betaFeedback

WorksPhysics and Astronomy

Interaction of Strong X-mode Waves with Magnetized Pair Plasmas

Abstract

The interaction of strong (a0>1a_0>1), superluminal X-mode electromagnetic waves with magnetized pair plasma falls into two regimes set by εB≡a0ωB/ω0\varepsilon_{\rm B}\equiv a_0ω_{\rm B}/ω_0, the ratio of the wave strength parameter to its frequency, in units of cyclotron frequency. For εB<1\varepsilon_{\rm B}<1 the wave propagates until induced scattering disrupts it. Using analytical theory and kinetic simulations, we show that the background magnetic field introduces a restoring frequency ΩBΩ_{\rm B} to scattering; when ΩBΩ_{\rm B} exceeds the scattering rate, the unattenuated length grows from ≃εp−2/3\simeq\varepsilon_{\rm p}^{-2/3} to ≃εB1/2/εp\simeq\varepsilon_{\rm B}^{1/2}/\varepsilon_{\rm p} wavelengths (εp≡a0ωp/ω0\varepsilon_{\rm p}\equiv a_0ω_{\rm p}/ω_0; ωpω_{\rm p} is the plasma frequency). When a02>εB>1a_0^2>\varepsilon_{\rm B}>1, the plasma current is nonlinear, and the wave is damped while riding a motional electric field, so one half-cycle appears attenuated. This framework has implications for the duration of fast radio bursts and whether they can escape their surroundings.

Introduction—Magnetized electron-positron pairs fill the magnetospheres of neutron stars and black holes [45, 8] which emit coherent radio signals with the highest brightness temperatures (∼1033−1041\sim10^{33}-10^{41} K) known: giant pulses [12] and fast radio bursts [FRBs; 23, 33, 13, 34]. Near their sources, such waves are ‘strong’: i.e., the Lorentz-invariant strength parameter a0≡eE0/mec ω0a_0\equiv eE_0/m_{\rm e}c\,\omega_0 greatly exceeds unity, so electrons oscillate relativistically in the wave field [16, 22]; here, E0E_0 is the amplitude of the wave electric field, ω0=2πc/λ0\omega_0=2\pi c/\lambda_0 its frequency, and mem_{\rm e} and ee are the mass and charge of the electron. Whether such waves emerge through the pair rich surroundings constrains where and how they are produced [3, 36, 6, 28, 39, 41, 43].

Besides a0a_0, the interaction of electromagnetic (EM) waves with magnetized pair plasmas depends on three frequencies, defined in the frame where the plasma ahead of the pulse is at rest. These are, ω0\omega_{\rm 0}, the plasma frequency ωp≡(8πn0e2/me)1/2\omega_{\rm p}\equiv(8\pi n_0e^2/m_{\rm e})^{1/2} (n0n_0 is the proper density of each species), and the cyclotron frequency ωB≡eB0/mec\omega_{\rm B}\equiv eB_0/m_{\rm e}c of the upstream field B0B_0. The plasma magnetization is σ≡ωB2/ωp2\sigma\equiv\omega_{\rm B}^2/\omega_{\rm p}^2. The cold X-mode dispersion relation in pair plasmas allows a subluminal fast-magnetosonic (FMS) branch at ω0<ωB\omega_{\rm 0}<\omega_{\rm B} and a superluminal, vacuum-like branch at ω02>ωp2+ωB2\omega_{\rm 0}^2>\omega_{\rm p}^2+\omega_{\rm B}^2, separated by a stop band [29].

The low-frequency band has been well studied: large-amplitude FMS waves can steepen into shocks [1, 26, 5], and decay into Alfvén waves through resonant three-wave interactions [14, 42]. For ω0≪ωB\omega_{\rm 0}\ll\omega_{\rm B}, the background field suppresses the particle quiver motion, and induced scattering is weakened by gyroradius and Debye-screening effects [31, 20, 32]; these kinetic theories assume non-relativistic particles in the wave field, a0ω0/ωB≪1a_0\omega_{\rm 0}/\omega_{\rm B}\ll1.

On the superluminal branch, εB≡a0ωB/ω0\varepsilon_{\rm B}\equiv a_0\omega_{\rm B}/\omega_{\rm 0} separates two regimes. For 1<εB<a021<\varepsilon_{\rm B}<a_0^2 (i.e., ωB/a0<ω0<a0ωB\omega_{\rm B}/a_0<\omega_{\rm 0}<a_0\omega_{\rm B}) the plasma current is a nonlinear function of the wave field [39] and the pulse can be absorbed by synchrotron absorption [25, 41], equivalent to the stochastic heating of test particles [4, 6]; the exact range of frequencies where this happens is debated. For εB<1\varepsilon_{\rm B}<1, the current is linear and the particles exhibit relativistic oscillation in the wave field when a0>1a_0>1 [39]. In the latter regime, parametric instabilities such as filamentation [37, 38, 19] and induced scattering [30, 43, 18], can still limit wave propagation. In unmagnetized pair plasmas, εp≡a0ωp/ω0\varepsilon_{\rm p}\equiv a_0\omega_{\rm p}/\omega_{\rm 0} controls the nonlinearity for all a0a_0: when εp<1\varepsilon_{\rm p}<1 induced Compton scattering limits the unattenuated pulse to ≃εp−2/3\simeq\varepsilon_{\rm p}^{-2/3} wavelengths; when εp>1\varepsilon_{\rm p}>1, the head of the pulse launches a shock and gets dissipated [43]. What dictates the nature of strong, superluminal EM wave propagation in magnetized plasmas when εB<1\varepsilon_{\rm B}<1? This regime applies to FRBs propagating in magnetar winds beyond ∼1012\sim10^{12} cm, where ω0>ωB\omega_{\rm 0}>\omega_{\rm B} in the magnetar wind-frame and εB<1\varepsilon_{\rm B}<1 [41]. Magnetized induced-scattering theories have thus far assumed ω0<ωB\omega_{\rm 0}<\omega_{\rm B}, weak waves [30, 31, 32], or omit plasma response [4, 35].

In this Letter, we combine analytical theory and kinetic simulations to study monochromatic X-mode pulses propagating along x^\hat x, with electric field oscillating along y^\hat y, across a uniform magnetic field along z^\hat z. This is akin to the geometry of a pulse propagating radially across the toroidal field of a neutron-star wind [9]. We derive the induced Compton scattering rate for a0>1a_0>1; the propagation length is found to depend on εp\varepsilon_{\rm p}, and the ratio of a magnetic restoring frequency ΩB\Omega_{\rm B} to the scattering rate η\eta. Kinetic plasma simulations spanning σ=10−2\sigma=10^{-2}–10410^{4} are performed to compare with our analytical scalings.

Analytical theory—We consider a monochromatic, linearly polarized X-mode EM pulse traveling along +x^+\hat{x} in pair plasma. Its oscillating electric field is along y^\hat{y}, its magnetic field is along z^\hat{z}, and so is the uniform background magnetic field, B0z^B_0\hat{z}. In the upstream plasma rest frame, its frequency and wavenumber are ω0\omega_0 and k0k_0. We adopt a cold two-fluid model for the plasma in which the electrons and positrons have the same number densities, the same velocities along x^\hat{x}, and move in the xyxy plane with opposite velocities along y^\hat{y}. We assume that all physical quantities depend only on xx and tt; taking c=1c=1, the governing equations reduce to [15, 39]

∂∂t(γn)+∂∂x(nux)=0,∂ux∂t+∂γ∂x=uynn0ωB,∂2uy∂t2−∂2uy∂x2+ωp2nn0uy=−ωB[∂∂t(nuxn0)+∂∂x(γnn0)]  ,\begin{align} & \frac{\partial}{\partial t}\left(\gamma n\right) + \frac{\partial}{\partial x}\left(nu_x\right) = 0, \tag*{(1)} \\ & \frac{\partial u_x}{\partial t} + \frac{\partial \gamma}{\partial x} = u_y\frac{n}{n_0} \omega_{\rm B}, \tag*{(2)} \\ & \frac{\partial^2 u_y}{\partial t^2} - \frac{\partial^2 u_y}{\partial x^2} +\omega_{\rm p}^2\frac{n}{n_0}u_y = -\omega_{\rm B}\left[\frac{\partial}{\partial t}\left( \frac{n u_x}{n_0}\right) + \frac{\partial}{\partial x}\left(\frac{\gamma n}{n_0}\right)\right] \;, \tag*{(3)} \end{align}

where nn is the proper density, uxu_x and uyu_y are the xx and yy components of the particles’ four-velocity with a Lorentz factor γ=(1+ux2+uy2)1/2\gamma=(1+u_x^2+u_y^2)^{1/2}.

