betaFeedback

WorksPhysics and Astronomy

Long-range correlations in the intrabeam scattering in relativistic beams

Abstract

The standard approach in the conventional theory of intrabeam scattering (IBS) is based on the analysis of binary Coulomb collisions of charged particles in the beam. This picture is only partially true: in reality, a given particle collides simultaneously with many partners and a more rigorous treatment should take this collective aspect of the collisions into account. A framework for such treatment has been known in plasma physics since beginning of the 1960s. In this work, we will show how it can be applied to IBS in relativistic beams for a particular problem of the growth of the beam energy spread. A preliminary conclusion of this work is that the conventional IBS theory might underestimate the effects of long-range correlations in Coulomb collisions in the beam.

Introduction

The Coulomb scattering in beams of charged particles is an important element of the beam physics in modern accelerators. In plasma, the Coulomb collisions are conventionally described by the collision integral in the Vlasov equation introduced by L. Landau in 1936 [1] (see also [2]). In accelerators, this process is known under the name of the intrabeam scattering (IBS) and has been developed in several papers in the last decades [3, 4, 5, 6].

The most straightforward derivation of the collision integral is based on the picture of binary particle collisions in the beam frame, as shown in Figure 1. Before the collision, the two particles are moving along straight orbits and the collision is characterized by the impact parameter bb. The interaction occurs in a relatively small region where the particles change the direction of motion and after the collision, asymptotically in time, move along straight orbits rotated by angle θ\theta relative to the original direction of motion. An important assumption in the derivation of the collision term is that the subsequent binary collisions of the particle are not correlated with each other. This allows one to introduce probabilities and statistical arguments in the analysis and to obtain a relatively simple expression for the collision integral.

Figure 1. Binary collision of two particles, 1 and 2, in the center-of-mass frame of reference.

An important simplification in the derivation of the Landau collision integral, and the corresponding expressions for IBS in relativistic beams1, comes from the observation that most of the collisions occur at small angles, θ≪1\theta\ll1. The use of the small-angle approximation limits the accuracy of the theory to the order of Λ−1\Lambda^{-1} where Λ\Lambda is the Coulomb logarithm,

Λ=ln⁡(bmax⁡bmin⁡),(1)\Lambda= \ln\left(\frac{b_{\max}}{b_{\min}}\right), \tag*{(1)}

with bmax⁡b_{\max} and bmin⁡b_{\min} the maximum and minimum impact parameters. At the minimum collision parameter bmin⁡b_{\min} the deflection angle θ\theta becomes of the order of π/2\pi/2 and the small-angle approximation breaks down. The situation with bmax⁡b_{\max} is more complicated. In plasma, bmax⁡b_{\max} is traditionally set to the Debye radius, rDr_D, at which the Coulomb interaction force becomes shielded due to the dynamic polarization of charges. This polarization establishes on the timescale that is much larger than the inverse plasma frequency ωp−1\omega_p^{-1}. For relativistic beams, the situation with bmax⁡b_{\max} is less clear. First, the time ωp−1\omega_p^{-1} transformed into the laboratory frame could likely correspond to such large distances that the beam motion is perturbed when it passes through the magnets in the lattice. In addition, a formally computed Debye length can be larger than the transverse size of the beam, and then it seems natural that the transverse size of the beam should play the role of bmax⁡b_{\max}. These, and some other arguments in the discussion of the choice of bmax⁡b_{\max}, can be found in Ref. [5].

While the picture of binary collisions provides a useful technique for the calculation of collision rates, it falls short of being fully satisfactory. The problem is that a given particle simultaneously participates in many collisions, making the whole process a collective effect2. A framework that takes the nature of the collective collisional interactions into account has been developed many years ago [7, 8, 9]: in it the collisions are treated as the result of the interaction of a particle with the fluctuating electric field created by its neighbors. In this approach, one can obtain a fully converging collision integral without an artificial truncation of the impact parameters by bmin⁡b_{\min} and bmax⁡b_{\max} (see, e.g., [2], §46), however, the result is extremely complicated and is rearly used in practice. The goal of this work is to apply the framework of the fluctuating electric field to the IBS taking into account that the transverse motion of the particles in the beam is bounded by the focusing of the magnetic lattice of the accelerator. This approach naturally incorporates the finite transverse dimensions of the beam and is expected to eliminate the uncertainty associated with the choice of bmax⁡b_{\max} in conventional theory. We will see however that it also adds new (and quite unexpected!) elements into the description of the Coulomb collisions which are due to the long-time correlations of the electric field fluctuations in the beam. Such correlations are completely ignored in the conventional treatment of the IBS.

The paper is organized as follows. In § 2, we formulate a particular IBS problem of energy diffusion that we address in this paper and the simplifying assumptions that we will use in our calculations. In Section III, we express the energy diffusion coefficient D\mathcal{D} in terms of the integrated correlator of the electric field in the beam. The latter is related to the correlator of the density fluctuations that are calculated in Appendices A and B. In Section IV, we derive a general expression for the averaged over the distribution function coefficient Dˉ\bar{\mathcal{D}} and in § 5 we study the limit of this coefficient when the beam energy spread tends to zero. Quite surprisingly, we find that in this limit the integral expression for Dˉ\bar{\mathcal{D}} diverges, although a convergent result can be obtained if one limits the contribution to the integral by small scales of the fluctuating field. The general case of non-zero energy spread is analyzed in Section VI. Section VIII summarizes the results of the paper.

We use the CGS system of units throughout this paper.

Formulation of the problem

In general, a complete IBS theory involves all three degrees of freedom of the particle. In this paper, however, we will focus only on the energy and will try to calculate the energy diffusion coefficient Dˉη\bar{\mathcal{D}}_{\eta} defined by the equation

dση2ds=2Dˉη,(2)\frac{d\sigma_{\eta}^{2}}{ds}=2\bar{\mathcal{D}}_{\eta}, \tag*{(2)}

where ση\sigma_{\eta} is the rms relative energy spread in the beam (in the lab frame), and the bar over D\mathcal{D} indicates that it is averaged over the beam distribution function. This coefficient can also be used in the diffusion equation for the energy distribution function f(η)f(\eta),

∂f(η)∂s=Dˉη∂2f(η)∂η2,(3)\frac{\partial f(\eta)}{\partial s}=\bar{\mathcal{D}}_{\eta}\frac{\partial^{2}f(\eta)}{\partial\eta^{2}}, \tag*{(3)}

where η=ΔE/E\eta= \Delta E/E is the relative energy deviation. The energy diffusion is important in linear accelerators of free electron lasers [10] where it increases the beam energy spread—a critical parameter that defines the gain length in the FEL. Recent experimental studies of the beam energy spread in several FEL facilities emphasized the importance of the correct account of the IBS in the electron injector [11, 12, 13, 14, 15] and in some cases indicated a discrepancy between the measured and expected values of the energy spread.

In addition to limiting our analysis to the energy spread only, we will take another important simplifying factor into account: a small ratio of the longitudinal and transverse temperatures, T∥/T⊥T_{\parallel}/T_{\perp}, of the beam in the beam frame. It is straightforward to estimate these temperatures as the transverse and longitudinal (relative to the direction of motion) energy spreads in the beam using the Lorentz transformations:

T∥∼mc2ση2,T⊥∼γmc2ϵnβ,(4)T_{\parallel} \sim mc^2\sigma_{\eta}^2,\qquad T_{\perp} \sim\gamma mc^2 \frac{\epsilon_n}{\beta}, \tag*{(4)}

where β\beta is the beta function of the magnetic lattice, ϵn\epsilon_n is the normalized beam emittance, and γ\gamma is the relativistic Lorentz factor. In many cases, the longitudinal beam temperature turns out to be much smaller than the transverse one, T∥≪T⊥T_{\parallel} \ll T_{\perp}. As an example typical for an x-ray FEL, let us take ση∼5×10−4\sigma_{\eta} \sim5 \times10^{-4}, ϵn=1 μm\epsilon_n = 1\ \mu\mathrm{m}, γ=2×103\gamma= 2 \times10^3 and β=10 m\beta= 10\ \mathrm{m}. This gives T∥≈0.12 eVT_{\parallel} \approx0.12\ \mathrm{eV} and T⊥≈100 eVT_{\perp} \approx100\ \mathrm{eV} with T∥/T⊥∼10−3T_{\parallel}/T_{\perp} \sim10^{-3}. To further simplify calculations, we will assume a coasting beam model in which the beam distribution function does not depend on the longitudinal coordinate zz.

In the limit of cold beam, T∥≪T⊥T_{\parallel} \ll T_{\perp}, the conventional IBS theory [16, 17] gives the following expression for the energy diffusion coefficient:

Dˉη=π4re2γ3ϵ3/2β1/2νΛ.(5)\bar{D}_{\eta} = \frac{\sqrt{\pi}}{4}\frac{r_e^2}{\gamma^3\epsilon^{3/2}\beta^{1/2}}\nu\Lambda. \tag*{(5)}

Here re=e2/mc2r_e = e^2/mc^2 is the classical electron radius, ν\nu is the linear particle density (the number of particles per unit length), ϵ\epsilon is the beam emittance and β\beta is the beta function (it is assumed that ϵx=ϵy=ϵ\epsilon_x = \epsilon_y = \epsilon and βx=βy=β\beta_x = \beta_y = \beta). Because the energy spread does not enter Eq. (5) this equation is also valid when the beam energy spread is equal to zero, ση=0\sigma_{\eta} = 0. We will use Eq. (5) as a reference for comparison of the result obtained in this work.

