You are on page 1of 12

1942

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

The Physics of the Nonlinear Optics of Plasmas at Relativistic Intensities for Short-Pulse Lasers
W. B. Mori (Invited Paper)

AbstractThe nonlinear optics of plasmas at relativistic intensities are analyzed using only the physically intuitive processes of longitudinal bunching of laser energy, transverse focusing of laser energy, and photon acceleration, together with the assumption of conservation of photons, i.e., the classical action. All that is required are the well-known formula for the phase and group velocity of light in plasma, and the effects of the ponderomotive force on the dielectric function. This formalism is useful when the dielectric function of the plasma is almost constant in the frame of the light wave. This is the case for Raman forward scattering (RFS), envelope self-modulation (SM), relativistic self-focusing (SF), and relativistic self-phase modulation (SPM). In the past, the growth rates for RFS and SPM have been derived in terms of wavewave interactions. Here we rederive all of the aforementioned processes in terms of longitudinal bunching, transverse focusing, and photon acceleration. As a result, the physical mechanisms behind each are made clear and the relationship between RFS and envelope SM is made explicitly clear. This allows a single differential equation to be obtained which couples RFS and SM, so that the relative importance between each process can now be predicted for given experimental conditions.

I. INTRODUCTION HE nonlinear optics of plasmas was developed extensively in the early 1970s [1][4]. This development resulted in the identication of numerous so-called parametric instabilities. Among these were Raman and Brillouin scattering, so-named because of their close connection to the processes which occur in unionized gases [5]. In these original analyses [1][4], the instabilities were formulated in terms of wavewave interactions and the ponderomotive force. Using this formalism, general dispersion relations were derived which can in principle be used to describe the evolution of arbitrary noise sources. These dispersion relations also described the well-known lamentation/self-focusing [4], [5], [7][9] and self-phase modulational instabilities [1][4], [9]. Mechanisms which occur only when the laser oscillates the electrons at relativistic velocities were also identied in the early work by Max et al. [4]. However, the relativistic analyses were conned to the weakly relativistic limit. Almost all of the early work was undertaken because of its importance to laser
Manuscript received August 22, 1996; revised July 1, 1997. This work was supported by DOE under Grant DE-FG0392ER40727, LLNL under Contract B291465 and Contract B335241, and the National Science Foundation under Grant DMS-9722121. The author is with the Department of Electrical Engineering and the Department of Physics and Astronomy, University of California of Los Angeles, Los Angeles, CA 90095 USA. Publisher Item Identier S 0018-9197(97)07828-7.

fusion which utilized lasers with relatively long pulse lengths and modest intensities. With the advent of short-pulse laser technology [10], new applications [11][13] have evolved and the relevant instabilities of the lasers in plasmas has changed. As a result, there has been a large amount of new work in the eld of the nonlinear optics of plasmas at relativistic intensities. For short-pulse lasers, the nonlinear optics of plasmas involves only electron motion because the ions are immobile during the transit time of the laser. This signicantly limits the number of instabilities which can occur. The most important instabilities for short-pulse lasers are Raman forward scattering (RFS) [14][30], relativistic self-focusing (SF)[4], [31][33], and relativistic self-phase modulation (SPM) [4], [27]. Each of these instabilities can be described conventionally in terms of wavewave interactions [1][4], [18], [19], [23][27]. In each case, an incident electromagnetic wave at frequency decays into two forward moving electromagnetic sidebands at (the Stokes wave) and (the antifrequencies corresponds to modulations Stokes wave). The frequency , where is to the index of refraction. In RFS, the plasma frequency, and an electrostatic plasma wave is generated. This instability is important because the phase velocity of the plasma wave is nearly the speed of light. This allows the plasma wave to accelerate electrons to relativistic , and energies [11][13]. In lamentation and SPM, the modulation at is caused by relativistic mass corrections to the electrons motion. Recently, the importance of RFS type instabilities for short pulses was clearly demonstrated in uid simulations [18][22]. In these simulations, nite-width pulses were found to break apart axially into beamlets separated in time by roughly . Associated with this breakup was the generation of a large amplitude plasma wave. This clearly indicates the occurrence of some form of RFS. In explaining these results, one group developed a theory which described an envelope self-modulation for nite-width pulses [28], [29], and the other used a conventional Raman wavewave analysis [18], [19]. The physical mechanism for the envelope self-modulation (SM) is as follows [13], [22], [28], [29]. First, the laser pulse creates a plasma wave wake noise source. Next, the density compressions and rarefactions of the plasma wave wake transversely focus and defocus laser energy. As a result, the laser spot size and hence intensity are modulated at nearly the plasma frequency and at nearly the wavenumber

00189197/97$10.00 1997 IEEE

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1943

. The ponderomotive force of the intensity modulation then excites a larger wake and the process feeds back on itself. Importantly, no mention is made of the light wave decaying into other light waves. It was subsequently shown that the differential equation and growth rate for this envelope SM are the same as those for RFS at small forward angles, [23][26]. It was also shown that the above physical picture is complicated by the fact that the intensity modulations and the density modulations are not resonantly in phase [23], [24] and that the envelope analysis cannot describe RFS in the exact or near forward direction, [23][26]. RFS in the near forward direction (very small angles) is distinct from SM by the phase relation between the density and intensity modulations. We will henceforth use RFS0 to mean scattering into the near forward direction. It has also been pointed out [23], [24], [34], [35] that plasma waves produced by RFS0 can also cause transverse focusing even when so-called SM has not occurred. In spite of all this new work to identify the various regimes of RFS, there is still some confusion and disagreement as to how RFS and envelope selfmodulation are related and which mechanisms are important for given experimental conditions. In this paper, we present a formalism which hopefully removes some confusion. We analyze RFS0, envelope selfmodulation (SM), self-focusing (SF), and self-phase modulation (SPM) all from the same set of physical phenomena. We do not use the concepts of Stokes and anti-Stokes waves, rather we calculate the modulation in laser intensity in terms of the physically intuitive phenomena of longitudinal bunching of laser energy, transverse focusing of laser energy, and photon acceleration [36], [37]. Each of these phenomena arise when modulations in the index refraction appear stationary in the light waves frame, i.e., the index of refraction has a relativistic phase velocity. The local index of refraction determines the lasers group and phase velocity. Longitudinal bunching (LB) is caused by longitudinal variations in the group velocity, transverse focusing (TF) is caused by transverse variations in the phase velocity, and photon acceleration (PA) is the change in local frequency caused by longitudinal variations in the phase velocity. Recall that both SF and envelope SM are typically explained in terms of transverse focusing of laser energy. We calculate the overall change in the waves amplitude from all of these effects, by assuming that photon number, i.e., action is conserved [1][4], [38]. An equation is then derived which relates the amplitude modulations to the modulations of the index of refraction. By combining these equations, we recover the exact growth rates for RFS0, SM, SF and SPM. Therefore, it is possible to precisely identify which phenomena causes which instability. In addition, because direct comparison between each phenomenon is now possible, we may now determine which instability is most important for given experimental conditions. We will argue that in all existing experiments [39][42] one-dimensional (1-D) and higher dimensional (2-D and/or 3-D) effects of RFS0 are more important than SM. In future experiments, in which shorter pulse lasers and lower plasma densities are used, then SM may dominate. Incidentally, these same physical phenemona can be applied to instabilities in other media so long as the index of

