequency component (5, n, E, B) viz. m ~= here
PefloT

m(u .V)u e(E+~xii)

_!~_

m((5-V)S) e (~x~)+-~(nV);

(11.1)

(11.2)

is the mean electron pressure. Obviously the three first terms in the right-hand side of (11.1) do not depend on a high frequency (laser) field. On the contrary the three last terms are non-vanishing averages of the non-linear parts of (2.1). They represent a low frequency non-linear force f~ which is readily calculated after linearized equations for the high frequency components, i.e.

9t
-

mn0

(11.3)

mc B= -VxS
e

Substitution into (11.1) yields

f~_m[((5.V)S)+

(SX

Vx 5)_~(V)].

(11.4)

For an electromagnetic wave in a uniform plasma the last term within the brackets vanishes. Furthermore in general n/no is small so that even if the oscillating kinetic energy and its gradient are not large with respect to the thermal energy and its gradient, the main contribution to f~ comes from the two first terms within the brackets. Finally for electrons one gets after some elementary vector calculus 2=-~ (11.5) F~=nof~=mnoV~2 2mw w 81r Pc 81T This force clearly derives from a potential which is the radiation pressure associated with the high

194

fL. Bobin, High intensity laser plasma interaction

frequency wave. We then write F0= dpL/dx in which, for an inhomogeneous plasma
2)
PL~J p (E Pc8~

(11.6)

I
Pc

pit f (E2)
dp, 81T (11.7)

Pu being the density of the overdense (radiation free) reference state. It behaves the same way for an electromagnetic (laser) wave or for a longitudinal plasma wave. In the following we are chiefly interested in the radiation pressure associated with a laser beam interacting with the driven flow. The ponderomotive force comes in the plasma dynamical equation as an external force. It acts on the electrons only. However through charge separation fields, it actually acts on the whole plasma, as one may easily find by the same token as used to set up equation (2.14). The new equation of motion is thus ~+(u~V)u=-~-V(p+pL). (11.8)
p

at

The energy equation is also modified. In a plane steady regime one gets the first integrals in V(= l/p)
V+p
+

and

p. = J2 V 0 + Pa
____

\2

yl

8ir

(11.9)

yl

Eliminating p between the two yields a quadratic equation for V. This can be cast in a non-dimensional form by setting (y = 2 Vop = p/po = V0/ V, ~P1~ = ~J 0 2) Vc (11.10)

PL

I+F

~E

W~

l5JpoV 5u8ir(15poV~)

w and 0 are not independent. They are linked by a relationship which depends on the interaction with the plasma. One need not enter into details here. In situations where absorption takes place (I + F)/J is negative so that 0 is locally positive for realistic interaction as shown in [431. The quadratic equation (11.11)

can be discussed with respect to the parameters to and 0 (>0). In the conditions corresponding to the impact of a laser beam onto a solid surface, the solution ~t(~JJ~) has the behaviour shown on fig. 11.1. We are interested in the ~.t > 1 branch only. The solution exists if and only if the mach number ~D?~ in

IL.

Bobin, High intensity laser plasma interaction

195

~ subsonic
I
0 V

sonic

~Y /
~

supersonic 1

Fig. 11.1. Specific volume variation across the front versus Mach number in the overdense reference state.

the dense fluid is smaller than some value (<1). In order to ensure a stable connection with an unsteady rarefaction, the sonic point should be the final state. The structure is then entirely subsonic as in the linear case. Equation (7.15) is now replaced by P
PO(~U~g+1)PL 1 2 =~[po(~P?o+l)pL]

(11.12)

for the sonic point. Clearly, at the same temperature, the material pressure in the final state is smaller than in absence of radiation pressure (section 7). The final sonic state is thus necessarily underdense. Since ~ is a function of tu and 0, the density in the final state turns out to be intensity dependent.

12. Resonant propagation and absorption The laser driven plasma is inhomogeneous. In section 7 a density profile was determined for the overdense region which is connected through a sonic point to an inhomogeneous absorption zone which is part of an isothermal rarefaction. The equations dealt with in the previous section are also expected to yield a non-uniform density profile consistent with the electromagnetic field energy density. In one-dimensional geometry the light waves propagate in the direction of the density gradient. In most experiments a laser beam is highly concentrated with a wide aperture optical device (a lens or mirror) so that interaction takes place near a focus and light may impinge on to the density gradient at oblique incidence. It has been shown long ago by Denisov [44]that oblique incidence has specific consequences, which have to be reviewed before determining the self consistent profiles resulting from the ponderomotive effects. Let us recall first the lowest order approximation for the plasma dielectric constant which depends upon the electron density according to
~

1w~/w2 1n/ne,

(12.1)

where absorption has been neglected together with unimportant higher order terms. In the plasma, Maxwells equations read

196

fL. Bobin, High intensity laser plasma interaction

cVXE=aB/at,

cVxB=~aE/at.

(12.2)

Now, consider an electromagnetic wave

E = E exp{i(wt k r)},

B = B exp{i(wt k r)}.

(12.3)

It obeys the propagation equations V2E+ ~~EV(V~E)=() w2 1 V2B+~~B+(V~~)xVxB=()


C

(12.4)

(12.5)

in which, for an inhomogeneous medium, ~ is space dependent. The simplest case is a linear profile, fig. 12.1: at x = 0, n = n~ and at x ~(x)=x/L.

L, n

0. Then
(12.6)

A plane electromagnetic wave comes from the vacuum side with an angle of incidence 0. Assume it is linearly polarized. When the electric field is perpendicular to the plane of incidence (the so-called S-polarization) one gets standing waves (fig. 12.2) with reflection at the mirror point such that
flflcC0520

(12.7)

The figure represents the energy density in the standing wave. The amplitude increases towards the inside of the plasma. It is mathematically described by an Airy function. The situation is rather different when E is in the plane of incidence. Indeed there exists a longitudinal component cL E~=Bstn0 x (12.8)

.....~

Fig. 12.1, Oblique incidence onto a linear density profile.

IL. Bobin, High intensity laser plasma interaction

197

(E2)

.--~~~

n~

(E~

-LOx
Fig. 12.2. (E2> profile for S-polarization. L is much longer than the wavelength AL of the light, Fig. 12.3. (E2) profile for P-polarization. The resonance height is limited by damping.

where the magnetic field of the wave whose direction is along Oz satisfies the differential equation
d2B dB

w2/x

(12.9)

dx

xdx

The solution of this equation is close to an Airy function with the property that B is non-zero for x = 0. Then (12.8) shows a divergent E~ for x = 0 (n = n~, = 0). Actually E~ does not go to infinity since the waves are damped (absorption). Hence the standing wave profile displayed on fig. 12.3. The high amplitude oscillation excited at the critical density is purely longitudinal. It results from tunneling across the electron density gradient. Obviously, this effect takes place if and only if the electric field has a component along Ox. In the case of a linear density gradient the solution can be calculated analytically. The well-known properties of Airy functions allow one to evaluate for every abcissa the value of B as a function of the incident wave electric field amplitude E 0 (outside of the plasma, i.e. to the left of fig. 12.3). One then gets approximately: 2 0) 0.92 Eo(wL/c)6, (12.10) B(n~ cos wL -1/2 2wL
~ ~

B(n~) 0.92 E

0 exp{_ ~ so that in the vicinity of X = 0

()

sin

3 o}

(12.11)

E 30 0L /wL\ t~ 2 wL x \cI -1/6 sin 0 exp~ I 3c sin J E ~ 0.92 ()

~.

(12.12)

The field E~ = 0.92E iwLV116


0 ()

2wL 30 t3c sin 0exp~ sin J


I

(12.13)

is called the pump field. Thus in the resonance E=E~/~=LE~/x. (12.14)

198

IL. Bobin, High intensity laser plasma interaction

E0 is better written as
E0= F0 1/2~(T) (2irwL/c) (12.15)

where ~ is the function 3) ~ Texp(~r of the variable r (wL/c)113 sin 0.


(12.17)

(12.16)

Equation (12.14) was written in the absence of damping. It exhibits an unphysical divergence at the origin. Now, denoting by i. an effective collision frequency which processes others than actual electronion contribute to,
(12.18)

w(w

lv)

with the consequence


E(nc)=~Ep,

(12.19)

a finite value. The absorbed intensity in this situation (the so-called resonant absorption) is
0

1=2 I -dx.

E2

J 8ir~~

(12.20)

In a linear profile
~

(12.21)
LL \

L/wi

L2

L1w2

so that using (12.15) together with a constant v E2


___

2 (12.22) Now, as shown on fig. 12.4, /(r) has a maximum close to 1 when r is in the vicinity of 0.7. Then the fractional absorption I/Ia, which is independent from v, may be as large as 50% provided the angle of incidence, which determines r, is suitably chosen. Measurements of the electron temperature [45] or of

IL.

Bobin, High intensity laser plasma interaction

199

c:~(~t)

Te IA

Experimenis P Polcirizahon

0.7T
Fig. 12.4. Plot of b(r) versus r. Fig. 12.5. Results of T~ or I/Jo measurements versus 0, for both polarizations.

the fractional absorption [46,47] clearly evidence the effect of the resonance. The general behaviour of the experimental results is sketched on fig. 12.5.

13. Electron bunching Cavities The ponderomotive force term (11.5) applies to both electromagnetic (laser) and longitudinal electron plasma waves. The force acts directly on the electrons. Thanks to the charge separations field which pulls back the electrons, it eventually acts on the plasma as a whole. One, of course, has to distinguish between high frequency phenomena in which ions may be considered as a motionless neutralizing background and low frequency processes in which both species move together. In the two cases the electron density profile has to be calculated consistently with the field, the ponderomotive force providing the coupling. Let us first look at the standing wave pattern found in the previous section. It results from an electromagnetic wave impinging onto a linear density gradient. Electrons feel the ponderomotive force which pushes them towards the nodes as shown on fig. 13.1. Obviously, the linear density profile is inconsistent with the standing wave. In order to investigate what is to happen, consider first the gas-dynamical equations (13.1) 3u
\3t

0U\

ax!

n ~E2 n~8irax

9n

3x EtN perturb~ n
~

(13.2)

Fig. 13.1. Electron bunching towards the nodes of a standing electromagnetic wave.

Fig. 13.2. Evolution of an electron density perturbation near cut-off. Cavity formation.

2(X)

J.L. Bobin, High intensity laser plasma interaction

in which ~udenotes either the electron or the ion mass. Moreover the averaging bracket () has been definitely dropped. The ponderomotive forces in the right-hand side of (13.2) involves a field which is given by the normal incidence standing wave equation
~2

a2E/ ax2 + w2(1

n/n~)E = 0.

(13.3)

(13.1) to (13.3) are a complete set of coupled equations. It can be linearized around a density profile N(x) and a zero initial velocity. One thus gets for the density perturbation
a2n

at2

iaE2aN 8lTflc \ ax ax
1

a2E~

a2N

ax2 /

ax2

(13.4)

Any perturbation is to grow up. To be more specific let N(x) be linear. Then a2N/ax2 is zero. At cut-off 8E2/ax is negative and 82E2/8x2 is zero so that a2n/at2 is negative. At points where aE2/ax is zero, i.e. the extrema of F2, a2n/ at2 have the sign of a2E2/ax2: negative for the maxima, positive for the minima. Hence cavities build up in the electron density profile specially in the vicinity of the critical density. Since the density at the closest mode may grow up to overdense values, the cavity may then trap radiation. In order to investigate transient behaviours, one has to use a time dependent equation for the field. This can be done by setting
E = E(t)

ei~0t

(13.5)

in which the amplitude E(t) is a slowly varying function of time, whose second derivative can be neglected. One then gets instead of (13.3) 2iw oE/at + a2 32E/ax2 +
w2(1

n/n~)E = 0.

(13.6)

Here a2 stands either for c2 or for 3v~, electron thermal velocity squared, so that transverse and longitudinal waves are simultaneously dealt with; eq. (13.6) is a Schrodinger equation. Since it is coupled to (13.1), (13.2) the whole system is non-linear. The behaviour of certain classes of solutions of (13.6) can be evidenced by playing again with a factorization of the (13.5) type, let then E = A(x,
t)

exp{iO(x,

t)};

(13.7)

substituting into (13.6), the equation splits into a real part


2~A +
~

(0~)2A + (i

A =0

(13.8)

and an imaginary part A a2 a2 (13.9)

0+O~A~+0~~A=0. w 2w

.J.L.

Bobin, High intensity laser plasma interaction

201

Introducing a hybrid velocity


2

u=~O~,
w

(13.10) 0, and linearizing again around N(x), one gets (13.11)

assuming A has a slow x dependence: ~ Du u 3u 1 aN =+u= Dt at ax 2N~ax

(13.12) Equation (13.11) is similar to the Lagrangian equation of motion of a particle in a gravitational field. The corresponding particle is decelerated when moving towards higher densities and accelerated in the opposite direction. According to (13.12), the amplitude A of the electric field perturbation with velocity u increases logarithmically all the way down the density gradient. Such results were obtained by Morales and Lee [48]using numerical simulation. The electron density profile is initially linear. The longitudinal field results from a capacitor model. When ions are not allowed to move, dips in electron density associated with soliton-like pulses of electric field, are generated at cut-off and propagate towards lower densities. Recent investigations [49]about this generation of solitons show for given values of the pump field a periodic birth of equal amplitude solutions at resonance. By decreasing the pump a double period shows up. A further decrease causes some chaos to set up: random amplitudes separated by random time intervals. The behaviour changes when using movable ions. A cavity begins to form at cut-off and is progressively transformed into a shelf. This effect was also evidenced by De Groot and Tull [50].It is illustrated on fig. 13.3.

Fig. 13.3. Unstable density formation: (a) initial density profile, (b) cavity at cut-off, (c) final state.

14. Steady structures, profile steepening In order to investigate which structures dominated by the ponderomotive force may take place, one has to start from the first integrals (11.9) in which the thermal flux F is neglected, an assumption to be justified a posteriori in section 16. The radiative pressure is given by (11.3). The electric field of the laser wave satisfies the time independent Schrdinger equation

202 C 2

fL. Bobin. High intensity laser plasma interaction

d2E

wdx

I+

fix!
~

1) _-_JE

Vci V

(14.1)

in which absorption is accounted for. Several steps are needed to determine qualitatively the density profile. First it is not unreasonable to assume that the field is to have a standing wave structure with zero field points (nodes) as in the linear profile dealt with in section 12. Then p is eliminated between the first integrals (11.9), yielding a quadratic equation in V (y =
4J2~5(J2V+pp)V+J2~2+5PV2~_E

~~=0. J
4ir

(14.2)

The knowledge of E(x), PL(X), 1(x) allows one to calculate V(x). The shape of the profile can be inferred by noting that absorption is important only in the vicinity of the critical density. Furthermore the points such that E(x) = 0 are singularities of the system. They are located in the PLV plane on the hyperbola whose equation is 2V~-5Pi:uV~= 0. (14.3) 0+p~~ pL)V+ J This is the locus of singularities which appear as cusp points of the curve pL(V). They accumulate when

4J2V2

5(J2 V

approaching the intersection of the curve (14.3) with the sonic line 8J2V+5(pLJ2V~po)=0. The starting point
PL =

(14.4)

0, V= V

0, being subsonic, so is the whole steady structure. The function pL(V) is sketched on fig. 14.1. There exists a limiting specific volume, an upper boundary for a subsonic profile. A steep gradient corresponds to the first arch of the field starting from state 0. Due to the existence of Vjim, there will be a minimum value for the density. Hence an underdense modulated shelf as shown on fig. 14.2. Density profile steepening due to ponderomotive pressure is a well-known effect which was widely observed in both numerical (Valeo and Kruer [51]; de Groot and Tull [501;Forslund et al. [52]) and laboratory (Azechi et al. [53]; Atwood et al. [54])experiments. Some computer simulations also exhibit

Vlpm~
.

Fig. 14.1. Ponderomotive pressure versus specific volume.

Fig. 14.2. Density and field profiles.

IL.

Bobin, High intensity

laser plasma interaction

203

a shelf. Since they have the periodicity of E2, the modulations have a wavelength equal to half that of the incoming radiation. The maximum length of the shelf is expected to match the coherence length. However high impinging intensities are usually associated with short pulses. One has then to check that the structure can be built up in a time smaller than the laser pulse duration. In order to do so, one needs further information specific to the problem at hand: e.g. mass and degree of ionization of the ions in the plasma. Whenever an underdense structure is evidenced through interferometric diagnostics (Fedosejevs et al. [55]; Raven and Willi [56]) it extends over a small number of wavelengths, thus indicating an ion sound velocity
Csi = (ZTeIM)~~ (14.5)

not greater than 10~ cm sec1 which might be due to a lagging ionization (small Z in the formula) and/or two-dimensional expansion. In the above quoted experiments, light absorption is in the range 2040%. The discussion in section 10 emphasizes the role of re-radiation to lower the useful flux, i.e. that part of the incident flux which actually drives the flow. Another cause for this effect is the creation of superthermal electrons with mean free paths much longer than the scale length of the structure. Part of them are a mere loss to be added to 4~ in the equations. Since superthermal electrons are the result of non-linear effects, re-radiation, by increasing the ratio E2/n~T, also enhances their production. Thus in the experiments the useful absorbed flux is indeed a small fraction (<10%) of the ingoing one. Coming again to the solution of the system of equations (11.7), (11.9), (14.1), the next step is a consistency check. The postulated standing wave pattern has to fulfill the wave equation (14.1). Assume p(E) is known. Then (14.1) may be rewritten as c2d2E dE dH(p,E) ~-~-~+F(v, w)-~--+ dE = 0
(14.6)

which is an equation of motion for a fictitious particle with mass c2/w2 in a potential well H(p, E). E plays the role of the coordinate, x the role of time and F is a damping term accounting for light absorption. Since the arches of E in the standing wave pattern have different amplitudes, the potential H which appears in (14.6) is multi-valued. It is of course negative if its value in the high-density radiation free state is chosen equal to zero. The first minimum of the multiple potential well occurs when the density is critical for a non-zero value of F. Others are obtained when E is zero in the underdense shelf. The particle whose motion is described by (14.6) falls into the successive potential wells along the dotted trajectory of fig. 14.3 which is drawn for a small damping, i.e. a strong influence of the ponderomotive

Fig. 14.3. Potential well for eq. (14.6). Case of weak damping.

204

fL. Bobin, High intensity laser plasma interaction

pressure. An opposite case of strong absorption is depicted in fig. 14.4. It corresponds to a single peak. This is an unrealistic situation for an electromagnetic field. However, eq. (14.6) also holds for longitudinal electron thermal velocity squared. It is then clear that fig. 14.4 may correspond to the resonant peak associated with an obliquely incident p-polarized wave. The corresponding density and longitudinal field are displayed in the figure. Finally, if only the first part of the trajectory, down to the first well, is taken into account, it can be identified with the leading edge of a subsequently uniform field: see fig. 14.5. One then recovers the case dealt with in the absence of non-linearities: absorption at cut-off (Fauquignon and Fbux [6]) with a skin-effect structure. So far, density and pressure profiles have been discussed for a single-component fluid. The corresponding temperature profile is then readily obtained. In the subsonic low-density shelf, the temperature is high and undergoes only small amplitude variations. It can be considered as uniform. Now, owing to the strong density gradient in the structure, ion heating through the relaxation with electrons is expected to be ineffective, in contrast to the case of heat transport dominated flows. There it was shown (see section 7) that Te/TI is, at most, of order 1.5. As confirmed by numerical simulations (Forslund et al. [49]), it is not unreasonable to assume that in the shelf of a ponderomotive pressure dominated flow, the electron temperature greatly exceeds the ion temperature, i.e.
~Te~T.

(14.7)

Consequently, the sonic point is such that the velocity is the ion sound velocity (14.5). Hence the subsequent rarefaction, if isothermal, has an exponential density profile, p = Pc exp(xlc1t).

(14.8)

E2

T
Fig. 14.4. Resonance absorption: (a) potential well, (b) density and 2) profiles. (E Fig. 14.5. Strong absorption in the overdense plasma: (a) potential well, (b) density and (F2) profiles.