The kinetic equation, fluctuations and energy diffusion

As was pointed out above, we will use the method of fluctuating electric fields to calculate IBS in a relativistic beam. In contrast to the conventional treatment of Ref. [7] where the effect of such fields in plasma is computed assuming that particles move along straight orbits, we include in our analysis the curvilinear trajectories of the particles executing betatron oscillations. This naturally takes into account the finite transverse size of the beam and rigorously defines the effective upper limit bmax⁡b_{\max} in the Coulomb logarithm (1).

We start from the kinetic equation for the beam in the laboratory frame of reference. The betatron oscillations of the particles are characterized by the action-angle variables JxJ_x, ϕx\phi_x and JyJ_y, ϕy\phi_y in the horizontal and vertical directions, respectively,

x(s)=2βx(s)Jxcos⁡[ψx(s)+ϕx],y(s)=2βy(s)Jycos⁡[ψy(s)+ϕy],(6)x(s) = \sqrt{2\beta_x(s)J_x}\cos[\psi_x(s) + \phi_x], \qquad y(s) = \sqrt{2\beta_y(s)J_y}\cos[\psi_y(s) + \phi_y], \tag*{(6)}

where x(s)x(s) and y(s)y(s) are the horizontal and vertical coordinates of the particle as a function of the longitudinal coordinate ss along the beam line, βx(s)\beta_x(s) and βy(s)\beta_y(s) are the beta functions in xx and yy directions, and ψx(s)\psi_x(s) and ψy(s)\psi_y(s) are the corresponding betatron phases. The variables ϕx,ϕy\phi_x,\phi_y are defined in such a way that they, as the actions Jx,JyJ_x,J_y, are integrals of motion (see, e.g., [18]).

The Vlasov equation is formulated for the 6-dimensional distribution function of the beam, f(Jx,ϕx,Jy,ϕy,z,η,s)f(J_x,\phi_x,J_y,\phi_y,z,\eta,s), where ss is the longitudinal coordinate along the reference orbit, z=s−v0tz=s-v_0t is the coordinate that measures the distance inside the beam with v0v_0 the nominal beam velocity, and η=ΔE/E\eta=\Delta E/E is the relative energy deviation. The distribution function is normalized so that

∫0∞dJx∫02πdϕx∫0∞dJy∫02πdϕy∫−∞∞dη∫0Ldz f=1,(7)\int_0^\infty dJ_x \int_0^{2\pi} d\phi_x \int_0^\infty dJ_y \int_0^{2\pi} d\phi_y \int_{-\infty}^{\infty} d\eta\int_0^L dz\, f = 1, \tag*{(7)}

where LL is the bunch length. Because Jx,ϕx,Jy,ϕyJ_x,\phi_x,J_y,\phi_y are integrals of motion, they do not enter the Vlasov equation for ff which can be written as

∂f∂s+eδEzγmc2∂f∂η+ηγ2∂f∂z=0.(8)\frac{\partial f}{\partial s}+\frac{e\delta E_z}{\gamma m c^2}\frac{\partial f}{\partial\eta}+\frac{\eta}{\gamma^2}\frac{\partial f}{\partial z}=0. \tag*{(8)}

The electric field δEz\delta E_z in this equation is a fluctuating electric field generated by the random motion of the beam particles; we will calculate its statistical properties in what follows. In our analysis, we ignore the steady state electric field in the beam (the space charge effect) as not relevant to collisions, as well as the fluctuating transverse components of the electric field that would be responsible for the IBS emittance growth effects. The last term in Eq. (8) takes into account the longitudinal slippage of the particles caused by the energy deviation η\eta. This term can be formally eliminated if instead of zz we use the variable ζ\zeta,

z=ζ+sη/γ2,(9)z = \zeta+ s\eta/\gamma^{2}, \tag*{(9)}

and consider ff as a function of this variable ζ\zeta instead of zz, f(Jx,ϕx,Jy,ϕy,η,ζ,s)f(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s). This function satisfies a somewhat simpler equation,

∂f∂s+eδEzγmc2∂f∂η=0.(10)\frac{\partial f}{\partial s}+\frac{e\delta E_z}{\gamma m c^{2}}\frac{\partial f}{\partial\eta}=0. \tag*{(10)}

Of course, all partial derivatives in this equation are taken keeping all other variables constant.

In the next step, we represent ff as a sum of the averaged over fluctuations part f0f_0 and a fluctuating (microscopic) part of the distribution function, δfM\delta f_{\mathrm{M}},

f=f0+δfM.(11)f=f_0+\delta f_{\mathrm{M}}. \tag*{(11)}

To guarantee a unique decomposition of ff into the two components, we require ⟨δfM⟩=0\langle\delta f_{\mathrm{M}}\rangle=0, where the angular brackets denote averaging over the fluctuations3. The smallness of the fluctuations is expressed as ∣δfM∣≪f0|\delta f_{\mathrm{M}}|\ll f_0. Together with δfM\delta f_{\mathrm{M}}, we treat the fluctuating field δEz\delta E_z as a small, first order, quantity, with the averaged value of δEz\delta E_z equal to zero, ⟨δEz⟩=0\langle\delta E_z\rangle=0. Substituting Eq. (11) into Eq. (10) we obtain

∂f0∂s+∂δfM∂s+eδEzγmc2∂f0∂η=−eδEzγmc2∂δfM∂η.(12)\frac{\partial f_0}{\partial s}+\frac{\partial\delta f_{\mathrm{M}}}{\partial s}+\frac{e\delta E_z}{\gamma m c^{2}}\frac{\partial f_0}{\partial\eta}=-\frac{e\delta E_z}{\gamma m c^{2}}\frac{\partial\delta f_{\mathrm{M}}}{\partial\eta}. \tag*{(12)}

Averaging Eq. (12) over the fluctuations and taking into account that ⟨δfM⟩=⟨δEz⟩=0\langle\delta f_{\mathrm{M}}\rangle=\langle\delta E_z\rangle=0, gives the Vlasov equation for the average distribution function f0f_0 with account of collisions,

∂f0∂s=Stη,(13)\frac{\partial f_0}{\partial s}=\mathrm{St}_{\eta}, \tag*{(13)}

where the energy collision integral Stη\mathrm{St}_{\eta} is,

Stη=−eγmc2∂∂η⟨δEzδfM⟩.(14)\mathrm{St}_{\eta}=-\frac{e}{\gamma m c^{2}}\frac{\partial}{\partial\eta}\langle\delta E_z\delta f_{\mathrm{M}}\rangle. \tag*{(14)}

Subtracting Eq. (13) from Eq. (12) gives an equation for the first order terms,

∂δfM∂s+eδEzγmc2∂f0∂η=0,(15)\frac{\partial\delta f_{\mathrm{M}}}{\partial s}+\frac{e\delta E_z}{\gamma m c^2}\frac{\partial f_0}{\partial\eta}=0, \tag*{(15)}

that defines the fluctuating part of the distribution function δfM\delta f_{\mathrm{M}}}.

We will solve Eq. (15) by integrating it along the particles’ trajectories. The integration is straightforward if we express δEz(x,y,z,s)\delta E_z(x,y,z,s) as a function of the action-angle variables,

δEz(x,y,z,s)→δEz(2βx(s)Jxcos⁡(ψx(s)+ϕx),2βy(s)Jycos⁡(ψy(s)+ϕy),ζ+sηγ2,s)≡δE~z(Jx,ϕx,Jy,ϕy,η,ζ,s),(16)\begin{aligned} \delta E_z(x,y,z,s)\to\delta E_z\left(\sqrt{2\beta_x(s)J_x}\cos(\psi_x(s)+\phi_x),\sqrt{2\beta_y(s)J_y}\cos(\psi_y(s)+\phi_y),\zeta+\frac{s\eta}{\gamma^2},s\right) \\ &\equiv\delta\widetilde{E}_z(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s), \tag*{(16)} \end{aligned}

where δE~z\delta\widetilde{E}_z is the electric field expressed in terms of the new set of variables. Then from Eq. (15) we obtain,

δfM(Jx,ϕx,Jy,ϕy,η,ζ,s)=−eγmc2∂f0∂η∫0sds′ δE~z(Jx,ϕx,Jy,ϕy,η,ζ,s′).(17)\delta f_{\mathrm{M}}(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s)=-\frac{e}{\gamma m c^2}\frac{\partial f_0}{\partial\eta}\int_0^s ds'\,\delta\widetilde{E}_z(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s'). \tag*{(17)}

Note that in this equation we have ignored the dependence of function f0f_0 versus ss and took the derivative ∂f0/∂η\partial f_0/\partial\eta outside the integral—this is justified by the fact that the dependence of f0f_0 on ss is given by Eq. (13) in which the right-hand side, according to Eq. (14), is of the second order of smallness. Substituting Eq. (17) into Eq. (14) we obtain

Stη=(eγmc2)2∂∂η(∂f0∂η∫0sds′⟨δE~z(s)δE~z(s′)⟩),(18)\mathrm{St}_{\eta}=\left(\frac{e}{\gamma m c^2}\right)^2\frac{\partial}{\partial\eta}\left(\frac{\partial f_0}{\partial\eta}\int_0^s ds'\left\langle\delta\widetilde{E}_z(s)\delta\widetilde{E}_z(s')\right\rangle\right), \tag*{(18)}