refraction appears stationary in the light waves frame, i.e., the Stokes and anti-Stokes waves are both important. II. PHYSICAL PHENOMENA In this section, we begin by stating the basic assumptions of this formalism and then describe how the waves amplitude can be changed by modulations to the index of refraction. A. Assumptions In an unmagnetized plasma, the index of refraction of a linearly polarized light wave is where , is the plasma density, and is the normalized vector potential of the incident laser, . A useful formula is 0.85 10 (W/cm ) ( m) where is the laser intensity and is the laser wavelength; is commonly referred to as . It is clear that the index of refraction can be altered by either modulating the plasma density, the laser amplitude, or the laser frequency. In this paper, we consider only small modulations and weakly relativistic pumps. Therefore, the index of refraction can be expanded as (1) where is the normalized density perturbation, and represents averaging over the fast laser oscillations. In general, , so an expression for is (2) and an expression for is [43], [44] (3) and can be modulated Therefore, it is clear that both through changes in either density, laser intensity, or frequency. Note that when the relativistic term is included [43], [44]. In addition, we assume that within a local volume the photon number, i.e., the classical action, is conserved. The conservation of the classical action in laser-plasma instabilites has been discussed extensively in the literature [1][4], [38], [45]; however, for completeness, we provide a derivation in Appendix A. Conservation of action can be stated as constant (4)

where represents averaging over the fast oscillations, is the spot size, and is some initial longitudinal extent. The lasers vector potential can therefore be modulated by only three ways: 1) modulate 2) modulate 3) modulate longitudinal bunching; transverse focusing; photon acceleration.

1944

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

The overall change in

can therefore be expressed as (5)

is to be modulated by either of these methods, then If the index of refraction must be stationary in the light waves frame. Therefore, we will henceforth use the so-called speed of light frame variables and . We next show how and can be calculated from variations in and with respect to the variable . B. Longitudinal Bunching The change in separation, , between two positions, 1 and 2, during a time is given by (6) If the two positions are near each other, then (7) where the fact that was used because we are considering regions which are moving at the speed of light. Therefore, the normalized change in due to modulations in can be expressed as (8) where i.e., represents a change in time while keeping xed, . Therefore, (7) can also be rewritten as (9) which is the rate that two positions bunch toward each other because of variations to the group velocity in the speed of light variable . C. Transverse Focusing Calculating the change in spot size, , because of transverse variations in takes two steps. First, we calculate the rate at which the wavefronts curve. Since energy ows normal to the phase front, i.e., in the direction of the Poynting vector, the outer part must curve forward in order for focusing to occur. In a time , the angle that the wavefront bends forward is given by (10) where position 2 is located on the axis and position 1 is located at the edge of the beam a distance from the axis. This is depicted in Fig. 1. Obviously, is required for focusing to occur, or in terms the index of refraction, . Second, we calculate the rate at which energy is focused inwards. The transverse component of the energy velocity is . So if is small then the transverse energy is focused

Fig. 1.

The bending of wavefronts during transverse focusing.

inwards with a velocity . This velocity is equal to the time rate of change of the spot size . Therefore, (11) Differentiating with respect to time gives (12) which is the acceleration of the spot size caused by tranverse then the spot size variations in . Note that if increases, i.e., defocusing occurs. D. Photon Acceleration In photon acceleration, the local frequency changes because of longitudinal variations in . Consider two phase fronts located at two different positions, and , as shown in Fig. 2. After a time the phase fronts will have moved to positions and , given by (13a) and (13b) and are the phase velocities at the respective where positions of each wave front. Subtracting from gives (14) Therefore, the time rate of change of is (15) where we now use the speed of light variables Rewriting this in terms of frequency gives and .

(16) Therefore, as a photon moves along in an index of refraction gradient which it views as stationary, its frequency increases if the slope with respect to is positive. We call this process photon acceleration because the waves frequency is directly related to the waves group velocity [36]. Therefore, if photon number is conserved then the group velocity of each photon is increasing in analogy with the kinetic velocity of each particle increasing in a potential gradient. The frequency can also change in a relativistic ionization front [46], [47]. However, in this case, photon number is not conserved, so strictly speaking this is not photon acceleration.

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1945

Fig. 2. A physical picture for photon acceleration.

Fig. 3. A physical picture for RFS0. RSF0 is due to equal amounts of longitudinal bunching and photon acceleration.