Because we consider superluminal waves (i.e., ω0/k0>1\omega_{\rm 0}/k_0>1), we can boost along x^\hat{x} with a velocity k0/ω0k_0/\omega_{\rm 0}. In this ‘wave frame’, the EM wave vector (and the spatial dependence in the equations) vanishes and the wave oscillates with frequency ω=ω02−k02\omega=\sqrt{\omega_0^2-k_0^2} [11]; however, the spatial dependence of any perturbations remains. Retaining variations on a timescale ∼1/ω\sim1/\omega in the wave frame and linearizing Eqs. (1)-(3) yields the X-mode dispersion relation [29, 39],

ω2=Aωp2;     A≡ω2ωp2=11−ωB2/ω02.\begin{align} \omega^2 = {\cal A}\omega_{\rm p}^2; \ \ \ \ \ {\cal A} \equiv \frac{\omega^2}{\omega_{\rm p}^2} = \frac{1}{1-\omega_{\rm B}^2/\omega_{\rm 0}^2}. \tag*{(4)} \end{align}

A{\cal A} ranges from 1 to 1+σ1+\sigma for the superluminal branch. The unperturbed solution of Eqs. (1)-(3) in the wave frame is independent of xx, and can be written as, γˉnˉ=γ0n0\bar{\gamma} \bar{n}=\gamma_0n_0, uˉx=−u0+χ(t)\bar{u}_x = -u_0 + \chi(t), and uˉy=a^2(e−iωt+eiωt)\bar{u}_y=\frac{\hat{a}}{2}(e^{-i\omega t}+e^{i\omega t}). Here, u0=k0/ωu_0=k_0/\omega, γ0=1+u02=ω0/ω\gamma_0=\sqrt{1+u_0^2}=\omega_{\rm 0}/\omega, a^=Aa0\hat{a}= {\cal A}a_0, and γˉ=1+uˉx2+uˉy2\bar{\gamma}=\sqrt{1 + \bar{u}_x^2 + \bar{u}_y^2}. Given this, Eq. (2) becomes,

∂χ∂t=ωB(γ0γˉ)uˉy,\begin{align} \frac{\partial\chi}{\partial t} = \omega_{\rm B}\left(\frac{\gamma_0}{\bar{\gamma}}\right)\bar{u}_y, \tag*{(5)} \end{align}

where,

γ0γˉ≈1+u0γ0χγ0+12(3u02γ02−1)χ2γ02−uˉy22γ02+⋯\begin{align} \dfrac{\gamma_0}{\bar{\gamma}} \approx 1 + \frac{u_0}{\gamma_0}\frac{\chi}{\gamma_0} + \frac{1}{2}\left(\frac{3u_0^2}{\gamma_0^2} - 1 \right)\frac{\chi^2}{\gamma_0^2} - \frac{\bar{u}_y^2}{2\gamma_0^2} + \cdots \tag*{(6)} \end{align}

This yields the leading order solution of χ(t)=ωBωa^sin⁡ωt\chi(t) = \frac{\omega_{\rm B}}{\omega}\hat{a}\sin{\omega t}. Eq. (3) becomes,

∂2uˉy∂t2+ωp2γ0γˉuˉy=−ωBγ0∂∂t(−u0+χγˉ)  .\begin{align} \frac{\partial^2 \bar{u}_y}{\partial t^2}+\omega_{\rm p}^2\frac{\gamma_0}{\bar{\gamma}}\bar{u}_y = -\omega_{\rm B}\gamma_0\frac{\partial}{\partial t}\left( \frac{-u_0+\chi}{\bar{\gamma}}\right) \;. \tag*{(7)} \end{align}

We introduce the following small perturbations to the solutions to study the stability of the unperturbed solution: γn=γ0n0+δρ\gamma n=\gamma_0n_0+\delta\rho, ux=uˉx+δuxu_x = \bar{u}_x + \delta u_x, and uy=uˉy+δuyu_y = \bar{u}_y + \delta u_y. Inserting this in Eq. (1)-(3) and upon linearizing, we get,

∂∂t(δργ0n0)+∂∂x[1γ0δ(nuxn0)]=0,∂δux∂t+∂δγ∂x=ωBδ(nuyn0),∂2δuy∂t2−∂2δuy∂x2+ωp2δ(nuyn0)=−ωB[∂∂tδ(nuxn0)+∂∂xδρn0],  \begin{align} & \frac{\partial}{\partial t} \left(\frac{\delta\rho}{\gamma_0n_0}\right) + \frac{\partial}{\partial x}\left[\frac{1}{\gamma_0}\delta\left(\frac{nu_x}{n_0}\right)\right] = 0, \tag*{(8)}\\ & \frac{\partial \delta u_x}{\partial t} + \frac{\partial \delta\gamma}{\partial x} = \omega_{\rm B}\delta\left(\frac{nu_y}{n_0}\right), \tag*{(9)}\\ & \frac{\partial^2\delta u_y}{\partial t^2} - \frac{\partial^2\delta u_y}{\partial x^2} + \omega_{\rm p}^2 \delta\left(\frac{nu_y}{n_0}\right) = -\omega_{\rm B}\left[ \frac{\partial}{\partial t}\delta\left(\frac{nu_x}{n_0}\right) + \frac{\partial}{\partial x}\frac{\delta\rho}{n_0} \right], \; \tag*{(10)} \end{align}

where,

δ(nuxn0)=γ0[uˉxγˉδργ0n0+(1+uˉy2)δux−uˉxuˉyδuyγˉ3]≈γ0[−vxδργ0n0+2(1−vx)γ0δux+vxuˉyγ02δuy],\begin{align} & \delta\left(\frac{n u_x}{n_0}\right) = \gamma_0\left[ \frac{\bar{u}_x}{\bar{\gamma}}\frac{\delta \rho}{\gamma_0n_0} + \frac{(1+\bar{u}_y^2)\delta u_x - \bar{u}_x\bar{u}_y\delta u_y}{\bar{\gamma}^3} \right]\nonumber\\ & \approx \gamma_0\left[-v_x\frac{\delta \rho}{\gamma_0 n_0} + \frac{2(1-v_x)}{\gamma_0}\delta u_x + \frac{v_x\bar{u}_y}{\gamma_0^2}\delta u_y\right], \tag*{(11)} \end{align}
δ(nuyn0)=γ0[uˉyγˉδργ0n0+(1+uˉx2)δuy−uˉxuˉyδuxγˉ3] ≈γ03γˉ3(1−2u0χγ02+χ2γ02)δuy+uˉyγ0γˉ(δργ0n0+vxγˉδux),δγ=uˉxδux+uˉyδuyγˉ≈−vxδux+uˉyγ0δuy.\begin{align} & \delta\left(\frac{nu_y}{n_0}\right) = \gamma_0\left[ \frac{\bar{u}_y}{\bar{\gamma}}\frac{\delta \rho}{\gamma_0n_0} + \frac{(1 + \bar{u}_x^2)\delta u_y - \bar{u}_x\bar{u}_y\delta u_x}{\bar{\gamma}^3}\right]\ \nonumber\\ & \approx \frac{\gamma_0^3}{\bar{\gamma}^3}\left(1-\frac{2u_0\chi}{\gamma_0^2} + \frac{\chi^2}{\gamma_0^2}\right)\delta u_y + \bar{u}_y\frac{\gamma_0}{\bar{\gamma}}\left(\frac{\delta \rho}{\gamma_0 n_0} + \frac{v_x}{\bar{\gamma}}\delta u_x\right), \tag*{(12)}\\ &\delta \gamma = \frac{\bar{u}_x \delta u_x + \bar{u}_y\delta u_y}{\bar{\gamma}} \approx -v_x\delta u_x + \frac{\bar{u}_y}{\gamma_0}\delta u_y. \tag*{(13)} \end{align}

In order to determine vxv_x, let us first expand uˉx/γˉ=\bar{u}_x/\bar{\gamma}=

−[1+1+uˉy2(u0−χ)2]−1/2≈−1+1+uˉy22(u0−χ)2+O((1+uˉy2)2(u0−χ)4)\begin{align} -\left[1+\frac{1+\bar{u}_y^2}{(u_0-\chi)^2}\right]^{-1/2} \approx -1+\frac{1+\bar{u}_y^2}{2(u_0-\chi)^2} + {\cal O}\left(\frac{(1+\bar{u}_y^2)^2}{(u_0-\chi)^4}\right) \tag*{(14)} \end{align}

and expand

1(u0−χ)2≈1γ02[1+2χγ0+3χ2γ02+⋯ ].\begin{align} \frac{1}{(u_0-\chi)^2} \approx \frac{1}{\gamma_0^2}\left[1+2\frac{\chi}{\gamma_0}+3\frac{\chi^2}{\gamma_0^2}+\cdots\right]. \tag*{(15)} \end{align}