(here and below we omit all the arguments in δE~z\delta\widetilde{E}_z except for ss). We expect that when s→∞s\to\infty this integral converges to a finite value, which does not depend on ss, due to the vanishing correlations between δE~z(s′)\delta\widetilde{E}_z(s') and δE~z(s)\delta\widetilde{E}_z(s) when the difference between ss and s′s' becomes large. With this assumption, we can change the integration variable, s′′=s′−ss''=s'-s, extend the lower limit of integration over s′′s'' to −∞-\infty and then set s=0s=0:

Stη=(eγmc2)2∂∂η(∂f0∂η∫−∞0ds′′⟨δE~z(0)δE~z(s′′)⟩).(19)\mathrm{St}_{\eta}=\left(\frac{e}{\gamma m c^2}\right)^2\frac{\partial}{\partial\eta}\left(\frac{\partial f_0}{\partial\eta}\int_{-\infty}^{0}ds''\left\langle\delta\widetilde{E}_z(0)\delta\widetilde{E}_z(s'')\right\rangle\right). \tag*{(19)}

Comparing this equation with Eq. (3), we obtain an expression for the energy diffusion coefficient,

Dη=(eγmc2)2∫−∞0ds⟨δE~z(0)δE~z(s)⟩.(20)D_{\eta}=\left(\frac{e}{\gamma m c^2}\right)^2\int_{-\infty}^{0}ds\left\langle\delta\widetilde{E}_z(0)\delta\widetilde{E}_z(s)\right\rangle. \tag*{(20)}

Note that this diffusion coefficient DηD_{\eta} is a function of six variables, Jx,ϕx,Jy,ϕy,η,ζJ_x,\phi_x,J_y,\phi_y,\eta,\zeta, that characterize particle’s trajectory in the beam.

The energy diffusion coefficient

Our next step is to calculate the fluctuating electric field δEz\delta E_z in the beam and to find the correlator ⟨δE~z(s)δE~z(s′)⟩\langle\delta\widetilde{E}_z(s)\delta\widetilde{E}_z(s')\rangle. We will do this by expressing δEz\delta E_z through the beam density fluctuations δn\delta n. If the density perturbation is known as a function of coordinates xx, yy, zz and the variable ss, δn(x,y,z,s)\delta n(x,y,z,s), it is convenient to work with its Fourier transform δn^k(s)\delta\widehat{n}_{k}(s),

δn^k(s)=∫−∞∞dx dy dz eik⋅rδn(x,y,z,s),(21)\delta\widehat{n}_{k}(s)=\int_{-\infty}^{\infty}dx\,dy\,dz\,e^{i\boldsymbol{k}\cdot\boldsymbol{r}}\delta n(x,y,z,s), \tag*{(21)}

where k=(kx,ky,kz)\boldsymbol{k}=(k_x,k_y,k_z). Similar to Eq. (21), we also define the Fourier components of the electric field δE^z,k(s)\delta\widehat{E}_{z,k}(s). The relation between δE^z,k\delta\widehat{E}_{z,k} and δn^k\delta\widehat{n}_{k} is derived in Appendix A and is given by Eq. (54). For the correlator ⟨δE^z,k(s)δE^z,k′(s′)⟩\langle\delta\widehat{E}_{z,k}(s)\delta\widehat{E}_{z,k'}(s')\rangle this relation gives the following equation:

⟨δE^z,k(s)δE^z,k′(s′)⟩=(4πi)2e2kzkz′[γ2(kx2+ky2)+kz2][γ2(kx′2+ky′2)+kz′2]⟨δn^k(s)δn^k′(s′)⟩.(22)\left\langle\delta\widehat{E}_{z,k}(s)\delta\widehat{E}_{z,k'}(s')\right\rangle= \frac{(4\pi i)^2 e^2 k_z k'_z}{\left[\gamma^2(k_x^2+k_y^2)+k_z^2\right]\left[\gamma^2(k_x'^2+k_y'^2)+k_z'^2\right]} \left\langle\delta\widehat{n}_{k}(s)\delta\widehat{n}_{k'}(s')\right\rangle. \tag*{(22)}

Finally, making the inverse Fourier transform of this expression we can find the field correlator in coordinates xx, yy, zz, ss

⟨δEz(x,y,z,s)δEz(x′,y′,z′,s′)⟩=1(2π)6∫−∞∞dkx dky dkz dkx′ dky′ dkz′⟨δE^z,k(s)δE^z,k′(s′)⟩e−i(kxx+kyy+kzz+kx′x′+ky′y′+kz′z′).(23)\left\langle\delta E_z(x,y,z,s)\delta E_z(x',y',z',s')\right\rangle= \frac{1}{(2\pi)^6}\int_{-\infty}^{\infty}dk_x\,dk_y\,dk_z\,dk'_x\,dk'_y\,dk'_z \left\langle\delta\widehat{E}_{z,k}(s)\delta\widehat{E}_{z,k'}(s')\right\rangle e^{-i(k_xx+k_yy+k_zz+k'_xx'+k'_yy'+k'_zz')}. \tag*{(23)}

To find this correlator as a function of the action-angle variables (which we indicate by the tilde notation), on the right-hand side of this equation we express the x,y,zx,y,z through Jx,ϕx,Jy,ϕy,η,ζJ_x,\phi_x,J_y,\phi_y,\eta,\zeta, and x′,y′,z′x',y',z' through Jx′,ϕx′,Jy′,ϕy′,η′,ζ′J'_x,\phi'_x,J'_y,\phi'_y,\eta',\zeta' variables using Eqs. (6) and the relation z=ζ+sη/γ2z=\zeta+s\eta/\gamma^2.

An important aspect of fluctuations, as proved in the general theory of collisions in plasma [2], is that calculating δn\delta n and δEz\delta E_z we can ignore the influence of the electric field on particles’ orbits and assume that the particles execute free betatron oscillations as they propagate through the magnetic system. The derivation of the correlator ⟨δn^kδn^k′⟩\langle\delta\widehat{n}_k\delta\widehat{n}_{k'}\rangle for this case is carried out in Appendix B using the formalism of the Klimontovich distribution functions [19]. In that derivation, we assume that the averaged distribution function f0f_0 of the beam is a Gaussian distribution in energy and exponential distributions in JxJ_x and JyJ_y,

f0(Jx,Jy,η)=f0x(Jx)f0y(Jy)f0z(η),(24)f_0(J_x,J_y,\eta)=f_{0x}(J_x)f_{0y}(J_y)f_{0z}(\eta), \tag*{(24)}

where

f0x(Jx)=12πϵxe−Jx/ϵx,f0y(Jy)=12πϵye−Jy/ϵy,f0z(η)=1L2πσηe−η2/2ση2,(25)f_{0x}(J_x)=\frac{1}{2\pi\epsilon_x}e^{-J_x/\epsilon_x},\qquad f_{0y}(J_y)=\frac{1}{2\pi\epsilon_y}e^{-J_y/\epsilon_y},\qquad f_{0z}(\eta)=\frac{1}{L\sqrt{2\pi}\sigma_\eta}e^{-\eta^2/2\sigma_\eta^2}, \tag*{(25)}

with ϵx\epsilon_x and ϵy\epsilon_y the emittance in xx and yy directions, respectively, and ση\sigma_\eta the rms relative energy spread. Note that f0f_0 does not depend on ζ\zeta and hence represents a (locally) uniform bunch distribution. The auxiliary parameter LL, representing the system length, does not appear in the final result which is given by Eq. (63).

To further simplify the calculations, in what follows, we will assume a smooth focusing channel. This means that the beta functions βx\beta_x and βy\beta_y are constant, and the betatron phases are linear functions of the distance, ψx(s)=s/βx\psi_x(s)=s/\beta_x and ψy(s)=s/βy\psi_y(s)=s/\beta_y. Substituting Eqs. (22) and (B8) into Eq. (23) then yields,

⟨δE~z(s)δE~z(s′)⟩=e2ν2π3∫dkxdkydkx′dky′ G(kx,ky,kx′,ky′,s−s′)×exp⁡[i(kx2βxJxcos⁡(s/βx+ϕx)+ky2βyJycos⁡(s/βy+ϕy))]×exp⁡[i(kx′2βxJxcos⁡(s′/βx+ϕx)+ky′2βyJycos⁡(s′/βy+ϕy))]×exp⁡[−12ϵxβx(kx2+kx′2+2kxkx′cos⁡[(s′−s)/βx])]×exp⁡[−12ϵyβy(ky2+ky′2+2kyky′cos⁡[(s′−s)/βy])],(26)\begin{aligned} \langle\delta\widetilde{E}_z(s)\delta\widetilde{E}_z(s')\rangle ={}&\frac{e^2\nu}{2\pi^3}\int dk_xdk_ydk'_xdk'_y\,G(k_x,k_y,k'_x,k'_y,s-s')\\ &\times\exp\left[i\left(k_x\sqrt{2\beta_xJ_x}\cos(s/\beta_x+\phi_x)+k_y\sqrt{2\beta_yJ_y}\cos(s/\beta_y+\phi_y)\right)\right]\\ &\times\exp\left[i\left(k'_x\sqrt{2\beta_xJ_x}\cos(s'/\beta_x+\phi_x)+k'_y\sqrt{2\beta_yJ_y}\cos(s'/\beta_y+\phi_y)\right)\right]\\ &\times\exp\left[-\frac{1}{2}\epsilon_x\beta_x\left(k_x^2+k_x'^2+2k_xk'_x\cos[(s'-s)/\beta_x]\right)\right]\\ &\times\exp\left[-\frac{1}{2}\epsilon_y\beta_y\left(k_y^2+k_y'^2+2k_yk'_y\cos[(s'-s)/\beta_y]\right)\right], \tag*{(26)} \end{aligned}