III. GROWTH RATES In this section, we consider specic modulations to and derive the growth rates for each instability. A. Raman Forward Scattering In order to derive growth rates, we need to identify how the index of refraction is being modulated. In the 1-D limit, the laser intensity can only be modulated from either longitudinal bunching or photon acceleration. This is illustrated in Fig. 3. Therefore, can be written as (17) and evolves in time as (18) Substituting (9) and (16) into (18) gives (19) In RFS0, the modulations to , i.e., and , are solely the result of modulations to . Therefore, from (2) and (3) we can rewrite (18) as (20) It is of interest to note that longitudinal bunching and photon . We acceleration contribute equally to the modulation of dene and where depends slowly with both and , to obtain (21) are out of Therefore, in RFS0 the modulations to phase with the density response, . In order to derive a growth rate, we need an equation which describes how and

modulations to cause density perturbations. This is well known to be a harmonic oscillator equation of the form [1][4] (22) where the right-hand side is the divergence of the ponderomotive force. In the speed of light variables, this becomes (23) which can be rewritten as (24) where

and depends slowly on . In reducing (23) to (24), we are not considering the strongly coupled (others call it the shortpulse) regime of RFS0 [13], [18], [19], [23][26], [28][30], because we are interested in the portions of the pulse with the most -foldings. Combining (24) with (21) gives

(25) . This is identical to [25, eq. (10)] where which was obtained from the conventional wavewave analysis using Stokes and anti-Stokes sidebands. It describes the so-called four-wave resonant regime. Therefore, we have shown that in RFS0 the density modulations of the plasma wave are out of phase with , and the perturbations to are caused equally by longitudinal bunching and photon acceleration. The longitudinal bunching is caused by the changes to . Note that the changes to from photon acceleration also change but these modulations do not have the correct phase to be reinforced.

1946

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

Differentiating (28) with respect to time leads to (29) The evolution of with time is given by (16) provided that the modulations in phase velocity are the result of the relativistic term in (2). This results in (30) which upon substitution into (29) gives (31) If grows in time as then

Fig. 4. The physical picture for relativistic SPM, which is photon acceleration followed by longitudinal energy bunching.

Equation (24) can be solved in closed form for specied boundary and initial conditions [23][26]. Here, we simply give its asymptotic solution which is (26) As noted earlier, this growth corresponds to four-wave resonant RFS0 and more research is required to obtain the four-wave nonresonant growth of RFS0 using these simple physical pictures. B. Relativistic Self-Phase Modulation In relativistic self-phase modulation, the index of refraction is modied by the relativistic term in (2). This leads to a modulation in from photon acceleration. However, which results from this unlike RFS0, the modulation to does not have the correct phase to reinforce the original perturbations of . Instead the modulation of causes to vary, causing energy to bunch longitudinally. This is illustrated in Fig. 4. This is a two-step process and this is why SPM has a lower growth rate than RFS. To calculate the growth rate, we rst write the change in from only longitudinal bunching,

(32) Thus, it would appear that grows exponentially in

which increases indenitely time with a growth rate with . Note that if this is times smaller than the RFS0 growth rate. In reality, there is a term which balances this growth for large . To understand this, we must, for the moment, realize that the intensity modulation can be and represented as the superposition of a pump at frequency two sidebands at . The two sidebands cannot both be exactly resonant because the linear dispersion relation of light is quadratic, i.e., . Therefore the beat pattern of is not stationary in the light waves frame, but appears to have a real frequency. A real frequency term would appear on the right-hand side of (32) but with the opposite sign. To calculate the effective frequency, we assume that is of the form (33) where same, and that and . We assume the s are the to obtain (34)

Therefore,

evolves in time as (27a)

Using a trigonometric identity, this can be rewritten as

Substituting (9) into (27a) gives (27b) Assuming that the changes in (3) results in are due to the term in Using the denitions for the s gives

(35)

(36a) and (36b)

(28) where we have substituted . consists of a beat pattern moving at the Therefore, speed of light times a temporal oscillation which means that in

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1947

the speed of light frame, oscillates at . Since is the same for both the Stokes and anti-Stokes waves, we can Taylor expand about to nd (37a) and (37b) Therefore, dispersion leads to (38) Since in the absence of any growth frequency, we can add must oscillate at this

This is not the whole story since is the absence of variations the spot size increases because of diffraction. Therefore, we need to add the diffractive term on the right-hand side of (42). The diffractive term in SF is analogous to the dispersion term in SPM. The evolution of the spot size from diffraction for a Gaussian beam is well known to be [5] (43) where is the Rayleigh or diffraction time, is the laser wavelength, and is the spot size at the focus. In Section II-C, we considered nearly planar wavefronts, i.e., regions near the focus. Near the focus, (43) can be differentiated twice to get (44)

(39) Adding this term to the right-hand side of (42) yields to the right-hand side of (32) to obtain the nal equation for relativistic SPM (45) Therefore, self-focusing occurs if the term in brackets is negative. The threshold condition for SF is therefore (40) Therefore, the threshold for instability is the maximum growth rate is and (46) This is identical to the result obtained from more formal and more complicated derivations using source-dependent expansions [48], [49] or variational techniques [50]. We emphasize that in the derivation presented in Section II-C the position of the outer position of the wavefront was somewhat arbitrarily assumed to be one spot size away from the axis. The more rigorous derivations must therefore be done at least once. The threshold condition is in terms of the product , which is proportional to the laser power. Therefore, (46) can be viewed as a power threshold condition where the critical power is [31] GW Note that if only transverse focusing occurs that constant, i.e., the laser power is conserved. D. Envelope SM (Laser Sausaging) Transverse focusing can also be caused if the index of refraction is varied by the density pertubation term in (12). In this case, (12) reduces to (48) Note that if is negative, focusing occurs while, if is positive, defocusing occurs. In order to couple the evolution of the spot size to that of the density pertubation, we need to relate to . From (5) (49) (47) is a

(41) and it occurs for

These results are in agreement with [4]. Therefore, the physical picture of SPM is as follows. It requires rst for the frequency, i.e., the phase, of the wave to be modulated and second for the laser energy to bunch longitudinally because of the group velocity modulations associated with the changes in . It is due to the relativistic corrections to , not to . This physical picture was given qualitatively in [4]. Note that the other mechanisms which modulate the laser amplitude do not have the correct phase to be reinforced. In particular, the relativistic terms in do not lead to an instability, but rather to group velocity steepening. C. Relativistic Self-Focusing In the next two sections, we consider the effects of transverse focusing. We begin with relativistic self-focusing and show that our physical arguments for deriving an equation for the evolution of the spot size yield the exact SF equation. We start from (12) and assume that varies only from the relativistic term in (2) to get (42)