Then, we can take ⟨uˉy2⟩≈a^2/2\langle\bar{u}_y^2\rangle\approx\hat{a}^2/2, ⟨χ⟩=0\langle\chi\rangle=0, ⟨χ2⟩/γ02=(AεB)2/2\langle\chi^2\rangle/\gamma_0^2=({\cal A}\varepsilon_{\rm B})^2/2, ⟨χ(1+uˉy2)⟩=0\langle \chi (1+\bar{u}_y^2)\rangle=0, ⟨(1+uˉy2)χ2⟩=(a^ωB/ω)2(12+a^28)\langle(1+\bar{u}_y^2)\chi^2\rangle=(\hat{a}\omega_{\rm B}/\omega)^2\left(\frac{1}{2}+\frac{\hat{a}^2}{8}\right) and γ0≃u0\gamma_0\simeq u_0, to obtain the following phase-averaged quantity,

⟨uˉxγˉ⟩≈−vx+O[1+a^2/2γ02(AεB)2+(1+a^2/2γ02)2],\begin{align} \left\langle\frac{\bar{u}_x}{\bar{\gamma}}\right\rangle \approx -v_x + {\cal O}\left[\frac{1+\hat{a}^2/2}{\gamma_0^2}({\cal A}\varepsilon_{\rm B})^2 + \left(\frac{1+\hat{a}^2/2}{\gamma_0^2}\right)^2\right], \tag*{(16)} \end{align}

from which we see that vx=(1−1+a^2/22γ02)v_x=\left(1 - \dfrac{1 + \hat{a}^2/2}{2\gamma_0^2}\right); the correction terms in the following quantity,

⟨γ03γˉ3(1−2u0χγ02+χ2γ02)⟩=1+3u02−γ022γ04⟨χ2⟩−3⟨uˉy2⟩2γ02+⋯\begin{align} \left\langle\frac{\gamma_0^3}{\bar{\gamma}^3}\left(1-\frac{2u_0\chi}{\gamma_0^2} + \frac{\chi^2}{\gamma_0^2}\right)\right\rangle = 1 + \frac{3u_0^2-\gamma_0^2}{2\gamma_0^4}\langle\chi^2\rangle - \frac{3\langle\bar{u}_y^2\rangle}{2\gamma_0^2} + \cdots \tag*{(17)} \end{align}

are of the order O[(AεB)2+a^2/γ02]≪1{\cal O}\left[({\cal A}\varepsilon_{\rm B})^2 + \hat{a}^2/\gamma_0^2\right]\ll1, and can be ignored. From Eqs. (15)-(16), we see that the fast-timescale-averaged longitudinal velocity changes (χ\chi-dependent terms) are suppressed in a relativistic mean motion in the regime of (AεB)2≪1({\cal A}\varepsilon_{\rm B})^2\ll 1, thus allowing us to neglect the longitudinal correction term (O(χ2/γ02){\cal O}(\chi^2/\gamma_0^2)) while retaining the transverse correction term (O(a^2/γ02){\cal O}(\hat{a}^2/\gamma_0^2)). Also, since uˉy/γˉ=uˉyγ0(1+u0χ/γ02)+⋯\bar{u}_y/\bar{\gamma} = \frac{\bar{u}_y}{\gamma_0}(1+u_0\chi/\gamma_0^2)+\cdots (where χuˉy=(a^2ωB/2ω)sin⁡2ωt\chi \bar{u}_y=(\hat{a}^2\omega_{\rm B}/2\omega)\sin{2\omega t}) does not alter the leading coefficient at frequency ω\omega, we consider uˉy/γˉ≃uˉy/γ0\bar{u}_y/\bar{\gamma}\simeq\bar{u}_y/\gamma_0. Taking into account the approximations mentioned above and substituting Eqs. (?)-(13) into Eqs. (8)-(10), we get the following continuity equation:

(∂t−vx∂x)(δργ0n0)+2(1−vx)γ0∂xδux+vxuˉyγ02∂xδuy=0,\begin{align} (\partial_t - v_x\partial_x)\left(\frac{\delta \rho}{\gamma_0n_0}\right) + \frac{2(1-v_x)}{\gamma_0}\partial_x\delta u_x + \frac{v_x\bar{u}_y}{\gamma_0^2}\partial_x\delta u_y=0, \tag*{(18)} \end{align}

longitudinal momentum equation:

(∂t−vx∂x)δux+uˉyγ0∂xδuy=ωB[δuy+uˉy(δργ0n0+vxγ0δux)],\begin{align} (\partial_t - v_x\partial_x)\delta u_x + \frac{\bar{u}_y}{\gamma_0}\partial_x\delta u_y = \omega_{\rm B}\left[\delta u_y + \bar{u}_y\left(\frac{\delta \rho}{\gamma_0n_0} + \frac{v_x}{\gamma_0}\delta u_x\right)\right], \tag*{(19)} \end{align}

and the transverse wave equation:

(∂t2−∂x2)δuy+ωp2[δuy+uˉy(δργ0n0+vxγ0δux)]=−ωBγ0[−vx∂tδργ0n0+2(1−vx)γ0∂tδux+vxγ02(uˉ˙yδuy+uˉy∂tδuy)+∂x(δργ0n0)].\begin{align} & (\partial_t^2-\partial_x^2)\delta u_y + \omega_{\rm p}^2\left[\delta u_y + \bar{u}_y\left(\frac{\delta \rho}{\gamma_0n_0} + \frac{v_x}{\gamma_0}\delta u_x\right)\right] = \nonumber \\ & -\omega_{\rm B}\gamma_0\bigg[-v_x\partial_t\frac{\delta \rho}{\gamma_0n_0} + \frac{2(1-v_x)}{\gamma_0}\partial_t\delta u_x \nonumber\\ &+\frac{v_x}{\gamma_0^2}\left(\dot{\bar{u}}_y\delta u_y + \bar{u}_y\partial_t\delta u_y\right) + \partial_x\left(\frac{\delta \rho}{\gamma_0n_0}\right)\bigg]. \tag*{(20)} \end{align}

The periodic coefficients of Eqs. (18)-(?) couple to perturbations having different frequencies. We look for solutions of the form,

δρ/γ0n0=ei(kx−Ωt)[Q0+Q+e−iωt+Q−eiωt+⋯ ]δux=ei(kx−Ωt)[U0+U+e−iωt+U−eiωt+⋯ ]δuy=ei(kx−Ωt)[A0+A+e−iωt+A−eiωt+⋯ ].\begin{align} \delta\rho/\gamma_0 n_0 &= e^{i(kx - \Omega t)}\left[Q_0 + Q_+ e^{-i\omega t} + Q_- e^{i\omega t} + \cdots\right] \nonumber \\ \delta u_x &= e^{i(kx - \Omega t)}\left[U_0 + U_+ e^{-i\omega t} + U_- e^{i\omega t} + \cdots\right] \nonumber \\ \delta u_y &= e^{i(kx - \Omega t)}\left[A_0 + A_+ e^{-i\omega t} + A_- e^{i\omega t} + \cdots\right]. \tag*{(21)} \end{align}

Substituting these expressions into Eqs. (18)-(?) and using Eqs. (?)-(13), we can isolate the coefficients corresponding to e−iΩte^{-i\Omega t} to obtain the following equations:

DvQ0=2(1−vx)γ0kU0+a^vx2γ02k(A++A−),DvU0=a^2γ0k(A++A−)+iωB(A0+a^2Σ).\begin{align} D_{\rm v} Q_0 & = \frac{2(1-v_x)}{\gamma_0} kU_0 + \frac{ \hat{a}v_x}{2\gamma_0^2} k \left(A_+ + A_-\right) , \tag*{(22)} \\ D_{\rm v} U_0 & = \frac{\hat{a}}{2\gamma_0} k \left(A_+ + A_-\right) + i\omega_{\rm B}\left(A_0 + \frac{\hat{a}}{2}\Sigma\right). \tag*{(23)} \end{align}

Comparing the above two equations with Eqs. (8)-(9), we get δ(nux/n0)∣Ω=γ0ΩQ0/k\delta(nu_x/n_0)|_\Omega=\gamma_0\Omega Q_0/k and δ(nuy/n0)∣Ω=A0+a^Σ/2\delta(nu_y/n_0)|_\Omega = A_0 + \hat{a}\Sigma/2, where Σ≡Q++Q−+vx(U++U−)/γ0\Sigma\equiv Q_+ + Q_- + v_x(U_+ + U_-)/\gamma_0. Substituting these in Eq. (10) yields,