where

G(kx,ky,kx′,ky′,s−s′)=∫−∞∞dkzkz2exp⁡[−12ση2kz2(s−s′)2/γ4][γ2(kx2+ky2)+kz2][γ2(kx′2+ky′2)+kz2].(27)G(k_x,k_y,k'_x,k'_y,s-s')=\int_{-\infty}^{\infty}dk_z\frac{k_z^2\exp\left[-\frac{1}{2}\sigma_\eta^2k_z^2(s-s')^2/\gamma^4\right]}{\left[\gamma^2(k_x^2+k_y^2)+k_z^2\right]\left[\gamma^2(k_x'^2+k_y'^2)+k_z^2\right]}. \tag*{(27)}

and we have carried out the integration over kz′k'_z using the delta function δ(kz+kz′)\delta(k_z+k'_z) in Eq. (63). Substituting Eq. (26) into Eq. (20) will give us the diffusion coefficient for a particle trajectory that is characterized by the parameters JxJ_x, ϕx\phi_x, JyJ_y, ϕy\phi_y. While this detailed knowledge might be of use in some particular problems, in many applications it makes sense to average it using the equilibrium distribution function (24). As a first step in this calculation, we average Eq. (26) over the phases ϕx,ϕy\phi_x,\phi_y. Given an explicit dependence of the integrand of Eq. (26) on ϕx\phi_x and ϕy\phi_y through the exponential function, this averaging is straightforward. We denote the result of this averaging by double angular brackets:

⟨⟨δE~z(s′)δE~z(s)⟩⟩=e2ν2π3∫dkxdkydkx′dky′G(kx,ky,kx′,ky′,s−s′)×J0(2βxJx[kx2+kx′2+2kxkx′cos⁡(s−s′)/βx])×J0(2βyJy[ky2+ky′2+2kyky′cos⁡(s−s′)/βy])×exp⁡[−12ϵxβx(kx2+kx′2+2kxkx′cos⁡[(s′−s)/βx])]×exp⁡[−12ϵyβy(ky2+ky′2+2kyky′cos⁡[(s′−s)/βy])].(28)\left\langle\left\langle\delta\widetilde{E}_z(s')\delta\widetilde{E}_z(s)\right\rangle\right\rangle= \frac{e^2\nu}{2\pi^3}\int dk_xdk_ydk'_xdk'_yG(k_x,k_y,k'_x,k'_y,s-s') \\ \times J_0\left(\sqrt{2\beta_xJ_x\left[k_x^2+k_x'^2+2k_xk'_x\cos(s-s')/\beta_x\right]}\right) \\ \times J_0\left(\sqrt{2\beta_yJ_y\left[k_y^2+k_y'^2+2k_yk'_y\cos(s-s')/\beta_y\right]}\right) \\ \times\exp\left[-\frac{1}{2}\epsilon_x\beta_x\left(k_x^2+k_x'^2+2k_xk'_x\cos\left[(s'-s)/\beta_x\right]\right)\right] \\ \times\exp\left[-\frac{1}{2}\epsilon_y\beta_y\left(k_y^2+k_y'^2+2k_yk'_y\cos\left[(s'-s)/\beta_y\right]\right)\right] . \tag*{(28)}

where J0J_0 is the Bessel function of the zero order. In the next step, we average this expression over the action variables integrating it over JxJ_x and JyJ_y with the weight factors ϵx−1e−J/ϵx\epsilon_x^{-1}e^{-J/\epsilon_x} and ϵy−1e−J/ϵy\epsilon_y^{-1}e^{-J/\epsilon_y}. This integration can be carried out using the formula

∫0∞dx e−xJ0(ax)=e−a2/4,(29)\int_0^\infty dx\,e^{-x}J_0\left(a\sqrt{x}\right)=e^{-a^2/4}, \tag*{(29)}

valid for a>0a>0. Using the long overline for the result, we obtain

⟨⟨δE~z(s′)δE~z(s)⟩⟩=e2ν2π3∫dkxdkydkx′dky′G(kx,ky,kx′,ky′,s−s′)×exp⁡[−ϵxβx(kx2+kx′2+2kxkx′cos⁡[(s′−s)/βx])]×exp⁡[−ϵyβy(ky2+ky′2+2kyky′cos⁡[(s′−s)/βy])].(30)\left\langle\left\langle\delta\widetilde{E}_z(s')\delta\widetilde{E}_z(s)\right\rangle\right\rangle= \frac{e^2\nu}{2\pi^3}\int dk_xdk_ydk'_xdk'_yG(k_x,k_y,k'_x,k'_y,s-s') \\ \times\exp\left[-\epsilon_x\beta_x\left(k_x^2+k_x'^2+2k_xk'_x\cos\left[(s'-s)/\beta_x\right]\right)\right] \\ \times\exp\left[-\epsilon_y\beta_y\left(k_y^2+k_y'^2+2k_yk'_y\cos\left[(s'-s)/\beta_y\right]\right)\right] . \tag*{(30)}

Our result can be even more simplified if we make an additional assumption of equal beta functions, βx=βy=β\beta_x=\beta_y=\beta, and equal emittances, ϵx=ϵy=ϵ\epsilon_x=\epsilon_y=\epsilon. We then have

⟨⟨δE~z(s′)δE~z(s)⟩⟩=e2ν2π3∫dkxdkydkx′dky′G(kx,ky,kx′,ky′,s−s′)×exp⁡[−ϵβ(kx2+ky2+kx′2+ky′2)]exp⁡[−2ϵβ(kxkx′+kyky′)cos⁡[(s′−s)/β]].(31)\left\langle\left\langle\delta\widetilde{E}_z(s')\delta\widetilde{E}_z(s)\right\rangle\right\rangle= \frac{e^2\nu}{2\pi^3}\int dk_xdk_ydk'_xdk'_yG(k_x,k_y,k'_x,k'_y,s-s') \\ \times\exp\left[-\epsilon\beta\left(k_x^2+k_y^2+k_x'^2+k_y'^2\right)\right]\exp\left[-2\epsilon\beta\left(k_xk'_x+k_yk'_y\right)\cos\left[(s'-s)/\beta\right]\right] . \tag*{(31)}

Two more integrations can be carried out if we use the cylindrical coordinate system and write kx=k⊥cos⁡αk_x=k_\perp\cos\alpha, ky=k⊥sin⁡αk_y=k_\perp\sin\alpha and kx′=k⊥′cos⁡βk'_x=k'_\perp\cos\beta, ky′=k⊥′sin⁡βk'_y=k'_\perp\sin\beta,

⟨⟨δE~z(s′)δE~z(s)⟩⟩=e2ν2π3∫k⊥dk⊥ dα k⊥′dk⊥′ dβ G(k⊥,k⊥′,s−s′)exp⁡[−ϵβ(k⊥2+k⊥′2)]×exp⁡[−2ϵβk⊥k⊥′cos⁡(α−β)cos⁡[(s′−s)/β]],(32)\left\langle\left\langle\delta\widetilde{E}_z(s')\delta\widetilde{E}_z(s)\right\rangle\right\rangle= \frac{e^2\nu}{2\pi^3}\int k_\perp dk_\perp\,d\alpha\,k'_\perp dk'_\perp\,d\beta\,G(k_\perp,k'_\perp,s-s')\exp\left[-\epsilon\beta\left(k_\perp^2+k_\perp'^2\right)\right] \\ \times\exp\left[-2\epsilon\beta k_\perp k'_\perp\cos(\alpha-\beta)\cos\left[(s'-s)/\beta\right]\right] , \tag*{(32)}

where we have used the fact that GG only depends on the absolute values of k⊥k_\perp and k⊥′k'_\perp. The integration over α\alpha and β\beta can be carried out using the formula

∫02πdϕ e−acos⁡(ϕ)=2πI0(a).(33)\int_0^{2\pi}d\phi\,e^{-a\cos(\phi)}=2\pi I_0(a). \tag*{(33)}

The result is,

⟨⟨δEz(s′)~‾δEz(s)~‾⟩⟩=2e2νπ∫0∞dk⊥dk⊥′G(k⊥,k⊥′,s−s′)×k⊥k⊥′exp⁡[−ϵβ(k⊥2+k⊥′2)]I0(2ϵβk⊥k⊥′cos⁡[(s′−s)/β]).(34)\left\langle\left\langle\overline{\widetilde{\delta E_z(s')}}\overline{\widetilde{\delta E_z(s)}}\right\rangle\right\rangle= \frac{2e^2\nu}{\pi}\int_0^\infty dk_\perp dk'_\perp G(k_\perp,k'_\perp,s-s') \times k_\perp k'_\perp\exp\left[-\epsilon\beta\left(k_\perp^2+k_\perp'^2\right)\right] I_0\left(2\epsilon\beta k_\perp k'_\perp\cos\left[(s'-s)/\beta\right]\right). \tag*{(34)}

We now introduce the dimensionless variables κ=ϵβ k⊥\kappa=\sqrt{\epsilon\beta}\,k_\perp and ξ=s/β\xi=s/\beta, which cast Eq. (34) into the following one

⟨⟨δEz(s′)~‾δEz(s)~‾⟩⟩=2νe2πγ(ϵβ)3/2∫0∞dκ dκ′ G(κ,κ′,ξ−ξ′)×κκ′exp⁡[−(κ2+κ′2)]I0(2κκ′cos⁡(ξ′−ξ)),(35)\left\langle\left\langle\overline{\widetilde{\delta E_z(s')}}\overline{\widetilde{\delta E_z(s)}}\right\rangle\right\rangle= \frac{2\nu e^2}{\pi\gamma(\epsilon\beta)^{3/2}}\int_0^\infty d\kappa\,d\kappa' \,G(\kappa,\kappa',\xi-\xi') \times\kappa\kappa'\exp\left[-\left(\kappa^2+\kappa'^2\right)\right] I_0\left(2\kappa\kappa'\cos(\xi'-\xi)\right), \tag*{(35)}

with

G(κ,κ′,ξ)=∫−∞∞dqzqz2exp⁡(−12ση2qz2ξ2β/(ϵγ2))(κ2+qz2)(κ′2+qz2),(36)G(\kappa,\kappa',\xi)=\int_{-\infty}^{\infty}dq_z\frac{q_z^2\exp\left(-\frac{1}{2}\sigma_\eta^2q_z^2\xi^2\beta/(\epsilon\gamma^2)\right)}{(\kappa^2+q_z^2)(\kappa'^2+q_z^2)}, \tag*{(36)}

where qz=ϵβkz/γq_z=\sqrt{\epsilon\beta}k_z/\gamma.