1948

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

from which it follows (50) Equation (50) combined with (23) completely describes envelope SM, and as before we are not considering the strongly coupled regime. Using the same denitions as before for the eld envelopes gives (51a) (51b) Combining (51a) and (51b) gives a single equation for : (52) We rewrite the right-hand side in terms of and to get (53) This is identical to the equation obtained by differentiating [28, eq. (4)] with respect to , which was derived using the sourcedependent expansion method. This differential equation has no closed form solution. However, using saddle point integration techniques, the asymptotic form for the solution can be shown to be [23][26], [28], [29] (54) Before proceding, there are several points to make. First, appears in the differential equation because of the choice for normalizing the parameters; the relativistic SF terms have been neglected. We will comment more on this shortly. Second, an identical differential equation and hence asymptotic solution was derived using a conventional wavewave RFS analysis [23], [24] for scattering angles of . Therefore, the scattered light is not in the direct forward direction but is at well-dened angles. This results in the generation of higher order Gaussian modes. Third, in SM, only transverse focusing and defocusing occurs. The intensity modulations arise from reductions and increases in the spot size, so sometimes SM is referred to as laser sausaging due its analogy with an electron beam instability. Fourth, the asympotic solution also has an imaginary part to the phase, which means that the density pertubation and intensity modulations are not out of phase [23], [24], and the phase velocity of the plasma wave gradually decreases in time [25], [26]. The decrease in phase velocity means that SM is less likely to generate ultrarelativistic electrons. Last, transverse focusing is not equivalent to SM, it can also occur during RFS0. Since in RFS0, and are still out of phase, transverse focusing leads to contour shapes in which one-half is narrow and the other half is wide. These shapes have been refered to as inverse D shapes [24], [34], [35]. This is shown in Fig. 5, and such shapes cannot be generated in the SM process. Envelope SM is directly related to RFS in the so-called four-wave nonresonant regime and is distinct from RFS0. The term nonresonant refers to the fact that because of the extra phase shift, the plasma wave does not oscillate at exactly . The difference between four-wave nonresonant RFS0 and SM lies in which term dominates in the mismatch quantity of [23], [24], i.e., if then RFS dominates. Similar conclusions were independently reached by Andreev et al. [25], [26]. In the nonresonant regime of RFS, multiple cascading is less likely to occur [25], [26], [34], [35], [51]. Therefore, an experimental diagnostic of RFS0 is the observation of strong multiple cascading. In the original SM paper by Esarey et al. [28], [29], it was indicated that relativistic SF was critical to envelope SM. They stated that needed to be larger than roughly 1/2 (depending on the sharpness of the rise time) in order for strong SM to occur. We believe this to depend on the parameter space under consideration. In the preceding analysis and in the conventional wavewave analyses [23][26], the relativistic SF terms have been neglected. This assumption is valid so long as the spot size changes little from SF or diffraction before the instability saturates. In particular, it is obvious from (53) that for any given value of a pulse length can be chosen such that SM (or RFS0) can still occur within a Rayleigh time. In PIC simulations, beam breakup is observed even for 0.5 [51][53]. The SF and diffraction terms can easily be included by combining (45) and (48) to get (55) Combining (54) with (49) now gives
Fig. 5. The effects of TF from RFS0 without SM. If n and p are =2 out of phase and the wave is moving to the right, then intensity contours become inverse D-shaped, so a signature of transverse focusing in RFS0 is inverse D-shaped contours.

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1949

If a matched beam is assumed, i.e., , or equivalently a preformed channel is used [20], [21], then the equations given earlier in this section are still valid. The physical mechanism of the instability is not due to relativistic SF, but rather is completely due to transverse focusing in plasma waves. The inequality, , can be important if relativistic guiding is required to increase the interaction length beyond several Rayleigh lengths. IV. COMPARISON BETWEEN RFS ENVELOPE SELF-MODULATION

be , and that TF dominates. To be more precise, we next derive a differential equation which couples both the 1-D and 2-D effects and then calculate its asymptotic response. This equation does not include the nonresonant regime of RFS0, which is reasonable if , i.e., small [23], [26].If this inequality is not satised, then RFS0 always dominates. We start by dening , where the subscript stands for total. Therefore, (58) and we substitute (21) and (50) into the right-hand side to get

AND

In the previous section, we carefully described the differences between longitudinal bunching (LB), photon acceleration (PA), and transverse focusing (TF). We also showed that when these processes are caused by plasma waves, they lead to RFS0 and SM. In this section, we make detailed comparison between the relative importance of the 1-D phenomena (LB and PA) and the 2-D phenomenon (TF), with particular emphasis given to discerning which processes dominate for given experimental conditions. For simplicity, we begin by assuming a constant amplitude plasma wave and compare the relative contributions to from 1-D and 2-D effects. This simple exercise is very illustrative and it has direct implications to beat wave excitation of plasma waves where the plasma wave amplitude changes little in . Focusing in constant amplitude plasma waves is related to cascade focusing [54]. If is constant in at a value , then (21) and (50) can be integrated in to give (56a) and (56b) grows as while It is immediately clear that grows as . The reason is that TF is a two-step process which requires the wavefronts to rst curve before energy can be focused. Therefore, the 1-D effects always dominate early in time while transverse focusing dominates late in time. In terms of the magnitudes of (note the phases are different), the transition occurs when (57) For tenuous plasmas, i.e., , this is typically a is generally so small number. In beatwave excitation, low, and the interaction time so large that TF almost always dominates. From (56b), 100% modulations to occur within a Rayleigh length from TF when is near unity. This could have deleterious consequences to the multiple pulse excitation process [55][58] since it requires accurate phase relationships to be maintained between each subsequent pulse. On the other hand, in single-frequency experiments, the RFS0 instability may have saturated before the transition to the TF dominated regime occurs. In this case SM, which is entirely the result of TF, may not occur. We reiterate that TF can occur in RFS0. The signature for the SM regime is not that TF occurs, but that the phase shift between and

(59) Substituting the earlier denitions for into (59) gives and and (24)