(k2−Ω2+ωp2)A0=−iωBγ0k2−Ω2kQ0−ωp2a^2Σ.\begin{align} (k^2 - \Omega^2 + \omega_{\rm p}^2)A_0 = -i\omega_{\rm B}\gamma_0\frac{k^2-\Omega^2}{k}Q_0 - \frac{\omega_{\rm p}^2\hat{a}}{2}\Sigma. \tag*{(24)} \end{align}

In a similar way, we can isolate the coefficients corresponding to e−i(Ω±ω)te^{-i(\Omega\pm\omega)t} to obtain the following continuity and longitudinal momentum equations:

(Dv±ω)Q±=2(1−vx)kγ0U±+vxa^k2γ02A0,(Dv±ω)U±=a^k2γ0A0+iωB[A±+a^2(Q0+vxU0γ0)],\begin{align} (D_{\rm v}\pm\omega)Q_\pm & = \frac{2(1-v_x)k}{\gamma_0}U_\pm + \frac{v_x\hat{a}k}{2\gamma_0^2}A_0, \tag*{(25)} \\ (D_{\rm v}\pm\omega)U_\pm & = \frac{\hat{a}k}{2\gamma_0}A_0 + i\omega_{\rm B}\left[A_\pm + \frac{\hat{a}}{2}\left(Q_0 + \frac{v_xU_0}{\gamma_0}\right)\right], \tag*{(26)} \end{align}

where Dv≡Ω+vxkD_{\rm v} \equiv \Omega + v_x k. Comparing the above two equations with Eqs. (8)-(9), we get δ(nux/n0)∣(Ω±ω)=γ0(Ω±ω)Q±/k\delta(nu_x/n_0)|_{(\Omega\pm\omega)}=\gamma_0(\Omega\pm\omega) Q_\pm/k and δ(nuy/n0)∣(Ω±ω)=A±+a^(Q0+vxU0/γ0)/2\delta(nu_y/n_0)|_{(\Omega\pm\omega)} = A_\pm + \hat{a}(Q_0+v_xU_0/\gamma_0)/2. Substituting these in Eq. (10) yields,

D±A±=−iωBγ0k(D±+ωp2)Q±+ωp2a^2(Q0+vxU0γ0),\begin{align} D_\pm A_\pm & = - \dfrac{i\omega_{\rm B}\gamma_0}{k}(D_\pm + \omega_{\rm p}^2)Q_\pm + \dfrac{\omega_{\rm p}^2\hat{a}}{2} \left(Q_0 + \dfrac{v_xU_0}{\gamma_0} \right), \tag*{(27)} \end{align}

where

D±≡(Ω±ω)2−k2−ωp2.\begin{align} D_\pm & \equiv \left(\Omega\pm \omega \right)^2 -k^2-\omega_{\rm p}^2. \tag*{(28)} \end{align}

Eqs. (22)-(27) constitute a system of nine equations and nine amplitude variables; requiring the corresponding coefficient matrix to have a zero determinant determines the dispersion relation. Using the above equations, we can solve for the coefficients,

A0=−iωBγ0Q0kf0−a^2(1−f0)Σ;A±=ωp2a^2D±(Q0+vxU0γ0)−iωBγ0k(D±+ωp2)D±Q±;A++A−=ωp2a^2(Q0+vxU0γ0)(1D++1D−)−iωBγ0k[D++ωp2D+Q++D−+ωp2D−Q−].\begin{align} A_0 & = -i\omega_{\rm B}\gamma_0\frac{Q_0}{k}f_0 - \frac{\hat{a}}{2}(1-f_0)\Sigma; \tag*{(29)} \\ A_\pm & = \frac{\omega_{\rm p}^2\hat{a}}{2D_\pm} \left(Q_0+\frac{v_xU_0}{\gamma_0}\right) - \frac{i\omega_{\rm B}\gamma_0}{k}\frac{(D_\pm+\omega_{\rm p}^2)}{D_\pm}Q_\pm; \tag*{(30)} \\ A_+ + A_- & = \frac{\omega_{\rm p}^2\hat{a}}{2}\left(Q_0 + \frac{v_xU_0}{\gamma_0}\right)\left(\frac{1}{D_+} + \frac{1}{D_-}\right) \nonumber\\ & - \frac{i\omega_{\rm B}\gamma_0}{k}\left[\frac{D_++\omega_{\rm p}^2}{D_+}Q_+ + \frac{D_-+\omega_{\rm p}^2}{D_-}Q_-\right]. \tag*{(31)} \end{align}

We can eliminate A0A_0 by inserting Eq. (29) into Eq. (23); the resultant equation can then be substituted into a form of Eq. (22) multiplied by DvD_{\rm v}. This yields,

(Dv2−f0ΩB2)Q0=a^k2γ02[2(1−vx)k+vxDv](A++A−)+ikωBa^f0γ0(1−vx)Σ.\begin{align} (D_{\rm v}^2 - f_0\Omega_{\rm B}^2)Q_0 & = \frac{\hat{a}k}{2\gamma_0^2}\left[2(1-v_x)k + v_x D_{\rm v}\right](A_+ + A_-) \nonumber \\ & + \frac{ik\omega_{\rm B}\hat{a}f_0}{\gamma_0}(1-v_x)\Sigma. \tag*{(32)} \end{align}

Likewise, after eliminating the other amplitude variables, we obtain the following dispersion relation,

Dv2−f0ΩB2≃2η3ωk(Ω+k)(1D++1D−).\begin{equation} D_{\rm v}^2 - f_0\Omega_{\rm B}^2 \simeq \frac{2\eta^3}{\omega} k\left(\Omega+k\right)\left(\frac{1}{D_+}+\frac{1}{D_-}\right). \tag*{(33)} \end{equation}

Here, we define the magnetic restoring frequency, ΩB≡2(1−vx)ωB=1+a^2/2(ωB/γ0)\Omega_{\rm B}\equiv \sqrt{2(1-v_x)}\omega_{\rm B}= \sqrt{1+\hat{a}^2/2}(\omega_{\rm B}/\gamma_0), the scattering rate η≡ω(Aεp/2)2/3\eta\equiv\omega({\cal A\varepsilon_{\rm p}}/2)^{2/3}, and f0≡k2−Ω2k2−Ω2+ωp2→1f_0\equiv\frac{k^2-\Omega^2}{k^2-\Omega^2+\omega_{\rm p}^2} \rightarrow 1 in the limit ∣k2−Ω2∣≫ωp2|k^2-\Omega^2|\gg\omega_{\rm p}^2. In the following, we will consider that ∣Dv∣≪∣Ω+k∣|D_{\rm v}|\ll|\Omega + k|.

Substituting the resonance conditions k=kmk=k_{\rm m} and Ω=Ωm+ΔΩ\Omega=\Omega_{\rm m}+\Delta\Omega into Eq. (33) yields the maximum growth rate of the instability. Here, kmk_{\rm m} and ΔΩ\Delta\Omega are determined by the conditions (i) Dv=0D_{\rm v}=0 and D±=0D_\pm=0 when ΩB≪η\Omega_{\rm B}\ll\eta (strong coupling), which gives Ωm=−vxkm\Omega_{\rm m}=-v_xk_{\rm m}, and (ii) Dv2=ΩB2D_{\rm v}^2=\Omega_{\rm B}^2 and D±=0D_\pm=0 when ΩB≫η\Omega_{\rm B}\gg\eta (weak coupling), which gives Ωm=−vxkm−ΩB\Omega_m=-v_xk_m-\Omega_{\rm B}. For the Dv±ΩB=0D_{\rm v}\pm\Omega_{\rm B}=0 branches, this yields,

ΔΩ2(ΔΩ∓2ΩB)=η3.\begin{align} \Delta \Omega^2(\Delta\Omega\mp2\Omega_{\rm B}) = \eta^3. \tag*{(34)} \end{align}

But, one can see by perturbing Dv=−ΩB+ΔΩD_{\rm v}=-\Omega_{\rm B}+ \Delta\Omega that only the Dv=−ΩBD_{\rm v}=-\Omega_{\rm B} solution to ΔΩ\Delta\Omega yields an imaginary value for all ΩB\Omega_{\rm B}, and has the largest growth rate. Note that the aforementioned conditions D+=0D_+=0 and D−=0D_-=0 do not hold simultaneously. The most unstable wave number is,

km={−2γ02ω1+a^2/2(ΩB≪η)−2γ02ω1+a^2/2(1−ΩBω)(ΩB≫η),\begin{align} k_{\rm m} = \begin{cases} - \dfrac{2\gamma_0^2\omega}{1+\hat{a}^2/2} & (\Omega_{\rm B}\ll\eta)\\ - \dfrac{2\gamma_0^2\omega}{1+\hat{a}^2/2}\left(1-\dfrac{\Omega_{\rm B}}{\omega}\right) & (\Omega_{\rm B}\gg\eta) \\ \end{cases} , \tag*{(35)} \end{align}

and the growth rate of the instability, Im(ΔΩ){\rm Im}(\Delta\Omega), is,

Γgr={32η(ΩB≪η)η3/22ΩB(ΩB≫η).\begin{align} \Gamma_{\rm gr} = \begin{cases} \dfrac{\sqrt{3}}{2}\eta & (\Omega_{\rm B}\ll\eta)\\ \dfrac{\eta^{3/2}}{\sqrt{2\Omega_{\rm B}}} & (\Omega_{\rm B}\gg\eta).\\ \end{cases} \tag*{(36)} \end{align}

Eqs. (35)-(36) are valid for any a0a_0 provided a^/γ0=A3/2εp≪1\hat{a}/\gamma_0={\cal A}^{3/2}\varepsilon_{\rm p}\ll 1, vx≈1v_x\approx1, AεB≪1{\cal A}\varepsilon_{\rm B}\ll 1, and ΩB≪ω\Omega_{\rm B}\ll\omega.