IL. Bobin, High intensity laser plasma interaction

205

The comparison with the high energy part of experimentally observed ion spectra (see, for example, Campbell et al. [57]) indicates guiding electron temperatures of the order of several keV. However, actual spectra have a structure more complicated than the simple law (14.8). They have been dealt with in recent numerical (Shvarts et al. [58]) and analytical (Mora and Pellat [59])investigations. The case of the shelf proper is also of interest. When setting up an equation of motion, the electrical neutrality of the plasma was assumed to hold everywhere in the flow Zn~(x) = ne(x). (14.9)

The actual link between ions and electrons is the electrostatic potential . which obeys a Poissons equation
= 4[l(~e

Zn1).

(14.10)

Since both sides of (11.8) vanish, the electrons are Boltzmannian


fleflee~

(14.11)

with the reduced potential


2IfleTe (14.12) ecos/Te E and n ~is a reference density. Now, V~ appears explicitly in (2.12) which for a plane one-dimensional flow is readily integrated to give
i/i =

(PS =

J2 ZeM 1n~ (14.13)

Since ions only carry inertia, the density modulations in the shelf are modulations of n1. Hence a more deeply modulated ~ pattern. Then from (14.12) one deduces cli and the corresponding electron density profile which exhibits, in the nodes of E, the expected bunching due to the ponderomotive force: fig. 14.6.

X a

Fig. 14.6. Electron and ion density profiles (a) and potential wells (b) in a ponderomotive pressure dominated standing wave pattern.

206

IL. Bobin, High intensity laser plasma interaction

The situation thus involves BGK-type equilibria [60]with cold ions and electron trapping. There is some analogy with what happens in the structure of collisionless shock waves (Montgomery and Joyce [61]). However, in the present case, one has to deal with multiple potential wells which is more complicated.

15. Isothermal regimes Some other properties of laser driven steady flows are determined after a very simple model in which an isothermal plasma is assumed. In the one-dimensional plane case, the first integrals reduce to pu=J
p+PLF/p=po+J2/pt).

(15.1) (15.2)

This system can be solved for PL as a function of V. i.e.


PL=

J2VpoVo/V+po+J2V 0. (15.3)

On fig. 15.1 are shown several special cases selected for different values of the mass flux J. The maximum value of PL decreases from p~to zero as J increases from zero to Po V~ (sonic conditions in state 0). When the mass flux J is zero, PL is a monotonic function of V. It increases from zero (V = V0) and tends asymptotically towards p0(V~ x). A round trip starting from the point (V0, 0) along the curve, describes a cavity inside the plasma. The cavity is filled with resonant high frequency electromagnetic field. The tuning depends on the cavity size. Of course the cavity maybe generated by the mechanism described in section 13 (fig. 13.2). 2) is bivalued. So is partly V(pL). A cavity is compatible When flow is present (J of 0), only witha plasma the smaller amplitudes PL. V(E Then V(p~) remains single valued. Beyond a critical J dependent value of PL, the only allowed profile is the one consisting of a steep gradient followed by an underdense modulated shelf. Such profiles were calculated both analytically and numerically in [62]. The transition from the cavity to the shelf (fig. 13.3) was observed in numerical simulations. E p
2

Fig. 15.1. Field energy density and ponderomotive pressure versus the specific volume.

IL. Bobin, High intensity laser plasma interaction

207

In the isothermal regime, first integrals can be found also for the one-dimensional spherical geometry. They read ?pu = J
~2 p L~ U2 0 +c~In+-=.

(15.4) (15.5)

U2

Po

8TTp

2, r(p/po) is The spatiala dependence of a the density p branch is then are readily obtained. For a given of E separates a bivalued: subsonic and supersonic found. A discontinuity in value the flow radiation free region (F2 = 0) from a region with a given value of E2. It can be a subsonic to a supersonic transition: fig. 15.2, or a supersonic one: fig. 15.3. Both were investigated analytically and numerically. The subsonic density hump has been evidenced in simulations [62].The underdense part has a modulated profile. Indeed standing waves may set up. The detailed structure can be found in [63] for the subsonic to supersonic case and in [64]for the supersonic one. In both, there is no special reason for the point at critical density to be sonic.

subsoni F/P(O)
-~---

~ PiP(O) ~

-supersonic
Fig. 15.2. Subsonic to supersonic transition.

~ supersoni~~~~
r
Fig. 15.3. Supersonic to supersonic transition.

16. Anomalous transport In the linear regime, the overdense region is dominated by the thermal transport flux density (7.1). This term was neglected in the investigation of ponderomotive force dominated structures. These occur at high laser irradiance. They are associated with low absorption and still with high electron temperatures, for which the standard theory of heat transport in plasmas [25,26] would predict an increase of the thermal flux. Now there exist experimental evidences for an inhibited transport in laser interaction with solid surfaces. For instance, Pearlman and Anthes [65] measured the velocity of ions recorded on both sides of a laser irradiated thin foil as a function of the incident intensity: fig. 16.1. Beyond a threshold of 7 x 1013 W/cm2 most of the energy goes to the plasma on the laser side thus demonstrating transport inhibition. Furthermore, Lagrangian computer codes devised for laser driven implosions, do not restitute the experimental results for high intensity irradiance, whenever (7.1) is used. The agreement is much better if a significantly lower flux is introduced in the calculations. This is usually done phenomenologically by applying a flux reduction factor to the free streaming limit

208

IL. Bobin. High intensity laser plasma interaction

velocity

Ofl

Foil ~ R

baser

id

iO~~ \N.crrj~

Fig. 16.1. Ion velocities on both sides of a laser irradiated foil.

nT T FL=fFFS=f(-) 4m
FFS

1/2

(16.1)

is the maximum energy which can be transported in one direction by a density n of electrons with temperature T Using (16.1) guarantees that no energy greater than that available in the Lagrangian cell is carried away owing to large gradients. A hybrid transport term such that 1 xT52(aT/ax) (16.2) FL

is the simplest way of accounting for limited transport in numerical codes. It turns out that the best agreement between experiment and simulation is obtained when in (16.1), f has the magic value
fm0.03.

(16.3)

The limited free streaming flux FL may also be expressed in terms of gas dynamical variables, i.e. FL = hpc~ with a limiting factor
h

(16.4)

~ 4(Z+ 1)3/2

~m)
FL

(165 is also

To fm corresponds a magic value of h of order 1, so that


1/2

FL~nT( M

) =nTc

5.

(16.6)

Thus, electrons in the transport process carry their energy T with at most the ion acoustic velocity. Ion acoustic waves do not involve charge separation. The semi-empirical formula (16.6) may then tell us that

IL. Bobin, High intensity laser plasma interaction

209

no charge separation takes place in electron thermal transport. This could be a guiding idea for investigations of the flux limited transport problem. A further discussion of the point is to be given in the next section. For the moment, it suffices to observe that no definitely convincing theory is available yet. Here, it will be simply assumed that formulas such as (16.1), (16.4) hold. Then, following Cowie and McKee [66], consequences on a steady gas dynamical regime will be examined. First the energy equation (7.2) is rewritten
= 0. V~ [pu(~u2+ ~c~)+ hpc~c~] (16.7)

Hence the first integral ~Y?(~P?2+5)+h =0 (16.8)

in which, as in section 7, the integration constant is set equal to zero. It results from (16.8) that the Mach number is a constant across the whole structure. This was first pointed out by Cowie and McKee in an astrophysical context. In the plane one-dimensional case, the first integral of the equation of motion associated with (16.8) only yields the trivial solution: all quantities are constants. Or alternatively, there exists a discontinuity in the flow [67]. It separates a subsonic state from a sonic state. When the ponderomotive force is accounted for, its structure is the one investigated in section 14. For the spherical case, using p = pc~, (7.10) becomes
~Pl2pr2=JU,

(16.9)

yielding
p/pa =

(rIro)_2m2~n21).

(16.10)

By the same token, after (7.23)


22)1(m?l). (16.11) 0 = (r/ro)4/s1), ~ = (T/T0)2(m Such profiles have slopes much steeper than those found in section 7, assuming a so-called classical transport. The flux limited transport term is used in gas dynamical equations. It was assumed in section 14 that FL is much smaller than the absorbed light intensity I. Hence the condition

T/T

fi

nT T~2 (;;-) <~

(16.12)

It is readily inferred from the energy equation, that for a constant absorbed intensity, the Mach number in the structure is a constant. The dimensional analysis done in the first chapter still holds. When absorption takes place at a sonic point, one may still use (5.3). In the usual case V 0 4 V~, isothermal rarefaction,

210

fL. Bobin, High intensity laser plasma interaction

1=

a. 2(yl) p
3/2
~

(16.13)

Then, the inequality (16.13) is rewritten


p32 4
m p

(16.14)

Accordingly a condition for f is obtained


f<

12(m/M)2.
f

(16.15)

It does not depend on the laser wavelength. Obviously the magic flux limiting factor

m fulfills the condition. Thus, the assumption made in section 14 is justified. Physically, a dominant ponderomotive force and a strongly inhibited transport apply together in laser plasma interaction. The influence of the thermal transport is a small correction which does not reduce the steep gradients in the vicinity of the critical density.

17. Possible physical origins of flux limitation It is of course a rather unsatisfactory situation, that an ad hoc flux limiter be introduced without referring to any physical process. Accordingly, attempts have been made towards an understanding of the physics of heat transport in laser irradiated plasmas. The problem can be stated as follows: given a region in a density gradient, how does heat propagate towards higher density zones? Furthermore the driving electron distribution normally has two temperatures and the regime may be unsteady. A first question arises: which physical processes should be accounted for? In a thermal conductivity such as (7.1), only electron collisions with charged particles are considered. The usual theory is steady state and one dimensional. In the plasma kinetic theoretical calculations by Spitzer and Harm [25] or Braginski [26], it is assumed that the gradient scale length is larger than the mean free path. A local theory then applies. Now, Spitzer and Harm derivation of transport coefficients do include flux limitation. Indeed thermoelectric effects should be also dealt with, since the electrons which participate in the heat transport generate an electric current. One then has to solve coupled phenomenological equations for the current density j and the heat flux F, which read with positive coefficients juEaVT
F=f3EKVT

(17.1) (17.2)

in which, of course due to the laws of irreversible thermodynamics, the coefficients are not independent. The main point here is that for steady state, j should be zero, otherwise electrical neutrality would no longer prevail and the electric field would increase without limits. Setting j = 0 in (17.1) yields a local electric field. The heat flux is accordingly

IL. Bobin, High intensity laser plasma interaction

211

F=

K(l

af3IuK)VT

(17.3)

in which a reduction factor due to thermoelectric effects shows up. However, this is still too large to be consistent with experimental results on laser driven plasma dynamics. Then, one may argue that mechanisms not accounted for in collisional theory should be investigated and their effects on heat transport evaluated. First of all, as usual in plasma physics one might think about instabilities. The heat transport problem may be stated as: given any heat flow, it can be proved after the theory of moments (up to the 14th) that one can always find a distribution function which carries the specified flux [68]. Such a distribution includes a beam and is strongly unstable. The maximum allowed heat flow in the problem at hand is then the one for which the electron distribution is still stable. Since as shown in (16.6) there seems to be some connection between thermal flux reduction and ion sound, it is no surprise that the onset of ion acoustic turbulence was the instability most widely investigated. Long ago, Fried and Gould [69] showed that an electron drift is able to trigger an ion acoustic instability. They calculated a critical drift velocity as a function of the ratio Tell1. Whenever the drift velocity is greater than critical, the instability sets in and the energy goes to turbulence instead of being transported. The way one-dimensional electron transport results into a drift is illustrated on fig. 17.1. With respect to equilibrium, the electron distribution function has a larger tail in the direction of the energy flow. In order to ensure a zero net current, the bulk of the skewed distribution function is displaced towards the opposite direction. The homogeneous growth rates of the instability were calculated by kinetic theory both for collisional [70, 71] and non-collisional [72, 73] situations. They coincide in the limit Te ~ T1. Now, the non-collisional growth rate turns out to be density dependent [74]. In a ponderomotive force dominated laser driven plasma flow, the growth rate is higher in the vicinity of the critical density where steep gradients prevail and are likely to inhibit the instability. For the underdense shelf, the growth rate is much smaller, so that in this case also the instability does not set up easily. However there has been experimental evidence of ion acoustic turbulence in plasmas created with comparatively low laser intensities [75,76]. The ion acoustic turbulence also acts directly on the transport coefficient through an effective collision frequency. The point is far from being clarified yet and large discrepancies exist between theoretical evaluations (see [77]). So the role of electron heat flux driven ion acoustic turbulence, in reducing transport is not firmly evidenced. Another source of flux limitation is also found in a kind of electromagnetic turbulence involving Weibel microinstabilities. Indeed, since the temperature gradient causes an anisotropy in the electron r(v) AcFual ~ ~Mawellian

O~v Energy Flow


Fig. 17.1. Heat carrying electron distribution compared to Maxwellian.

212

IL. Bobin, High intensity laser plasma interaction

distribution function, low frequency electromagnetic modes become unstable [781. The chaotic magnetic field thus generated was shown to effectively reduce the electron mean free path [79]. Saturation is expected from mode coupling. At the saturation level, the electron mean free path turns out to be consistently smaller than the temperature gradient scale length L. The limitation factor in (16.1) is accordingly
f-~ (c/Lw~)~

(17.4)

with the effective collision frequency


Ve LW Peff~~(~)
1/3

(17.5)

which appears to be larger than the values calculated for the ion acoustic turbulence. One more candidate for transport inhibition is the self generated D.C. macroscopic magnetic field (see section 20). Across such a field whose order of magnitude can be megagauss, heat transport is mainly due to ions whose diffusion coefficient is expected to be somewhere between classical (oB2) and Bohm-like (ceB1). The influence of magnetic fields on electron transport was recently demonstrated [80] in an experiment involving several laser beams impinging onto different locations regularly spaced on the surface of a rod-like target. X-ray emission from the hot plasmas thus created, allows pinhole imaging. Hot spots are naturally observed at the laser foci. Some others also appear at discrete intermediate locations corresponding to zero field point in the magnetic field pattern resulting from laser interaction in such a geometry. Energetic electrons emitted by the plasmas concentrate again in such points as shown on fig. 17.2. The magnetic field both contributes to the inhibition of heat transport perpendicularly to the lines of force and favours energy transport to definite places by superthermal electrons. Details on the influence of D.C. magnetic fields on electron transport are given in the review by C. Max [77]. In all the above investigations, the plasma is considered classical and dilute. This might not be always the case in laser interaction. When a frequency converted Nd-laser beam impinges into silica or usual metal targets, the resulting temperatures (0.51 keV) at critical density correspond to an average ion charge number (Z) of about 10. Thus the ion plasma parameter
F, =

(Z)2 e2(ne)3/ T

(17.6)

.~

O~\

/)

Fig. 17.2. Electron trajectories (~) in presence of magnetic fields (0

) generated by laserplasma

interaction.

IL. Bobin, High intensity laser plasma interaction

213

T(kev) 1 10

decreasing Z ~
-

cutoFf

~
incr~osing

Id

schocked region fl~cm 1025

10

1023

Fig. 17.3. Temperature/density profile in the conduction zone of a 0.25 ~smlaser irradiated high Z material. The effective ion charge may decrease when going from the critical layer to the ablation front. However, the ion plasma parameter is still high.

a value which is conserved all along the profile on the high density side (fig. 17.3). Consequently heat transport takes place in a strongly correlated medium, the potential energy between neighbouring ions being of the same order as the kinetic energy. The influence of strong correlations on ion thermal conductivity was studied numerically by molecular dynamics in a one component plasma, i.e. in which the electrons are a neutralizing fluid [81]. Of course, as before, the electric current is zero in the one-dimensional problem. It turns out that for a F of 0.5 or 1 the transport coefficient is significantly reduced (roughly by a factor of 2) with respect to the usual values for the same densities and temperatures. Transport in the region of laser driven plasma flows might also be due to the spreading of thermal radiation. Indeed radiative phenomena are expected to be important for high Z target irradiated by high frequency lasers. Soft X-ray emission is widely used as a diagnostic [82]. Reportedly [83]indirect implosion schemes use thermal radiation. The corresponding transport problems are scarcely dealt with in the open literature. However some results are available. For instance it was shown in [84] that lateral radiative transport in layered targets induces a pressure smoothing. Moreover any important transverse energy transport contributes to flux reduction in the direction of the laser beam axis. 18. Flux limitation from transport theory The usual SpitzerHrm derivation of transport coefficients includes a flux limiter thanks to thermoelectric effects. However, assumptions were made which do not apply to laser driven plasma flows: thick (with respect to the electron mean free path) conducting zone, gentle gradients and small distortions of the equilibrium (Maxwellian) electron distribution function. Extensive research has been recently carried on transport theory. The starting point is the kinetic equation for electrons (18.1)
a~ a~-

3v

t~

in which t9f/otI~is any realistic convenient collision term.

214

fL. Bobin, High intensity laser plasma interaction

A standard technique for the numerical solution of this equation uses a splitting of the distribution function into an isotropic part f(0) (Maxwellian) and an anisotropic part
(18.2)

over the where ~t = vjv is a small parameter. The heat flow can be calculated by integrating v3~1~ entire velocity space

v3f~(v) d3v = 4ir

v5 J~(v) dv.

(18.3)

It was shown by Gray and Kilkenny [85]that in the SpitzerHarm theory, the heat is mainly transported electrons in the high energy tail (fig. 18.1) and the total electron distribution function is not positive for all ,a. This is unacceptable. In order to avoid it, the anisotropic part of the distribution function has to be smaller than the isotropic one, i.e.
by (18.4)

for those velocities which correspond to an important heat flow. The condition implies a smaller flux. However, using a spherical harmonic expansion and considering f(O), f(l) as the moments of the distribution function, one gets [86] fluid dynamical-like equations

(~+ ~ t9t

v)t0

a[g

_~O)]

(18.5)
+

(-~-+ vp1~ v)f~=

-~{V 1 vJ~ f~-f~f~} agf~}

2); g (>0) is a given source, and a is a relaxation in which the indices to condition the components ff constant. Then if therefer initial is Lf11 <of1f(l) everywhere, it cannot be subsequently greater than 1. Indeed in the r.h.s. of the second equation (18.5), the last term acts to decrease ~ in magnitude whereas when Lf~11I* 1 the first term vanishes as 1 {f(l)J2 The equations are automatically flux limiting. Furthermore, they have the same free streaming solutions as the exact transport equations in the limit of a very thin plasma layer. Many attempts have recently been made in order to get quantitative results and a large comL F~

~)

~V

Fig. 18.1. SpitzerHrm values for electron energy distribution, heat flow and the ratio f,/fo. L is the gradient scalelength. A the electron mean free path.

IL. Bobin, High intensity laser plasma interaction

215

putational effort is going on [87].Most of it is grounded on a kinetic theoretical approach, the collision term in (18.1) being of the FokkerPlanck type. Usually one uses a Legendre polynomial expansion: f(x, v, ~i,
t) =

~ f~(x, v, t) P~(~)

(18.6)

with a truncation at some specified fN~ The resulting set of coupled equations for the f~ is then integrated numerically with a zero current constraint and given initial and boundary conditions, examples of which are displayed on fig. 18.2. The transport is one dimensional but of course the distribution functions are 3-D. Initial conditions with a temperature step inside a uniform density plasma slab were used by Bell et al. [88]. Their computation shows an evolution towards a temperature gradient which extends over the whole slab. The heat flux is plotted at each grid point for a given time as a function of the local ratio of the gradient scale length to the electron mean free path LIA. The maximum heat flux is always found for LIlt <10, almost an order of magnitude below the corresponding SpitzerHrm value: fig. 18.3. Using the same underlying theory, Matte and Virmont [89] started from initial conditions of the fig. 18.2b type with different values of the ratio LIlt at t = 0. Then results are also plotted on fig. 18.3 evidencing the influence of the initial density gradient on the flux reduction. A slightly different approach by Shvarts et al. [90] uses a kinetic equation with a simple relaxation term in the right-hand side and the prescriptions
fi~fo,

FF1~~.