(60) and . Note that where we have dened (no TF) we recover (25) while if (no LB if or PA) we recover (52). Equation (60) cannot be solved exactly. However, we can calculate its asymptotic behavior from stationary phase arguments. We assume solutions of the form for and obtain the dispersion relation (61) The stationary phase conditions are (62a) and (62b) The solutions to (62b) can be seperated into two regimes is large depending on whether the parameter or small compared to unity. In the small limit (63) Substituting (63) and (62a) into the phase factor leads to the asymptotic behavior (64) In the opposite limit where is small, (65)

1950

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

Substituting this form for asymptotic behavior

into the phase factor gives the

V. IMPLICATIONS

OF

EXPERIMENTS

(66) In the limit that , we recover the RFS0 growth while in the opposite limit, , we recover both the SM growth and imaginary phase terms. Note that early and late in time. Consequently, just as in the constant example, 1-D effects dominate early and TF dominates late in time. The transition occurs when . At this point in time, (64) gives the number of -foldings which have already occurred

Bearing in mind the above assumptions, we are now in a position to make remarks regarding recent acceleration experiments [39][42]. We assume that an instability is saturated after 5 -foldings based on PIC simulation results which show wavebreaking to occur at this time. Based on this assumption and (68), RFS0 dominates if (71a) and from (70) SM dominates when (71b) In what follows, we dene . For intermediate values of 3 to 0.4, RFS0 is still occuring along with TF. The denition for SM is that the phase shift between and is and this regime is relevant when inequality (71b) is satised. Note that these inequalities are actually energy not power thresholds. We now use the parameters of several recent self-trapped electron acceleration experiments in inequalities (71a) and (71b). In each experiment, multiple cascade peaks and multimegaelectronvolt electrons are seen only before inequality (71b) is satised. The correlation between the occurrence of anti-Stokes satellites and of megaelectronvolt electrons was demonstrated rst in [39]. In this experiment, a short-pulse laser was incident on a helium gas jet. The laser tunnel ionized the gas to form a uniform underdense plasma which was fully ionized. In this experiment, one or two anti-Stokes satellites and 2-MeV electrons were observed when 3.5, 0.6, 1, and 2.3. For these parameters 16, so inequality (71a) was exceeded by a factor of ve. On the other hand, when was dropped to around unity the anti-Stokes and electron signals disappeared; in this case, 2.1. This illustrates the strong dependence of on density. In [40], a similar experimental setup was used and once again anti-Stokes and megaelectronvolt electrons were detected. In this case, the laser had more energy and the length of the gas (now hydrogen) jet, i.e., the plasma, was increased. As a result, the number of anti-Stokes satellites was signicantly greater and copious amounts of electrons were observed to energies as high as 43 MeV. These observations occurred for 1020, 0.8, 1 and 0.51.4. Over this range of parameters, is between 1.4 and 128. Recently, another underdense experiment was carried out at the University of Michigan [42]. They also found a correlation between anti-Stokes satellites and megaelectronvolt electrons. In this experiment, the parameters were 17, 0.4, 1, and 0.52.3. As in the previous experiments, anti-Stokes and megaelectronvolt electrons were only observed when exceeded 13. Therefore, we argue RFS0 was important in each experiment. In fact, in all three experiments, inequality (71a) can be used as a rough threshold condition (it is within a factor of 2) for the observation of cascading and megaelectronvolt electrons. The fact that RFS0 was important is consistent with the fact that SM is a nonresonant interaction, so it is less likely to generate multiple cascading and multi-megaelectronvolt

(67) We evaluate (67) at the back of the pulse where the number of -foldings is largest and rewrite it as (68) where is the power in units of TWs, is the laser pulse duration in units of s, is the wavelength in units of ms and is the density in units of 10 cm . We will comment on the implication of (68) to experiments shortly. As exceeds unity and TF begins to dominate then the asymptotic growth changes. However, this is not equivalent to a transition into the SM-dominated regime. The condition for this is that the imaginary phase term in (66) reduce to that in (54). To be more precise, recall that the denition for RFS0 is there be a phase shift between and , while in SM there is an additional phase shift of for a total phase shift of . To clearly see the transition between RFS0 and SM, we use (66) in (24) to determine that for , the phase shift between and , is (69) . where The SM-dominated regime occurs when the term is small compared to . We dene the transition into the SM regime to be when this term is . By this denition the transition occurs when exceeds 6.69. The number of foldings at the transition is obtained by substituting the value of corresponding to this value of in (66), giving (70) and if the number of -foldings is large then the instability will have saturated before the transition into the SM-dominated regimes occured. We note that others have given similar scaling laws [25], [26] (different coefcient), however, these were based on comparing asymptotic growth rates, while here we have an expression which provides a more precise transition between the various RFS0 and SM regimes.

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1951

electrons because the phase velocity of the plasma wave it generates has a spatio-temporal-dependent phase velocity. We close this section with a few comments. First, this analysis is only stricly valid for and for times before saturation occurs. In some of these experiments, , and this can lower the growth rate [23][26]. Second, the asymptotic solutions have restrictions. The front part of the pulse, i.e., smaller values of , makes a transition to SM earlier than the back of the pulse. This will obviously effect how the back of the pulse evolves. The asymptotic solutions given here only include the stationary phase part. Saddle point integration provides other factors to the asymptotic solutions. Third, the theory assumes diffraction-limited beams, while the beam quality varies from experiment. Last, the time of saturation and the transverse prole depends on the noise source. There have already been several types of noise sources which have been identied [22], [25], [26], [34], [35], [38], [51][53]. Therefore, to understand current experiments, PIC simulations [34], [35], [51][53] are required. These indicate that RFS0 plays an important role and that the dominant saturation mechanism is wavebreaking. It is worthwhile to note that some of these PIC simulations were carried out before many of the above experiments and that many of their predictions have been born out in the subsequent experiments.