From the head of the EM pulse, the distance where induced scattering significantly affects the pulse profile (in the wave frame) is approximately 1/Γgr1/\Gamma_{\rm gr}. In the reference frame where the plasma ahead of the pulse is at rest, the corresponding distance is,

ℓe/λ0≃ω0/γ02πΓgr≃{22/3π3(Aεp)−2/3(ΩB≪η)2ΩB/ωπAεp→a^≫121/4πAεBεp(ΩB≫η).\begin{align} \ell_{\rm e}/\lambda_0 \simeq \frac{\omega_0/\gamma_0}{2\pi\Gamma_{\rm gr}}\simeq \begin{cases} \dfrac{2^{2/3}}{\pi\sqrt{3}}({\cal A}\varepsilon_{\rm p})^{-2/3} & (\Omega_{\rm B}\ll\eta)\\ \dfrac{\sqrt{2\Omega_{\rm B}/\omega}}{\pi{\cal A}\varepsilon_{\rm p}} \xrightarrow[]{\hat{a}\gg1} \dfrac{2^{1/4}}{\pi\sqrt{{\cal A}}}\dfrac{\sqrt{\varepsilon_{\rm B}}}{\varepsilon_{\rm p}} & (\Omega_{\rm B}\gg\eta).\\ \end{cases} \tag*{(37)} \end{align}

Eq. (37) represents the e-fold growth length for the strong and weak coupling regimes. The strong and weak coupling regimes meet when ΩB=η\Omega_{\rm B}=\eta at a critical frequency of (ω0/ωB)⋆=1+a^2/2(Aεp/2)−2/3(\omega_{\rm 0}/\omega_{\rm B})_\star=\sqrt{1+\hat{a}^2/2}({\cal A}\varepsilon_{\rm p}/2)^{-2/3}. Beyond the linear propagation length from its head, the EM pulse is attenuated because the energy of the forward-propagating pump wave is transferred into compressing the background magnetic fields, the scattered waves, and plasma density/velocity fluctuations. Ref. [27] derives exact nonlinear dispersion relations for strong circularly polarized waves propagating along a background magnetic field, but not induced-scattering rates, so a direct comparison with our results is not possible. Further, as a consistency check, Eqs. (35)-(37) reduce to Eqs. (14)-(16) of Ref. [43] in the limit of ωB→0\omega_{\rm B}\rightarrow0.

Figure 1, panel 1Figure 1, panel 2Figure 1, panel 3

Figure 1. Snapshots of the EM pulse-plasma interaction region for a simulation with a0=3a_0=3, σ=10\sigma=10 and ω0/ωp=100\omega_{\rm 0}/\omega_{\rm p}=100, i.e., in the relativistic/linear regime where ω0>a0ωB\omega_{\rm 0}>a_0\omega_{\rm B}. Different columns represent different stages of evolution (the entire pulse is in vacuum at time Tω0=0T\omega_{\rm 0}=0). [a] The wave’s electric field (black), and phase-averaged magnetic field (orange) normalized by their respective incident-wave amplitudes, [b] simulation-frame particle number density in units of the upstream density, and cell-averaged four-velocity of electrons [c] along the x^\hat{x} direction (normalized by a02a_0^2) and [d] along the y^\hat{y} direction (normalized by a0a_0). The orange dashed vertical lines in the middle and right columns enclose the linear propagation length, ℓlin\ell_{\rm lin}. The distance from the head of the EM pulse, x−xheadx-x_{\rm head}, is expressed in units of the vacuum wavelength.

Figure 2

Figure 2. Left: Linear propagation length in units of the vacuum wavelength as a function of a0a_0; grey line-segments denote reference slopes. Right: Linear propagation length in units of the vacuum wavelength, multiplied by (Aεp)2/3({\cal A}\varepsilon_{\rm p})^{2/3}, as a function of ΩB/η\Omega_{\rm B}/\eta (Eq. (37)). The markers denote numerical simulation results (belonging to the linear current regime with ω0>a0ωB\omega_{\rm 0}>a_0\omega_{\rm B} for a0>1a_0>1); their shape and color represent ω0/ωp\omega_{\rm 0}/\omega_{\rm p} and σ\sigma, respectively. The black dotted curve denotes the one e-fold analytical scaling across the strong (ΩB<η\Omega_{\rm B}<\eta) and weak coupling (ΩB>η\Omega_{\rm B}>\eta) regimes (see Eq. (34)).

Numerical simulations—We perform fully kinetic particle-in-cell (PIC) simulations with OSIRIS [46]. Our effectively one-dimensional setup follows that of Ref. [43], but with an added uniform field B0z^B_0\hat z throughout the computational domain. We launch an EM pulse from vacuum along +x^+\hat{x} with fields Eyy^E_y\hat{y} and Bzz^B_z\hat{z}, hence an X-mode. Its amplitude gradually rises over 10-100 wavelengths to its maximum value (corresponding to a0a_0), and stays flat for long enough to capture nonlinear effects. The cold plasma is initially at rest in the simulation frame. It enters the computational domain through a density ramp at the right boundary, and leaves from the left edge, while the window follows the EM pulse along x^\hat{x} at cc (more in the End Matter). We vary a0=[0.1,1,3,10]a_0=[0.1,1,3,10], ω0/ωp=[10,100,333,1000]\omega_{\rm 0}/\omega_{\rm p}=[10,100,333,1000], and σ=[10−2,1,10,102,103,104]\sigma=[10^{-2},1,10,10^2,10^3,10^4], retaining only combinations above the superluminal X-mode cutoff ω0/ωp>1+σ\omega_{\rm 0}/\omega_{\rm p}>\sqrt{1+\sigma}.

Flux compression—We find that the background magnetic flux is frozen-in with the plasma, which is compressed by the EM pulse in both the linear and the nonlinear regimes. For a plasma moving at ∼c\sim c, the phase-averaged quantities yield ⟨Bz⟩−⟨Ey⟩=B0\langle B_z\rangle-\langle E_y\rangle=B_0. The plasma drifting at βD\beta_{\rm D} introduces a motional electric field ⟨Ey⟩≃βD⟨Bz⟩\langle E_y\rangle\simeq\beta_{\rm D}\langle B_z\rangle. In the steady state, the background magnetic field in the pulse is compressed by a factor ⟨Bz⟩/B0≃1+a^2/4\langle B_z\rangle/B_0\simeq1+\hat{a}^2/4. We show an example of this with the orange curve in the top row of Figure 1: within the linear head of the pulse, ⟨Bz⟩/B0\langle B_z\rangle/B_0 saturates over time to a value of ∼3.3\sim3.3, close to 1+a^2/4=3.251+\hat{a}^2/4=3.25 for the simulated a0=3a_0=3. At a given distance from the head of the pulse, the saturation is reached after a time ∝σ\propto\sigma, because the frozen-in field adds inertia to the plasma [7]. The steady compression is O(a^2)\mathcal O(\hat{a}^2) in the upstream frame but only O(a^2/γ02)\mathcal O(\hat{a}^2/\gamma_0^2) in the wave frame (Eqs. 36-44 of Ref. [39]). Therefore, induced scattering growth rate and ℓe\ell_{\rm e} only weakly depend on whether the σ\sigma-dependent relaxation has finished.

The linear regime—When εB<1\varepsilon_{\rm B}<1, we verify that the particles’ velocity is relativistic and the plasma current is linear in the wave field [41]. In Figure 1, we present an example of this case with a0=3a_0=3, σ=10\sigma=10 and ω0/ωp=100\omega_{\rm 0}/\omega_{\rm p}=100 at different stages of interaction. At later stages of the simulation (middle and right columns), the vertical orange dashed lines enclose and highlight the linear portion of the interaction, which evolves only modestly with time once identified. In all our simulation runs with εB<1\varepsilon_{\rm B}<1, the initialized wave propagates through the plasma, albeit for a finite length; the activation of induced Compton scattering introduces large amplitude fluctuations in density and momenta a distance behind the pulse head, and inhibits propagation of the trailing part of the pulse.