(18.7)

The results also appear on fig. 18.3. The same authors recently developed a non-local hybrid model in which the electrons are a single fluid except that the heat flow is calculated through a multi-group
F/F~5

~SpitzerHarm
1
N

Free streaming

Fig. 18.2. Initial conditions for transport calculations: (a) Bell et al. [88],(b) Matte and Virmont [89]; L can be varied; in both cases, the density is spatially uniform,

Fig. 18.3. Local heat flux across the conductive layer. Each point along the temperature profile is referred to by the ratio of the local temperature gradient length L to the electron mean free path A. Computational results: after Bell et al., after Matte and Virmont, 000 after Shvarts et al.

216

IL. Bobin, High intensity laser plasma interaction

diffusion procedure for those electrons whose velocity exceeds some prescribed value v (-~2ve). The temperature profile is then qualitatively different (fig. 18.4), and very likely is more realistic. Indeed although experiments are seldom clear, it can be inferred that comparable energy fluxes are transported with mean free paths which differ by an order of magnitude. In laser irradiation of thin foils, two-sided measurements such as those at Rochester [91] indicate a stronger flux limitation than when measurements are made on the laser side only [92].Now, according to D. Shvarts, in a temperature profile such as the case b) of fig. 18.4, half of the energy leaving the interaction region is deposited in the steep gradient and contributes to imparting momentum to the target. The other half is deposited far beyond point A. Its role is mainly preheating. Measurements on the laser side indicate which (limited) flux is leaving the interaction region whereas on the opposite side many properties depend on the (twice as much limited) flux deposited in the steep gradient zone. The next step in the theory deals with strongly inhomogeneous plasmas. A lot of computational work is presently being done [87], little of which is so far published. An alternative numerical approach to heat transport across steep density gradients uses particle simulation or rather a mixture of particle and fluid dynamics [93]. A Monte-Carlo method is used to predict the behaviour of the particles. In such a simulation, coronal decoupling of the electrons in the density gradient and the subsequent electric field which drive the return currents are automatically accounted for. The comparison with flux limited calculation is better made for the temperature profile, fig. 18.5. There is no value of the flux limitation factor to match the results of the Monte-Carlo simulations over the whole heat conduction region. The thermal flux at some given point inside a thin slab depends upon the distribution function at different places throughout the temperature gradient. This non-local behaviour is automatically accounted for in the above numerical computations. On the contrary, in the Lagrangian fluid dynamical codes extensively used to describe laser plasma interaction, the thermal flux is modeled by means of a local formula of the (16.1), (16.2) type. In order to reconcile the actual non-local character of the heat flux with a simple formulation acceptable by fluid codes, Luciani et a]. [94] proposed the following expression F(x) = J K(x, x) FSH(x) dx. (18.8)

Te

Te fUi

// _______

Flux limi~ed lranspor~ Monte Carlo calculations


Fig. 18.5. Comparison of Monte-Carlo calculations after Mason [93] with temperature profile resulting from transport theory with different flux limiting factors.

Fig. 18.4. Temperature profiles obtained computationally: (a) usual flux limited transport, (b) non-local hybrid model,

IL. Bobin, High intensity laser plasma interaction FSH

217

is the SpitzerHarm flux, and K is a temperature dependent kernel (18.9)

1 / ~ n(x) dx \ K(x, x)= 2A( ~)exP~ ~ A(X1) n(x))~ A is an effective electron range aT2 4irn(Z+ 1)12e4lnA

(18.10)

For gentle temperature and density gradients, the kernel K(x, x) reduces to a 8 function and one gets the usual Spitzer.-Harm value. On the contrary, in the case of a step-like temperature gradient associated with a uniform density, the exponential in (18.9) can be approximated by 1 at the location of the temperature jump. A straightforward integration of (18.8) yields a maximum heat flux which is a fraction of the free steaming value, provided the adjustable constant a is conveniently chosen: a value of 32 restitutes the results of FokkerPlanck simulations [89]. Furthermore the model predicts a precursor and except for this particular feature is in agreement with a flux limitation factor f of about 0.1.

19. Filamentation and self focusing So far, mainly one-dimensional situations were considered. We have seen how in high intensity interaction, the ponderomotive force acting in the direction of light propagation at normal incidence results in density profile modifications. The wave nature of electromagnetic radiation is essential. The structure of the plasma flow matches the standing wave pattern due to the reflexion at cut-off. Through coupling via the ponderomotive force, E.M. waves and the plasma flow behave consistently with each other. Of course one may expect similar results in two (or more) dimensional situations. For instance, when light impinges obliquely into a density gradient, the ponderomotive force acts in both directions parallel and perpendicular to the gradient. The effect was investigated by means of numerical simulations which show the building up of a diffraction grating matching the wave properties of the incoming radiation [73]: fig. 19.1. Such a behaviour cannot be considered as an instability. On the contrary, a plane electromagnetic wave propagating in a plasma is unstable against modulations perpendicular to the direction of propagation. Growing density fluctuations end up in light filaments. The effect has also been predicted and observed in ordinary dielectrics [95]. In contrast to weakly non-linear dielectrics, the non-linear dielectric constant of a plasma saturates when filaments turn into empty channels. They trap radiation whose ponderomotive pressure balances the material pressure of the surrounding plasma. In order to deal with the transverse components of the ponderomotive force, we start from a non-linear Schrodinger equation of the form ~ (19.1)

218

fL. Bobin, High intensity laser plasma interaction

~Thr~

~ ~

L_
~

..
-~-

.~

nc

x
[731.

Laser
Fig. 19.1. Density contours in high intensity interaction: 2-D computational results after Lindman

Here, E represents the slowly varying amplitude of a field oscillating with a high frequency w. In the equation, the density p contributes to the non-linearity. The Laplacian operator V2 is three dimensional. Now assume the plasma is uniform and at equilibrium. The electron distribution is Boltzmannian
(~5)

flefle

exp

_____
Te

(19.2)

where q5~is the ponderomotive potential


W2 VE2 F~=~=V~
w

8ir

0.

(19.3)

The ion distribution is also Boltzmannian (Z= 1)


lie =
~~O)

exp(e4IT1).

(19.4)

Eliminating ~ yields
=

Telfl(fle/fl~) T~ ln(n1/n~~).

(19.5)

For distances large with respect to Debye shielding fle = n, and


fle

n~0)exp(_T~+71)

(19.6)

Linearizing and substituting into (19.1), one gets 2 w2 W2 475 _E 2i c 5+~V2E+(1__~+~ )E=o
W W
(i)

(19.7)

W Te+l1

IL. Bobin, High intensity laser plasma interaction

219

in which E has been factorized into a part with slow space variation and a rapidly oscillating one. ~ is the reduced coordinate in the reference frame in which E has no time variation. Since E satisfies the linear dispersion relation, (19.7) then reduces to (19.8)

t4)p

8ITW

in which V~ is the transverse Laplacian and E~was neglected. Equation (19.8) is a cubic non-linear Schrodinger equation, the longitudinal coordinate ~ playing the role of the time in more usual situations. This equation has soliton-like solutions which correspond to oblique filaments. On a macroscopic scale, the intensity, inside the cross section of a laser beam, is inhomogeneous: typically Gaussian. Due to the balance between material and ponderomotive pressure, the regions of higher light intensity correspond to lower electron densities. Since the plasma dielectric constant is in first approximation
= 1 W~/W~,

(19.9)

a low electron density (higher refractive index) surrounded by denser zones, acts as a converging lens resulting in self focusing of the laser beam (fig. 19.2) provided the incident power is larger than an instability threshold. The condition for the instability was given in [96]using a perturbative approach to solve an equation similar to (19.8). Let ~E be the field perturbation with respect to the incident value E 0 (in the y direction). The non-linear dielectric constant is then written ~= ~()+~~~_+ 1E~E (19.10) w 8irT and the instability grows if
W2
2 C

_____________ + (w~/W2)(Eo2/8 ITT)


~

~E~

(19.11)

where k~= + k~ is the transverse wave number squared. As expected high intensities favour the instability. So do small k 1, i.e. transverse mode with long wavelengths. For self focusing distances of the order of beam cross section diameter, the inequality is transformed into a condition for the incoming power.

Fig. 19.2. Beam global self focusing. Arrows indicate refractive index gradients.

220

IL. Bobin, High intensity laser plasma interaction

Numerical simulations [97]show that when the instability builds up, the laser light decays into lower frequency modes. The electrons are consequently heated and then contribute to the development of narrow filaments which trap the incoming radiation. More can be derived about self focusing when adding to (19.8) terms accounting for absorption (damping) and pump (forced oscillations inside a laser active medium). One then gets an equation which with properly reduced variables reads: E~iE~~2ig2JEI2E=FEEoexp(iko~) very similar to the one investigated by Kaup and Newell corresponds an energy equation
[981

(19.12)

in a different context. To this equation

-~

EE* d~ = 2F

EE* dy

[EO

exp(iko~)J E* dy + c.c.].

(19.13)

One then looks for a solution of the form


E

2~ exp{2i(cr + ~4)}~ ch2ij(yy 0)

(19.14)

Substituting into (19.13), one obtains the following system


=

2Fi~+

E0.
IT

Slfl

(19.15)

2 ~koX+2o OF Xeko4~ in the phase space mx; singular points may exist on the lines
= (k~/2)11~

(19.16)

provided E 0 is larger than a threshold given by 2 E~>~-k 4F 0.


IT

(19.17)

One thus finds saddle points for i~ cos x <0 and attractors (nodes or foci) for s~ cos x > 0. An example of phase space trajectories is shown on fig. 19.3. Note that the threshold is proportional to the spatial frequency k0 which vanishes at the critical density. One then is to expect local lowering of the threshold whenever the density is close to cut-off. When the pump intensity is above threshold i~ increases with In the case 2 of a strong damping, after an overshoot. attractors of (19.15) are nodes and ~ reaches its asymptotic value (k0/2)~ Furthermore since there is a sine in (19.15), ~j is periodic along the f-axis as shown on fig. 19.4. As the damping decreases, nodes of (19.15) become foci and the solutions i~(~) oscillate around the asymptotic value. The solution is, of course, still periodic along the f-axis.
~.

IL. Bobin, High intensity laser plasma interaction

221

_O~\X

Fig. 19.3. Phase space plot for eq. (19.15); E above threshold, (a) small damping; attractors are foci, (b) large damping.

Fig. 19.4. Amplitude of the self focused field along the direction of propagation in case of strong damping.

So, accounting for damping, i.e. absorption in the non-linear theory of the propagation of a cylindrical laser beam, results in a pattern of repeated foci along the beam axis. Long filaments are thus expected with periodic light concentration. The theory might also describe the occurrence of bubbles regularly located along the axis of an amplifying laser rod occasionally observed when operating a Nd-glass cascade.

20. Self generated magnetic fields Magnetic fields were observed quite early in the investigations of laser product plasmas [99]. Strong magnetic fields, associated with rapid plasma flows are also commonplace in astrophysics. However, since in laser interaction there is no initial B field, one cannot use 1100] the dynamo theory which applies to astrophysical problems. In section 11, the low frequency motion of the electrons was derived by averaging over one period high frequency terms in the equations. If the low frequency electric field E is not curl free, a magnetic field may set up according to MaxwellFaraday and MaxwellAmpere laws

cVxE=oB/at,

cVxB=4ITf.

(20.1)

A solenoidal electric field exists first when density and temperature gradients are present simultaneously in different directions. Indeed, starting from the electron equation of the section in which inertial terms are, as usual, neglected, one gets the steady electrostatic field associated with the pressure gradient

222

IL. Bobin, High intensity laser plasma interaction

E=~-~for a plasma with local density n and conductivity u. Substituting into (20.1), one gets

(20.2)

(20.3)

The first term in the right-hand side is the source of a toroidal magnetic field associated with a fountain shaped flow of electrons (fig. 20.1). Such currents and fields were actually observed in experiments in which laser irradiation of surfaces takes place in presence of a background gas [101]. Furthermore, the current was proved to flow inside the target material [102]. Now the second term in the r.h.s. of (20.3) is rewritten after some algebra as
C

VTXVn.

en

This is another source for the building up of a magnetic field provided the density and temperature gradients are at an angle. It was shown in part 1, that longitudinal density gradients exist together with an almost uniform temperature in a laser driven plasma flow. Since the laser beam has a limited transverse extension, lateral temperature gradients combined with the longitudinal density gradient generate a toroidal Vn x VT magnetic field (fig. 20.2) whose magnitude can be evaluated by balancing the growth rate with convective losses, -~-VnXVT~VX(vXB).
(20.4)

Assuming that the convective velocity is the isothermal velocity of sound (5.2) and that all gradients have the same scale length L
c /4 B=_(~ 1)
1/2

T112
~

(20.5)

LASER~ BEAM

LASER BEAM

plane
Fig. 20.1. Magnetic field generation associated with fountain shaped electron currents. Fig. 20.2. Magnetic field generation associated with density and temperature gradients.

IL. Bobin, High intensity laser plasma interaction

223

Putting numbers in the formula, one easily finds megagauss fields for T 1 keV, L -~ 10 p.m. Another source for an electric field which is not curl free stems from collisions through the ponderomotive force. Indeed, let a simple phenomenological collision term m~(v v1) be added to the right-hand side of the equation of motion of the electrons. v is an effective collision frequency and v, is the ions velocity (v, ~ v). Equation (11.1) is then replaced by m ~= m(u .V)u e (E+~xB)~m((i3 .V)) eK~x B) m ~-(n(v v~)). (20.6)

The ponderomotive force which is calculated after the averaged terms in the right-hand side of (19.3) now reads F~= _mn{V~~+ [V. v(ar)] where ~ir denotes the electron excursion in their high frequency motion:
(20.7)

~r=Jvdt.

(20.8)

The ponderomotive force mainly acts on the electrons and thus induces a charge separation field E5=F~Ien (20.9)

such that taking the curl of (20.6), in which electron inertia is neglected, yields a low frequency magnetic field satisfying the differential equation

Vx
49t

uXB

2B = S
~

4( r, t) + S~(r, t)

(20.10)

(~

5R are the thermoelectric and radiative source terms respectively. They are linked to the where S~ andradiation pressure tensors through the formulas electron and

S~=

e
C

V x -~-V~ (Pc),

(n) 1

S~=

VX
e

(n)

V~ (PR)

(20.11)

In the case of thermal equilibrium, one has a scalar pressure and S~ reduces to the only one surviving

224

IL. Bobin, High intensity laser plasma interaction

term

S~=VTxVn.
The collisions and kinetics will manifest themselves as anisotropic terms in S~ :~ SR is also
SR

(20.12)

On the other hand

~-V X (n x [Vx v(~ir)] + [V. (~dr)Ji~)

(20.13)

in which the collision frequency ii appears explicitly. The complete equation for B field generation was given for the first time by Braginskii [1031.Such an equation of the (20.10) type can be solved computationally as was done in [104] in order to account for the experimental observation of megagauss fields [1051. A general discussion can also be made analytically following Mora and Pellat [1061. Obviously, according to the importance of collisions, several regimes are possible for magnetic field generation. Furthermore, the way collisions are dealt with in kinetic equations might also influence the final result. In the right-hand side of (19.2) a simple relaxation term accounts for electronion collisions and will be the only one to be used in the following. Roughly, one has first a strongly collisional regime in which the magnetic field evolution is given by (19.7) together with ~(kLL)1 SR ~ (20.14)

where kL is the laser wave number and L is a characteristic gradient length. In usual experimental situations, the ratio (19.11) is much larger than 1 so that SR may be neglected, i.e. the magnetic field due to the non-linear anisotropy of the electron pressure dominates the field due to the radiation pressure tensor. However the reverse may be true in the case of very strong density gradients (small L) such as the ones discussed in section 14. In the weakly collisional case, one has to compare L to the electron mean free path A~ and also to the Debye length AD. If L ~ (AeAD)112,
(20.15)

collisions may be completely neglected. The weakly collisional limit corresponds actually to the inequalities
(AeAD)12

~ L ~ Ac.

(20.16)

Then S~ is still the dominant part in the right-hand side of (20.10) which is now replaced by

9t

T m

ac2ala2B

8e3i ox n On \

8x w~ 9x

8x2

1T~82 v m Ox

(20.17)
/

IL. Bobin, High intensity laser plasma interaction

225

in which A is an integral over the Maxwellian distribution function at temperature Te:


A
~J~fM(V~ Te)dV.

(20.18)

The strongly [107,108, 109] and the weakly [110]collisional regimes were independently investigated. In order to encompass all possible cases in one and the same discussion, Mora and Pellat [106] introduced the following set of dimensionless parameters: normalized collision rate normalized thermal velocity normalized time scale normalized convective velocity
= p1w ~ =

vjwL ~ 1 2(wr)~ = (kLL)


= kLL U/C.

8H

Now, in the equation for B, (19.7) or (19.13), the left-hand sides differ only by the third term. In (19.7), the more general, it scales as 3.8/Ot whereas in (19.13), weakly collisional limit, it scales as 8~8~1. In all cases OB/Ot scales as 8~ and V X v X B as 8H~ If 8~,6H .:~ Sup(&, 8~6~.1), the diffusive term is dominant and the saturation of the field is due to resistive diffusion if 6~
~

(20.19)
~,

or

thermal effects if 6~ 6~.If


~

6c, ~

(20.20)

OB/Ot is the dominant term and no saturation level is reached. Finally if 8H~8o,6~, 6~6~, the convective term is dominant and ensures saturation. The various possible regimes are displayed on fig. 20.3, in which
B &,

(20.21)

i.e. the collision rate, has been

~- Convective short time or

saIurahon~

~ ~H (I~Li1

Fig. 20.3. Regimes of magnetic field generation.

226

fL. Bobin, High intensity laser plasma interaction

chosen as the relevant parameter. At both ends the maximum field does not depend on the collision frequency: regimes (1), collisionless, and (6), strongly collisional. In between, several situations appear. Regimes (2) and (3) are weakly collisional. However there are enough collisions to generate a magnetic field which in case (3) may saturate because of the fluid dynamical convection or alternatively because of a short plasma lifetime. Regimes (4) and (5) are collision dominated. Saturation comes in (4) as in (3) whereas in (5) it is due to collisions. In (6) the laser wave is totally absorbed and the magnetic field results from momentum transfer from the light beam to the plasma. In absence of convective or short time saturation, the maximum possible value of the field (3,,, = 6,.
~o 6H0)iS

Bmax

(f).

(20.22)

Hence in regimes (3) and (4)


B

win

3,4~--

6}-Je

()
C

~2

wm or 80e

()
C

(20.23)

Experiments were carried out on laser plasmas, locally probing the underdense region with a weak laser beam and measuring the Faraday rotation [105,1111. Megagauss fields were observed with a toroidal geometry. Electron trajectories in such configurations were investigated in [112] where the connection with transport inhibition and fast ion generation is established.

III. Wave Couplings 21. Purely growing modes and 3-wave instabilities A second source of non-linearities was identified in section 2 with the term
(9 in.).

at

It appears in wave propagation equations. n and v may include components with different frequencies. It can always be expanded as (9 0u1
flV

On1

0u1

On2

at

fl0+U1+fl1+U!~+~

at

at

at

at

(21.1)

Indices 0, 1, 2,... denote zero, first, second.... order quantities respectively. The zero-order (fluid) velocity is chosen to be zero which has no consequence on the following. The zeroth-order particle density corresponds to the unperturbed neutral plasma. Its spatial variations are due either to the flow or to ion acoustic oscillations. They have a very low frequency. In a first

IL. Bobin, High intensity laser plasma interaction

227

illustrative step, they are taken as purely spatial with a wave vector k. Then, assume that an electromagnetic wave (laser) acts on this spatial modulation with an electric field K0 parallel to k. Furthermore the laser wavelength is very large, i.e. the wave number is k0(0) 4 k so that E0 can be taken as spatially uniform: dipole approximation (fig. 21.1). Under the influence of the electromagnetic wave, electrons are displaced inducing a periodical charge density. The charge separation electrostatic field E~(4Eo)tends to pull the electrons back. One thus gets an oscillator whose eigenfrequency w~ satisfies the BohmGross dispersion relation 2v~. (21.2) = w~ + 3k Of course, a resonance occurs for We = w 0 (laser frequency). In the vicinity of the resonance one has the usual phase relationships. For w~ < We, E~ is in phase with the driving oscillation whereas for w0> We, Ee is out of phase. What happens can be understood using again the ponderomotive force which in the present case results from the superposition of both fields. Furthermore, since Ee is small and E0 is uniform, one has 2) ii (Eo)d(Ee) 213 F ii d ((Eo+Ee) n,dx 8ir fle 47T dx ( . ) At resonance, E 0 and Ee have the same frequency and are in quadrature so that F0 is zero. The perturbation cannot grow nor propagate. Close to the resonance, one has the two situations displayed on fig. 21.2 [113]. For w <We the ponderomotive force acts in such a way that it amplifies a motionless perturbation. On the contrary, when w is greater than We, the force acts the opposite way so that the perturbation propagates. Electrostatic and ion acoustic waves are launched in opposite directions. The former mechanism, a purely growing mode is a two stream instability as shown by Nishikawa [114].The latter is an unstable 3-wave coupling in which an electromagnetic wave induces an electron plasma and an ion acoustic wave: parametric decay. Momentum and energy conservation require that the following selection rules
=

w~ +

Wac,

ke + kac

(21.4)

be satisfied by the waves participating in the parametric decay. Furthermore, since oscillations are quantized a quantum of electromagnetic radiation (photon)

~
0
WO( ~e
Fig. 21.1. Interaction geometry.