analysis, it seems worthwhile to try and extend it to the fully relativistic limit, . This will require precise knowledge of the fully nonlinear phase and group velocities of light in plasmas. To date we have not made a successful treatment for . Another area to investigate is the role of photon acceleration during SPM. In particular, a modulation to could lead to a from PA, which in turn could lead to further modulations to from PA. This two-step process was neglected in Section II-B, but it may be important. Still another area for future research is implementing hosing and the nonresonant regime of RFS0 into these physical pictures. Last, we point out that this analysis not only allows one to rederive known growth rates, it also makes it possible to identify new processes and to properly analyze old problems. For example, it is possible for SF to occur because of transverse variations to caused by PA [61], and the phenomenon of resonant relativistic self-focusing [33] can be coupled to SM [62]. Note Added in Proof: The eld of nonlinear optics of plasmas is still developing rapidly, as illustrated by the numerous experimental results on electron acceleration which have been obtained since the original submission data, e.g., [63][65]. APPENDIX A LOCAL ACTION CONSERVATION In this paper, we have assumed that the action, i.e., photon number, is conserved locally. In this appendix, we derive a local conservation law for the action using similar reasoning as others when they derived a global conservation law [38], [45]. We begin with the basic quasi-static equation for the normalized vector potential (A1) where and are the speed of light frame variables dened in the paper. Next, we assume to be of the to obtain the envelope equation form (A2) We next dene the operator

VI. SUMMARY

AND

FUTURE WORK

In this paper, we have given simple physical pictures for the nonlinear optics of plasmas at relativistic intensities. Here, relativistic refers to either the electrons oscillating at relativistic energies or the plasma dielectric function having a relativistic phase velocity. In particular, Raman forward scattering (RFS0), envelope self-modulation (SM), relativistic self-focusing/lamentation (SF), and relativistic self-phase modulation (SPM) were analyzed. The analyses used only the concepts of longitudinal bunching (LB), transverse focusing (TF), and photon acceleration (PA), together with the assumption of conservation of photons, i.e., the classical action. Using just these concepts and the well-known phase and group velocity of light, the growth rates for RFS0, SM, SF, and SPM were rederived. Direct comparison between each instability is now possible because they were derived from the same set of physical phenomenon (although the derivations are not rigorous). It was shown that RFS0 is caused by equal amounts of LB and PA (both are 1-D effects). On the other hand, SPM is caused by PA followed by LB. It is therefore a two-step process so it has a lower growth rate than RFS0. It was also shown that SF is caused by TF from relativistic mass corrections to the light waves phase velocity. Finally, SM is the result of TF in a nonresonant plasma wave. Importantly, it was shown that TF in RFS0 generated plasma waves is not the same as SM. A comparison between RFS0 and SM showed that in self-trapped electron experiments RFS0 plays a key role. This could change in future experiments. We close with a few remarks regarding future work. This analysis is only strictly valid in the weakly relativistic limit, . Based on the success of this weakly relativistic

and take the linear combination of (A2)

to get

(A3) To obtain a global conservation law, (A3) can be integrated ; but to obtain the more powerful over all space, i.e., local conservation law, we need to rewrite several terms in

1952

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 11, NOVEMBER 1997

terms of complete derivatives. The rst two terms of (A3) can be combined as (A4) which can be rewritten as

In conclusion, in this Appendix, we have derived the local conservation law

(A13) and have given arguments that it is a conservation law for the action. ACKNOWLEDGMENT The author acknowledges useful discussions with K.-C. Tzeng and Dr. T. Katsouleas, Dr. C. D. Decker, Dr. C. Joshi, Dr. J. M. Dawson, Dr. E. Esarey, and Dr. G. Shvets. REFERENCES
[1] D. W. Forslund, J. M. Kindel, and E. L. Lindman, Theory of stimulated scattering processes in laser-irradiated plasmas, Phys. Fluids, vol. 18, pp. 10021016, 1975 and references therein. [2] , Plasma simulation studies of stimulated scattering processes in laser-irradiated plasmas, Phys. Fluids, vol. 18, pp. 10171030, 1975. [3] J. F. Drake, P. K. Kaw, Y. C. Lee, G. Schmidt, C. S. Liu, and M. N. Rosenbluth, Parametric instabilities of electromagnetic waves in plasmas, Phys. Fluids, vol. 17, pp. 778785, 1974. [4] C. Max, J. Arons, and A. B. Langdon, Self-modulation and selffocusing of electromagnetic waves in plasmas, Phys. Rev. Lett., vol. 33, pp. 209212, 1974. [5] Y. R. Shen, The Principles of Nonlinear Optics. New York: Wiley, 1984. [6] P. K. Kaw, G. Schmidt, and T. Wilcox, Filamentation and trapping of electromagnetic radiation in plasmas, Phys. Fluids, vol. 16, pp. 15221525, 1973. [7] B. I. Cohen and C. E. Max, Stimulated scattering of light by ion modes in a homogeneous plasma: Space-time evolution, Phys. Fluids, vol. 22, pp. 11151131, 1979. [8] C. E. Max, Strong self-focusing due to the ponderomotive force in plasmas, Phys. Fluids, vol. 19, pp. 7477, 1976. [9] R. Bingham and C. N. Lashmore-Davies, Self-modulation and lamentation and lamentation of electromagnetic waves in a plasma, Nucl. Fusion, vol. 16, pp. 6772, 1976; and references therein. [10] D. Strickland and G. Mourou, Compression of amplied chirped optical pulses, Opt. Commun., vol. 56, pp. 219221, 1985. [11] T. Tajima and J. M. Dawson, Laser electron accelerator, Phys. Rev. Lett., vol. 43, pp. 267270, 1979. [12] C. Joshi, W. B. Mori, T. Katsouleas, J. M. Dawson, J. M. Kindel, and D. W. Forslund, Ultrahigh gradient particle acceleration by intense laser-driven plasma density waves, Nature, vol. 311, pp. 525529, 1984. [13] E. Esarey, P. Sprangle, J. Krall, and A. Ting, Overview of plasma-based accelerator concepts, IEEE Trans. Plasma Sci., vol. 24, pp. 252288, 1996; and references therein. [14] D. W. Forslund, J. M. Kindel, W. B. Mori, C. Joshi, and J. M. Dawson, Two-dimensional simulations of single-frequency and beat-wave laserplasma heating, Phys. Rev. Lett., vol. 54, pp. 558561, 1985. [15] C. Joshi, T. Tajima, J. M. Dawson, H. A. Baldis, and N. A. Ebrahim, Forward Raman instability and electron acceleration, Phys. Rev. Lett., vol. 47, pp. 12851288, 1981. [16] K. Estabrook and W. L. Kruer, Theory and simulation of one dimensional raman back and forward scatter, Phys. Fluids, vol. 26, pp. 18921903, 1983. [17] W. B. Mori, Computer simulations on the acceleration of electrons by fast large-amplitude plasma waves driven by laser beams, Masters thesis, Univ. California at Los Angeles, 1984. [18] T. M. Antonsen, Jr., and P. Mora, Self-focusing and Raman scattering of laser pulses in tenuous plasmas, Phys. Rev. Lett., vol. 69, pp. 22042207, 1992. [19] , Self-focusing and Raman scattering of laser pulses in tenuous plasmas, Phys. Fluids B, vol. 5, pp. 14401452, 1993. [20] P. Sprangle, E. Esarey, J. Krall, and G. Joyce, Propagation and guiding of intense laser pulses in plasmas, Phys. Rev. Lett., vol. 69, pp. 22002203, 1992.