Linear propagation length—We follow the same procedure as Ref. [43] to measure the linear propagation length, ℓlin=Nℓe\ell_{\rm lin}={\cal N}\ell_{\rm e}, which is a multiple of the instability e-fold length, based on the departure of Γβy\Gamma\beta_{\rm y} from eEy/mecω0eE_y/m_{\rm e}c \omega_{\rm 0}. The left panel of Figure 2 shows the measured ℓlin/λ0\ell_{\rm lin}/\lambda_0 against a0a_0: at fixed (σ,ω0/ωp)(\sigma,\omega_{\rm 0}/\omega_{\rm p}), Eq. (37) predicts that ℓlin/λ0∝a0−1\ell_{\rm lin}/\lambda_0\propto a_0^{-1} (for a^≪1\hat{a}\ll1) and ∝a0−1/2\propto a_0^{-1/2} (for a^≫1\hat{a}\gg1) in the weak coupling regime (ΩB≫η\Omega_{\rm B}\gg\eta), and ℓlin∝a0−2/3\ell_{\rm lin}\propto a_0^{-2/3} in the strong coupling regime (ΩB≪η\Omega_{\rm B}\ll\eta). These slopes are similar, making it challenging to identify this transition from the measurements (left panel). On the other hand, Eq. (37) also implies that (Aεp)2/3ℓlin/λ0({\cal A}\varepsilon_{\rm p})^{2/3}\ell_{\rm lin}/\lambda_0 depends on ΩB/η\Omega_{\rm B}/\eta alone. We test this in the right panel of Figure 2: the colored markers denote simulation measurements and the black dotted curve is a solution to Eq. (34) for its complex root with Im(ΔΩ)>0\Delta\Omega)>0 as a function of ΩB/η\Omega_{\rm B}/\eta. In our PIC simulations, we measure the distance when the instability becomes visible above intrinsic particle noise, ℓlin\ell_{\rm lin}. From Figure 2, we measure ℓlin/ℓe\ell_{\rm lin}/\ell_{\rm e} to be N∼4{\cal N}\sim4 e-folds. We note that the numerical measurements of (Aεp)2/3ℓlin/λ0({\cal A}\varepsilon_{\rm p})^{2/3}\ell_{\rm lin}/\lambda_0 are consistent with our analytical scaling across six decades in σ\sigma, following a flat slope for ΩB≪η\Omega_{\rm B}\ll\eta as in the unmagnetized case [43] and rising as (ΩB/η)1/2(\Omega_{\rm B}/\eta)^{1/2} for ΩB≫η\Omega_{\rm B}\gg\eta.

Figure 3, panel 1Figure 3, panel 2Figure 3, panel 3

Figure 3. Snapshots of the EM pulse-plasma interaction region for a simulation with a0=10a_0=10, σ=10\sigma=10 and ω0/ωp=10\omega_{\rm 0}/\omega_{\rm p}=10, i.e., in the relativistic/non-linear regime where a02>εB>1a_0^2>\varepsilon_{\rm B}>1. Different rows [a-d] denote the same quantities as in Figure 1.

The nonlinear regime—We identify in our PIC simulations, the particles’ velocity to be relativistic and their response to the wave field to be nonlinear when a02>εB>1a_0^2>\varepsilon_{\rm B}>1 [39]. In Figure 3, we show the spatial structure of the EM pulse-plasma interaction region (for an example case with a0=10a_0=10, σ=10\sigma=10 and ω0/ωB≈3.2\omega_{\rm 0}/\omega_{\rm B}\approx3.2) at different stages of interaction (columns). Behind the head of the pulse, the plasma settles into a compression front [7]. The particles exhibit E×B\mathbf E\times\mathbf B drift along x^\hat x in this compressed field rather than a coherent quiver motion (mean transverse momentum Γβy≃0\Gamma\beta_y\simeq0, row [d]). As discussed previously, EyE_y acquires a motional component due to the drifting plasma on which the oscillation rides. Empirically, we find that the amplitude of the oscillations is damped to a value ≃E0/εB\simeq E_0/\varepsilon_{\rm B}, where the local resonance condition of a≃ω0/ωBa\simeq\omega_{\rm 0}/\omega_{\rm B} switches on [41]. We interpret the apparent attenuation of the half-cycle with Ey<0E_y<0 as a damped oscillation riding on a positive motional field, and not as an attenuation of one phase. Unlike previous works [1], this wave does not steepen into strong shocks in each wavelength. For a vacuum-like wave with electric (and magnetic) field amplitudes EwE_{\rm w} propagating on these mean fields, B2−E2=(⟨Bz⟩+⟨Ey⟩+2Ew)(⟨Bz⟩−⟨Ey⟩)B^2-E^2=(\langle B_z\rangle+\langle E_y\rangle+2E_{\rm w})(\langle B_z\rangle-\langle E_y\rangle), where ⟨Bz⟩−⟨Ey⟩=B0\langle B_z \rangle - \langle E_y\rangle=B_0 is the upstream value. Without field compression, phases with Ew<−(⟨Bz⟩+⟨Ey⟩)/2=−B0/2E_{\rm w}<-(\langle B_z \rangle + \langle E_y\rangle)/2=-B_0/2 violate the MHD condition B2>E2B^2>E^2 and dissipate at shocks in each wavelength [1, 5]. Once the compressed flux attains ⟨Bz⟩/B0≃1+a^2/4\langle B_z \rangle/B_0\simeq1+\hat{a}^2/4, ⟨Bz⟩+⟨Ey⟩\langle B_z\rangle+\langle E_y\rangle increases to (1+a^2/2)B0(1+\hat{a}^2/2)B_0. Therefore, the condition for not producing electric-dominated shocks becomes 2E0/εB≲(1+a^2/2)B02E_0/\varepsilon_{\rm B}\lesssim(1+\hat{a}^2/2)B_0, so the wave in Figure 3 is absorbed resonantly rather than through strong shocks at each wave cycle.

Conclusions—We have studied, analytically and with fully kinetic simulations, the interaction of strong (a0>1a_0>1) linearly polarized superluminal X-mode EM waves with magnetized pair plasmas. The interaction is nonlinear when a0≳1a_0\gtrsim1 in electron–proton plasmas [21] and when εp≳1\varepsilon_{\rm p}\gtrsim1 in unmagnetized pair plasmas [40, 43]. In magnetized pair plasmas (σ>1\sigma>1), this regime separation is set by the parameter εB≡a0ωB/ω0\varepsilon_{\rm B}\equiv a_0\omega_{\rm B}/\omega_{\rm 0}.

(1) For εB<1\varepsilon_{\rm B}<1 the EM pulse propagates until induced Compton scattering disrupts it. The background field imparts an effective restoring frequency to the instability at ΩB\Omega_{\rm B}, which competes with the wave scattering rate η\eta. For ΩB≪η\Omega_{\rm B}\ll\eta, the instability proceeds as in unmagnetized plasmas, and the linear propagation distance in units of the pulse wavelength is ℓlin/λ0∝(Aεp)−2/3\ell_{\rm lin}/\lambda_0\propto({\cal A}\varepsilon_{\rm p})^{-2/3}. For ΩB≫η\Omega_{\rm B}\gg\eta the instability growth slows and ℓlin/λ0\ell_{\rm lin}/\lambda_0 increases to ∝(ΩB/ω)1/2/(Aεp)\propto(\Omega_{\rm B}/\omega)^{1/2}/({\cal A}\varepsilon_{\rm p}), which tends to εB/εp\sqrt{\varepsilon_{\rm B}}/\varepsilon_{\rm p} for a^≫1\hat{a}\gg1. Kinetic plasma simulations spanning σ=10−2\sigma=10^{-2}–10410^4 and a0=0.1a_0=0.1–10 confirm the predicted analytical scalings.

(2) For a02>εB>1a_0^2>\varepsilon_{\rm B}>1, the plasma current is nonlinear [39]. Behind the head of the pulse, the plasma relaxes into a compression front [7] where the particles exhibit a longitudinal E×B\mathbf E\times\mathbf B drift with subdued transverse momenta. In the presented case, flux compression ensures strong shocks are not formed at each wave. The EM pulse rides on top of a motional electric field and is gradually damped, consistent with synchrotron absorption [25, 4, 41] (so one half-cycle appears attenuated).