0
~o> ~e

Fig. 21.2. Effects of an electromagnetic wave on plasma neutral density spatial oscillations.

228

IL. Bobin, High intensity laser plasma interaction

generates a quantum of electronic plasma oscillation (plasmon) plus a quantum of ion acoustic oscillation (phonon). One then has for the total number of quanta of each species involved
NoNeNac. (21.5)

Now the energy of any kind of wave is


WNhw
(21.6)

hence the following relationships (ManleyRowe) [1151 Wo/wo = We/We


=

Wad Wac

(21.7)

which apply to any 3-wave couplings. Now, due to the expansion (21.1), the wave propagation can always be written in the form Dispersive l.h.s. for wave I
=

source term depending on waves 2 and 3.

(21.8)

The possible coupling always involve an electromagnetic (laser) wave, which can be associated with any other kind of waves leading to instabilities classified in table 21.1, together with their frequency selection rule. Anticipating on section 26, the table includes scattering processes in which another electromagnetic wave is generated. An important feature linked with these effects is of course the resonance which takes place whenever an electromagnetic wave frequency is equal to the local plasma frequency or is twice this value just as in the case of parametric excitation of mechanical oscillators (Lord Rayleigh [116]). Accordingly, in a density gradient two main resonance zones appear, one at the critical density Pc, the other at ,o~/4. Each zone has specific properties and is a privileged place for non-linear effects listed on fig. 21.3. There is no resonance for Brillouin scattering, which however deserves a special emphasis, see

Table 21.1 Coupling of an electromagnetic wave with: Generated wave electromagnetic electron plasma

ion acoustic wave electron plasma wave

Brillouin scattering
(05

Raman scattering W~w~ anti-Stokes Raman scattering ws= W~-~W~ 2nd harmonic generation w, 2w~

= =

parametric decay

2 plasmon parametric resonance

w= 2w~

electromagnetic wave w~ in the presence of a gradient

IL. Bobin, High intensity laser plasma interaction

229

mirror point

for

c...)~/2

resonance for 2 ptasmon Roman 2 harmonic ~

,.-

mirror point / for c.i 4, resonance for parametric decay Osc. 2 stream ontistokes Raman
2c.so

harmonic

2nd nonLinear
zone

l5tnon(inear zone

Fig. 21.3. Resonance regions in a density gradient.

sections 26 and ff. Let us just note that the corresponding ManleyRowe relations Wo/wo = W5/w. = Wac/Wac show that, since one always has
Wac4W0,

(21.9)

(21.10)

the backscattering process is energetically very unfavourable whenever the saturation level of the instability is high, i.e. W0 is only slightly smaller than the incident energy. 22. Instability dynamics In order to get detailed information about 3-wave coupling, a general approach is necessary. In order to do so, let q1 be any one of the variables ~,i5~,E1 describing the wave 112 with (longitudinal index i. The characteristic waves). The velocity is c, which may be general equal form to C (E.M. waves) or to (3TIm) propagation equation has the ~ We now apply the multiple scale method to both time and space. Let then
to=t,

(22.1)

t 1et,...

x0x,

x1

cx,...
(22.2)
/
~2 ~2 ~

0 0 3 =+s+... Ot 0t0 3t1


i02

3 0 3 =-+e+... Ox Ox0 Ox1


~2

P~P0+sP1+~lC~----+D~)+2e( C~ \Ot~ Ox~ I \Ot0 at1 Ox0 Ox1!

230

IL. Bobin, High intensity laser plasma interaction

The q are expanded:


q 2q2~ + eq~ + e
=

Re a~ exp{i(kxt 1

~ }
wto)}

(22.3)

where a~ is a function of the slow variables t1 and x1. In first order in c, one finds
=

(w~ + C~k~ + f2~) a ~ exp{i(k1x~1i w.t0)}

(22.4)

which has to be zero in order to be consistent with the propagation equation. Hence the linear dispersion relation D1(w1, k1)
=

w~+ C~k~+ Q~ = 0.

(22.5)

One may then write p1 nsi ( 2w+2C~k-) = i (+-). \ 0t1 Ox1! \Ow, Ot~ Ok, Ox1! In second order in c, one finds
P0q~ 2~ = 0
/

0\

iOD.0

OD.0\

(22.6)

Oto
=

1q~q~~ P1q~
Wk)tO]}

~i(w1+ wk)A~aJakexp{i[(~+ kk)xO~ (w~ +


i

/ODO OD. 9~ (k + a ~exp{i(k1x0~ w~t~)}. Ow, Ot1 Ok, Ox1

(22.7)

2), one has to set P 2) 0. A first consequence is: the two In order to prevent secular of q( restitutes the selection 0q( rules, exponentials should a be equal.behaviour This condition
W(Wj+Wk+Z~W,

kI=kJ+kk+z~k

(22.8)

in which frequencies and wave vector mismatches are introduced for completeness. As a second consequence, one gets a partial differential equation for a which is better rewritten after noticing that

(~/(~)

Vgi

(22.9)

the group velocity of the wave i, and setting


(22.10)

IL. Bobin, High intensity laser plasma interaction

231

which defines the coupling constant g,. The same procedure can be applied to waves j and k, and yields the following system of coupled equations:
/0 Ot +
Vg~ _)a

O\

Ox

1 =

g,(wj + wk)a)ak exp{l(i~k x Aw t)} wk)a~a~ exp{i(i~k x E~w t)} w1)a1a~exp{i(~k x ~w t)}
.

(-f- + v~ ~-)a= gj(wi

(22.11)

(~+

Vgk

= gk(wa

Here the wave labelled i is the pump. It may vary in time, a possibility, whose consequences will be investigated in further sections. Time and space derivatives are with respect to the slow variables. If necessary, damping rates as well as frequency and wave number mismatches can be added to the equations. To begin with, let us deal with the simplest situation in which the pump is time and space independent on the slow scales, the so-called dipole approximation. We are left with two equations (including damping) and perfect matching: (.~-+ vi -~-+i~) a1 = K1a0a~
O 0
.

(22.12)

(~-+v2_+F2) a2= K2a0a~ Denoting by u a flow velocity, introducing a similarity variable


=

at O/Ox=d/d~,}
(22.13)

O/Ot=ud/d~,

and looking for a real solution of the form a


=

a exp(a~)

(22.14)

one gets an algebraic system whose determinant should vanish. Solving for a yields 2 + 4K 2 = 11(v1 u) + F2(v2 u) {[F1(v2 u) 12(v1 u)] 1K2a~(v1 u)(v2 u)}~ 2(v 2 u)(v1 u)

(22.15)

If a is real and positive, instability will take place. Hence, the threshold condition a~ > f112/K1K2 with the supplementary condition (22.16)

232

IL.

Bobin, High intensity laser plasma interaction

(v

1u)(v2u)>0,

(22.17)

i.e. in the reference frame moving with velocity u, both waves propagate energy in the same direction. They are amplified in space at the same rate. If we put together the dispersion relations obtained in first (22.5) and second-order for both waves a1 and a2 the frequency and the wave number are complex quantities, a situation typical of a convective instability [117]. The waves grow while propagating. If inequality (22.17) is reversed, one gets oscillations on the slow time and space scales. However, in the case v~v2<0,
(22.18)

there is the possibility of a local temporal growth. Indeed let u = 0 in the above derivation; a has an imaginary part whenever ir r \2 ~..1 2V1 I a~> 4K1K21 V1

(22.19)

Now, assume that the wave numbers are real. When looking for solutions of (22.12) in the form
a = a exp{i(Dt Kx)}

(real

K)

(22.20)

and setting the determinant of the subsequent algebraic system equal to zero, it is readily seen that ~Qis a complex quantity, thus including a growth rate. This is the situation, a complex frequency associated with a real wavenumber, of an absolute instability whose threshold is given by (22.19) for a homogeneous plasma. Let us apply the above results to the parametric decay instability. Then a1 is for instance the electric field Ee of the electron plasma wave while a2 is the amplitude nac of the density perturbation in the ion acoustic wave. The equations then read on the slow scale:
dEe
+ FeEe

d~

wE0 nac 4n~1


l

0
(22.21)

dflac +Tacflac d~

w0E0 67TMV~acV~e Ee0

Here
=

w/k,

V 4,ac =

(TIM)

112

(22.22)

are the phase velocity of the plasma and ion acoustic wave respectively. The way they appear is due to momentum conservation in the actual 3-wave coupling in the plasma with unperturbed density n 0. Now, the damping coefficients result from both collisions (dominant in dense plasmas) and Landau damping (dominant in dilute plasmas). Collision frequencies are of the form (3.1) whereas Landau damping rates are easily found in textbooks on Plasma Physics. One then has the effective frequencies

IL. Bobin, High intensity laser plasma interaction


IT ~ Fac = I/ti + j

233

112 / ZTe /ZT~\ l~~) exp~~


(22.23)

Fe

irw~ / 1 Pei~, ~ exp~,.,,2~2


L

lve/tD

Far from threshold, (Fe, Fac negligible) the growth rate is


= 1 dEe(ti) = W~~

1~

11/2

(22.24)

E~(t 1) dt1 2w0 Lflo.M~CVq,acVcei in which the E.M. wave intensity (the quantity which is actually measured)
2 IoCE 0/41T (22.25)

was used preferably to E0. The convective threshold is


l6FacFefloMVcacVceC 16FacFefloCTe
WpWac

(22.26)

since the electron plasma wave and the ion acoustic wave have almost the same k, and use was made of (22.22). Using these formulas, the thresholds for the parametric decay instability were calculated in the case of Nd-glass laser light for several values of the ratio Tell1 [118]. The result is shown on fig. 22.1. On the same graph is the temperature/intensity line corresponding the regime. It experimencrosses the 1o 2 X 2 and Te 0.5to keV inlinear agreement with boundary of the stability domain at 10t3 W cm tal measurements. The parametric decay instability can be absolute. When the pump frequency w 0 is close to the plasma frequency, k0 is almost zero. Momentum conservation implies k1 = k2, the electron plasma wave and the ion acoustic wave propagates in opposite directions. Formulas for growth rates and thresholds obtained in the section can be put in a simpler form by using the electron oscillation velocity, v0 = eEIwom.
11 Te/Ti Teke (22.27)

1012

1013

W~/cm2

Fig. 22.1. Parametric decay instability thresholds. L and C denote the regions where respectively Landau and collisional damping is dominant.

234

IL. Bobin, High intensity laser plasma interactIon

For instance in the case of the parametric decay instability

V~e~V

5oac

(22.28)

Wac

and since
WO~Wp~We,

(22.29)

one finds for the growth rate


= ~ (~,~~)t/2.

(22.30)

2 Ve By the same token the convective threshold is put in dimensionless form


Jo l6Fe[ac ____ n~CT~
=

(22.31)

WacWp

where ~ is the ratio between the pump energy density and the electron thermal energy in the plasma. The absolute threshold is
2/~~c. ~A (Fe~+Fac)

(22.32)

Similar transforms apply to the oscillating two stream instability and the parametric resonance. It was shown in the latter case that k 0 0 is a necessary condition for growth [119] and the instability may become absolute [120].Table 22.1 displays the corresponding results.
Table 22.1 Parametric instabilities growth rate convective parametric decay absolute purely growing mode
-~-(w~u~)
1)5

~ threshold
16JcJ~ac
WpW

2v~

~~)l/2

(r. \/ + r~) /&~c


4T~
(Op

112 /vok\23 (\~) Wp /m\ M 2wp


~~)

2 plasmon

resonance

wi,

c (~) ()
2 2

IL. Bobin, High intensity laser plasma interaction

235

In laser plasma interaction, the plasma is usually strongly inhomogeneous. The convective or absolute nature of an instability is accordingly of paramount importance. Indeed waves propagate beyond the zone they were created in. They reach regions where resonance conditions no longer hold. Spatial growth cannot take place. Thus, only absolute instabilities are expected. However their threshold might also depend upon density gradients. The discussion on this point is postponed until section 26.

23. Influence of the pump tuning and intensity It was found in the previous section that the growth rate of an instability is proportional to the pump intensity. Actually this result was obtained after simplifying assumptions, and apply only at low intensity, near the threshold. In order to get a better picture of the dependence on the pump characteristics, one has to take up a slightly different approach. Neglect the thermal motion of the particles, and start from the fluid dynamical equations which, for one species with charge q and mass j.t, reduce to On/Ot+V.(nv)=0
dt/Ot + (v V)v = qEljs

~ (23.1)

Linearizing around n = n0, v = 0 and going to the second derivatives, one gets for the density perturbation , the equation
2 ~aOx ~=q--~. Ot
(23.2)

Now, assume that a spatially uniform equilibrium plasma is under the influence of an external field E 0 sin wt. From the second equation (23.1), after linearization, all particles have the uniform velocity qE0 v=-coswt
/1W

(23.3)

and excursion r=~~sinwt. (23.4)

In a non-equilibrium situation particles undergo traveling oscillations with wavenumber k. These create a charge separation field, the one in the right-hand side of (23.2) given by the Poisson equation. In such modes, densities are of the form exp(ik. r) and it is readily shown [121]that the functions
N =
~-

n exp(_iq

kE0

sin Wt)

(23.5)

236

IL. Bobin, High intensity laser plasma interaction

are for electrons (charge e) and ions (charge +e) solution of the differential system:
d2Ne/dt2 + ~~[Ne
+

N, exp(ia sin

Wi)]

0 0

(23 6)

d2N 1Idt

2 + w~t[Nt + Ne exp(ia sin wt)]

in which the ion plasma frequency is


=

4irnoe2lM

(23.7)

and 1\k~E 0 k~E0 2 mw2 . a=eii (23.8) \m MI w Now expanding the exponential up to the first order in the small parameter a restitutes the approximation dealt with before. Considering that a is no longer small, one has to use the Bessel identity
/1

exp(ia sin Wi) =

J,(a) et.

(23.9)

Growth rates and thresholds have been calculated after (23.6) using (23.8) [122]. Results are expressed in terms of Bessel functions. Denote by
=

w~/w2 1

(23.10)

a convenient detuning parameter. Growth rates are then functions of a (i.e. the pump amplitude) and 5. For the oscillating two stream instability (w <w 0) one finds (23.11)

whereas in the case of the parametric decay (w

>

w0)

eIaT~

Fig. 23.1. Growth rate for the parametric decay and the oscillatory two stream instabilities as functions of w~Iw.

Fig. 23.2. Instability domains (hatched areas). The frequency tuning range reduces to discrete values for the zeros of Ji(a).

IL.
~2

Bobin, High intensity laser plasma interaction

237

32

1/2

y2.w2[(J2(a)W1~1)

i].

(23.12)

The dependence on 8 is shown on fig. 23.1. Numerical simulations [123]confirm such results. Furthermore, the intensity dependence is found to agree with the dependence of J1(a) upon the argument a. Figure 23.2 shows the instability domain 1. which was determined computationally [124] in the plane a, w~ 24. Pump depletion Since part of the incoming wave energy is converted into different modes of oscillation the pump is to be depleted. The instability growth is of course expected to saturate accordingly. Equations (22.10) are suited to investigate such a process. For convenience rewrite the system as
(O/Ot+ v 0/Ox) a, = iKla

3ak
(OIOt+ v10IOx)a1 = iKa~a~ . 12k 0/Ox) a,., = iKka~a~ (O/Ot+
(24.1)

Use will be made again of the transform ~=xut and a solution of the form a=a(~)e4 is to be looked for. Now setting
= (24.3) (24.2)

K/(v u),

O(~)= q5 1(~:)_ b1(~) ~k(S~)


(24.4)

the real quantities a(~) and da1/d~ = ~i,aJaksin 0, da1/d~ = LI1a ~ak sin 0, dak/d~ = Akalaf sin 0, A first integral is aa,ak cos 0 = const.

4(e) are readily found to be solutions of the differential system


a db1/d~ = LI ajak cos 0 a1 d41/d~ = LIJa,.ak cos 0 a,., dcbk/d~ = LIkala) cos 0
.

(24.5)

(24.6)

Then a stationary solution in a reference frame moving with the velocity u has been found [125].It is expressed in terms of Jacobian elliptic functions sn(xlm), cn(xlm), dn(xlm) with

238

IL. Bobin, High intensity laser plasma interaction

in
~

,~
\1/2 ii~\ ? 2_

ak~u, ,

-I a,.,

a1

and the boundary condition a1(0) = 0. Then one gets


a~ =

j~-~a~(0) sn2(xlm)
(24.8)

a~=a~(0)cn2(xJm) . 2(xJm) = a~(0)dn The result is shown on fig. 24.1 for u = 0. In the special case u = 0, m = 1 since sn(xIl)= tanh x, cn(xIl) = dn(xIl) = sech x,

(24.9)

one gets a soliton instead of a periodic solution as shown on fig. 24.2. The above solutions which clearly evidence the saturation of the instability due to pump depletion, apply to conservative systems: no damping. Now, a more general situation would imply linear growth and damping together with mismatches, so that (24.1) is to be replaced by (0/Ot + v~ 0/Ox + F 1)a1 = iK,afak exp{i(i~k x + ~w t)}
(O/0t+ v1O/Ox+F1)a1 = iK1a1a~exp~i(~kx~wt)}.

(24.10)

(O/Ot+ vk0/Ox+Fk)ak

iKkaIa~ exp{i(z~kxz~swt)}

It was shown in the previous sections that waves j and k have the same spatial and temporal growth rates. We may then restrict the problem to waves with equal amplitudes and group velocities. The system reduces then to two coupled equations:
(0/Ot + Vtt

0/Ox + F0) a0 = iKa~ exp{i(~k x + ~w t)} 0/Ox


+

(OIOt+

Vt

F1) at = iKa0a~ exp{ i(~kx ~w t)}

(24.11)

Consequently the following apply only to convective processes. The system (24.11) is then transformed by using (24.2) together with

Fig. 24.1. Periodic amplttudes of non-linearly coupled waves.

Fig. 24.2. Soliton resulting from non-linearly coupled waves.

IL.

Bobin, High intensity laser plasma interaction

239

y = F/(v u),

LI = K/(v u)

and

u = ~w/i~k

(24.12)

so that the problem is now investigated in a reference frame moving with the group velocity of the pump. Thus we get dao/d~ + y 0a0 = i LI a~ exp{i(~k)~} da1/d~+y1a1 = iLI a0a~ exp{i(~k)~} for which solutions of the type (24.3) are to be sought. There are three real independent variables A0 = (a0), satisfying
dA0/d~ = LI A~ sin 0 70A0
dA1/d~= LI

(24.13)

A1 = Ja1J,

0=

4o

(i~k)~

(24.14)

A0 A1 sin 0 y1A1 d~
\A0 I

(24.15)

d~51 iA~ = do0 2 +~k= LI(2A1


dO

d~ d~

Icos 0+

A0

Such systems of coupled non-linear differential equations have been recently studied [125]. They are usually cast into a dimensionless form more familiar to mathematicians by setting x=A0sinO, Yo z=jA~, LI Yo so that dx/dr=x+3yz+2y2
dy/dr = y 3x

LI

y=A0cosO (24.16)

LI

2xy

(24.17)

dz/dT = 2yz 2xz


in which two parameters are left y=yi/yo, i5r_~k/yo. (24.18)

Varying both yields dramatic changes in the behaviour of the solution. In the parameter space with both y and 5 positive, one can find unstable solutions, stable equilibria, limit cycles, strange attractors. A detailed mapping of the y, 5 plane is given in [127].Fig. 24.3 shows what may happen when varying y at given 5.