(A5) The third term of (A3) can also be rewritten as a complete derivative (A6) Recombining (A5) and (A6) gives a local conservation law in the coordinates

(A7) Integrating (A7) over law and gives the global conservation

(A8) The term in brackets can be identied as the action by rewriting ; substituting this form for into the term in brackets in (A8) gives

(A9) which is times the action, , since is the instantaneous frequency. A local conservation of law of a quantity, , has the generic form (A10) where is the ux of and Therefore, we can interpret is the ow velocity of .

(A11) and (A12) Note, to lowest order the velocity of the action is equal to the group velocity of light, so in the speed of light frame and . Using these denitions, it is straightforward to verify that interpretations (A11) and (A12) are correct.

MORI: THE PHYSICS OF THE NONLINEAR OPTICS OF PLASMAS FOR SHORT-PULSE LASERS

1953

[21] J. Krall, A. Ting, E. Esarey, P. Sprangle, and G. Joyce, Enhanced acceleration in a self-modulated laser wakeeld accelerator, Phys. Rev. E, vol. 48, pp. 21572161, 1993. [22] N. E. Andreev, L. M. Gorbunov, V. I. Kirsanov, A. A. Pogosova, and R. R. Ramazashvili, Resonant excitation of wakeelds by a laser pulse in a plasma, JETP Lett., vol. 55, pp. 571576, 1992. [23] W. B. Mori, C. D. Decker, D. E. Hinkel, and T. Katsouleas, Raman forward scattering of short-pulse high-intensity lasers, Phys. Rev. Lett., vol. 72, pp. 14824185, 1994. [24] C. D. Decker, W. B. Mori, T. Katsouleas, and D. E. Hinkel, Spatial temporal theory of Raman forward scattering, Phys. Plasmas, vol. 3, pp. 13601372, 1996. [25] N. E. Andreev, L. M. Gorbunov, V. I. Kirsanov, A. A. Pogosova, and A. S. Sakharov, Theory of the resonance modulational instability of short laser pulses in a plasma and plasma channels, Plasma Phys. Rep., vol. 22, pp. 379389, 1996. [26] A. S. Sakharov and V. I. Kirsanov, Theory of Raman scattering for a short ultrastrong laser pulse in a rareed plasma, Phys. Rev. E, vol. 49, pp. 32743282, 1994. [27] C. J. McKinstrie and R. Bingham, Stimulated Raman forward scattering and the relativistic modulational instability of light waves in rareed plasma, Phys. Fluids B, vol. 4, pp. 26262633, 1992. [28] E. Esarey, J. Krall, and P. Sprangle, Envelope analysis of intense laser pulse self-modulation in plasmas, Phys. Rev. Lett., vol. 72, pp. 28872890, 1994. [29] P. Sprangle, J. Krall, and E. Esarey, Hose-modulation instability of intense laser pulses in plasmas, Phys. Rev. Lett., vol. 73, pp. 35443547, 1994. [30] G. Shvets and J. S. Wurtele, Instabilities of short-pulse laser propagation through plasma channels, Phys. Rev. Lett., vol. 73, pp. 35403543, 1994. [31] P. Sprangle, C. M. Tang, and E. Esarey, Relativistic self-focusing of short-pulse radiation beams in plasmas, IEEE Trans. Plasma Sci., vol. PS-15, pp. 145153, 1987. [32] G.-Z. Sun, E. Ott, Y. C. Lee, and P. Guzdar, Self-focusing of short intense pulses in plasmas, Phys. Fluids, vol. 30, pp. 526532, 1987. [33] W. B. Mori, C. Joshi, J. M. Dawson, D. W. Forslund, and J. M. Kindel, Evolution of self-focusing of intense electromagnetic waves in plasma, Phys. Rev. Lett., vol. 60, pp. 12981301, 1988. [34] C. D. Decker, W. B. Mori, and T. Katsouleas, Particle-in-cell simulations of Raman forward scattering from short-pulse high-intensity lasers, Phys. Rev. E, vol. 50, pp. R3338R3341, 1994. [35] C. D. Decker, W. B. Mori, K. C. Tzeng, and T. Katsouleas, Evolution of ultra-intense, short-pulse lasers in underdense plasmas, Phys. Plasmas, vol. 3, pp. 20472056, 1996. [36] S. C. Wilks, J. M. Dawson, W. B. Mori, T. Katsouleas, and M. E. Jones, Photon accelerator, Phys. Rev. Lett., vol. 62, pp. 26002603, 1989. [37] E. Esarey, J. Krall, and P. Sprangle, Frequency shifts induced in laser pulses by plasma waves, Phys. Rev. A, vol. 42, pp. 35263531, 1990. [38] S. V. Bulanov, I. N. Inovenkov, V. I. Kirsanov, N. M. Naumova, and A. S. Sakharov, Nonlinear depletion of ultrashort and relativistically strong laser pulses in an underdense plasma, Phys. Fluids B, vol. 4, pp. 19351941, 1992. [39] C. A. Coverdale, C. B. Darrow, C. D. Decker, W. B. Mori, K. C. Tzeng, K. A. Marsh, C. E. Clayton, and C. Joshi, Propagation of intense subpicosecond laser pulses through underdense plasmas, Phys. Rev. Lett., vol. 74, pp. 46594662, 1995. [40] A. Modena, Z. Najmudin, A. E. Dangor, C. E. Clayton, K. A. Marsh, C. Joshi, V. Malka, C. B. Darrow, C. Danson, D. Neely, and F. N. Walsh, Electron acceleration from the breaking of relativistic plasma waves, Nature, vol. 337, pp. 606608, 1995. [41] K. Nakajima, D. Fisher, T. Kawakubo, H. Nakanishi, A. Z. Ogata, Y. Kato, Y. Kitagawa, R. Kodama, K. Mima, H. Shiraga, K. Suzuki, K. Yamakawa, T. Zhang, Y. Sakawa, T. Shoji, Y. Niahisida, N. Yugami, M. Downer, and T. Tajima, Observation of ultrahigh gradient electron acceleration by a self-modulated intense short laser pulse, Phys. Rev. Lett., vol. 74, pp. 44284431, 1995.