Recalling Eq. (20), we now have the averaged diffusion coefficient Dη‾\overline{D_\eta},

Dη‾=2νre2πγ3ϵ3/2β1/2∫0∞dκ dκ′∫0∞dξ G(κ,κ′,ξ)×κκ′exp⁡[−(κ2+κ′2)]I0(2κκ′cos⁡(ξ)).(37)\overline{D_\eta}=\frac{2\nu r_e^2}{\pi\gamma^3\epsilon^{3/2}\beta^{1/2}}\int_0^\infty d\kappa\,d\kappa'\int_0^\infty d\xi\,G(\kappa,\kappa',\xi) \times\kappa\kappa'\exp\left[-\left(\kappa^2+\kappa'^2\right)\right]I_0\left(2\kappa\kappa'\cos(\xi)\right). \tag*{(37)}

Cold beam limit and small-scale fluctuations

As we mentioned in the Introduction, Eq. (5) gives the averaged energy diffusion coefficient for a cold beam derived from the conventional IBS theory. It is natural to try to reproduce that formula in our formalism setting ση=0\sigma_\eta=0 in Eq. (37). In this limit, the function GG (36) simplifies,

G(κ,κ′,ξ′)=∫−∞∞dqzqz2(κ2+qz2)(κ′2+qz2)=πκ+κ′,(38)G(\kappa,\kappa',\xi')=\int_{-\infty}^{\infty}dq_z\frac{q_z^2}{(\kappa^2+q_z^2)(\kappa'^2+q_z^2)}=\frac{\pi}{\kappa+\kappa'}, \tag*{(38)}

and gives for the diffusion coefficient (37) the following expression:

Dη‾=2νre2πγ3ϵ3/2β1/2∫0∞dξ∫0∞dκ dκ′κκ′κ+κ′exp⁡[−(κ2+κ′2)]I0(2κκ′cos⁡(ξ)).(39)\overline{D_\eta}=\frac{2\nu r_e^2}{\pi\gamma^3\epsilon^{3/2}\beta^{1/2}}\int_0^\infty d\xi\int_0^\infty d\kappa\,d\kappa'\frac{\kappa\kappa'}{\kappa+\kappa'}\exp\left[-\left(\kappa^2+\kappa'^2\right)\right]I_0\left(2\kappa\kappa'\cos(\xi)\right). \tag*{(39)}

Unfortunately, the integral on the right-hand side of the equation diverges at ξ→∞\xi\to\infty because I0(2κκ′cos⁡(ξ))I_0(2\kappa\kappa'\cos(\xi)) is a periodic function of the variable ξ\xi and does not vanish at infinity. We will return to the discussion of physical mechanism of this divergence in the next section.

Nevertheless, it turns out that we can obtain the diffusion coefficient (5) from Eq. (39) if we limit the integration in Eq. (39) by the wavenumbers that are much larger than the inverse bunch size (ϵβ)−1/2(\epsilon\beta)^{-1/2}. In other words, we will only take into account small scale fluctuations of the electric field. In terms of our dimensionless variables κ\kappa and κ′\kappa', it means that we limit the integration by the region κ,κ′≫1\kappa,\kappa' \gg1. In this limit, we can use the asymptotic expression of I0I_0 for a large argument,

I0(x)≈(2π∣x∣)−1/2e∣x∣.(40)I_0(x) \approx(2\pi|x|)^{-1/2}e^{|x|}. \tag*{(40)}

Substituting this expression into the internal integral in Eq. (39) we obtain,

∫dκ dκ′ κκ′κ+κ′exp⁡[−(κ2+κ′2−2κκ′∣cos⁡(ξ)∣)]14πκκ′∣cos⁡(ξ)∣.(41)\int d\kappa\,d\kappa'\,\frac{\kappa\kappa'}{\kappa+\kappa'}\exp\left[-\left(\kappa^2+\kappa'^2-2\kappa\kappa'|\cos(\xi)|\right)\right]\frac{1}{\sqrt{4\pi\kappa\kappa'|\cos(\xi)|}}. \tag*{(41)}

In this integral, we do not specify the integration limits and will return to their choice below. The exponential function in the integrand is exponentially small for most values of κ≫1\kappa\gg1 and κ′≫1\kappa' \gg1, unless κ2+κ′2\kappa^2+\kappa'^2 is almost cancelled by 2κκ′∣cos⁡(ξ)∣2\kappa\kappa'|\cos(\xi)|. This happens when κ′\kappa' is close to κ\kappa: κ′=κ+q\kappa'=\kappa+q with ∣q∣≪κ|q|\ll\kappa, and ξ\xi is close to πn\pi n, where nn is an integer (for κ′=κ\kappa'=\kappa and ∣cos⁡(ξ)∣=1|\cos(\xi)|=1, the argument of the exponential function is equal to zero). Each such occurrence can be considered as an analog of a two-particle collision in the framework of the binary collisions. Let us consider one value of nn, and write ξ=πn+δξ\xi=\pi n+\delta\xi with ∣δξ∣≪1|\delta\xi|\ll1. We then have

∣cos⁡(ξ)∣=∣cos⁡(πn+δξ)∣≈1−12δξ2,(42)|\cos(\xi)|=|\cos(\pi n+\delta\xi)|\approx1-\frac{1}{2}\delta\xi^2, \tag*{(42)}

and the integral Eq. (41) is approximated as

14π∫dκ dκ′ exp⁡[−(κ2+κ′2)+2κκ′(1−12δξ2)]=14π∫dκ∫−∞∞dq exp⁡[−q2−κ2δξ2]=14∫dκ exp⁡(−κ2δξ2).(43)\begin{aligned} \frac{1}{4\sqrt{\pi}}\int d\kappa\,d\kappa'\,\exp\left[-\left(\kappa^2+\kappa'^2\right)+2\kappa\kappa'\left(1-\frac{1}{2}\delta\xi^2\right)\right] &=\frac{1}{4\sqrt{\pi}}\int d\kappa\int_{-\infty}^{\infty}dq\,\exp\left[-q^2-\kappa^2\delta\xi^2\right]\\ &=\frac{1}{4}\int d\kappa\,\exp\left(-\kappa^2\delta\xi^2\right). \tag*{(43)} \end{aligned}

Note that since we assume κ≫1\kappa\gg1, the integral over qq in this equation converges over the scale q∼1q\sim1 which is much smaller than κ\kappa, and for this reason we extended the integration over qq from −∞-\infty to +∞+\infty.

We now need to substitute this expression to the integral in Eq. (39). To compare the result with the standard collision term, we consider only the “first collision”, that is n=0n=0 in Eq. (42). The integration over ξ\xi in Eq. (39) is replaced by the integration over δξ\delta\xi and because the integral converges rapidly when δξ→∞\delta\xi\to\infty, the integration interval can be extended to infinity:

Dˉη=νre22πγ3ϵ3/2β1/2∫dκ∫0∞dδξexp⁡(−κ2δξ2)=π4νre2γ3ϵ3/2β1/2∫dκκ.(44)\bar{D}_{\eta} = \frac{\nu r_{e}^{2}}{2\pi\gamma^{3}\epsilon^{3/2}\beta^{1/2}} \int d\kappa\int_{0}^{\infty} d\delta\xi\exp\left(-\kappa^{2}\delta\xi^{2}\right) = \frac{\sqrt{\pi}}{4}\frac{\nu r_{e}^{2}}{\gamma^{3}\epsilon^{3/2}\beta^{1/2}} \int\frac{d\kappa}{\kappa}. \tag*{(44)}

This result coincides with Eq. (5) if we limit the integration over κ\kappa between some κmin⁡\kappa_{\min} and κmax⁡\kappa_{\max} and associate the Coulomb logarithms with such integral:

Λ=∫κmin⁡κmax⁡dκκ.(45)\Lambda= \int_{\kappa_{\min}}^{\kappa_{\max}} \frac{d\kappa}{\kappa}. \tag*{(45)}

Since we assumed in our derivation κ≫1\kappa\gg1, κmin⁡\kappa_{\min} should also be large, κmin⁡≫1\kappa_{\min} \gg1. This means that we take into account the fluctuations in the beam with characteristic scales much smaller than the beam size. Comparing Eq. (45) with Eq. (1), we see that κmin⁡\kappa_{\min} can be associated with the inverse maximum impact parameter bmax⁡b_{\max} (normalized by the beam size), κmin⁡∼ϵβ/bmax⁡\kappa_{\min} \sim\sqrt{\epsilon\beta}/b_{\max}. This also tells us that κmax⁡∼ϵβ/bmin⁡\kappa_{\max} \sim\sqrt{\epsilon\beta}/b_{\min}. The fluctuations on this scale correspond to the large-angle collisions and cannot be properly treated by our theory that assumes a small δEz\delta E_z and uses a perturbation theory based on this smallness.