240

IL. Bobin, High intensity laser plasma interaction

unstable

stabLe
equiLtbrtum c

Limit

strange
32n cycles

9cLes attractor

sequence of bifurcations

Fig. 24.3. Various behaviours of solutions of (24.17).

The occurrence of strange attractors indicates that the solution has a chaotic behaviour. Amplitudes change randomly and the relative phase locking is lost. It is to be noted that the pump should have a linear growth in the reference frame moving with the velocity u. Such a situation may be obtained in the trailing edge of a laser pulse since according to (24.12), u is the group velocity of the pump wave.

25. Coherence and incoherence


So far, electromagnetic or plasma oscillations were represented by plane waves with well-defined amplitudes and phases. Only in the cases with the onset of chaos, the phase was lost. Now, whenever phases are randomized, complex amplitudes are functions of the wave vectors. Averages over large times are such that t~a1a2a3)0

(a1a2)= a~5(kik2)J

(25.1)

and the rate equations which were used so far no longer hold. Expansions should be extended to higher orders or alternatively one has to move to a quantum description since energy exchanges between oscillators are quantized. This fact was used when setting up the ManleyRowe relation (section 21). It is well known that the quantum aspect of electromagnetic radiation very seldom appears as such [128]. However, a quantized picture is often useful. Wavewave coupling is thus very conveniently described through conspicuous schemes: consider for instance the parametric decay and the Brillouin backscattering as shown on fig. 25.1.

Parametric decay

Raman backscattering
Fig. 25.1. Quantum schemes for 3-wave couplings.

IL. Bobin, High intensity laser plasma interaction

241

In general, a plasma may be represented as a quantum system whose energy is in its oscillation modes: plasmons and phonons. Energy levels correspond to N plasmon states. They are equally spaced. Each level is split into N phonon states. (This description bears some resemblance with electronic and vibrational molecular states.) Accordingly another description of parametric effects is: by absorption of a photon a N plasmon state goes to a N + 1 or N + 2 plasmon state as shown on fig. 25.2. Similarly, in a scattering process, a photon is absorbed, another is emitted. The intermediate state is virtual. It may however coincide with a real one (resonance). Raman and Brillouin processes are represented this way on fig. 25.3. In order to calculate the time behaviour of the number of quanta of different species, one has to perform a detailed balance of gains and losses. Consider a process in which a pump wave i induces two waves j and k. All quanta are bosons. Then denoting by w a characteristic constant, the probability of the direct process is

P = wN,(k~) [N1(k1)+ 1] [Nk(kk)+ 1]

(25.2)

in which the number 1 in the brackets [ ] accounts for spontaneous emission which was overlooked in the previous sections. The probability of the inverse process is

P = w [Ni(k,) + 1] N1(k,) Nk(kk).


Hence for N,, N1, N,, the rate equations [129]:

(25.3)

J w[N,(N1+N,,+ dN1/dt = J w[N


dN,Idt=

3kJd3kk 1)N)Nk]d 3k, d3k,,


N.Nk] d 3k, d3k 1
(25.4)

1(N1

N,.

1)

dNk/dt =

J w [N,(N

1+ N,., + 1) N1Nk] d

in which 1 is usually neglected. The following form of the ManleyRowe relation immediately results from this system - ~-

f N1(k,) d

3k 1=

~-JN1(k1) d

3k 1=
~-

J Nk(kk) d

3kk.

(25.5)

N+1 pLasmOflS-i-~....... N+2 pLasmonS emitted phonon N+1 pLasinonS


_____

N+1 pLasmons
N pLasmons

N+1 phonons Nphonons RAMAN

______

N plasmons
b
a

b BRILLOUIN

Fig. 25.2. Energy level diagram for (a) parametric decay, (b) parametric resonance.

Fig. 25.3. Energy level diagram for (a) Raman scattering, (b) Brillouin scattering.

242

fL. Bobin. High intensity laser plasma interaction

From the system (25.4) it is possible to calculate the growth rate independently from any phase relationship between the quanta. Such results make sense if and only if the pump spectral width is larger than the growth rate, i.e. (25.6) then bL~wIy27rLXP/y>2rr, (25.7)

and any phase is allowed. In the opposite case, one gets a phase locking of the different oscillations. Now, in an electromagnetic wave the number of photons and electric field are not simultaneously measurable. Hence an uncertainty relationship between the number of photons and the phase, ~N . (25.8)
~-= ~ ~

2~

4ir~t

In case of equality (minimum uncertainty) one has the coherent states. Then in the limit ~N-~X,
~*()

(25.9)

which corresponds to the classical representation of a wave with a well-defined phase. The proper growth rates are to be calculated according to the prescriptions of section 22. In a plasma, coherent processes are often associated with resonances. It is the case, e.g. for the coherent absorption of n photons in which the plasma undergoes a transition from the N plasmon level to the N+p plasmon level (fig. 25.4).

Fig. 25.4. Resonant absorption of n photons.

So far, only 3-wave processes were considered. The quantum representation is also well suited for the case of 4-wave processes such as the Raman anti-Stokes which plays an important role in harmonic generation.

26. Scattering instabilities in homogeneous and inhomogeneous plasmas We now deal with three wave processes in which two electromagnetic waves, the pump and a scattered wave, are involved. When the third is an electron plasma wave the mechanism is called Raman scattering. If alternatively an ion-acoustic wave is excited, the mechanism is called Brillouin scattering. The names come from analogies with effects known previously in molecular and solid state

IL. Bobin, High intensity laser plasma interaction

243

physics. Other types of scattering, for instance on random density fluctuations (resistive quasimode scattering) will not be reviewed here. Growth rates and thresholds can be calculated using the procedures of section 22. There is no resonance for Brillouin scattering. The growth rate depends on the wave vectors and turns out to be proportional to (kacIko)~2. Now, the k matching, illustrated on fig. 26.1, is such that kac/ko, hence the growth rate is maximum for backscattering.

k~
Fig. 26.1. k-matching for Brillouin scattering. k

$
,,

0~

k~I k~CIis angle dependent with a maximum kacI = 2~ko~ for

backscattering.

Raman scattering is resonant and turns out to be an absolute instability at the quarter critical density where 2We. (26.1) (00 2Wp 2W~ Homogeneous thresholds and growth rates for backscattering are cast into the simpler form used in section 22. Hence the table 26.1 which is completed by the values corresponding to a purely growing ion acoustic mode (c. and Wp are given in (5.2) and (23.7) respectively). In section 21 three wave couplings were introduced using a simple physical model. A similar picture with a local ponderomotive force can also be set up for backscattering instabilities. Going back to the general expression (11.4), it is then obvious that all terms in (n . V), i.e. (E V) vanish. The ponderomotive force is thus entirely due to the surviving v X B terms, F~=~(v 0XB~+v~XB0).
Table 26.1 Backscattering instabilities growth rate convective Raman absolute vo c Brillouin purely growing mode
(~)i/2

(26.2)

~ threshold

(wou~)/2
C

V.,

V.,

(OO)p

w~

c
V.,

(F~(v~/c)/3 Vei)~

3 ~V2 w f Vo w~~Y COO

2 (~) ~ 20,. i.,,

w0w~V3 1~wov., m 1

244

fL.

Bobin. High intensity laser plasma interaction

Fig. 26.2. Ponderomotive force resulting from v

XB

terms.

The electromagnetic waves do not induce charge separation for density perturbations in the direction of the k0 and k~ vectors. An analysis of the phase relationships resulting from (26.2) [113] shows that F0 contributes to the growth of the perturbation. When the density perturbation is purely electronic, one gets stimulated Raman scattering. When the perturbation is due to an ion acoustic mode or to the density profile modification associated with standing wave patterns as found in section 14, one gets stimulated Brillouin scattering. When the fluctuation is not associated with a normal mode of the plasma, instability may still take place. Indeed the field of the pump wave can be large enough to maintain the density perturbation against diffusion. Furthermore if the phase velocity of the perturbation is the electron or ion thermal velocity, there can be an unstable interaction with resonant particles. One then has electron or ion stimulated Compton scattering. The relevant growth rates and threshold are given in [130]. A strong backscattering goes together with a poor energy coupling between the incident laser beam and the plasma. Accordingly backscattering in a density gradient received a great deal of attention. Extensive investigations have been performed on this specific problem, see e.g. [131, 132]. The propagation equation of an electromagnetic or an electron plasma wave contains a term in w~ which is independent of the coordinate if and only if the plasma is spatially uniform. When instead there exists a density gradient with the characteristic length in the direction of propagation, the term has to be replaced by w~(1 + x/L) which is also the first-order expansion of an exponential density profile around a reference density. Then
1 ldN

Nd

(26.3)

The substitution has to be made in equations like (22.1). Now, waves transport energy with their group velocity Vg. The time necessary for leaving the coupling zone in which the selection rules are fulfilled, is

r,=J~.

(26.4)

t. The threshold is expected to raise accordingly. This is the same as a supplementary damping rT Furthermore, the wave vector matching conditions hold only locally. Then one has a space dependent phase mismatch,

IL. Bobin, High intensity laser plasma interaction

245

K = ko(x)

k1(x) k2(x).

(26.5)

Assume that waves 1 and 2 have their k vectors in opposite directions along the k-axis, and that the pump has a constant amplitude. Equations are to be replaced by
x
~-+ vi,.~-+yiai=yoa2exp{iJKdx}
ox

Oa1

Oa1

(26.6)

0a2
+

Oa2 V~+y~a2 yoaiexp{iJKdx}

with V1 V2 <0. In order to investigate this system, Liu, Rosenbluth and White [131] used one of the most successful methods of plasma physics. Laplace transforming in the time and setting

at(p) = a1 exp{~ K dx),

a2(p) = a2 exp{

K dx),

(26.7)

one gets the algebraic system in 12 (p+y1)a1+iO1+ V1= yoa2+1(0) 2 Ox (p+y2)2+i~a2+


V2~_2= yoOi+2(0)

.KV1

0a~

(26.8)

choosing 1(0) = a2(0) = 0 and eliminating a2 yields a=atexP( 2a O Ox2 1~ 4L P+yt V1


. + P+Y2

V2

x )~ (26.9)

~p+y~ P~Y2\]2 / yo~ ,dK\ ____ II a+i \ V~ V 2 Ii \ V1 V2 dx I

The second equation (26.9), is of the well-known type 2a/Ox2 + ip(p, x) a = 0. (26.10) O Once p is known, the wave amplitude is calculated by reversing the Laplace transform. A necessary condition for instability is the existence of a solution a(x) for Rep > 0. More precisely, (26.10) should have a well-behaved solution around a pole Po with Re Po >0. In the general case (26.10) can be solved only approximately (Liouville [133]).It turns out that the

246

fL. Bobin. High intensity laser plasma interaction Table 26.2 Instability Density range
n~~ T 0 2~1/2

Growth rate n., 4


<

Threshold
/Vo\ 2
~)

Raman backscatter

convective
absolute

4\mc n -. n.,/4 n., T

<n

V~

k11L>

(~) (knL)4> I
/2

Bnllouin backscatter

(:2)

<n <n.,

(2cv~.,)2

VoWp~

(~) (~) k
- Wp

0L>

density profile, through the x dependence of K, plays an important role in determining the solution, as initially shown by Rosenbluth [134]. In [131] W.K.B. approximations were calculated for various assumptions concerning K(x) (see also Pellat [135]). A temporal growth rate is found when the pump intensity exceeds the absolute threshold (22.19) provided K is a quadratic function of x, table 26.2. From the saturation value of the instability, one can deduce a reflectivity coefficient for the incoming laser light. This parameter can also be very crudely evaluated [136] in the case of a slab of uniform underdense plasma with density n <n., and length 1. Assume the density perturbation, associated with the ion acoustic wave, is limited in amplitude by ion trapping. The condition for the average ion to be brought to rest in a frame moving with the wave, is
Zeq5 +

(c. v\/3)2 =0

(26.11)

where ~ is the potential that ions undergo in the wave. The corresponding value of the relative density fluctuation is from the Boltzmann law:
Z/TC = ~(VZ+ 3T1/Te
V3T1/Te)2.

(26.12)

Now in steady state, equations of the type (22.21) reduce to dE0


dx

=aE1,
n

dE1
dx

aE0
n

~n

(26.13)

in which E1 is the electric field amplitude of the backscattered wave, and a is


2

(26.14)

a = 4woc(l n/n) Then going to second-order derivatives in (26.13), and assuming that E 1(l) is much smaller than E0(0) one gets 2
=

tanh2(a ~i).

(26.15)

IL. Bobin, High intensity laser plasma interaction

247

In the case of an inhomogeneous plasma with gradient scale length L, Max and Estabrook [132] obtained an implicit formula for the reflectivity 91: 91(191) ~ exp{Q(1 91)} 91 E~ where E~ is the noise level of the backscattered wave and
=

(26.16)

(L~(fl~

E~ ~

(26.17)

\A0I \n0I 8fleTe Fac 1 + 3T1/ZTe 1 n/ne

The calculations for threshold also apply to decay type instabilities. Using a slightly different approach, Perkins and Flick, in an earlier paper gave threshold conditions combining homogeneous and inhomogeneous parts [138]. Their results are summarized in table 26.3 (Pe and i.~are electron and ion collision frequencies respectively; yj = 1 + 3T1/Te).
Table 26.3 Instability purely growing mode at Threshold conditions 2 / Te\ /41., 2 >11 + II + E 4srnT \ T/\w 0 k0L 8y~/ p \ 1/2 I I + 3.2 E > 4irnT k 0L ~ ~

n.,

parametric decay at n.,

27. Filamentation revisited So far, the main concern was Brillouin backscattering. This process involves ion motion (oscillations). On the other hand, it was seen in section 19 that coupling electromagnetic waves to fluid motion via the ponderomotive force leads to a solution in which light propagates obliquely along filaments with finite width. It is readily inferred from eq. (19.5) that a lower density is associated with the filaments. Now, transverse standing ion acoustic oscillations may act as a diffraction grating bending the trajectories of the light waves. The corresponding basic mechanism is shown on fig. 27.1, in which the standing modes actually couple 4 electromagnetic waves. This is just another way of looking at the origin of filamentation. The relevant equations are naturally derived from (24.10). The ion acoustic frequency ~ac neglected so that the time dependence of both the upgoing and the downgoing density perturbations (5n)~
Nac(x, t)

exp(ikacy),

(Sn)d =

N~,(x, t) exp(ikacy)

(27.1)

is ignored. The amplitude of the pump as well as the density fluctuation depend upon both k1 and k~ scattered waves. The equations are then

248

J.L. Bobin, High intensity laser plasma interaction

Fig. 27.1. 4-wave coupling through standing ion acoustic modes.

(-~-+ V~

r0) F0 = i Ko(NacEi
=

N~E~)exp{i(z~k x + z~w t)}

(f+ V1-~--+r~)E1iK1N~E0exp{i(Akx+z~wt)}
(27.2)

(~+ V1~+F1)E~ = i K1 NacEoexp{i(L~kx +~wt)}


Nac = -

[E0E1 exp{i(iXk x + ~w t)} + E~ E~ exp{i(z~k x + ~w t)}].

Substituting the last one into the others yields the following system for E0, E1 and E

(-~-+ %7~

r0) E0 = i go[JE1~ 0 +
2 E
=

2E

2E !E~I

0 + 2E O*E1E~exp{2i(~k x + ~w t)}] (27.3)

(-~-+ V1 -~+ r1) E1 i g1 [!E0j 1 + E~E*exp{-2i(~k x + ~w t)}] (-~-+ V~ -~+ r1) E


2 E~]
=

ig1[E~E~ exp{2i(~k x + z~iwt)}+ JE01

in which the non-linearities are now cubic, just as it was in the non-linear Schrodinger equation used in the description of section 19. Again, a similarity transformation is to be used together with real variables: amplitudes and phases. Let thus
=

Ut,

U =

z~w/L~k, 5 = ~k,

E~exp(2i5~) = a

1, and the complex amplitudes E1a1 be replaced by E0


=

(27.4)

a0 exp(iqSo),

E1

a1 exp(i~1),

a1

a exp(i~);

(27.5)

as usual the relative phase is 02q~0q51q5~. Let us first look at the subsequent first-order ordinary differential equations in a1 and a~: (27.6)

IL. Bobin, High intensity laser plasma interaction

249

(v1

u) da1Id~ = F1a1 g1a~a~ sin ~ 2 (v1u)da~Id~ F1a~g1a 0a1sinO


(277)

They are readily combined to give


2) = V1 U (a~a

2F

1 (a~ a~

2)

(27.8)

which shows that the difference between the two amplitudes goes exponentially to zero. We may then safely take up identical amplitudes and phases for the waves k 1 and k~,and concentrate on the following system of three ordinary differential equations 2 sin 0 (vo u) dao/d~ = F0a0 + 2g0a0a (vu)da/d~=Fagao~a sin 0

(27.9)

dz

VU

V 0U VU

The case with no linear growth nor damping was investigated in detail by Bingham and LashmoreDavies [139]. Periodic solutions were found for u 0. They are expressed in terms of Jacobian elliptic functions a~) = a~(0) + 2g0(V Urn)

[a2(0)-

Urn]

sn2[13(u)

~t(u)] (27.10)

(VoUm)

a2(~)= a2(0)_ [a2(0)_

Urn]

sn2[$(u)~(u)]

2/d~ = 0, and $(u) and in which Urn is the intermediate zero of da 1a(u) are combinations of the three zeros of this equation. In the special case u = 0, i.e. when the density corresponds to resonance conditions for the pump, one gets as solution 2(/3 a~) = a~(0) +2 ~ a2(0) tanh 0~)
gV0
.

(27.11)

2(/3o~)
a2(~)=

a2(0) sech

This is another way of describing the cavity already found in section 13. Now, assume linear growth and damping are accounted for in the equations, namely the pump has a linear growth F~(<0)in a reference frame moving with its group velocity, as in the trailing edge of a laser pulse, whereas the wave a undergoes damping 1>0. Let then
y
= 1/10,

a~,

2a2

(27.12)

normalized in such a way that we get the dimensionless system:

251)

IL. Bobin. High intensity laser plasma interaction

dp!d~=2p(1+qsinO)
dq/d~=2q(-y+psinO) d0/d~ = 2[ S
+ .

(27.13)

(q p) (1

cos 0)]

This dynamical system can be investigated in the parameter space y (>1, a necessary condition for volume contraction in phase space), 8 (<0, anti-Stokes forward scattering). For given 5, increasing y causes the solution to evolve from tending towards a limit cycle to complete chaos through a set of period doubling bifurcations which obeys Feigenbaums [140]universal behaviour: fig. 27.2. In most of the parameter space, the system goes to a more or less complicated limit cycle, i.e. tends to behave periodically, although often in a rather involved way. In well-defined narrow domains which are not yet all identified, the period is infinitely long and p, q and 0 vary chaotically along a so-called strange attractor. The trajectory in p, q, 0 space is perfectly determined at any time provided the calculation has been performed with the required accuracy. However the behaviour is unpredictable from the knowledge of the initial conditions. For instance the magnitude of the variable q exhibit a random sequence as shown on fig. 27.3. Since 0 also varies randomly, it means that the relative phase between waves is lost. The interaction is no longer coherent. The beam breaks up. This phenomenon also occurs in the laser itself, the linear growth coming from the gain of the active medium. Small scale self focusing may be thought about as guided waves [141] for which the above analysis may apply. Experimentally in laser media with high second-order refractive index such as neodymium glass, one observes the initially coherent beam exploding into a large number of apparently point-like incoherent sources [142].