(3) In both the regimes, the EM pulse compresses the background field to ⟨Bz⟩/B0≃1+a^2/4\langle B_z\rangle/B_0\simeq1+\hat{a}^2/4 over a time ∝σ\propto\sigma [7]. The compression is only O(a^2/γ02){\cal O}(\hat{a}^2/\gamma_0^2) in the wave frame [39], so ℓlin\ell_{\rm lin} is weakly sensitive to it.

These results could be applicable to large amplitude, linearly polarized radio pulses such as FRBs and giant radio pulses, which eventually cross the predominantly toroidal field of a neutron star’s wind. There, the wind-frame cyclotron frequency falls below GHz frequencies outside ∼1011\sim10^{11} cm, and εB<1\varepsilon_{\rm B}<1 beyond ∼1012\sim10^{12} cm [41]. Since σ∼102\sigma\sim10^2–10310^3 in such winds [7], εp=εB/σ1/2≪1\varepsilon_{\rm p}=\varepsilon_{\rm B}/\sigma^{1/2}\ll1. For a^≫1\hat{a}\gg1, (ΩB/η)3≃2 AσεB(\Omega_{\rm B}/\eta)^3\simeq\sqrt2\,{\cal A}\sigma\varepsilon_{\rm B}, so induced scattering there is in the weak-coupling regime wherever σεB≳1\sigma\varepsilon_{\rm B}\gtrsim1. There, magnetization lengthens ℓlin\ell_{\rm lin} relative to the unmagnetized estimate by a factor of ≈1.2(ΩB/η)1/2\approx1.2(\Omega_{\rm B}/\eta)^{1/2}, i.e., 2–4×4\times for σεB∼10\sigma\varepsilon_{\rm B}\sim10–10310^3. The predictions of this theory could be tested with next-generation multi-petawatt laser and accelerator experiments [10, 2] capable of producing dense pair plasmas. Broadband pulses [24, 31], multidimensional, and radiative effects remain to be studied.

Acknowledgments—This work benefits from discussions with Roger D. Blandford, Lorenzo Sironi, and Emanuele Sobacchi. N.S. acknowledges the support of the Simons Foundation (grant MP-SCMPS-00001470) and the National Aeronautics and Space Administration (grant 80NSSC26K0946). The computing for this project was performed on Stanford University’s Sherlock cluster. Anthropic’s Claude was used to assist with plotting in Matplotlib [47] and checking algebra, under author’s supervision. N.S. thanks Stanford University and their Research Computing Center for providing computational resources and support that contributed to this research.

Data availability—The data supporting the findings of this article are not publicly available upon publication because it is not technically feasible and/or the cost of preparing and hosting the data would be prohibitive within the terms of this research project. The data are available from the author upon reasonable request.

End Matter

Numerical simulation details—Our numerical simulations use the Higuera & Cary particle pusher [17], and the shape of the macroparticles is represented with quadratic splines. The EM field solver employs a second-order finite-difference method using an extended 4-coefficient stencil along the wave-propagation direction (x^\hat{x}), which improves dispersion errors. This choice requires the computational domain to be two-dimensional, and the quadratic particle splines require at least 5 cells along the y^\hat{y} direction. The simulation setup overall follows Ref. [43]; we repeat it here for completeness. At t=0t=0 the EM pulse is initialized in a vacuum region, located to the left of the space filled with plasma. The entire domain is initialized with a uniform magnetic field along the z^\hat{z} direction. The amplitude of the EM pulse is gradually increased and decreased over a length of ℓrise=ℓfall=10−100\ell_{\rm rise} = \ell_{\rm fall} = 10-100 wavelengths using a smooth 5th-order polynomial profile function of the form f(x)=10l(x)3−15l(x)4+6l(x)5f(x) = 10l(x)^3 - 15l(x)^4 + 6l(x)^5, where l(x)=(x−xhead)/ℓrisel(x) = (x - x_{\rm head})/\ell_{\rm rise} for the rising phase of the pulse (right) and l(x)=(xhead+ℓflat−x)/ℓfalll(x) =(x_{\rm head} + \ell_{\rm flat} - x)/\ell_{\rm fall} for the falling phase (left). Here, xheadx_{\rm head} is the position of the right-edge of the pulse, and ℓflat=102−104\ell_{\rm flat}=10^2-10^4 wavelengths is length of the ‘flat’ part of the pulse (with strength parameter a0a_0), which is chosen to be sufficiently long to capture nonlinear propagation effects. The computational domain along x^\hat{x} is chosen to encompass the entire pulse. We resolve the wavelength of the EM pulse with ≥100\ge 100 cells, the electron skin depth c/ωpc/\omega_{\rm p} with 160−16000160-16000 cells, and both the cold Larmor radius and the Debye length λD=θ0(c/ωp)\lambda_{\rm D}=\sqrt{\theta_0}(c/\omega_{\rm p}) with at least one cell for our cold plasma with a normalized temperature of θ0=10−4\theta_0=10^{-4}. Our simulations employ 50 particles per cell (equally divided between electrons and positrons). In the transition zone between the vacuum and the plasma, the density profile smoothly increases from 0 to 1 using a hyperbolic tangent function of the form g(x)=0.5[1+tanh⁡(x−1.1xhead)]g(x) = 0.5\left[1 + \tanh{(x-1.1 x_{\rm head})}\right]. The plasma entering the domain through the right boundary exit through an absorbing boundary modeled as a perfectly matched layer [44] on the left. The choice of a smooth profile in the amplitude of the EM pulse and in the plasma density results in a gradual increase of the current and minimizes potential numerical artifacts during the early stages of interaction. The plasma is initially at rest, and the simulation window follows the pulse as it propagates through the plasma.