Calculation of the diffusion coefficient for ση≠0\sigma_{\eta} \ne0

We now drop the assumption ση=0\sigma_{\eta} = 0 and consider the general case given by Eq. (37) which we write as

Dˉη=2νre2πγ3ϵ3/2β1/2J,(46)\bar{D}_{\eta} = \frac{2\nu r_{e}^{2}}{\pi\gamma^{3}\epsilon^{3/2}\beta^{1/2}}J, \tag*{(46)}

where

J=∫0∞dκ∫0∞dκ′∫0∞dξ G(κ,κ′,ξ)κκ′exp⁡[−(κ2+κ′2)]I0(2κκ′cos⁡(ξ))=∫0∞dκ∫0∞dκ′∫0∞dξ∫−∞∞dqz qz2exp⁡(−τqz2ξ2)(κ2+qz2)(κ′2+qz2)κκ′exp⁡[−(κ2+κ′2)]I0(2κκ′cos⁡(ξ)),(47)\begin{aligned} J = \int_{0}^{\infty} d\kappa\int_{0}^{\infty} d\kappa' \int_{0}^{\infty} d\xi\,G(\kappa,\kappa',\xi)\kappa\kappa' \exp\left[-\left(\kappa^{2}+\kappa'^{2}\right)\right] I_{0}\left(2\kappa\kappa'\cos(\xi)\right) \\ &= \int_{0}^{\infty} d\kappa\int_{0}^{\infty} d\kappa' \int_{0}^{\infty} d\xi\int_{-\infty}^{\infty} dq_{z}\,\frac{q_{z}^{2}\exp\left(-\tau q_{z}^{2}\xi^{2}\right)}{\left(\kappa^{2}+q_{z}^{2}\right)\left(\kappa'^{2}+q_{z}^{2}\right)}\kappa\kappa' \exp\left[-\left(\kappa^{2}+\kappa'^{2}\right)\right] I_{0}\left(2\kappa\kappa'\cos(\xi)\right), \tag*{(47)} \end{aligned}

with τ=12ση2β/ϵγ2\tau= \frac{1}{2}\sigma_{\eta}^{2}\beta/\epsilon\gamma^{2}. Even though we set the upper limits of the integration over κ\kappa and κ′\kappa' to infinity, we will later need to limit it by some finite value κmax⁡\kappa_{\max}, as we did in Eq. (44). For numerical calculations of the integral in Eq. (47), we will use the cylindrical coordinates in (κ,κ′)(\kappa,\kappa') space: κ=Kcos⁡α\kappa= K\cos\alpha, κ′=Ksin⁡α\kappa' = K\sin\alpha, dκ dκ′=K dK dαd\kappa\,d\kappa' = K\,dK\,d\alpha, and represent the four-dimensional integral JJ as a combination of one and two-dimensional ones:

J=∫0∞dK P(K),(48)J = \int_{0}^{\infty} dK\,P(K), \tag*{(48)}

where

P(K)=∫0∞dξ Q(ξ,K),(49)P(K) = \int_{0}^{\infty} d\xi\,Q(\xi,K), \tag*{(49)}

and

Q(ξ,K)=K3∫0π/2dα∫−∞∞dqz qz2exp⁡(−τqz2ξ2)(K2cos⁡2α+qz2)(K2sin⁡2α+qz2)×cos⁡(α)sin⁡(α)e−K2I0[2K2cos⁡(α)sin⁡(α)cos⁡(ξ)].(50)Q(\xi,K) = K^3 \int_{0}^{\pi/2} d\alpha\int_{-\infty}^{\infty} dq_z\,\frac{q_z^2\exp\left(-\tau q_z^2\xi^2\right)}{\left(K^2\cos^2\alpha+q_z^2\right)\left(K^2\sin^2\alpha+q_z^2\right)}\times\cos(\alpha)\sin(\alpha)e^{-K^2}I_0\left[2K^2\cos(\alpha)\sin(\alpha)\cos(\xi)\right]. \tag*{(50)}

The function QQ can be interpreted as the correlation of fields with the wavenumber ∼K\sim K at the distance s=βξs=\beta\xi.

Unfortunately, not much can be done analytically with the integral Eq. (48). Of course, for a given set of beam and lattice parameters, the diffusion coefficient can be calculated numerically. To illustrate such a calculation, we consider an electron beam with the energy corresponding to γ=2000\gamma=2000, the normalized transverse emittance ϵn=1 μm\epsilon_n=1\ \mu\mathrm{m} and the relative energy spread ση=5×10−4\sigma_\eta=5\times10^{-4}. The beam propagates in a focusing channel with β=8 m\beta=8\ \mathrm{m}. For the dimensionless parameter τ\tau, this gives τ=5×10−4\tau=5\times10^{-4}.

First, we calculate the function Q(ξ,K)Q(\xi,K) using the two-dimensional integral Eq. (50). The plot of this function for three different values of K=10,20,30K=10,20,30 is shown in Figure 2. We see that the field correlations quickly decrease almost to zero near the origin, at ξ≪1\xi\ll1, but then reappear as sharp spikes at ξ≈π,2π,3π,…\xi\approx\pi,2\pi,3\pi,\ldots. The amplitudes of the subsequent spikes become smaller, with the spikes corresponding to larger values of KK decaying faster than those with smaller values of KK. The plot of QQ in the region of the first spike, at ξ<0.3\xi<0.3 is shown in Figure 3.

Figure 2. The plot of the function QQ versus ξ=s/β\xi= s/\beta for a beam with τ=5×10−4\tau= 5 \times10^{-4} for three different values of KK.

Figure 3. The function QQ at small values of ξ=s/β\xi= s/\beta.

For comparison, in Figure 4 we show the function QQ for the same beam parameters, except for the value of ση\sigma_\eta set to zero, ση=0\sigma_\eta=0, that is in the limit of the cold beam. As we see, in this case the amplitude of the subsequent spikes remains constant. Since the function QQ needs to be integrated over ξ\xi from zero to infinity, it is clear that such an integral would diverge. This is the situation that we have discussed in § 5.

Figure 4. The plot of the function QQ versus ξ=s/β\xi=s/\beta for a cold beam with τ=0\tau=0 for three different values of KK.

The large-scale correlations with small values of KK decay on a relatively large distance. This is illustrated by Figure 5 which shows the function QQ for K=1K = 1. We see that it takes many betatron periods for QQ to decay to small values in this case.

Figure 5. The plot of the function QQ versus ξ=s/β\xi=s/\beta for a beam with τ=5×10−4\tau=5\times10^{-4} for K=1K=1.

Finally, Figure 6 shows numerically calculated function P(K)P(K) for our case (ση=10−4\sigma_{\eta} = 10^{-4}). Theoretical analysis of Eqs. (48) and (50) reveals that at large values of KK, P(K)P(K) decays as π3/2/8K\pi^{3/2}/8K, which is in agreement with numerical calculations shown in Figure 6. According to Eq. (48), to find JJ, we need to integrate P(K)P(K) over KK from zero to infinity. This can be done by integrating the numerically calculated P(K)P(K) from zero to some large value K0K_0 and then using the asymptotic formula P(K)≈π3/2/8KP(K) \approx\pi^{3/2}/8K for K>K0K > K_0. Of course, such an integral logarithmically diverges at infinity and, as it was done in § 5, we need to truncate the integration interval by some value Kmax⁡≫1K_{\max}\gg1. Carrying our this calculation, we obtain the following result:

Figure 6. The plot of the integral P(K)P(K) (black line). At large values of KK this integral decays as π3/2/8K\pi^{3/2}/8K shown by the blue dashed line.

J=33.3+π3/28log⁡(Kmax⁡ϵβ).(51)J=33.3+\frac{\pi^{3/2}}{8}\log\left(K_{\max}\sqrt{\epsilon\beta}\right). \tag*{(51)}

The second term in this expression substituted to Eq. (46) gives the energy diffusion coefficient that (apart of somewhat different definition of the log term) coincides with Eq. (46) and Eq. (5). The first term in Eq. (51) adds a noticeable correction to the Coulomb logarithm due to the long-range correlations of the fluctuating fields in the beam.

Summary and discussion

In this work, we derived the energy diffusion coefficient due to the intra-beam scattering in a relativistic beam. In contrast to the conventional approach, we used the framework of the fluctuations of the electric field in the beam caused by random motion of beam particles. We also included the effect of betatron oscillations on the collisions in our analysis.

In Section IV, we derived a general expression for the energy diffusion coefficient Dˉ\bar{D} averaged over the distribution function of the beam. Quite surprisingly, an attempt to calculate its value for a beam with zero energy spread resulted in a diverging integral. In § 5, we have shown that the situation can be rectified if we limit the integration domain by wavenumbers of the fluctuating fields that are much smaller than the transverse beam size. In this case, we obtained the result that coincides with the conventional formula (5) with a logarithmic factor that mimics the Coulomb logarithm. Such truncation of the integration range means that Eq. (5) ignores the long-range fluctuations that exist in the beam and amplifies the diffusion caused by the short-range fluctuations.