[42] D. Umstadter, S. Y. Chen, A. Maksimchuk, G. Mourou, and R. Wagner, Nonlinear optics in relativistic plasmas and laser wakeeld acceleration of electrons, Science, vol. 273, pp. 472475, 1996. [43] C. D. Decker and W. B. Mori, Group velocity of large-amplitude electromagnetic waves in a plasma, Phys. Rev. Lett., vol. 72, pp. 490493, 1994. , Group velocity of large-amplitude electromagnetic waves in a [44] plasma,Phys. Rev. E, vol. 51, pp. 13641375, 1995. [45] T. Antonsen and P. Mora, Kinetic modeling of intense, short laser pulses propagating in tenuous plasmas, Phys. Plasmas, vol. 4, pp. 217229, 1997. [46] W. B. Mori, Generation of tunable radiation using an underdense ionization front, Phys. Rev. A, vol. 44, pp. 51185121, 1991. [47] E. Esarey, G. Joyce, and P. Sprangle, Frequency up-shifting of laser pulses by copropagating ionization fronts, Phys. Rev. A, vol. 44, pp. 39083911, 1991. [48] P. Sprangle, A. Ting, and C. M. Tang, Analysis of radiation focusing and steering in the free-electron laser by use of a source-dependent expansion, Phys. Rev. A, vol. 36, pp. 27732781, 1987. [49] P. Sprangle, E. Esarey, A. Ting, and G. Joyce, Laser wakeeld acceleration and relativistic optical guiding, Appl. Phys. Lett., vol. 53, pp. 21462148, 1988. [50] D. Anderson and M. Bonnedal, Variational approach to nonlinear selffocusing of Gaussian laser beams, Phys. Fluids, vol. 22, pp. 105109, 1979. [51] C. D. Decker, W. B. Mori, K.-C. Tzeng, and T. C. Katsouleas, Modeling single-frequency laser-plasma acceleration using particle-incell simulations: The physics of beam breakup, IEEE Trans. Plasma Sci., vol. 24, pp. 379392, 1996. [52] K.-C. Tzeng, W. B. Mori, and C. D. Decker, Anomalous absorption and scattering of short-pulse high-intensity lasers in underdense plasmas, Phys. Rev. Lett., vol. 76, pp. 33323335, 1996. [53] K.-C. Tzeng and W. B. Mori, Self-trapped electron acceleration from wakeelds generated by short-pulse lasers, Bull. Amer. Phys. Soc., vol. 40, no. 11, p. 1660, 1995. [54] P. Gibbon and A. R. Bell, Cascade focusing in the beat-wave accelerator, Phys. Rev. Lett., vol. 61, pp. 15991562, 1988. [55] D. Umstadter, E. Esarey, and J. Kim, Nonlinear plasma waves resonantly driven by optimized laser pulse trains, Phys. Rev. Lett., vol. 72, pp. 12241227, 1994. [56] V. I. Berezhiani and I. G. Murusidze, Relativistic wakeeld generation by an intense laser pulse in a plasma, Phys. Rev. A, vol. 148, pp. 338340, 1990. [57] S. Dalla and M. Lontano, Large amplitude plasma wave excitation by means of sequences of short laser pulses, Phys. Rev. E, vol. 49, pp. R1819R1822, 1994. [58] G. Bonnaud, D. Teychenne, and J. L. Bobin, Wake-eld effect induced by multiple laser pulses, Phys. Rev. E, vol. 50, pp. R36R39, 1994. [59] T. Katsouleas and W. B. Mori, Wave-breaking amplitude of relativistic oscillation in a thermal plasma, Phys. Rev. Lett., vol. 61, pp. 9093, 1988. [60] W. B Mori and T. Katsouleas, Wavebreaking of longitudinal plasma oscillations, Physica Scripta, vol. T30, pp. 127133, 1990. [61] W. B. Mori, unpublished. [62] W. B. Mori, unpublished. [63] A. Ting, C. I. Moore, K. Krushelnick, C. Manka, E. Esarey, P. Sprangle, R. Hubbard, H. R. Burris, R. Fischer, and M. Baine, Plasma Wakeeld generation and electron acceleration in a self-modulated laser Wakeeld accelerator experiment, Phys. Plasmas, vol. 4, pp. 18891898, 1997. [64] R. Wagner, S.-Y. Chen, A. Maksimchuk, and D. Umstadter, Electron acceleration by a laser Wakeeld in a relativistic self-guided channel, Phys. Rev. Lett., vol. 78, pp. 31253128, 1997. [65] D. Gordon, K.-C. Tzeng, C. E. Clayton, A. E. Dangor, V. Malka, K. A. Marsh, A. Modens, W. B. Mori, P. Muggli, Z. Najmudin, and C. Joshi, Observation of electron energies beyond the linear dephasing limit from a laser-excited relativistic plasma wave, unpublished.

W. B. Mori, photograph and biography not available at the time of publication.

You might also like