Fig. 27.2. limit cycleTransition with twiceto the chaos: period, 8 is (c) fixed. y3> (a) Y2 strange yi limit attractor. cycle, (b)

Y2>

y~

~~lll~~II lll~I~~l~IIlIl Fig. 27.3. q(t) for 8 I~l~~lllllllI~ 2. y 12.

IL. Bobin, High intensity laser plasma interaction

251

28. Laser interaction with an inhomogeneous plasma flow In laser interaction with an inhomogeneous plasma, many mechanisms may take place. Some of them were actually evidenced. However none of them sets up independently from the others. Couplings, some of which are to be discussed in part IV, are possible. Also competition between non-linear processes is expected. The overall absorption of course depends upon the result of the competition for given maximum intensity and pulse shape. As it was shown in this review we have on the one hand, ponderomotive force effects which are proportional to the light intensity and tend to steepen the density gradient around cut-off. On the other hand 3-wave couplings either absorptive or reflective, also are increasing functions of the intensity, but their threshold is gradient dependent. A steep gradient rises the threshold and eventually inhibits the instability. Our concern is thus with the inequality k0L> 1 in which the gradient scale length L results from ponderomotive effects coupled to gas dynamical motion. It was shown in section 11 that in the whole radiation driven structure the inequality PLIP0<i
(28.1)

in which Po is the fluid pressure in the overdense region holds. In the underdense shelf the ratio is very small. At the sonic point (with ~ 1) one has after (11.12)
ppL/2p012

(28.2)
-~-~ Po12

4Po implies p which for P1. PL/P<i.

and accordingly
(28.3)

This is also i~ratio of the pump energy density to the electron thermal energy density in the plasma (see (22.31)). Consequently low order expansions of the type used throughout part III are justified in the underdense regions of the laser induced plasma flow. The situation is different around the critical density. Indeed in this vicinity, the electron temperature is smaller than in the shelf. This compensates the increasing density. Furthermore, the electric field squared pattern, as described in section 11, exhibits field swelling close to cut-off even in the absence of resonance. Of course, in case of resonance 2 E2/v2 may take exceedingly high values. The ratio PL/P thus can be close to 1. Now, the profile is determined by the balance of forces: radiation pressure force versus gas dynamical pressure force. Then one has PLL/PA

-= 1

(28.4)

i.e. the gradient scale length compares with the wavelength and k 0L -~ 1. It results from the formulas of tables 26.1 and 26.2 that when k0L which is the same as L/A decreases the threshold value (V~ -~ Eo~ Jo) increases accordingly. The effect is even stronger in the absence of flow, i.e. when the light impinges onto a surface (wall). Then the field energy density gradient is evaluated after the skin depth ~ ~ w):

252
~ / S= ~)
(Op (1)

IL. Bobin, High intensity laser plasma interaction


2P\
1/2

(28.5)

The balance between forces requires ---1 8rrn~T S

E2L

(28.6)

and since at the same time, absorption is very weak, the fluid pressure is much less than the field pressure so that one has L~8<A, (28.7)

and it is expected that no parametric instability can take place in such extreme conditions. Harmonic generation experiments (see part IV) have shown that this is actually the case. Coming back to the IA2 scaling of fig. 5.1, we may now distinguish between domains where a given type of interaction is dominant. This is done in table 28.1. It is also evident from fig. 5.1 that the slope 2/3 indicates the special importance of the cut-off density whenever IA2 is between 1012 and 1015 W p~2cm2. Then the particle velocity for the critical density is the ion sound velocity which depends on the electron temperature (the higher one in case of a two-temperature distribution). The region in which parametric instabilities are dominant is also characterized by two-electron temperatures. The electron distributions are strongly perturbed. Indeed laser induced longitudinal waves trap electrons whose velocity is close to the phase velocity. These electrons themselves thermalize yielding the high temperature component. Following Kruer and Dawson [123], one may then define an anomalous heating rate v*Eo2/87r. Now the energy transfer to the electron plasma waves takes place at a rate

-~-

dt4i~

(~)
=

(28.8)

where y is the oscillation growth rate. When the instability saturates, one gets the effective collision frequency 4yE~/E~ (28.9)

which can be used in crude gas dynamical calculations, i.e. without analysing the structure of the profile. Since the maximum velocity of a trapped electron is 2(eE~/mk)12,the number of trapped electrons
Table 28.1 J,~2
<lOt2W ~e2cm2 1O2<1A2<3x 1013 W~cm 3 x 1013 < j~2 < io~~ w ~ cm2 1Q15 W ~ cm <

Dominant process volume collisional absorption collisional absorption at cut-off parametric instability at cut-off f ponderomotive force

Sonic point

if at cut-o underdense

IL. Bobin, High intensity laser plasma interaction

253

increases with Ee. It may occur that the entire distribution be trapped and then no low temperature component is left. It turns out in experiments [21] that the low temperature component disappears for laser intensities slightly above 1014 W p~2cm2. Beyond 1015 W p.2 cm2, profile steepening is the dominant effect. The corresponding flow profiles are those described in section 14. The sonic point at the end of the underdense modulated shelf, has a density smaller than critical and with no simple proportionality to the wavelength dependent cut-off value. Parametric decay is strongly inhibited. Two plasmon and absolute Raman backscattering may occur if the density of the shelf is higher than quarter critical. Brillouin backscattering is allowed whatever the density of the shelf is. It was already noticed in section 14 that the modulations have twice the wave number of the incoming light. This is very favourable for the onset of the Brillouin instability, especially in the case of long pulses.

29. Wave breaking High laser intensities drive longitudinal waves which may grow under conditions which were discussed above. One also has in case of oblique incidence with the electric field in the plane of incidence, resonance propagation (section 12) which is a different way of driving large amplitude longitudinal oscillations. In such oscillations, wave breaking may occur and contribute to the generation of high energy particles (electrons) escaping from the plasma. Let there be a longitudinal plane wave in which an electron oscillates around an equilibrium position x 0. At any time t the electron is in x(t)= x0+ h(x0,
t).

(29.1)

Since h depends on x0, electrons coming from different locations x0 and x1 may be spatially inverted as shown on fig. 29.1. This is the origin of wave breaking. Assume first that it does not occur yet. At time t, the oscillation creates an excess of negative charges for h >0 en0h

an excess of positive charges for h <0

en0h.

The resulting electric field is given by Gauss theorem E=4iren0h


(29.2)

h(x~,~) x xo
Fig. 29.1. Spatial inversion of

xl

electrons in a longitudinal oscillation.

254

f.L. Bobin. High intensity laser plasma interaction

hence the equation of motion d2h/dt2 = eE/m whose solution is h(x 0)= h~~cos(w0t+t/). The spatial order of the electrons is conserved whenever the following inequality is fulfilled: h(x0) h(x0 + ~x0) < ~x0. (29.5) (29.4)
=

w~h

(29.3)

Expanding the left-hand side, one finds in lowest order the condition (first given by Dawson [143]): dh/dx0> 1 dh/dxo < 1 no wave breaking wave breaking (29.6)

Now, consider the resonance peak dealt with in section 12. The resonant field is 5)
E
=
-~-~

(29.7)

cos(wt + q

where F

0 is the pump field defined by (12.14). The dielectric constant is


(29.8)

2/L~+ (1 x/L)2 v2/w2]112 [x assuming the peak is at the origin and x increases with the density. For the phase one has
=

~(1+x/L) tang*~=

(in/IL

w cos~=,

t x sin~=__(i.+_). w~ L

(29.9)

The electron equation of motion is d2h/dt2+


p

dh/dt+ 4h

eE/m

(29.10)

whose solution h= eEP() m~w allows one to calculate Oh Ox 2eE


m~2w2L2c05~t+~

(29.11)

(29.12)

IL. Bobin, High intensity laser plasma interaction

255

At the resonance peak maximum, q~ = irI2 so that the minimum value of Oh/dx is obtained for wt = 3ir/2. The wave breaking condition is thus
= eE~/mv2L2> 1. eE~/m~2w2L2

(29.13)

In such conditions (29.11) shows that Oh/Ot is negative which means that, should wave breaking occur, electrons escape on the low-density side. The maximum absolute value of the oscillating electron velocity is Oh/OtI = eE~w/m~2 and in phase space one gets the graph shown on fig. 29.2. (29.14)

-/

Fig. 29.2. Wave breaking in electron phase space.

In case of wave breaking, part of the electrons oscillating resonantly lose the phase relationship with the driving wave and escape towards lower densities. They leave the critical zone with the kinetic energy corresponding to the velocity (29.14). It can be very large for weakly collisional plasmas. Furthermore the wave breaking condition (29.13) depends on the density squared through the factor p2. Since it occurs at cut-off, the threshold scales as A4 when Te is given. The above discussion applies to any longitudinal wave excited at resonance. So far L was arbitrary. Actually the ponderomotive force associated with the wave induces profile steepening. Accordingly a smaller L tends to enhance wave breaking. The combination of both effects was investigated in numerical simulations [144,145]. The qualitative behaviour is basically the one shown on fig. 29.2. Profile steepening also increase fast electron losses. The resulting charge separation field may also drive fast ions out of the plasma. Some connection between fast ion detection and wave breaking can be inferred from the comparison of experiments with laser light impinging either onto solid surfaces or onto gas jets. In the former case strong density gradients are present favouring the wave breaking. The fast ions show up in most of the experiments. In the latter case on the contrary, density gradients are gentle, inhibiting wave breaking. No fast electrons are observed [146].

IV. Harmonic Generation


30. v X B non-linearities Harmonic generation in a plasma was predicted very early by the theorists of non-linear optics [11]. Indeed the v x B term in the electron equation of motion naturally generates harmonics of the

256

f.L. Bobin, High intensity laser plasma interaction

impinging electromagnetic frequency. Consider a gas of free electrons. Each one undergoes the Lorentz force. The equation of motion is
/ v dr d2r m~=e(E+xB)mp. di \ c / di
~

(30.1)

Let E be along the y-axis and propagate along the x-axis: E~ = B~ = E exp{i(kx Neglecting r(w) =
v/c wt)} +

E* exp{i(kx wt)},

k=

(0/C.

(30.2)

(~1)in (30.1) one gets the linear approximation in steady state


eE exp{i(kx
wt)}

m(w +lvw)

y,

Jy~= 1

(30.3)

If this solution is substituted into (30.1), the lowest order non-linear approximation is
r(2w)=
l

e2E2 exp{2i(kx wt)} x, m c(4w+2iv)(w +lvw)


2 2

x~ = 1.

(30.4)

A further substitution yields r(3w) polarization is then

or j3

and so on. For a plasma with electron density n 11


~

n~,the

21+ (30.5) P= noer(w)+noer(2o4+ .. . =xt(w)E(w)+x2(2w)E in which the susceptibilities at any order are readily determined after (30.3), (30.4) and similar higher order equations. One striking feature is that all even harmonic motions r(2w), r(4w),... are along the x-axis, i.e. along the direction of propagation. Therefore, in a uniform plasma, no radiation will be generated at these frequencies. Only odd order harmonics are permitted. In order to get even order harmonics, a density gradient along the x-axis is necessary. Let then the continuity and the momentum equation for the electron be linearized as in section 13 around a density profile N(x) and a null initial velocity: aN +V~VN=0,
=-(E+xB).
Ot C

Ov

e m

(30.6)

The perturbed quantities n, v, are expanded up to second order


flflt+fl2+~,

v=vt+v 2+~~.

(30.7)

In second order one thus gets a current density [147] J 2 [N (VN~E)Ei e 2=Nv2+n~v1=l VE+ mw 4 lW~/W i
2 2 2 2

(30.8)

IL. Bobin, High intensity laser plasma interaction

257

Both terms in the right-hand side oscillate with the frequency 2w. The first one which comes from v x B in the Lorentz force is identical to (30.4) but for the damping. It is still purely longitudinal and cannot radiate at 2w. On the contrary, the second term gives rise to the harmonic 2w provided one has an oblique incidence. Furthermore it is resonant at cut-off (w = wv). Similar considerations apply to all even order harmonics. Thus the whole series of integer order harmonics is expected when a laser beam impinges onto an inhomogeneous plasma provided the gradient is at an angle with the wave vector. However, the v x B force is not the only one to participate in harmonic generation. It turns out actually that coupling with longitudinal waves is a much more efficient process.

31. Oscillating mirrors Another simple way of accounting for harmonic generation in laser plasma interaction considers phase variations when light travels up to the critical density (mirror point) and back. Non-linear couplings of electromagnetic waves to longitudinal modes amplify the latter which induce significant oscillations of the mirror point. Figure 31.1 represents the situation of a linear profile with length L. Near the critical density, the profile is perturbed under the influence of longitudinal plasma waves with frequency we and/or ion acoustic waves with frequency Wac. Both are generated in the parametric decay process. The density near cut-off is
N = N~ + ~fle cos w~t + ~flaC
COS Wart

(31.1)

where the ~n denote the perturbation amplitudes. Part of the incoming laser beam (normal incidence) is reflected at cut-off. At the point x = L it goes out of the plasma and with respect to the incident wave, its phase is delayed. Denoting by ~(x) the refractive index, the phase delay is for a linear steady profile

~o=-~J

~(x)dx=-~L.

(31.2)

Now, if instabilities at cut-off are accounted for, two things happen. First the index of refraction is

Fig. 31.1. Geometry of moving mirrors.

258

IL. Bobin, High intensity laser plasma interaction

modified and is now approximately


=

[a~ COS Wet + aac COS

Wact

(1

ae COS Wet

aac

cos Wart)]

1/2

(31.3)

where
= ~fle/Ne,

aac

~flac/Nac.

(31.4)

Second, the upper boundary of the integral (31.2) is oscillating so that

(31.5)

in which ~ is given by (31.3) and ~ix= ac COS Wet+ aac ~OS Wart.
(31.6)

Integrating (31.5) yields ~ as a function of q5o and of the different oscillations of the electron density. The exponential phase factor contains a trigonometric function. As usual it is expanded in terms of Bessel functions. The time dependence of the outgoing wave is accordingly of the form [148]: exp{i(wt+ b)}= exp{i(wt+ bo)x} x
m~.oo~Jm

J~(ae~o)exp{i Ifl(We_

Wac)t

~J}
(31.7)

o)exp{_i (mwact m

~)}.

The reflected wave is thus dressed with all the harmonics of the slightly red shifted incident frequency. Every harmonic (whose orderis an integer) is itself split into ion acoustic satellites. In (31.7) the J 1 are Bessel functions whose magnitude generally decreases rapidly when the order I increases. One finally gets the spectrum shown on fig. 31.2. The same kind of analysis can be applied to Raman backscattering which is an absolute instability in the second non-linear zone near N~/4 where the (moving) mirror point for w/2 is. The location of the

-~

3w

Fig. 31.2.

Harmonics induced

by parametric decay instability at cut-off.

IL. Bobin, High intensity laser plasma interaction

259

source at w/2 oscillates under the influence of both, the Raman effect itself, and the two plasmon parametric resonance. The phase of the backscattered light is then
I,
I,

wi
=

ci

~(w) dx

wsf

ci

q(w~) dx

(31.8)

-L

where w. is the scattered frequency (w12) and 1 = 3L/4+ Sx. (31.9)

Both terms in the right-hand side of (31.8) are at the origin of harmonics with frequency w~I2 with different amplitude distribution as shown on fig. 31.3. The resulting overall spectrum is the superposition of spectra given by phase factors (31.5) and (31.8). It has integer harmonics nw and half integer harmonics pw/2 which may be split into satellites, the whole picture originating in longitudinal modes of the plasma.

O~

3~t2~

o ~

Fig. 31.3. Amplitude distributions of hannonics for the two terms in the right-hand side of (31.8).

32. Raman upconversion. Coupling and cascades The oscillations at the origin of moving mirrors are longitudinal and come from laser interaction with the non-uniform plasma. In this approach harmonic generation appears as a by-product of all mechanisms by which the incoming laser light induces longitudinal oscillations in the plasma. Now, it is well known that effects such as Raman and Brillouin scattering go both ways: Transverse E.M. wave~Longitudinal wave + Transverse E.M. wave. (32.1)

So far, only the direct process was considered except in section 26 where inverse processes had to be accounted for in the calculations dealing with incoherent regimes. From the viewpoint of harmonic generation, the inverse process of Raman scattering, i.e., Raman anti-Stokes upconversion deserves a special emphasis. Only resonant situations are of interest here. The basic schemes are shown on fig. 32.1. More generally, all resonances are of the type

260

fL. Bobin, High intensity laser plasma interaction

u+i

--~-~--

N+2 plasmons N plasmons N-i plasmons ~

a)

~~Up

(b) (J~2(.Jp
0/2.

Fig. 32.1. Raman anti-Stokes transitions leading to the generation of (a) harmonic 2o~.(b) harmonic 3w

w~=nw~,

w~(n+1)w~w0. n

(32.2)

Now, the unconverted wave with frequency w~.may also resonantly interact with electron plasma waves in the inhomogeneous plasma giving rise to new frequencies: (1 + p/n)wo, (1 + 1/n) (1 + I/p)wo, ii and p being integers. In such cascading harmonic generation, the intensity decreases rapidly as n and p increase. All these interactions involve a coupling with an electron plasma wave, i.e. since these originate from instabilities, combinations of non-linear processes. The first possibility comes from the resonance peak described in section 12. When the electric field or an obliquely incident electromagnetic wave is in the plane of incidence, the longitudinal component induces a resonance peak at the critical density, thanks to tunneling from the classical mirror point. The subsequent coupling of the transverse wave to the longitudinal oscillations generates the harmonic 2w~.The linear conversion mechanism was investigated theoretically [149], for the harmonic generation both far from and near to the resonance peak. It turns out that the latter is dominant, in agreement with recent experiments [150] which show characteristic variation of the 2W~intensity as a function of the angle of incidence. For S and P polarization the behaviour is very similar to the one shown on fig. 12.5. Longitudinal waves are also excited in three wave couplings, parametric or reflexive. Figure 32.2 presents the basic couplings that are expected to take place in laserplasma interaction. Resonances occur at different locations in the inhomogeneous plasma: near the quarter critical density for the
~ VVW1 A/i.) t

~ 2 ~~2W0~man

2 Roman upconversion with 2 Plasmon resonance upconversion


Rarnan

,,.

scattering

k
0

CJOi!Wi.fV1M/iMAL)~a2(,J0..(

Roman upconversion with parametric decat~

k~ k

~~/

L)~wv~M~pvvw ~
w0vtivvu~QQ---Q--Q

Roman upconversion

with cascade

kR

Fig. 32.2. Couplings of Raman upeonversion to laser induced instabilities. is a photon, . is a plasmon, is a phonon.

Fig. 32.3. k-matching in 3o~/2 generation. (a) Coupling with 2 plasmon resonance; (b) coupling with Raman backseattering. The angle between the two k vectors depends on the local density.

IL. Bobin, High intensity laser plasma interaction

261

coupling of Raman upconversion with either the two plasmon decay or the absolute Raman scattering. The selection rules in k together with the linear dispersion relations imply a one-to-one correspondence between the local density and the angle of the two incident electromagnetic quanta involved in both coupled mechanisms. The situation is illustrated on fig. 32.3. Obviously focused beams accommodating a large range of relative angles near the focal point, favour the 3w~/2harmonic generation. Now, the dispersion relation for an electron plasma wave with frequency w close to ~ and wave vector k, parametrically excited by a laser wave with frequency w~ close to 2w~,and wave vector k0 is [151] w = w~/2 + ~(k . ko) v~/w~.
(32.3)

In the corresponding two-plasmon decay illustrated on fig. 32.4, the longitudinal waves are emitted in almost opposite directions. Owing to the sign of the scalar product k . k0, the wave in the forward direction has a frequency
WB>

(i)o/2

(32.4)

whereas for the wave in the backward direction


WR <

w~J2.

(32.5)

The subsequent Raman upconverted lines are shown on fig. 32.5. The one built up in the forward

V9-~c
CJ~

x
Fig. 32.4. Asymmetric 2-plasmon resonance: k vector matching with blue (B) and red (R) components. Fig. 32.5. Raman upeonversion associated with the asymmetric 2plasmon resonance. (a) Mirror reflection of the waves at their cut-off density; (b) expected spectrum in the backwards direction.