A numerical example presented in Section VI clearly demonstrated the presence of such correlations in the field correlator QQ. For a finite energy spread shown in Figure 2, these correlations decay with distance, however, in the limit ση=0\sigma_\eta= 0 they persist without damping as shown in Figure 4. In the latter case, these repeated spikes in the correlations lead to a divergent integral in the expression for Dˉ\bar{D}. For ση≠0\sigma_\eta\ne0, the extended correlations manifest itself in the increase of the effective Coulomb logarithm as demonstrated by Eq. (51).

In our calculations, we have assumed a smooth focusing with the equal betatron oscillation frequencies in xx and yy directions. In this situation, the microscopic positions of the particles arrange themselves into the same pattern after each betatron period (and the positions are mirror symmetric after half a period). This is illustrated in Figure 7. In the language of binary

Figure 7. The plot of the positions of the test particle (red dot) and the source particles (black dots) executing betatron oscillation. The red line is the trajectory of the test particle and the blue lines are the trajectories of the source particles. The positions of all particles repeat every betatron period.

collisions, such patterns mean that a collision between two particles will be repeated every betatron period, and the result of such collisions adds up in a regular fashion instead of being accumulated as random encounters. This also means that in this case the diffusion-like scaling in the growth of the energy spread in the beam due to IBS, Δση∝s\Delta\sigma_\eta\propto s, is not valid; it will be replaced by Δση∝s2\Delta\sigma_\eta\propto s^2.

The persistence of the long-range correlations in the electric field of the beam has been demonstrated in this paper for very special model of the beam transport. In reality, in addition to the beam energy spread, there will be other mechanisms that will be suppressing these correlations, e.g., the lattice chromaticity and nonlinearity, quantum diffusion in the beam, etc. Their influence of these factors on the field correlations requires further studies..

Finally, we note that even though we considered only the energy diffusion in this paper, other elements of the collision integral will be affected by the long-scale correlations. In particular, on can probably see their effect in the emittance growth of the beam that is crucial in the modern light sources.

Acknowledgements

I would like to thank Z. Huang, R. Robles and S. Kladov for useful discussions.

This work was supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contracts No. DE-AC02-76SF00515.

This work

Calculation of the longitudinal field in the beam

We consider the beam moving with velocity v0=β0cv_0 = \beta_0 c along the zz axes. Let the charge density perturbation of the beam be eδn(x,y,z)e\delta n(x,y,z), then one can obtain the following equation for the longitudinal component of the electric field from Maxwell’s equations,

∇⊥δEz+1γ2∂zzδEz=4πeγ2∂zδn,(52)\nabla_{\perp}\delta E_z + \frac{1}{\gamma^2}\partial_{zz}\delta E_z = \frac{4\pi e}{\gamma^2}\partial_z\delta n, \tag*{(52)}

where z=s−v0tz = s - v_0t, ∇⊥\nabla_{\perp} is the transverse gradient operator, and ∂z\partial_z and ∂zz\partial_{zz} denote the first and the second partial derivatives along zz. Making the 3D Fourier transformation,

δn^k=∫−∞∞dx dy dz e−i(kxx+kyy+kzz)δn(x,y,z),δE^z,k=∫−∞∞dx dy dz e−i(kxx+kyy+kzz)δEz(x,y,z),(53)\begin{aligned} \widehat{\delta n}_k &= \int_{-\infty}^{\infty} dx\,dy\,dz\,e^{-i(k_xx+k_yy+k_zz)}\delta n(x,y,z),\\ \widehat{\delta E}_{z,k} &= \int_{-\infty}^{\infty} dx\,dy\,dz\,e^{-i(k_xx+k_yy+k_zz)}\delta E_z(x,y,z), \tag*{(53)} \end{aligned}

we obtain

δE^z,k=4πeikzγ2(kx2+ky2)+kz2δn^k.(54)\widehat{\delta E}_{z,k} = \frac{4\pi e i k_z}{\gamma^2(k_x^2+k_y^2)+k_z^2}\widehat{\delta n}_k. \tag*{(54)}

The energy modulation in Fourier representation (per unit length of path) is

δη^k=−eδE^z,k1meγc2=−4πe2ikzγ2(kx2+ky2)+kz21meγc2δn^k.(55)\widehat{\delta\eta}_k = -e\widehat{\delta E}_{z,k}\frac{1}{m_e\gamma c^2} = -\frac{4\pi e^2 i k_z}{\gamma^2(k_x^2+k_y^2)+k_z^2}\frac{1}{m_e\gamma c^2}\widehat{\delta n}_k. \tag*{(55)}

Density fluctuations in the beam

To calculate the density fluctuations, we have to start from the following microscopic (Klimontovich) distribution function describing a collection of non-interacting particles that execute betatron oscillations,

fM(Jx,ϕx,Jy,ϕy,η,ζ,s)=1N∑n=1Nδ(Jx−Jx,n)δ(ϕx−ψx(s)−ϕx,n)δ(Jy−Jy,n)δ(ϕy−ψy(s)−ϕy,n)δ(η−ηn)δ(ζ−ζn),(56)\begin{aligned} f_{\mathrm{M}}(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s) ={}& \frac{1}{N}\sum_{n=1}^{N}\delta(J_x-J_{x,n})\delta(\phi_x-\psi_x(s)-\phi_{x,n})\\ &\delta(J_y-J_{y,n})\delta(\phi_y-\psi_y(s)-\phi_{y,n})\delta(\eta-\eta_n)\delta(\zeta-\zeta_n), \tag*{(56)} \end{aligned}

where N≫1N \gg1 is the number of particles in the beam. This distribution function is normalized so that ∫dJx dJy dϕx dϕy dη dζ fM=1\int dJ_x\,dJ_y\,d\phi_x\,d\phi_y\,d\eta\,d\zeta\,f_{\mathrm{M}}=1.

The 6N6N parameters in Eq. (56), Jx,n,ϕx,n,Jy,n,ϕy,n,ζn,ηnJ_{x,n},\phi_{x,n},J_{y,n},\phi_{y,n},\zeta_n,\eta_n, are random variables that are distributed according to the averaged distribution function f0f_0. Specifically, the phases ϕx,n\phi_{x,n} and ϕy,n\phi_{y,n} are uniformly distributed in the interval [0,2π][0,2\pi], the longitudinal coordinate ζ\zeta is uniformly distributed over [0,L][0,L], and the variables Jx,nJ_{x,n}, Jy,nJ_{y,n} and ηn\eta_n are distributed according the distribution functions f0xf_{0x}, f0yf_{0y} and f0zf_{0z}, respectively. We will denote by the angular brackets ⟨…⟩\langle\ldots\rangle averaging over these variables:

⟨…⟩=∏n=1N∫0∞f0x(Jx,n) dJx,n∫02πdϕx,n∫0∞dJy,n f0y(Jy,n)∫02πdϕy,n∫0Ldζn∫−∞∞dηnf0z(ηn)…(57)\langle\ldots\rangle=\prod_{n=1}^{N}\int_{0}^{\infty}f_{0x}(J_{x,n})\,dJ_{x,n}\int_{0}^{2\pi}d\phi_{x,n}\int_{0}^{\infty}dJ_{y,n}\,f_{0y}(J_{y,n})\int_{0}^{2\pi}d\phi_{y,n}\int_{0}^{L}d\zeta_n\int_{-\infty}^{\infty}d\eta_n f_{0z}(\eta_n)\ldots \tag*{(57)}

Applying this averaging to fMf_{\mathrm{M}} gives the averaged distribution function, ⟨fM⟩=f0(J)\langle f_{\mathrm{M}}\rangle=f_0(J). In our calculations, we will use Eqs. (24), (25) for f0f_0. Note that the longitudinal distribution function is normalized by

∫0Ldζ∫−∞∞dηf0z(η)=L∫−∞∞dηf0z(η)=1,(58)\int_{0}^{L}d\zeta\int_{-\infty}^{\infty}d\eta f_{0z}(\eta)=L\int_{-\infty}^{\infty}d\eta f_{0z}(\eta)=1, \tag*{(58)}

where LL is the length of the bunch (in this analysis we assume a uniform longitudinal distribution of particles).