References

  1. [1]Chen, A. Y., Yuan, Y., Li, X., & Mahlmann, J. F. (2022). Propagation of a Strong Fast Magnetosonic Wave in the Magnetosphere of a Neutron Star. arXiv E-prints, arXiv:2210.13506. https://doi.org/10.48550/arXiv.2210.13506
  2. [2]Arrowsmith, C. D., Simon, P., Bilbao, P. J., Bott, A. F. A., Burger, S., Chen, H., Cruz, F. D., Davenne, T., Efthymiopoulos, I., Froula, D. H., Goillot, A., Gudmundsson, J. T., Haberberger, D., Halliday, J. W. D., Hodge, T., Huffman, B. T., Iaquinta, S., Miniati, F., Reville, B., … Gregori, G. (2024). Laboratory realization of relativistic pair-plasma beams. Nature Communications, 15, 5029. https://doi.org/10.1038/s41467-024-49346-2
  3. [3]Beloborodov, A. M. (2021). Can a Strong Radio Burst Escape the Magnetosphere of a Magnetar?. Astrophys. J., 922(1), L7. https://doi.org/10.3847/2041-8213/ac2fa0
  4. [4]Beloborodov, A. M. (2022). Scattering of Ultrastrong Electromagnetic Waves by Magnetized Particles. Physical Review Letters, 128(25). https://doi.org/10.1103/physrevlett.128.255003
  5. [5]Beloborodov, A. M. (2023). Monster Radiative Shocks in the Perturbed Magnetospheres of Neutron Stars. ApJ, 959(1), 34. https://doi.org/10.3847/1538-4357/acf659
  6. [6]Beloborodov, A. M. (2024). Damping of Strong GHz Waves near Magnetars and the Origin of Fast Radio Bursts. ApJ, 975(2), 223. https://doi.org/10.3847/1538-4357/ad698c
  7. [7]Beloborodov, A. M. (2026). Compression Fronts from Fast Radio Bursts. ApJ, 1000(2), 157. https://doi.org/10.3847/1538-4357/ae4692
  8. [8]Blandford, R. D., & Znajek, R. L. (1977). Electromagnetic extraction of energy from Kerr black holes.. MNRAS, 179, 433–456. https://doi.org/10.1093/mnras/179.3.433
  9. [9]Cerutti, B., & Beloborodov, A. M. (2017). Electrodynamics of Pulsar Magnetospheres. Space Sci. Rev., 207(1-4), 111–136. https://doi.org/10.1007/s11214-016-0315-7
  10. [10]Chen, H., & Fiuza, F. (2023). Perspectives on relativistic electron-positron pair plasma experiments of astrophysical relevance using high-power lasers. Physics of Plasmas, 30(2), 020601. https://doi.org/10.1063/5.0134819
  11. [11]Clemmow, P. C. (1974). Nonlinear waves in a cold plasma by Lorentz transformation. Journal of Plasma Physics, 12(2), 297–317. https://doi.org/10.1017/S0022377800025125
  12. [12]Cognard, I., Shrauner, J. A., Taylor, J. H., & Thorsett, S. E. (1996). Giant Radio Pulses from a Millisecond Pulsar. Astrophys. J., 457, L81. https://doi.org/10.1086/309894
  13. [13]Cordes, J. M., & Chatterjee, S. (2019). Fast Radio Bursts: An Extragalactic Enigma. ARA&A, 57, 417–465. https://doi.org/10.1146/annurev-astro-091918-104501
  14. [14]Golbraikh, E., & Lyubarsky, Y. (2023). On the Escape of Low-frequency Waves from Magnetospheres of Neutron Stars. ApJ, 957(2), 102. https://doi.org/10.3847/1538-4357/acfa78
  15. [15]Gruzinov, A. (2019). Nonlinear scattering of Fast Radio Bursts. arXiv E-prints, arXiv:1912.08150. https://doi.org/10.48550/arXiv.1912.08150
  16. [16]Gunn, J. E., & Ostriker, J. P. (1971). On the Motion and Radiation of Charged Particles in Strong Electromagnetic Waves. I. Motion in Plane and Spherical Waves. ApJ, 165, 523. https://doi.org/10.1086/150919
  17. [17]Higuera, A. V., & Cary, J. R. (2017). Structure-preserving second-order integration of relativistic charged particle trajectories in electromagnetic fields. Physics of Plasmas, 24(5), 052104. https://doi.org/10.1063/1.4979989
  18. [18]Iwamoto, M., & Ioka, K. (2026). Induced scattering of strong waves in pair plasmas. Phys. Rev. D, 114(6), 063022. https://doi.org/10.1103/9yxj-c8fr
  19. [19]Iwamoto, M., Sobacchi, E., & Sironi, L. (2023). Kinetic simulations of the filamentation instability in pair plasmas. MNRAS, 522(2), 2133–2144. https://doi.org/10.1093/mnras/stad1100
  20. [20]Kamijima, S. F., Nishiura, R., Iwamoto, M., & Ioka, K. (2026). One-dimensional PIC Simulation of Induced Compton Scattering in Magnetized Electron-Positron Pair Plasma. arXiv E-prints, arXiv:2601.01169. https://doi.org/10.48550/arXiv.2601.01169
  21. [21]Kruer, W. (2019). The physics of laser plasma interactions. crc Press.
  22. [22]Landau, L. D., & Lifshitz, E. M. (1975). The classical theory of fields. Oxford Pergamon Press.
  23. [23]Lorimer, D. R., Bailes, M., McLaughlin, M. A., Narkevic, D. J., & Crawford, F. (2007). A Bright Millisecond Radio Burst of Extragalactic Origin. Science, 318, 777. https://doi.org/10.1126/science.1147532
  24. [24]Lyubarsky, Y. (2008). Induced Scattering of Short Radio Pulses. ApJ, 682, 1443–1449. https://doi.org/10.1086/589435
  25. [25]Lyubarsky, Y. (2018). Interaction of the electromagnetic precursor from a relativistic shock with the upstream flow - I. Synchrotron absorption of strong electromagnetic waves. MNRAS, 474(1), 1135–1142. https://doi.org/10.1093/mnras/stx2832
  26. [26]Lyubarsky, Y. E. (2003). Fast magnetosonic waves in pulsar winds. MNRAS, 339(3), 765–771. https://doi.org/10.1046/j.1365-8711.2003.06221.x
  27. [27]Lyutikov, M. (2026). Relativistically strong electromagnetic waves in magnetised plasmas. Journal of Plasma Physics, 92(3), E88. https://doi.org/10.1017/S0022377826101342
  28. [28]Lyutikov, M. (2024). The escape of fast radio burst emission from magnetars. MNRAS, 529(3), 2180–2190. https://doi.org/10.1093/mnras/stae591
  29. [29]Melrose, D. B. (1997). Some properties of magnetized pair plasmas. Plasma Physics and Controlled Fusion, 39, A93–A100. https://doi.org/10.1088/0741-3335/39/5A/010
  30. [30]Nishiura, R., Kamijima, S. F., Iwamoto, M., & Ioka, K. (2025). Induced Compton scattering in magnetized electron and positron pair plasma. Phys. Rev. D, 111(6), 063055. https://doi.org/10.1103/PhysRevD.111.063055
  31. [31]Nishiura, R., Kamijima, S. F., & Ioka, K. (2026). Unified kinetic theory of induced scattering: Compton, Brillouin, and Raman processes in magnetized electron and positron pair plasma. Phys. Rev. D, 113(12), 123070. https://doi.org/10.1103/3n51-x65q
  32. [32]Nishiura, R., Kamijima, S. F., & Ioka, K. (2026). Induced Scattering of Fast Radio Bursts in Magnetar Magnetospheres. arXiv E-prints, arXiv:2601.18865. https://doi.org/10.48550/arXiv.2601.18865
  33. [33]Petroff, E., Hessels, J. W. T., & Lorimer, D. R. (2019). Fast radio bursts. A&ARv, 27(1), 4. https://doi.org/10.1007/s00159-019-0116-6
  34. [34]Petroff, E., Hessels, J. W. T., & Lorimer, D. R. (2022). Fast radio bursts at the dawn of the 2020s. A&ARv, 30(1), 2. https://doi.org/10.1007/s00159-022-00139-w
  35. [35]Qu, Y., & Kumar, P. (2026). Scattering of Strong Radio Waves by Particles in Strongly Magnetized Plasmas and Implications for Fast Radio Bursts. arXiv E-prints, arXiv:2606.31439. https://doi.org/10.48550/arXiv.2606.31439
  36. [36]Qu, Y., Kumar, P., & Zhang, B. (2022). Transparency of fast radio burst waves in magnetar magnetospheres. MNRAS, 515(2), 2020–2031. https://doi.org/10.1093/mnras/stac1910
  37. [37]Sobacchi, E., Lyubarsky, Y., Beloborodov, A. M., & Sironi, L. (2022). Filamentation of fast radio bursts in magnetar winds. MNRAS, 511(4), 4766–4773. https://doi.org/10.1093/mnras/stac251
  38. [38]Sobacchi, E., Lyubarsky, Y., Beloborodov, A. M., Sironi, L., & Iwamoto, M. (2023). Saturation of the Filamentation Instability and Dispersion Measure of Fast Radio Bursts. Astrophys. J., 943(2), L21. https://doi.org/10.3847/2041-8213/acb260
  39. [39]Sobacchi, E., Iwamoto, M., Sironi, L., & Piran, T. (2024). Escape of fast radio bursts from magnetars. A&A, 690, A332. https://doi.org/10.1051/0004-6361/202451725
  40. [40]Sobacchi, E., Iwamoto, M., Sironi, L., & Piran, T. (2024). Propagation of strong electromagnetic waves in tenuous plasmas. Physical Review Research, 6(4), 043213. https://doi.org/10.1103/PhysRevResearch.6.043213
  41. [41]Sobacchi, E. (2025). Absorption of strong electromagnetic waves in magnetized pair plasmas. \pre, 112(6), 065208. https://doi.org/10.1103/1b36-qn66
  42. [42]Solanki, S., Mahlmann, J., & Philippov, A. (2026). Nonlinear Decay of Fast Magnetosonic Waves through Weak Turbulence: Force-Free Electrodynamics Simulations. arXiv E-prints, arXiv:2606.19450. https://doi.org/10.48550/arXiv.2606.19450
  43. [43]Sridhar, N., Sobacchi, E., Sironi, L., Iwamoto, M., Grošelj, D., & Russell, B. K. (2026). Interaction of Strong Electromagnetic Waves with Unmagnetized Pair Plasmas. Phys. Rev. Lett., 137(4), 045102. https://doi.org/10.1103/h7fg-xjby
  44. [44]Vay, J.-L. (2002). Asymmetric Perfectly Matched Layer for the Absorption of Waves. Journal of Computational Physics, 183(2), 367–399. https://doi.org/10.1006/jcph.2002.7175
  45. [45]Goldreich, P., & Julian, W. H. (1969). Pulsar Electrodynamics. ApJ, 157, 869–+. https://doi.org/10.1086/150119
  46. [46]Fonseca, R. A., Silva, L. O., Tsung, F. S., Decyk, V. K., Lu, W., Ren, C., Mori, W. B., Deng, S., Lee, S., Katsouleas, T., & Adam, J. C. (2002). OSIRIS: A Three-Dimensional, Fully Relativistic Particle in Cell Code for Modeling Plasma Based Accelerators. Computational Science --- ICCS 2002, 342–351.
  47. [47]Hunter, J. D. (2007). Matplotlib: A 2D graphics environment. Computing in Science & Engineering, 9(3), 90–95. https://doi.org/10.1109/MCSE.2007.55

Paper details

vlnc:W_bwjqtp79
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Anthropic's Claude

View statement
Looking for related papers…
Contents