262

fL. Bobin. High intensity laser plasma interaction

direction is blue shifted with respect to ~ In a density gradient, it is to be reflected back when encountering its critical density. A red shifted line results from upconversion of the laser light back reflected at cut-off. Consequently, in laser interaction with a solid surface, the spectrum of the ~W( harmonic in the backwards direction is made of a doublet whose components correspond to the blue and red plasmon emitted in the parametric resonance. In recent experiments [152,153], laser interaction takes place on a thin plasma layer in which the maximum electron density is below the critical one for ~wo radiation. Then the blue shifted component disappears from the reflected light spectrum. Furthermore, evidence was found experimentally [154, 155] and computationally [97] that two plasmon decay occurs in filaments and the ~W() light source has a filamentary structure. A model was recently proposed to describe ~wo emission from filaments [156]. Experimentally observed asymmetric spectra and their angular dependence are thus satisfactorily accounted for. Many processes also contribute to 2w~harmonic generation, besides the already mentioned linear conversion. The coupling of Raman upconversion with parametric decay occurs near the critical density. The upconverted line is red shifted by Wac with respect to 2w~. The geometry of wavevector matching is shown on fig. 32.6. The coupling of the two 3-wave mechanism may take place in a parallel laser beam. So far Raman upconversion was considered as a three wave process. Now, when instabilities occur in laser plasma interaction, they may produce coherently ion acoustic waves: parametric decay at cut-off, Brillouin scattering in the underdense regions. These ion acoustic modes also take part in stimulated 4-wave upconversion as described on fig. 32.7. An electromagnetic wave and a longitudinal plasma wave combine to form some virtual intermediate state which in turn decays into an upconverted electromagnetic wave and an ion acoustic one. A strong background of ion acoustic waves should be present. An important effect of the ion acoustic quantum is to relax the wave vector matching conditions in the case of ~ generation, by carrying away some extra momentum. Although in the coupling of 3-wave upconversion to any other 3-wave mechanism, there are two entering quanta and two outgoing ones, this is not a 4-wave process. The non-linearities remain quadratic. On the contrary the superelastic anti-Stokes scattering of fig. 32.7 as well as the filamentation exhibit cubic non-linearities. Consequently its behaviour can be predicted by reference to the analysis made in section 27. (a)
photon~.~~~ plasmon
~~J_J~J

photon

~~phonon

(b)
4
,47

(c)
N+2plasmons

N~i N ptasmons_L...... p(asmOflS~~-~-..~........PhOflOn N-i

~ i

~ /

/kac

N-i

plasmons

ptasmons

k0 k0
Fig. 32.6. k-matching in second harmonic generation by Raman upconversion coupled to parametric decay. Fig. 32.7. 4-wave upconversion: (a) scheme, (b) transition at w (c) transition at w 2w~.

IL. Bobin, High intensity laser plasma interaction

263

33. Growth rates and thresholds Whenever Raman upconversion is at the origin of harmonic generation, it comes after another instability. It is thus expected that the growth rate depends on the properties of the initial process. By the same token the threshold value is at least the value found for the latter. Consider first the coherent coupling of three wave mechanisms with a constant pump and negligible damping. The relative phase is fixed so that one has only three equations dau/dt = Kuaoae dae/dt = Ke[aoan

aoa~]

(33.1)

da~/dt = Kaaoa~ where the a are the real amplitudes of the waves. The index 11 denotes an electron plasma wave (in 2w~) or an ion acoustic wave (in case of parametric decay at w~ w~) case of two-plasmon decay at w~ or an electromagnetic wave (in case of Raman scattering). Setting a = a exp y one readily gets either y = 0 (steady state) or (Kgj Ku)Kea~
(33.2)

which implies that in order to get an actual growth rate, the probability K~ of the first 3-wave process should be greater than the probability K~ of the subsequent Raman upconversion. It is usually the case. Incoherent processes can be investigated the same way. Then the reverse mechanisms are to be taken into account. This was done in [157]. The result is y = (2KD

K~)ao

(33.3)

except for an initial absolute Raman backscattering for which y is K~a 0V3. The 4-wave upconversion behaves differently. As in the filamentation investigated in section 27, one is left with only two equations with cubic non-linearities: da~/dt+F~a~ = ig~aoa~a~
daac/dt + racaac
=

(33.4)

igacaoaea ~

where the coupling constants have complicated frequency dependences given in [158]. The growth is
2}

~{(f~ + .fac) + [(fs 1~ac)2+ 4gugac]

(33.5)

provided the fields exceed the threshold a~a~ = FuFac/4g~g~~. Far from threshold, the growth is of the form
ycxlaoj~a~J
(33.7)

(33.6)

264

IL. Bobin, High intensity laser plasma interaction

in which lad is itself an exponentially growing function of time. In the non-linear phase ae (hence also a0 for this process) has linear gain. One may then expect some chaotic behaviour in harmonic generation for certain values of the growth and damping rates of ae and a~, aa. respectively. However there is no experimental evidence, so far, of such a regime. Experimentally, the generation of the harmonic ~Wo is the process for which the threshold was most clearly evidenced. Both Nd-glass laser and CO2 laser were used [159,160]. Typically the behaviour of the ~woline intensity is investigated as a function of the incoming laser intensity or energy (for given pulse width). The result is shown on fig. 33.1. Around a well-defined value (within at most a factor of 2) the harmonic line intensity jumps by several orders of magnitudes (3 to 6). Negligible at lower intensities, it becomes suddenly a sizable fraction of the laser flux (usually =-l%) and the subsequent dependence at higher intensities is in agreement with the coupling of three wave mechanisms described in section 32. Indeed in the overall process two incoming photons with frequency Wo combine to give rise to a single ~wo photon. Hence the intensity of the harmonic is proportional to the laser intensity squared. Such a dependence can be inferred [161] after the correct normalization of crude experimental data [162]. LflIH~

Fig. 33.1. Intensity

of harmonic line

near threshold.

34. Spectrum of harmonic lines Several mechanisms dealt with in the preceding sections predict harmonic lines which are shifted with respect to an exact multiple of the exciting frequency. This is for instance the case of Raman upconversion following parametric decay at cut-off. The result is a second harmonic line red shifted by some ion acoustic frequency. One may then expect that the spectrum of harmonic lines yield clues to determine which non-linear processes take place in the plasma. Furthermore in the above quoted case, the ion acoustic spectrum might be extracted from the line profile. However, real life is not that simple. Several effects indeed contribute to a given line. Actual spectra look very intricate. Unfolding them has not yet been satisfactorily done and may be an indomitable challenge as well. Linear dispersion relations contribute to the line shift. Consider Raman backscattering at quarter critical density. The waves involved in the process satisfy the following relationships 2, (V~= w~+ k~c2, w~ = w~,+3k~v~. (34.1) w~=w~+ k~c

IL. Bobin, High intensity laser plasma interaction

265

The scattered line is observable provided one has k. then has an absolute instability with k0 = ke = k and after (34.1)
2(c23v~) w~w~= k

0. The equality occurs at cut-off for v. = wi,. One

(34.2)

(34.3) (34.4)

u~ w~ = k2c2. Eliminating We yields


= ~Wo[1

~v~/c2]

(34.5)

i.e. a significant red shift which depends on the electron temperature. Such an effect has been observed [163].Consequently
We =

~w~(1 + ~v~/c2)

(34.6)

and the corresponding Raman upconverted line frequency is


(Vu =

(Vo +

We

~w~(1 + ~v~/c2),

(34.7)

i.e., a blue shift. Electron temperatures of a few keV are commonplace in laserplasma interaction. According to formulas (34.5) and (34.7), relative frequency shifts of a few percent are thus predicted. The actual experimental situation is far from being that simple. The occurrence and structure of the w 2 and the 3w~/2 lines are indeed indicators of the interaction mechanisms which occur in the 0/ non-linear region near the quarter critical density. In order to observe the subharmonic Wo/2, a high intensity is needed. When detected for the first time [164], the spectral resolution was insufficient to unravel any line structure. Specially designed recent experiments [165,166], on the contrary, do show the structure of the line which turns out to be an asymmetric doublet. It is evident from fig. 32.4 that the 2-plasmon resonance results in two longitudinal waves with opposite frequency shifts. Now, Raman scattering on the Stokes side might also be stimulated by such waves both for the incident and reflected laser light. After the wave-vector matching, the red shifted line in the wo/2 spectrum is attributed to Raman scattering of the incident radiation with the plasma wave with the higher frequency, whereas the blue shifted component is due to scattering on the lower frequency plasma wave, of the unabsorbed laser light reflected at the critical density [166].Accounting for the details of the line structure requires further theoretical studies. However, one may already state that the blue shifted peak should not extend beyond the Landau damping limit of the plasma waves (kA 0 0.3). Such a limited extension was actually observed in U.V. laser experiments [167]. Consider now the two-plasmon resonance with longitudinal waves emitted in almost opposite directions and the subsequent upconversion as shown on figs. 32.4, 32.5. The splitting of the doublet line around 3w~!2 observed in the back reflected spectrum was calculated by Gusakov [151]
--

266

iL. Bobin. High intensity laser plasma interaction

3w~3v~I where k

I6k~c2 (34.8)

1 is the component of the plasmon wave vector perpendicular to the direction of the laser light. As in (34.7) the shift is temperature dependent and its order of magnitude is very similar. Shifted harmonic lines are also broad. A first reason for that comes from the instability growth rates. Indeed let ~wo be the spectral width of the laser and 6 be the frequency mismatch in a three wave interaction. In section 23, it was seen that 6 defines the frequency range in which the parametric instability takes place. The relationship between 6 and the growth rate y can be put in a different form [168]and the maximum value for the onset of the instability is

r ~r
6rnax

FiT2

(34.9)

in which the F are the damping rates of the decay waves. Consequently the broadening of such waves is
L~w= (~w~+ 6~iax)u/2.

(34.10)

For the subsequent upconversion one gets


=

(~w~ + ~~2)u/2

(2~w~ + 6~iax)i~2.

(34.11)

In actual situations these broadenings are much smaller than the shifts associated with the dispersion relation. Another cause for comparatively large shifts and broadenings comes from ion acoustic waves which are part of the harmonic generation process. For instance Raman upconversion coupled to parametric decay at cut-off generates a second harmonic line red shifted by War as already shown on fig. 32.2. Experimentally, the 2W~~ line appears as an unshifted narrow peak with a sideband on the low frequency side: fig. 34.1 [169]. The rather broad sideband peaks at WM~~O.SW~t. ion plasma frequency at cut-off. This observation is consistent with Nishikawas prediction [114] of a highest growth rate for the parametric decay and represents the ion acoustic spectrum at cut-off whereas the narrow line at 2w~ ~5 due to the simple gradient created optical non-linearity dealt with in section 31. Furthermore on fig. 34.1, a satellite on the low frequency side is clearly apparent with a shift of about l.25w~.This corresponds to the cyclotron frequency of electrons spiralling in a 2.5 megagauss magnetic field. In such a field which might well be generated in laser plasma interaction (c.f. section 20), electron plasma modes are replaced by Bernstein modes [170] that subsequently participate in the upconversion. The point has been investigated in more detail by means of numerical simulation [171]. The case of the ~Wo line looks more puzzling. Indeed, the experimentally observed spectra are very intricate. One usually gets two broad peaks, one blue shifted, one red shifted. The intensities and shifts are either nearly equal (symmetrical doublet [172])or markedly different [156,173]. Fine structures were also evidenced [174].When observed for the first time and later on (depending on angle and laser intensity) a narrow peak at just ~ was obtainedwith a broad low frequencysideband [164].One is tempted to attribute a blue shifted peak to upconversion coupled to Raman backscattering, a doublet or an unshifted line to the coupling to two plasmon decay, a red shifted line to a 4 wave upconversion mechanism. Furthermore all these effects might be superimposed in the actual spectrum. A very murky situation.

IL. Bobin. High intensity laser plasma interaction

267

Brillouin backscaP~ering
a

.~o-(4

2u 0

~js~p-2&U 3t~ 2 T
Fig. 34.2. shiftsirradiated of (a) the Brillouin backscattered line, (b) the 3aw/2 line Red in a laser gas jet.

Fig. 34.1. Asymmetric broadening of the 2wo line.

However some clarity is being brought by specifically designed experiments. Consider for instance CO 2 laser radiation (10 p.m) impinging onto a gas jet [175].In such a situation, the whole laser created plasma is underdense. The maximum electron density is however larger than quarter critical. Thus the reflectivity is entirely due to Brillouin backscattering, with a mean red shift I~WR.The observed ~woline 2~WR from iwo: fig. in the back direction has only the red shifted component which peaks at a distance 34.2. It obviously comes from the upconverted backscattered light. The two plasmon resonance at n~/4 may be due either to the incoming or to the backscattered radiation. In both cases wave vectors in the backward direction can be produced either through 3-wave or 4-wave processes yielding the shifts (mac, 2Wae dominant), ~ ~~ac as shown on fig. 34.3.

Brillouin

k(s)

plasmon

Raman anpiSIokes
Fig. 34.3. Possible arrangements of wave couplings.

268

IL. Bobin, High intensity laser plasma interaction

35. Intensities of harmonic lines

In section 33, growth rates and thresholds were determined. Now the intensity of a given harmonic line saturates at some level which depends on the pump wave intensity. The saturation level is comparatively low. Indeed for instance, in a three wave process such as Raman upconversion, only a small part of the electron plasma wave goes to harmonic generation, an unlikely process compared to damping. Experimentally, it was found that the intensity of the strongest harmonic line is about 1% of the pump [134]. The dependence on the pump intensity is readily deduced from the coupling diagrams of fig. 32.2. It is a power law whose exponent is the number of incoming photons necessary to produce a single harmonic photon. In the case of Raman upconversion, whether it gives rise to the 2o or the ~ line, a square law is expected. Figure 35.1 shows the results of experiments in which the intensity is varied by defocusing the target. Then a constant flux impinges onto the plasma whose irradiated area S is such that 1 0S=const. Accordingly the harmonic intensity IHKI~S
1 K 0

(35.1)

(35.2)

and depends linearly on the pump. This behaviour was found to hold both for the ~w~j and the 2Wo line. In the latter case however, a kink in the experimental curve is observed when the pump is 2. In a more detailed analysis the spectrum was recorded at the same intensity time [176]: about 4 x 1013 w cm fig. 35.2. At intensities below 3x lOis Wcm2, a single broadened but unshifted line is obtained indicating a generation through the optical v X B non-linearity in a density gradient. For intensities between 3 x 10s and 3 x 10~ W cm2, the line splits into the unshifted component and a red shifted satellite due to Raman upconversion coupled to parametric decay. For intensities above 3 X 10~ W cm2, only the red shifted component survives. The gradient steepens inhibiting the v X B non-linearity, and also provoking a saturation at a lower pump level. The transition is not a sharp one and apparently takes place for intensities between 5 x 1013 and 3 x 10~ W cm~2. More information about the plasma and the interaction processes could possibly be inferred from the relative intensities of harmonic lines corresponding to different orders. In section 31, the oscillating mirror theory led to sets of harmonics whose intensities are given by squared amplitudes of Bessel functions. Consequently harmonic amplitudes are expected to decrease when the order increases.
~2

In 1
2c.~o

- _______________

mb

Fig. 35.1. Intensities of harmonic lines versus irradiation intensity.

IL. Bobin, High intensity laser plasma interaction

269

I ni

w 10 IOW/cmt
harmonic line.

10

13

10

Fig. 35.2. Relative intensity and spectrum of the ~

Fig. 35.3. Relative intensities of harmonic lines with integer order. Moderate incoming CO 2 laser intensity.

Experiments with CO2 laser allow one to detect high order harmonics with frequency in the visible. Two different kinds of results were thus obtained. At moderate intensities [177] the relative harmonic intensities decrease when the order increases: fig. 35.3. The tenth and eleventh harmonics are hardly distinguishable background noise. Themodifications situation is completely different at very high intensities: 2> 1016W p.2 from cm2.the Then dramatic profile occur. Due to ponderomotive pressure IA effects, a steep gradient sets in. Now, simple theories of harmonic generation predict that the more intense harmonics correspond to resonant frequencies at densities critical or quarter critical. In a gentle gradient, very few lines can be resonant. On the contrary, in a very steep gradient, resonant couplings take place as far as critical densities for the fundamental and order m harmonic are within a distance smaller than the latters wavelength Am. Since n~(m)=ir/A~ro where r 0 is the classical electron radius, the orders which can be resonant in a density gradient with maximum density flM and length L are such that

(35.3)

n~ n~(m) n~
Am

(35.4)

(fig. 35.4). Experiments with very high intensity CO2 laser beams show harmonic generation up to 46th order. Harmonic intensities then have a very weak dependence on the order [178,179]: fig. 35.5. The highest order observed in a given laser shot leads through (35.4) to an evaluation of the height of the density wall created by the non-linear laser interaction. Furthermore it can be inferred that since all harmonics up to order m are resonant, energy is to be equally shared among the whole set. However it is still not clear yet, which oscillations contribute to the harmonic generation process. Numerical simulations were carried on successfully with electron surface waves [180]. There is also a strong evidence of self focusing occurring at the same time [179].

270

IL. Bobin, High intensity laser plasma interaction

m=20

I
36. Conclusion

nc(Xm)/
I

V
623 505 392 353 311

m=46

265 230

~
Fig. 35.4. Critical density for harmonic m in a steep gradient.

Anm

Fig. 35.5. Relative intensities of harmonic lines with integer order. High incoming intensities (after 179]).