Subtracting from fMf_{\mathrm{M}} its average value gives the perturbation of the distribution function δfM\delta f_{\mathrm{M}},

δfM=fM−f0,(59)\delta f_{\mathrm{M}}=f_{\mathrm{M}}-f_0, \tag*{(59)}

where δfM\delta f_{\mathrm{M}} is small, ∣δfM∣≪f0|\delta f_{\mathrm{M}}|\ll f_0. The function δfM\delta f_{\mathrm{M}} is responsible for the density fluctuations in the beam. Being a function of random variables Jx,n,ϕx,n,Jy,n,ϕy,n,ζn,ηnJ_{x,n},\phi_{x,n},J_{y,n},\phi_{y,n},\zeta_n,\eta_n, it itself is a random variable. Integrating it over the whole phase space with the proper delta-functions that establishes a relation between particle’s coordinates x,y,zx,y,z and the action-angle variables (see Eqs. (6) and Eq. (9)), gives the density perturbation δn(x,y,z,s)\delta n(x,y,z,s),

δn(x,y,z,s)=N∫0∞dJx∫02πdϕx∫0∞dJy∫02πdϕy∫0Ldζ∫−∞∞dη δfM(Jx,ϕx,Jy,ϕy,η,ζ,s)×δ(x−2βx(s)Jxcos⁡(ψx(s)+ϕx))δ(y−2βy(s)Jycos⁡(ψy(s)+ϕy))×δ(z−ζ−snγ2).(60)\begin{aligned} \delta n(x,y,z,s)={}&N\int_{0}^{\infty}dJ_x\int_{0}^{2\pi}d\phi_x\int_{0}^{\infty}dJ_y\int_{0}^{2\pi}d\phi_y\int_{0}^{L}d\zeta\int_{-\infty}^{\infty}d\eta\,\delta f_{\mathrm{M}}(J_x,\phi_x,J_y,\phi_y,\eta,\zeta,s)\\ &\times\delta\left(x-\sqrt{2\beta_x(s)J_x}\cos(\psi_x(s)+\phi_x)\right)\delta\left(y-\sqrt{2\beta_y(s)J_y}\cos(\psi_y(s)+\phi_y)\right)\\ &\times\delta\left(z-\zeta-\frac{sn}{\gamma^2}\right). \tag*{(60)} \end{aligned}

As is δfM\delta f_{\mathrm{M}}, δn\delta n is a random function, and we are interested in calculation the averaged product ⟨δn(x,y,z,s)δn(x′,y′,z′,s′)⟩\langle\delta n(x,y,z,s)\delta n(x',y',z',s')\rangle. A straightforward although lengthy calculation that uses Eqs. (56)–(59) gives the following result:

⟨δn(x,y,z,s)δn(x′,y′,z′,s′)⟩=N×∫0∞dJx∫02πdϕxf0x(Jx)δ(x−2βx(s)Jxcos⁡[ψx(s)+ϕx])×δ(x′−2βx(s′)Jxcos⁡[ψx(s′)+ϕx])×∫0∞dJy∫02πdϕyf0y(Jy)δ(y−2βy(s)Jycos⁡[ψy(s)+ϕy])×δ(y′−2βy(s′)Jycos⁡[ψy(s′)+ϕy])×∫−∞∞dηf0z(η)δ(z−z′−(s−s′)ηγ2).(61)\begin{aligned} \langle\delta n(x,y,z,s)\delta n(x',y',z',s')\rangle={}& N \times\int_{0}^{\infty} dJ_x \int_{0}^{2\pi} d\phi_x f_{0x}(J_x)\delta\left(x-\sqrt{2\beta_x(s)}J_x\cos[\psi_x(s)+\phi_x]\right)\\ &\times\delta\left(x'-\sqrt{2\beta_x(s')}J_x\cos[\psi_x(s')+\phi_x]\right)\\ &\times\int_{0}^{\infty}dJ_y\int_{0}^{2\pi}d\phi_y f_{0y}(J_y)\delta\left(y-\sqrt{2\beta_y(s)}J_y\cos[\psi_y(s)+\phi_y]\right)\\ &\times\delta\left(y'-\sqrt{2\beta_y(s')}J_y\cos[\psi_y(s')+\phi_y]\right)\\ &\times\int_{-\infty}^{\infty}d\eta f_{0z}(\eta)\delta\left(z-z'-\frac{(s-s')\eta}{\gamma^2}\right). \tag*{(61)} \end{aligned}

The Fourier transform of the density fluctuations is defined by Eq. (21). For the correlator of the Fourier components, ⟨δn^k(s)δn^k′(s′)⟩\langle\delta\hat{n}_{k}(s)\delta\hat{n}_{k'}(s')\rangle, through which we have expressed the correlator of the electric field in Eq. (22), we have

⟨δn^k(s)δn^k′(s′)⟩=∫dx dy dz dx′ dy′ dz′ eik⋅r+ik′⋅r′⟨δn(x,y,z,s)δn(x′,y′,z′,s′)⟩.(62)\langle\delta\hat{n}_{k}(s)\delta\hat{n}_{k'}(s')\rangle= \int dx\,dy\,dz\,dx'\,dy'\,dz'\,e^{i\mathbf{k}\cdot\mathbf{r}+i\mathbf{k}'\cdot\mathbf{r}'}\langle\delta n(x,y,z,s)\delta n(x',y',z',s')\rangle. \tag*{(62)}

Substituting Eq. (61) into this equation and taking the integrals gives the final result:

⟨δn^k(s)δn^k′(s′)⟩=2πνδ(kz+kz′)exp⁡[−12ση2(kzs−s′γ2)2]×exp⁡[−12εx(kx2βx(s)+kx′2βx(s′)+2kxkx′βx(s)βx(s′)cos⁡(ψx(s)−ψx(s′)))]×exp⁡[−12εy(ky2βy(s)+ky′2βy(s′)+2kyky′βy(s)βy(s′)cos⁡(ψy(s)−ψy(s′)))].(63)\begin{aligned} \langle\delta\hat{n}_{k}(s)\delta\hat{n}_{k'}(s')\rangle={}& 2\pi\nu\delta(k_z+k'_z)\exp\left[-\frac{1}{2}\sigma_\eta^2\left(k_z\frac{s-s'}{\gamma^2}\right)^2\right]\\ &\times\exp\left[-\frac{1}{2}\varepsilon_x\left(k_x^2\beta_x(s)+k_x'^2\beta_x(s')+2k_xk_x'\sqrt{\beta_x(s)\beta_x(s')}\cos(\psi_x(s)-\psi_x(s'))\right)\right]\\ &\times\exp\left[-\frac{1}{2}\varepsilon_y\left(k_y^2\beta_y(s)+k_y'^2\beta_y(s')+2k_yk_y'\sqrt{\beta_y(s)\beta_y(s')}\cos(\psi_y(s)-\psi_y(s'))\right)\right]. \tag*{(63)} \end{aligned}

where ν=N/L\nu=N/L is the number of particles per unit length. For the correlation at one location, s=s′s=s', we obtain

⟨δn^kδn^k′⟩=2πνδ(kz+kz′)exp⁡[−12ση2(kzs−s′γ2)2]×exp⁡[−12σx2(kx+kx′)2]exp⁡[−12σy2(ky+ky′)2].(64)\begin{aligned} \langle\delta\hat{n}_{k}\delta\hat{n}_{k'}\rangle={}& 2\pi\nu\delta(k_z+k'_z)\exp\left[-\frac{1}{2}\sigma_\eta^2\left(k_z\frac{s-s'}{\gamma^2}\right)^2\right]\\ &\times\exp\left[-\frac{1}{2}\sigma_x^2(k_x+k_x')^2\right]\exp\left[-\frac{1}{2}\sigma_y^2(k_y+k_y')^2\right]. \tag*{(64)} \end{aligned}

References

  1. [3]A. Piwinski, in 9th International Conference on High-Energy Accelerators (1974) pp. 405–409.
  2. [4]J. D. Bjorken and S. K. Mtingwa, Particle Accelerators 13, 115 (1983).
  3. [5]K. Kubo and K. Oide, Phys. Rev. ST Accel. Beams 4, 124401 (2001).
  4. [6]S. Nagaitsev, Phys. Rev. ST Accel. Beams 8, 064403 (2005).
  5. [7]J. Hubbard, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences 260, 114 (1961).
  6. [8]J. Hubbard and G. P. Thomson, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences 261, 371 (1961), https://royalsocietypublishing.org/doi/pdf/10.1098/rspa.1961.0084.
  7. [9]O. Aono, Journal of the Physical Society of Japan 16, 2264 (1961).
  8. [10]Z. Huang, Intrabeam Scattering in an X-ray FEL Driver, Report LCLS-TN-02-08 (SLAC, 2002).
  9. [11]E. Prat, P. Dijkstal, E. Ferrari, A. Malyzhenkov, and S. Reiche, Phys. Rev. Accel. Beams 23, 090701 (2020).
  10. [12]S. Di Mitri, G. Perosa, A. Brynes, I. Setija, S. Spampinati, P. H. Williams, A. Wolski, E. Allaria, S. Brussaard, L. Giannessi, G. Penco, P. R. Rebernik, and M. Trovò, M, New Journal of Physics 22, 083053 (2020).
  11. [13]S. Tomin, I. Zagorodnov, W. Decking, N. Golubeva, and M. Scholz, Phys. Rev. Accel. Beams 24, 064201 (2021).
  12. [14]H. Qian, M. Krasilnikov, A. Lueangaramwong, X. Li, O. Lishilin, Z. Aboulbanine, G. Adhikari, N. Aftab, P. Boonpornprasert, G. Georgiev, J. Good, M. Gross, C. Koschitzki, R. Niemczyk, A. Oppelt, G. Shu, F. Stephan, G. Vashchenko, and T. Weilbach, Phys. Rev. Accel. Beams 25, 083401 (2022).
  13. [15]E. Prat, P. Craievich, P. Dijkstal, S. Di Mitri, E. Ferrari, T. G. Lucas, A. Malyzhenkov, G. Perosa, S. Reiche, and T. Schietinger, Phys. Rev. Accel. Beams 25, 104401 (2022).
  14. [16]V. Lebedev, AIP Conference Proceedings 773, 440 (2005).
  15. [17]G. Stupakov, in Proceedings of the 2011 FEL Conference (Shanghai, China, 2011) p. 49.
  16. [18]G. Stupakov and G. Penn, Classical Mechanics and Electromagnetism in Accelerator Physics, Graduate Texts in Physics (Springer International Publishing, 2018).
  17. [19]Y. Klimontovich, Kinetic Theory of Nonideal Gases and Nonideal Plasmas, International series in natural philosophy (Pergamon, 1982).

Paper details

vlnc:W_yz5judp5
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

Not recorded for this paper

View statement
Looking for related papers…
Contents