High intensity laserplasma interaction takes place through many mechanisms often occurring simultaneously. Most of them are non-linear. Thus, one may be interested in fundamental phenomena within the realm of non-linear physics. The difficulty lies in isolating a given process. Besides this aspect albeit attractive, many scientists have in mind the application to inertially confined thermonuclear fusion. The key issue then is: how to drive efficiently the implosion of the thermonuclear fuel. In this respect, since energy goes away from thermal pressure, non-linearities (i.e. instabilities) are unfavourable. One might be accordingly tempted to stick to linear regimes which are best understood and look easily controlable. They imply intensities below the parametric instability thresholds that are fairly known even in the presence of a density gradient. Furthermore, the wavelength scaling, both for absorption and instability thresholds, induces one to use laser light with the highest available frequency. However, there still remain intensity requirements on the target surface. Indeed, in order to handle manageable quantities of explosive energy, the pellet should not be made larger than a few tenths of a millimeter in diameter. Since focusable radiation beams are necessary, the light wavelength has to be at least 0.25 p.m. One then cannot take full advantage of the IA2 scaling to ensure the absence of non-linear effects. Nonetheless, on one point a conclusion can be stated. Since the absorption is very high in the U.V., the gross effects of the ponderomotive force (e.g. profile steepening) are minimal or even nonexistent. Other problems remain unsolved in general and also in the perspective of laser fusion. For instance, the role of scattering instabilities in the global energy balance is still not quantitatively determined. Moreover, at quarter critical density, there exists a competition between Raman scattering and 2-plasmon decay. Numerical simulations and analytical theories have given conflicting results. Experiments are scarce enough. One cannot tell which instability is likely to be dominant in given conditions. The question of preheat is far from being settled either, although some progress (numerical and theoretical as well) is being made on this important feature of laser driven flows. Many uncertainties also persist in magnetic field generation and their exact role in transport inhibition. The geometry is fairly complicated, they are associated with electric currents that are to be related to a high energy tail in the electron distribution. A final topic is the use of harmonic generation to get information on interaction processes. The starting point was the discovery of harmonic lines with half integer order and peculiar line structures [164]. Obviously, the physics are far from being mastered. The correspondence

IL. Bobin, High intensity laser plasma interaction

271

between observed features: order of the lines, spectra, relative intensities and the interaction mechanisms, is still largely uncertain. The subject as a whole is worth further theoretical, numerical and experimental studies in connection with the new generation of high power lasers. References
[1] A. Kastler, CR. Acad. Sc. Paris 238 (1963) 486. [2] N.G. Basov and ON. Krokhin, in: 3eme Conference dElectronique Quantique, Paris 1963, eds. P. Gnvet and N. Bloembergen (Dunod, 1964). [3] J.M. Dawson, Phys. Fluids 7 (1964) 981. [4] J. Nuckolls, L. wood, A. Thiessen and G. Zimmerman, Nature 239 (1972) 135. [5] A. Caruso, B. Bertotti and P. Giupponi, Nuov. Cim. 45B (1966) 176; A. Caruso and R. Gratton, Plasma Phys. 11(1969)839. [6] C. Fauquignon and F. Floux, Phys. Fluids 13 (1970) 390. [7] iL. Bobin, Phys. Fluids 14 (1971) 2341. [8] RE. Kidder, in: Physics at High Energy Density, ed. Caldirola (Academic, 1971); ON. Krokhin, bc. cit. [91 A.V. Gaponov and MA. Miller, Soviet Phys. J.E.T.P. 34 (1958) 242. [10] H. Hora, D. Pfirsch and A. Schlutter, Z. Naturforschung 22A (1967) 278. [11] N. Bboembergen, Non.linear Optics (Benjamin, 1964). [12] Yu. M. Aliev and V.P. Silin, Soviet Phys. J.E.T.P. 21(1965) 601. [13] D.F. Dubois and MV. Goldman, Phys. Rev. Lett. 14 (1965)143. [14] J.M. Dawson and C. Oberman, Phys. Fluids 5 (1962) 517. [15] J.M. Dawson, in: Advances in Plasma Physics, vol. 1(1968) p. 1. [161P.A.G. Scheuer, Month. Not. Roy. Astron. Soc. 120 (1960) 231. [17] J.P. Christiansen et a1., Computer Phys. Comm. 7 (1974) 271. [181G. Zimmerman, Livermore Rept. UCRL 50021-72-1 (1972). [19] J.L.Bobin and G. Tonon, Bull. Sc. Tech. C.E.A. 160 (1971). [20] H.M. Thompson, J.W. Daiber and R.G. Rehm, J. AppI. Phys. 42 (1971) 310. [21] F. Floux, D. Cognard, A. Saleres and D. Redon, Phys. Lett. 45A (1973) 483. [221H. Ahlstrom, in: Laser-Plasma Interactions, Les Houches 1980, eds. R. Balian and J.-C. Adam (North-Holland, 1982). [23] J.L. Bobin, in: Laser Interaction and Related Plasma Phenomena, Vol. 4B, eds. Schwarz and Hora (Plenum, 1977) p. 689. [24] P. Mora, Phys. Fluids 25 (1982) 1051. [25] L. Spitzer and R. Harm, Phys. Rev. 89 (1953) 977. [26] SI. Braginskii, Soviet Phys. J.E.T.P. 6 (1958) 358. [27] AR. Fraser, Proc. Roy. Soc. A245 (1960) 536. [28]J.L. Bobin, Report C.E.A. R4606 (1975). [29] F. Amiranoff et al., in: VIIth Intern. Conf. on Plasma Physics and Controlled Fusion, I.A.E.A. Vienna (1977); see also: C. Garban-Labaune et al., Phys. Rev. Lett. 48 (1982) 1018. [30] Ya. B. Zeldovich, J. of Phys. U.R.S.S. 10 (1940) 542. [31] J. von Neumann, Collected Works, VI 203, paper written in 1942. [32] W. Donng, Ann. der Phys. 6 (1949) 133. [33] C. De Michelis, I.E.E.E. J. Quantum Electron. QE-5 (1969) 188. [34] F. Morgan, L.R. Evans and C. Grey Morgan, J. Phys. D. (1971) 225. [35] Ostrovskaya and Zaideb, Soy. Phys. Uspekhi (1974). [36] Yu.P. Raizer, Soviet Phys. J.E.T.P. 21(1965) 1009. [37]M. Gravel, W.J. Robertson, AJ. Alcock, K. Buchi and MC. Richardson, Appl. Phys. Lett. 18 (1971) 75. [38] S.A. Ramsden and P. Savic, Nature 203 (1964) 1217. [39]J.L. Champetier, CR. Acad. Sc. Paris 261 (1965) 3954. [40] E. Fabre and C. Stenz, Phys. Rev. Lett. 32 (1974) 823. [41]E. Fabre, C. Popovics, C. Stenz and J. Virmont, in: Proc. Vth mt. Conf. on Plasma Physics and Controlled Fusion Research, I.A.E.A. Vienna (1974) vol. II, p. 435. [42]H.D. Shay et al., Phys. Fluids 21(1978)1634. [43]J.L. Bobin, J. Plasma Phys. 25 (1981) 193. [44] N.G. Denisov, Soviet Phys. J.E.T.P. 4 (1957) 344. [45] J.E. Balmer and T.P. Donaldson, Phys. Rev. Len. 39 (1977) 1054. [46] KR. Manes et al., Phys. Rev. Lett. 39 (1977) 281. [47] C. Garban-Labaune et al., J. de Phys. Lett. 41L (1980) 463.

272

IL. Bobin, High intensity laser plasma interaction

[48] G. Morales and Y.C. Lee, Phys. Rev. Lett. 33 (1974) 1016 and Phys. Fluids 19 (1976) 691). [49] J.C. Adam, A. Gourdin Servenire-Hron and G. Laval, Phys. Fluids 25 (1982) 376; A. Gourdin Servenire-Hron and J.C. Adam, Phys. Fluids 27 (1984) 2005. [50] J.S. DeGroot and J. Tull, Phys. Fluids 18 (1975) 672. [51] E. Valeo and W.L. Kruer, Phys. Rev. Lett. 33 (1974) 750. [52] D.W. Forslund, J.M. Kindel, K. Lee and E.L. Lindman, Phys. Rev. Lett. 36 (1976) 35. [53] H. Azechi, S. Oda, K. Tanaka, T. Norimatsu, T. Sasaki, T. Yamanaka and C. Yamanaka, Phys. Rev. Lett. 39 (1977) 1144. [54] D.T. Atwood, D.W. Swinney, J.M. Auerbach and P.H.Y. Lee, Phys. Rev. Lett. 40 (1978) 184. [55] R. Fedosejevs, w. Tomov, N.H. Burnett. GD. Enright and M.C. Richardson. Phys. Rev. Lett. 39 (1977) 932. [56] A. Raven and 0. Willi, Phys. Rev. Lett. 43 (1979) 921. [571P.M. Campbell, P. Hammerling, R.R. Johnson, J.J. Kubis. F.J. Mayer and D.C. Slater. in: Proc. 6th Intern. Conf. on Plasma Physics and Controlled Fusion Research, I.A.E.A. Vienna (1976) p. 227. [58] D. Shvarts, C. Jablon, lB. Bernstein, J. Virmont and F. Mora. NucI. Fusion 19 (1979) 1457. [59] P. Mora and R. Pellat, Phys. Fluids 22 (1979) 2300. [60] lB. Bernstein, J.M. Greene and M.D. Kruskal, Phys. Rev. 108 (1957) 546. [61] D.C. Montgomery and G. Joyce, J. Plasma Phys. 3 (1969) 1. [62] K. Lee, D.W. Forslund, iA. Kindel and E.L. Lindman, Phys. Fluids 20 (1977) 55; R.D. Jones, C.H. Aldrich and K. Lee, Phys. Fluids 24 (1981) 31t). [63] J. Virmont, R. Pellat and P. Mora, Phys. Fluids 21(1978) 567. [64] P. Mulser and C. Van Kessel, Phys. Rev. Lett. 38 (1977) 902. [65]J. Pearlman and J. Anthes, AppI. Phys. Lett. 27 (1975) 581. [66] L.L. Cowie and CF. McKee, Astrophys. J. 211(1977)135; CE. Max and CF. McKee, Phys. Rev. Lett. 39 (1977) 1396. [67] R. Fabbro, These Universit Paris-Sud (1982). [68] A. Decoster, private communication. [69] B.D. Fried and R.W. Gould, Phys. Fluids 4 (1961) 139. [70] D.w. Forslund, J. Geophys. Res. 75 (1970) 17. [71] Ri. Bickerton, Nucl. Fusion 13 (1973) 457. [72] W.M. Manheimer, Phys. Fluids 20 (1977) 265. [73] EL. Lindman, J. de Physique, Colboque C6, Supplement 38 (1977) C69. [74] N.M. Manheimer and D.G. Colombant, Phys. Fluids 21(1978)1818. [75] C. Stenz, These Universit Paris-Sud (1980). [76]DR. Gray, iD. Kilkenny et al., Phys. Rev. Lett. 39 (1977) 1270; see also ref. [112]. [77] CE. Max, in: Laser-Plasma Interaction, eds. R. Balian and i-C. Adam (North-Holland, 1982). [78] E. Weibel, Phys. Rev. Lett. (1959). [79] A. Ramani and G. Laval, Phys. Fluids 21(1978) 290. [801D.W. Forslund and J.U. Brackbill, Phys. Rev. Lett. 48 (1982) 1614. [81] B. Bernu and J.P. Hansen, Phys. Rev. Lett. 48 (1982) 1375. [82] For a review see e.g. N.J. Peacock, in: Laser Interaction, eds. Cairns and Sanderson (Scottish Universities, 1980). [83] S. Bodner, N.R.L. Memorandum Report 4453 (1981); see also ref. [22]. [84] iL. Bocher et al., Phys. Rev. Lett. 52 (1984) 823. [85] DR. Gray and J.D. Kilkenny, Plasma Phys. 22 (1980) 81. [861D. Kershaw, Flux limiting. Natures own way, U.C.R.L. 78378 (1976). [87] C.E.C.A.M. Reports on: heat flux instabilities and anomalous transport (1981), (1982). [88] AR. Bell, R.G. Evans and Di. Nicholas, Phys. Rev. Lett. 46 (1981) 243. [89] J.P. Matte and J. Virmont, Phys. Rev. Lett. 49 (1982) 1936. [90] D. Shvarts et al., Phys. Rev. Lett. 47 (1981) 247. [91] J. Delettrez, C.E.C.A.M. Workshop on Transport Problems in Laser Created Plasmas, Orsay (1982). [92] R. Fabbro, private communication. [93] Ri. Mason, Phys. Rev. Lett. 47 (1981) 652. [94] iF. Luciani, P. Mora and J. Virmont, Phys. Rev. Lett. 51(1983) 1664. [95] R. Chiao, E. Garmire and Ch. Townes, Phys. Rev. Lett. 13 (1963) 479; E. Garmire, R. Chiao and Ch. Townes, Phys. Rev. Lett. 16 (1966) 347; 5.A. Akhmanov, A.P. Sukhorukov and R.V. Khokbov, Soviet Phys. Usp. 10 (1968) 609. [96] P. Kaw, G. Schmidt and T. Wilcox, Phys. Fluids 16 (1973) 1572. A.B. Langdon and B. Lasinski, Phys. Rev. Lett. 34 (1975) 934. [98] Di. Kaup and AC. Newell, Proc. Roy. Soc. London 361(1978) 413. [99] V.N. Korobkin and R.V. Serov, J.E.T.P. Lett. 4 (1966) 70; GA. Askarian et al., J.E.T.P. Lett. 4 (1967) 93; J.A. Stamper et al., Phys. Rev. Lett. 26 (1967)1012.

[971

IL. Bobin, High intensity laser plasma interaction

273

[100] L. Biermann, Zs. f. Naturforsch. 5a (1950) 65. [101] F. S.chwirtzke, in: Laser Interaction and Related Plasma Phenomena, Vol. 3A, eds. Schwarz and Hora (Plenum, 1974) p. 213. [102] MG. Drouet, in: Laser Interaction and Related Plasma Phenomena, Vol. 4B, eds. Schwarz and Hora (Plenum, 1977) p. 737. [103] SI. Braginskii, Rev. Plasma Phys. 1(1966) 205. [104] TiM. Boyd, G.J. Humphrey-Jones and D. Cooke, Phys. Lett. 88A (1982) 140. [105] A. Raven, 0. Willi and PT. Rumsby, Phys. Rev. Lett. 41(1978) 554. [106] P. Mora and R. Pellat, Phys. Fluids 24 (1981) 2219. [107]lB. Bernstein, C.E. Max and K. Estabrook, Phys. Fluids 21(1978) 905. [108]P. Mora and R. Pellat, Phys. Fluids 22 (1979) 2408. [109] I.P. Shkarovsky, Phys. Fluids 23 (1980) 52. [110] B. Bezzerides, D.F. Dubois, D.W. Forslund and EL. Lindman, Phys. Rev. Lett. 38 (1977) 495. [111]i. Briand, J.L. Bourgade, ED. Nasser and B. Visentin, in: Rapport dActivite, G.R.E.C.O. Interaction Laser Matiere (1979) p. 101. [112] R. Fabbro and P. Mora, Phys. Lett. 90A (1982) 48. [113] F.F.Chen, in: Laser Interaction and related plasma Phenomena, Vol. 3A, eds. Schwarz and Hora (Plenum, 1974) p. 291. [114] K. Nishikawa, J. Phys. Soc. Japan 24 (1968) 916; 24 (1968)1152. [115]i.M. Manley and H.E. Rowe, Proc. IRE. 47 (1959) 2115. [116] Lord Rayleigh, Phil. Mag. (1883). [117] PA. Sturrock, Phys. Rev. 112 (1958) 1488. [118] C. Yamanaka et al., Phys. Rev. A 6 (1972) 2335. [119] E.A. Jackson, Phys. Rev. 153 (167) 235. [120]MN. Rosenbluth, RB. White and CS. Liu, Phys. Rev. Lett. 31(1973)1190; CS. Liu and MN. Rosenbluth, Phys. Fluids 19 (1976) 967. [121] V.P. Silin, Soy. Phys. J.E.T.P. 21(1965)1127. [122] V.P. Silin, VIlith I.C.P.I.G. Vienna (1967). [123]W.L. Kruer and J.M. Dawson, Phys. Fluids 15 (1972) 446. [124]J.P. Freidberg and B.M. Marder, Phys. Rev. A 4 (1973) 1549. [125]R. Bingham and C.N. Lashmore-Davies, Nucl. Fusion 16 (1976) 67. [126] AS. Pikovsky and MI. Rabinovich, Physica 2D (1981) 8; E. Ott, Rev. Mod. Phys. 53 (1981) 655. [127] C. Meunier, MN. Bussac and G. Laval, Physica 4D (1982) 236. [128] WE. Lamb and MO. Scully, in: Polansation, Matiere, Rayonnement (P.U.F., 1969); MO. Scully and M. Sargent, Physics Today (March 1972) 38. [129] V.N. Tsytovich, Non-linear Effects in Plasma (Plenum, 1970). [130] J.M. Dawson and AT. Lin, U.C.L.A. Report PPG 191 (1974). [131] CS. Liu, MN. Rosenbluth and RB. White, Phys. Fluids (1974). [132] B.I. Cohen and CE. Max, Phys. Fluids 22 (1979)115; K. Estabrook et al., Phys. Rev. Lett. 46 (1981). [133]J. Liouville, J. de Math. 2 (1837) 16; 2 (1837) 418. [134] MN. Rosenbluth, Phys. Rev. Lett. 29 (1972) 565. [135] R. Pellat, in: LaserPlasma Interactions, eds. R. Balian and J.-C. Adam (North-Holland, 1982). [136] W.L. Kruer, in: Laser Plasma Interaction, eds. Cairns and Sanderson (SUSSP Publications, Edinburgh, 1980) p. 388; Phys. Fluids 23 (1980) 1273. [137] CE. Max and K. Estabrook, Comments Plasma Phys. Cont. Fusion 5 (1980) 239. [138] F.W. Perkins and J. Flick, Phys. Fluids 14 (1971) 2012. [139] R. Binghain and C.N. Lashmore-Davies, Plasma Phys. 21(1979) 433. [140] Mi. Feigenbaum, J. Stat. Phys. 19 (1978) 25. [141]N.N. Rozanov and V.A. Smimov, Soviet Phys. J.E.T.P. 43 (1976)1075. [142] G. Bret, private communication. [143]J.M. Dawson, Phys. Rev. 113 (1959) 383. [144]P. Koch and J. Albritton, Phys. Rev. Lett. 32 (1974)1420. [145]W.L. Kruer, Phys. Fluids 22 (1979)1111. [146] J.A. Tarvin et al., Phys. Rev. Lett. 48 (1981) 256. [147]Y.F. Shen, Rev. Mod. Phys. 48 (1976)1. [148] J. Martineau and J.L. Bobin, Phys. Lett. 47A (1974) 43. [149] N.S. Erokin, S.S. Moiseev and V.V. Muklin, NucI. Fusion 14 (1974) 333. [150] OP. Banfi, PG. Gobbi and A.M. Malvezzi, Optics Comm. 44 (1983) 337. [151) V.V. Pustovalov, V.P. Silin and VT. Tikhonchuk, Soviet Phys. J.E.T.P. 38 (1974) 938; AL. Avrov et al., Soviet Phys. J.E.T.P. 46 (1977) 507; E.Z. Gusakov, Tech. Phys. Lett. (in Russian) 3 (1977)1219. [152]J. Bnand Ct al., XVIth Europ. Conf. on Laser Interaction with Matter, London (1983).

274

IL. Bobin, High intensity laser plasma interaction

[153] V. Aboites et al., XVIth Europ. Conf. on Laser Interaction with Matter, London (1983). [154] HA. Baldis and PB. Corkum, Phys. Rev. Lett. 45 (1980) 1260. [155] M.J. Herbst et al., Phys. Rev. Lett. 46 (1981) 328. [156] R.W. Short et al., Phys. Rev. Lett. 52 (1984) 1496. [157] iL. Bobin, Optics Comm. 14 (1975) 339. [158] Nguyen Due Long, K.J. Parbakhar and T.W. Johnston, Proc. Vllth Europ. Conf. on Plasma Phys. and NucI. Fusion, Lausanne 1 (1975) 73. [159] H.C. Pant, K. Eidman, R. Sachsenmaier and R. Sigel, Optics Comm. 16 (1976) 396. [160] C. Garban et al., J. de Phys. Lett. 39L (1978) 165; A.A. Offenberger et al., Phys. Rev. A18 (1978) 74.6. [161] iL. Bobin, in: Plasma Physics, ed. Wilhelmsson (Plenum, 1977). [162] A. Saleres et al., J. de Phys. 34C2 (1973) 9. [163] D.W. Phillion et al., Phys. Fluids 25 (1982) 1434. [164] iL. Bobin, M. Decroisette, B. Meyer and Y. Vitel, Phys. Rev. Lett. 30 (1973) 594. [165] MC. Richardson et al., Bull. Amer. Phys. Soc. 26 (1981) 954, paper 5G2. [166] RE. Turner et al., Phys. Fluids 27 (1984) 511. [167] S.M.L, Sim et al., quoted in ref. [166]. [168] D. Anderson and H. Wilhelmsson, Phys. Lett. 56A (1976) 37. [169]M. Decroisette, B. Meyer and Y. Vitel, Phys. Lett. 45A (1973) 443. [170] lB. Bernstein, Phys. Rev. 109 (1958) 10. [171] W.L. Kruer and K. Estabrook, in: Laser Interaction and Related Plasma Phenomena, Vol. 4B, eds. Schwarz and Hora (Plenum, 1977) p. 709. [172]Al. Avrov et al., J.E.T.P. Lett. 24 (1976) 293. [1731 H.C. Pant, K. Eidman, P. Sachsenmaier and R. Sigel, Report 1.P.P. Garching IV/85 (1975). [174]i. Briand, private communication. [175]A.A. Offenberger, A. Ng, MR. Cervenan, Can. J. Phys. 56 (1978) 381, A. Ng, L. Pitt, D. Salzmann and A.A. Offenberger, Phys. Rev. Lett. 42 (1979) 307. [176] A. Saleres, These Universit Paris-Sud (1977). [177]N.H. Burnett, HA. Baldis, MC. Richardson and GD. Enright, AppI. Phys. Lett. 31(1977)172. [178] R.L. Carman, D.W. Forslund and i.M. Kindel, Phys. Rev. Lett. 46 (1981) 29. [179]R.L. Carman, C.K. Rhodes and R.F. Benjamin, Phys. Rev. 24A (1981) 2649. [180]B. Bezzerides, RD. Jones and D.W. Forslund, Phys. Rev. Lett. 49 (1982) 202.

You might also like